Environ. Sci. Technol. 2005, 39, 6470-6476
Initial Stages of Bacterial Fouling during Dead-End Microfiltration† WENDONG XU AND SHANKARARAMAN CHELLAM* Department of Civil and Environmental Engineering, University of Houston, 4800 Calhoun Road, Houston, Texas 77204-4003
Constant pressure experiments were performed using track-etched polycarbonate membranes and rod-shaped bacteria (viz., Brevundimonas diminuta and Serratia marcescens) to study flux decline and backwashing during the early stages of microfiltration. The intermediate blocking law originally derived for spherical particles was modified to account for the approximate cylindrical shape of the selected bacteria. A deposition factor was introduced to empirically account for the morphology of bacterial deposits. The initial stages of flux decline prior to the secretion of new extracellular polymeric substances (EPS) was quantitatively described by the intermediate blocking law before transitioning to cake filtration at later times. Scanning electron microscopy (SEM) provided additional visual evidence that bacteria simultaneously deposited directly on the membrane and on each other during early stages of filtration as assumed by the intermediate blocking law. Empirical deposition factors decreased with initial permeate flux indicating its effect on bacteria deposition patterns, which was also confirmed by SEM. Bacteria were easily removed following short filtration times before significant secretion of new EPS by simply rinsing with ultrapure water, thereby completely restoring the clean membrane permeability. In contrast, this rinsing procedure did not completely recover the membrane permeability following longer durations when significant amounts of new EPS proteins and polysaccharides were secreted. Consequently, backwashing effectiveness during water and wastewater microfiltration will be high prior to EPS production whereas flux recovery may not be possible solely by hydrodynamic means once EPS are secreted.
Introduction Microfiltration (MF) and ultrafiltration (UF) membranes are being increasingly implemented to purify drinking water, to treat wastewater prior to discharge, and even to reclaim wastewater to augment existing potable water supplies (1, 2). However, membrane fouling continues to hamper their implementation. Much effort has already been spent on delineating the mechanisms by which colloids and natural organic matter (NOM) cause fouling (3-6). These results have allowed the development of pretreatment methods to alleviate colloidal and NOM fouling of low-pressure membranes (MF and UF) using coagulation, adsorption, or softening (7-11). †
This paper is part of the Charles O’Melia tribute issue. * Corresponding author phone: (713)743-4265; fax: (713)743-4260; e-mail:
[email protected]. 6470
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 17, 2005
In comparison, bacterial fouling mechanisms of MF/UF membranes are not yet well-understood even though biofouling of high-pressure membranes (reverse osmosis (RO) and nanofiltration (NF)) has been well-documented (ref 12 and references therein). However, results from NF/RO biofouling studies are not directly applicable to MF/UF membranes because of fundamental differences between operation and design of these types of membranes. For example, NF/RO membranes are always operated under a cross-flow whereas environmental MF/UF systems are preferentially operated in the dead-end mode (1). Importantly, the maximum element recovery allowed for NF/RO modules is only 10-15% (in other words only 10-15% of the flow is normal to the membrane). The cross-flow velocity induced by the remaining 85-90% of the flow being tangential to the membrane is ∼3-4 orders of magnitude greater than the permeate flux and reduces the propensity for bacteria accumulation near RO/NF membrane surfaces. In contrast, all micro- and ultrafilters employed for water and wastewater treatment are preferentially operated at 100% recovery during forward filtration. Hence, all rejected materials, including bacteria, will deposit on the membrane surface during forward filtration necessitating regular backwashing (at intervals of ∼15 min to 1 h) to restore their hydraulic permeability. The effective recovery of MF/UF systems (including water employed for backwashing) is typically in the range 90-95%. However, spiral-wound NF/RO membranes are not backwashed, and their permeability is restored by taking them off-line and chemically cleaning them. During the relatively long interval between NF/RO chemical cleaning events (∼3-6 months) (13), deposited bacteria can secrete extracellular polymeric substances (EPS) to maintain their microenvironments and establish synergistic relationships in order to enhance their survival (14, 15). EPS has also been shown to permit bacteria to better adhere to and colonize NF/RO membrane surfaces (12, 16). In contrast, because MF/UF membranes are backwashed more frequently (∼15 min to 1 h), the role of EPS on their fouling is not clear. Hence, biological fouling dynamics inferred from NF/RO systems cannot be directly applied to direct flow MF/UF membranes. Bacterial fouling studies have also been performed using MF/UF membranes to determine factors controlling cell deposition during cross-flow operation (e.g., refs 17-19) or to delineate dead-end filtration fouling mechanisms including measurements of specific cake resistances (20-22). Paradoxically, to date the vast majority of bacterial MF/UF studies have not explicitly considered EPS production (17, 19, 20, 23, 24). Only a limited number of investigations have included EPS effects on MF/UF fouling (21, 22). However, these studies did not link EPS production to the efficacy of backwashing procedures. The principal objective of this study is to investigate bacterial fouling during the initial stages of dead-end microfiltration. This research also aims to provide a link between the effectiveness of backwashing and EPS production by bacteria. Constant pressure experiments were performed using track-etched polycarbonate membranes and rod-shaped bacteria (viz., Brevundimonas diminuta and Serratia marcescens). The intermediate blocking law originally derived for spherical particles was modified to account for the approximate cylindrical shape of the bacteria. Transient flux decline data and scanning electron microscopy (SEM) were used to determine the mechanisms responsible for fouling at short times prior to the secretion of new EPS. 10.1021/es0500862 CCC: $30.25
2005 American Chemical Society Published on Web 05/19/2005
Theoretical Section Blocking laws (25) provide a convenient framework to quantitatively interpret fouling during dead-end microfiltration of bacteria prior to the secretion of significant amounts of EPS. During constant pressure unstirred filtration, the time (t) dependent behavior of cumulative filtrate volume (V) is related as:
d2t dt )k dV dV2
( )
n
(1)
Note that k and n are parameters that depend on the properties of the suspension and membrane characteristics as well as operational parameters. The exponent n characterizes the filtration mechanism, with n ) 0 for cake filtration, n ) 1 for intermediate blocking, n ) 1.5 for standard blocking, and n ) 2 for complete blocking. Blocking laws or their modifications have been used to interpret membrane fouling caused by natural organic matter (26, 27), colloids (28), and proteins (29). Only the intermediate blocking law will be discussed in this paper because it accurately quantified the very early stages of flux decline prior to substantial quantities of EPS production in all our experiments. This fouling model assumes that particles have an equal probability to deposit on previously deposited particles (which are already blocking membrane pores) or on the membrane surface itself and directly block pores. Under this assumption, it has been shown that (25)
Q)
Q0 1 + (σ∆P/µR)t
(2)
where Q is the instantaneous flow rate ()dV/dt), Q0 is the initial flow rate, σ is the membrane surface area blocked after filtration of a unit volume of the feed suspension, ∆P is the transmembrane pressure, µ is the absolute viscosity, and R is the clean membrane resistance. Rearranging and differentiating eq 2 with respect to V to appear in the same form as eq 1:
d 2t σ∆P dt ) dV2 µRQ0 dV
(3)
Cylindrical or rod-shaped bacteria of diameter d and length h appearing near the filter surface will effectively block an area ∼hd/Ψ of the membrane, where Ψ is a deposition factor that empirically accounts for the morphology of the bacterial deposit. Using Hermia’s approach (25), σ can be represented as 4γss/πγ0Ψd. Recognizing ∆PA0/µR as the initial flow rate, eq 3 becomes
4γss d2t dt ) 2 πγ0 dA0Ψ dV dV
(4)
where A0 is the initial membrane area, γs is the feed suspension density, γ0 is the bacteria density, and s is the mass fraction of bacteria in the feed. The deposition factor Ψ is the only unknown in eq 4 and was estimated by using it as an adjustable parameter to fit eq 4 to measured instantaneous flux decline data using experimental values of all other parameters. The value of Ψ obtained by this procedure represents an effective deposition factor that empirically describes the morphology of the entire deposit because all bacteria cannot be expected to accumulate in the same manner or geometry on the membrane surface.
Experimental Section Microorganisms. Brevundimonas (formerly Pseudomonas) diminuta and Serratia marcescens were used as model
bacteria primarily based on the American Society of Testing and Materials (ASTM) protocol, which is universally accepted by regulatory agencies to challenge 0.20 and 0.45 µm membranes, respectively. B. diminuta (American Type Culture Collection (ATCC) 19146) and S. marcescens (ATCC 14756) were grown according to ATCC and ASTM recommendations. Briefly, B. diminuta and S. marcescens were grown in batch culture using tryptic soy broth at 30 and 26 °C, respectively. Bacteria were harvested in stationary phase (30) by centrifugation at 5000g for 15 min at 4 °C and washed twice by phosphate-buffered saline solution (PBS, pH ) 7.3, 137 mM NaCl, 2.7 mM KCl, 4.3 mM Na2HPO4‚7H2O, 1.4 mM KH2PO4). Bacterial stock suspensions for filtration were prepared by resuspending in PBS at a concentration ∼4 × 108 cells/mL corresponding to an optical density at 600 nm of 0.56. Digital image analysis (Scion Image, Frederick, MD) of bacteria deposited on membranes revealed that both are rod-shaped with B. diminuta being 0.39 ( 0.03 µm in diameter and 1.54 ( 0.27 µm in length, whereas S. marcescens was 0.65 ( 0.10 µm in diameter and 2.21 ( 0.43 µm in length. (The value before the ( sign is the average, and the value after it is the standard deviation of at least 45 cell images.) Membranes. Track-etched polycarbonate membrane filters (Isopore Millipore Corp., Bedford, MA) with pore sizes of 0.2, 0.4, and 0.6 µm were employed as model microfilters. These hydrophilic membranes possess nearly cylindrical pores of very narrow pore size distributions facilitating fundamental investigations of blocking phenomena. Filtration Procedure. All experiments were performed at room temperature in the unstirred dead-end mode using commercially available cells accommodating 4.1 cm2 of effective membrane area (model 8010, Millipore Corporation, Bedford, MA). In most cases, parallel experiments were performed using two cells to increase the precision and validity of our experimental data. The pressure and cumulative mass of filtered water were monitored at a rate of 1 Hz using a computerized data acquisition system (LabVIEW, National Instruments, Austin, TX). Before filtering bacterial suspensions, membranes were fully wetted by first passing 500 mL of PBS through them in the pressure range of 7-172 kPa. During this time, the clean membrane resistance was also calculated by measuring flux at a minimum of three transmembrane pressures. More details of the filtration apparatus and protocols can be found in ref 31. In each experiment, ∼500 mL of bacterial suspension with concentration ∼2 × 106 cells/mL was filtered. For selected experiments, backwashing was simulated by carefully rinsing the membrane using a small volume (10 mL) of ultrapure water. Shorter backwashing intervals (15 min, 30 min, and 1 h) were obtained by terminating filtration at the corresponding time. To make a stronger correlation between EPS production and backwash efficacy, longer intervals (2.5-48 h) were also evaluated. These extended backwashing intervals were simulated by simply storing the membrane along with deposited bacteria in a covered Petri dish for the requisite duration. During storage, these samples were kept hydrated by periodically adding a few drops of ultrapure water. The membrane resistance after backwashing was also calculated by three measurements of flux in the pressure range of 7-172 kPa using PBS. EPS Extraction and Measurement. After backwashing, EPS was extracted using methods similar to those reported in the literature (e.g., refs 15, 32, and 33). The collected cell solution was harvested by centrifugation (5000g) for 15 min at 4 °C. The supernatant yielded the soluble or “free” EPS fraction. The pellets were resuspended in 2% NaCl solution (at pH 7.0) and extracted by shaking for 1 h at 30 °C in an incubator at 100 rpm. The “bound” EPS fraction was subsequently recovered from the supernatant by centrifugaVOL. 39, NO. 17, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
6471
FIGURE 1. Flux decline profiles for B. diminuta (a) and S. marcescens (b) at different initial fluxes. tion (5000g) also for 15 min at 4 °C. Finally, both supernatants were centrifuged again at higher speed (10000g) for 15 min at 4 °C to remove the residual cells. Polysaccharides were assayed by the phenol-sulfuric acid method, wherein 1 mL of phenol (5%, w/v) and 2.5 mL of concentrated H2SO4 were first added to 0.5 mL of the sample and mixed thoroughly. Samples were left to stand for 10 min at room temperature and then heated to 30 °C for 20 min in a water bath, which produced a yellow-brown color proportional in intensity to concentration. Polysaccharide concentration was measured against glucose standards using absorbance at 490 nm. Proteins were assayed by the Pierce Modified Lowry method (33) using a commercially available kit (Pierce Biotechnology, Rockford, IL, PI-23240) by producing a water-soluble tetradentate copper complex with characteristic blue color. Bovine serum albumin was used as calibration standards, and protein concentrations were measured using absorbance at 750 nm. Electron Microscopy. Selected experiments were also designed to visualize the morphology of bacterial deposits on membrane surfaces following filtration of different volumes of suspension. To better preserve the deposit structure, the experiment was terminated once the desired volume was filtered, and no visible liquid remained above the membrane. Samples were prepared using the method suggested in ref 34 before microscopic examination. First, the bacterial deposit was fixed on the membrane using 2.5% glutaraldehyde solution in ultrapure water for 1 h. The fixed sample was then sequentially dehydrated using increasing concentrations of ethanol (viz., 50, 75, 90, and 95% (v/v)) for 20 min each and finally 100% ethanol for 30 min. The specimens were further air-dried and observed using a field emission scanning electron microscope (JSM-6330F, JEOL, Peabody, MA) after sputter coating with a thin (10 nm) layer of gold.
Results and Discussion A general characteristic of microorganisms that differentiates them from all other foulants is their ability to produce EPS. The early stage of bacterial fouling is analyzed using blocking law in this paper because we explicitly considered very short filtration times during which interval B. diminuta and S. marcescens did not produce significant amounts of extracellular proteins and polysaccharides (see EPS Production and Backwashing Efficiency section and Figure 9). However, it should be noted that protein and polysaccharides were detected even in the bacterial feed suspension. This constitutes the EPS that are tightly bound to the cells and could not be removed by centrifugation while preparing the feed 6472
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 17, 2005
suspension (15, 33). In other words, similar to other researchers (15, 32, 33), our centrifugation and washing procedure for preparing bacterial feed suspension was only capable of removing loosely bound and soluble EPS produced during the incubation period. The bacteria sizes reported earlier in the Experimental Section, all blocking law calculations, and electron micrographs include this tightly bound layer. Evaluation of the Intermediate Blocking Law for the Initial Stages of Fouling. Representative flux decline profiles induced by B. diminuta and S. marcescens are shown in Figure 1a,b, respectively. Close agreement between eq 2 (solid lines) and the experimental data (discrete points) indicates that bacteria fouled membranes in a manner consistent with the intermediate blocking law. Note that transient changes in the flow across membranes at constant pressure were measured in the laboratory (Figure 1). However, we need the first and second derivatives of the time with respect to the cumulative volume in order to compare experimental results with eq 1. The first derivative (dt/dV) was obtained by numerical differentiation of exponentially smoothed mass data (damping factor 0.3). However, this procedure did not reduce the experimental noise sufficiently to obtain accurate estimates of the instantaneous second derivative (d2t/dV2). Therefore, two piecewise continuous interpolation polynomials of order 4-7 were obtained for the smoothed first derivative (29) ensuring an accurate fit over the entire duration of each experiment. Finally, a combination of smoothed and interpolated data was numerically differentiated using forward difference to obtain the second derivative. Note that experimental flux measurements transformed in this manner allow a direct evaluation of eq 1. Laboratory flux data transformed using this procedure for three flux profiles in Figure 1 are shown in Figures 2 and 3 for B. diminuta and S. marcescens, respectively. In all cases the initial portion of the curve (corresponding to a maximum of ∼1 h filtration time) increased in a straight line fashion with a slope very near the theoretical value of 1 (n ) 1 in eq 1) demonstrating that the early stages of flux decline induced by bacterial deposition quantitatively corresponded to the intermediate blocking law. As observed, the slope changed at later times for the higher flux conditions (when sufficient bacteria had deposited), indicating a change in the fouling mechanism, eventually reaching a zero value. The horizontal behavior corresponds to n ) 0 in eq 1 demonstrating that flux decline follows the cake filtration mechanism toward the later stages of microfiltration.
FIGURE 2. Evaluation of the intermediate blocking law for B. diminuta microfiltration.
FIGURE 3. Evaluation of the intermediate blocking law for S. marcescens microfiltration.
FIGURE 4. Progressive changes in the morphology of B. diminuta deposits on 0.4 µm membranes with increasing volume of suspension filtered at 138 kPa at an initial flux of 0.0071 m/s (
