Environ. Sci. Technol. 2006, 40, 6812-6817
Three-Component Competitive Adsorption Model for Fixed-Bed and Moving-Bed Granular Activated Carbon Adsorbers. Part II. Model Parameterization and Verification LANCE C. SCHIDEMAN, BENITO J. MARIN ˜ AS, AND VERNON L. SNOEYINK* Department of Civil & Environmental Engineering and Center of Advanced Materials for the Purification of Water with Systems, University of Illinois, Urbana, Illinois 61801 CARLOS CAMPOS Suez Environment, 75009 Paris, France
COMPSORB-GAC is a 3-component competitive adsorption kinetic model for granular activated carbon (GAC) adsorbers that was developed in Part I of this study, including a proposed procedure for determining model parameters in natural water applications with background natural organic matter (NOM). Part II of this study demonstrates the proposed parameterization procedure and validates the modeling approach by comparing predictions with experimental breakthrough curves at multiple emptybed contact times for both fixed-bed and moving-bed reactors. The parameterization procedure consists of a set of independent, short-term experimental tests with fresh and batch preloaded adsorbents and then data fitting using both classic and recently developed theoretical expressions. The model and parameterization procedure simplifies NOM into two fictive fractions (pore-blocking and strongly competing) and incorporates three competitive effects that vary both temporally and axially in a GAC column (direct competition for sites, intraparticle pore blockage, and external surface pore blockage). With all three competitive mechanisms accounted for, the model could accurately predict breakthrough profiles for column lengths and durations that were much longer than those used for model parameterization. Model predictions that ignored one or more of the competitive mechanisms showed that each mechanism was important for different regions of the breakthrough curve. The external surface pore-blockage effect was predominant for the prediction of early breakthrough data, whereas direct competition for sites and intraparticle pore blockage were prevalent when predicting higher breakthrough levels and data later in the column run.
Introduction Granular activated carbon (GAC) adsorption is an important water purification process that provides effective removal of * Corresponding author phone: +1 217 333 4700; fax: 1 217 333 6968; e-mail:
[email protected]. 6812
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 21, 2006
many trace organic contaminants but is also highly susceptible to competitive effects from background natural organic matter (NOM). NOM can significantly reduce both the equilibrium capacity and the kinetics of trace compound adsorption because it is present at relatively high concentrations and has slow adsorption kinetics (1-5). These characteristics usually cause some GAC to be exposed to NOM before adsorbing any trace contaminants, which reduces GAC effectiveness and complicates adsorption modeling. A strategy for minimizing preloading effects on trace compound adsorption is the use of moving-bed systems, where fresh adsorbent is added in small layers only when needed to contain the trace compound mass-transfer zone. Predictive models that can accurately describe competitive adsorption behavior in such systems are urgently needed to quantify the benefits and optimize design and operating procedures. Recent work on predictive GAC kinetic models has focused on using variable isotherm and intraparticle diffusion parameters that are empirical functions of preloading time to reflect the competitive effects of NOM (3, 6-8). However, this approach has some key limitations because spatial variations of preloading effects are ignored, and the empirical parameter determination tests generally need to have long durations for the model to remain calibrated. An alternative approach, referred to as 3-component adsorption modeling, recently developed for powdered activated carbon-ultrafiltration systems (1, 9, 10), is based on making adsorption parameters empirical functions of NOM surface loading. Such an approach provides a way to address the key limitations of GAC modeling noted above. Part I of this study developed the 3-component COMPSORB-GAC adsorption modeling approach that integrated several new features, including variable external film masstransfer coefficient as a function of NOM surface coverage, spatially varying equilibrium and kinetic parameters, and simulation of both fixed- and moving-bed GAC reactor performance. The specific objectives of the Part II are (i) to demonstrate a procedure for determining the COMPSORBGAC model parameters, (ii) to experimentally verify the model and parameterization procedure, and (iii) to evaluate the relative importance of the different competitive adsorption mechanisms of NOM.
Materials and Methods Part I of this study provides a description of the water sources, trace organic compound, analytical methods, and short-bed adsorber tests that were also used in Part II of the study. Additional details of the experimental materials and methods are provided below. Adsorbent. U.S. Standard 30/40 mesh fraction (515 µm mean diameter) of F-400 GAC (Calgon Carbon Corp., Pittsburgh, PA) was sieved out for column experiments. The bulk or apparent bed density of the GAC was calculated to be 0.50 g/cm3 on the basis of measurements of bed volume and carbon mass. To reduce the time needed for isotherm and batch kinetic tests, a pulverized GAC (PGAC) was obtained by crushing F-400 GAC with a concentric ring mechanical grinder until 99% passed a size 200 sieve. The average particle size of the resulting PGAC was found to be 10 µm from observations with a light microscope. Other characteristics of interest that were available from the manufacturer are a particle density of 0.76 g/cm3, a skeletal density of 2.3 g/cm3, and a BET surface area of 1110 m2/g. Isotherm Tests. To determine equilibrium adsorption capacity, the bottle-point isotherm technique (11) was used for both organic-free water (OFW) and a Central Illinois 10.1021/es060603w CCC: $33.50
2006 American Chemical Society Published on Web 09/22/2006
groundwater taken from a well below the Newmark Civil Engineering Laboratory (NCEL) with different initial atrazine concentrations. To develop an isotherm, the initial atrazine concentration was held constant, and the mass of PGAC was varied in a series of amber glass bottles that were mixed on a shaker table for 7 days, which was determined on the basis of homogeneous surface diffusion model (HSDM) simulations of equilibration time. Samples of each bottle were filtered through a 0.45 µm nylon membrane filter to remove the PGAC particles, and these filters were rinsed with OFW and sample water to avoid leaching or adsorption of organics. Samples were analyzed for either dissolved organic carbon (DOC) or atrazine to determine the equilibrium aqueous concentrations, and the solid-phase surface loadings were calculated by mass balance. Isotherm tests were conducted at room temperature, but all kinetic tests and the multiport column test described later were performed at 4 °C in a cold room. Sensitivity testing was used to verify that any discrepancy in the equilibrium isotherm capacity that resulted from the temperature differences had a negligible effect on the COMPSORB-GAC model predictions. Batch Kinetic Tests. Batch kinetic tests were carried out with both fresh and preloaded PGAC in NCEL water to define the effect of NOM on intraparticle diffusion rates. The test solution of interest and a known mass of prewetted PGAC were mixed vigorously (200 rpm) in a jar testing apparatus (Phipps & Bird, Richmond, Virginia). Samples were taken regularly, filtered, and analyzed for DOC or atrazine as needed. For preloaded batch kinetic tests, PGAC was first mixed with NCEL water for 5 days prior to the addition of atrazine. Bench-Scale Multiport GAC Column Test. A multiport GAC column was used to validate the proposed modeling approach, and it consisted of a series of glass chromatography column sections, each with a 1000 mg layer of GAC and a sampling valve that connected adjacent sections. The total GAC mass was 4000 mg, and the flow rate was 4.0 mL/min, which resulted in a total empty-bed contact time (EBCT) of 2.0 min and sampling points at each 0.5 min of EBCT. The multiport column was operated as a fixed-bed adsorber for 1400 h, and then operation continued until 3500 h in a moving-bed mode. This was accomplished by removing column sections with partially spent GAC layers from the reactor inlet and simultaneously adding new sections with fresh GAC layers at the outlet. Half of the total GAC (2000 mg) was replaced at 1400 and 2100 h, and a quarter of the total GAC (1000 mg) was replaced at 2800 h and 3150 h to maintain a constant carbon usage rate.
Results and Discussion Equilibrium Parameter Determination: K, 1/n, and C0. The COMPSORB-GAC model requires the Freundlich isotherm parameters and initial concentration for each of the three components to calculate equilibrium conditions and direct competition for adsorption sites. Figure 1 presents the singlesolute isotherm for the TR compound (atrazine), F-400 PGAC, and OFW, as well as three natural water isotherms for different initial atrazine concentrations, F-400 PGAC, and NCEL water. The equilibrium solid-phase concentrations, qe, were determined by the difference in liquid-phase concentrations as follows
qe )
C0 - Ce Cc
(1)
where C0 and Ce are the TR compound initial and equilibrium liquid-phase concentrations and Cc is the carbon concentration. The Freundlich equation parameter fits are presented
FIGURE 1. Atrazine isotherms in OFW and NCEL groundwater with F-400 PGAC in Figure 1 for the single-solute isotherm data and one natural water isotherm (C0 ) 9.9 µg/L). The single-solute isotherm provides the values of KTR and 1/nTR used in COMPSORBGAC. The natural water Freundlich isotherm parameters, although not needed for the 3-component model input, were used for some conventional HSDM kinetic parameter fittings in NCEL water experiments as described later. C0,TR was selected corresponding to the initial atrazine concentration used in GAC column tests (9.9 µg/L). Figure 1 also shows that NOM competition significantly reduced atrazine adsorption capacity, and the reduction was more severe at lower initial atrazine concentrations. However, it is not necessary to change the trace compound Freundlich parameters to model different initial concentrations with COMPSORB-GAC. Instead, the effect of initial concentration is accounted for by using the simplified ideal adsorbed solution theory (IAST) of eqs 9 and 10 from Part I to adjust the single-solute equilibrium relationship. The competitive TR compound isotherms in natural water were also used to quantify the SC compound equilibrium parameters needed for the 3-component model following the procedure of Ding et al. (10); this procedure begins with the simplifying assumption that the SC and TR compounds have the same molecular weight and singlesolute isotherm parameters. Then, the initial SC compound concentration, C0,SC, was determined using the equivalent background compound (EBC) approach (12, 13) and equating the EBC to the SC compound. With this approach, the IAST (eqs 7 and 8 in Part I) and eq 1 are used to search for the C0,SC value providing the best fit of the natural water isotherm data. With the assumptions noted above, the EBCIAST approach yielded natural water isotherms for the TR compound that were parallel to the single-solute OFW isotherm but with a lower capacity as shown by the dashed lines in Figure 1. C0,SC was determined to be 175 µg/L by fitting the isotherm with C0,TR ) 9.9 µg/L. Figure 1 also shows predictions of the other two natural water isotherms by the IAST with the previously determined SC parameters, which confirmed a reasonably good representation of the competitive adsorption equilibrium data. Table 1 summarizes the parameters determined for the SC fraction, including C0,SC also expressed as DOC by assuming that carbon accounts for 50% of the molecular weight, which is a reasonable approximation for most NOM sources (14). Note that any error introduced by assuming similarities between the TR and SC compound isotherm parameters becomes calibrated when C0,SC is fitted to the natural water isotherm data. Thus, the net impact on model predictions is modest, and predictive ability is not significantly hampered, as will be shown later. With the SC fraction equilibrium parameters determined, the PB fraction parameters could be calculated using the VOL. 40, NO. 21, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
6813
TABLE 1. Parameters for 3-component GAC Modeling with NCEL Water and F-400 GACa
K (µg/mg)(L/µg)(1/n) 1/n C0 (µg/L) kf,0 (cm/s) Ds,0 (cm2/s) kxf,min R (g/mg) qcr (mg/g) β (g/mg) a
TR compound atrazine
SC fraction of NOM
PB fraction of NOM
18.6 0.393 9.9 1.66 × 10-3 6.93 × 10-13 0.250 0.210 1.01 0.0244
18.6 0.393 175 (87.5 as C) 1.66 × 10-3 6.93 × 10-13 0.250 0.210 1.01 0.0244
1.65 × 10-4 1.77 2213 (as C) 1.58 × 10-3 5.20 × 10-13 n/a n/a n/a n/a
n/a ) not applicable.
FIGURE 3. Atrazine breakthrough in virgin short-bed adsorbers with conventional HSDM parameter fits and predictions
FIGURE 2. NOM adsorption isotherms for NCEL water and F-400 PGAC total NOM isotherm and the following mass balance equations (10)
qPB ) qNOM - qSC
(2)
CPB ) CNOM - CSC
(3)
First, the initial concentration of the PB fraction, C0,PB, was calculated by subtracting C0,SC from the total initial NOM concentration, CNOM, with all concentrations expressed as DOC (eq 3). Similarly, the PB isotherm data were calculated with eqs 2 and 3 by subtracting the SC contribution, which can be calculated for each data point using the previously determined SC parameters and eq 1. The total NOM, PB, and SC isotherms expressed in terms of DOC and the corresponding Freundlich equation fits are shown in Figure 2. The PB compound equilibrium parameters come from the isotherm curve furthest to the right in Figure 2 and are listed in Table 1. From this figure, it is readily apparent that the SC compound accounts for a very small fraction of the total NOM, and the PB isotherm is very similar to the total NOM isotherm. This is consistent with previous findings for another natural water and carbon by Ding et al. (10). Model sensitivity testing confirmed that any errors in the PB compound parameters from the SC parameter assumptions have a negligible impact on COMPSORB-GAC predictions. Initial Kinetic Parameter Determination: kf,0 and Ds,0. The 3-component GAC model requires initial values for the film mass-transfer coefficient, kf,0, and the intraparticle surface diffusion coefficient, Ds,0, for each of the three components (TR, SC, and PB). Both these coefficients can be determined using short bed adsorber (SBA) tests (15-17). However, Part I of this study showed that the kf values for the TR compound declined significantly during the period of a typical SBA test. This makes conventional constant parameter-fitting techniques difficult to use because parameter fits can be dependent on the time range of data 6814
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 21, 2006
chosen for analysis (16, 18). The difficulty associated with rapidly declining kf values was avoided by fitting kf on the basis of only the earliest breakthrough data before kf erodes and then fitting Ds on the basis of only the relatively late data after the influence of kf becomes negligible. This approach is illustrated in Figure 3, in which the initial breakthrough data were fit to determine the initial film mass-transfer coefficient, kf,0,TR (Table 1). Then Ds,0,TR was determined by fitting the data from 150 to 250 h, when the breakthrough profile was essentially independent of kf but before the intraparticle pore blockage caused a significant reduction in surface diffusion rates. Independence from the influence of kf is demonstrated by convergence of the solid and dashed lines in Figure 3 for conventional HSDM predictions with significantly different kf values. Minor intraparticle poreblockage effects were confirmed using the intraparticle poreblockage parameters developed later. A kinetic test with OFW might seem to be a preferable way to determine Ds,0 without complications from NOM competitive effects. However, several studies including this one have observed that OFW kinetic tests yielded slightly lower surface diffusion coefficients than natural water kinetic tests with the same adsorbate and adsorbent (10, 15, 19). This counterintuitive result can be explained by understanding that Ds,0 is an effective surface diffusion coefficient that is somewhat dependent on the other HSDM parameters because of deviations from the idealized conditions assumed in the HSDM. In this case, the much higher equilibrium capacity associated with OFW kinetic tests resulted in a lower effective surface diffusion coefficient. Thus, obtaining Ds,0 from an SBA test in natural water as described above was the preferred approach. On the basis of the simplifying assumption that the SC and TR compounds are similar in size, the kinetic parameters for the SC fraction (Table 1) were assumed to be the same as those determined for the TR compound. To obtain the kinetic parameters for the PB fraction, the SBA technique was applied to DOC breakthrough data. Because of the larger mass-transfer zone for DOC in comparison to atrazine, separate SBA tests with larger EBCTs of at least 1 min were needed. Otherwise, the effluent DOC concentration would approach the influent concentration too quickly to obtain good parameter fits. The breakthrough of the SC compound at these larger EBCTs was found to be negligible on the basis of conventional HSDM modeling with the previously determined SC parameters. Thus, the entire effluent DOC was attributed to the PB fraction of NOM. Figure 4 presents the PB compound breakthrough data along with a conventional HSDM fit of the PB kinetic parameters at 1 min of EBCT (Table 1), and the resulting prediction at 2 min of EBCT with the same parameters, which agreed fairly well with experimental data.
FIGURE 4. PB fraction of NOM breakthrough in virgin short-bed adsorbers with conventional HSDM parameter fits and predictions
FIGURE 6. Atrazine batch kinetic tests with various PGAC preloading conditions
TABLE 2. R2 Values for COMPSORB-GAC Model Predictions of Experimental Data Shown in Figures 7-10 with Various NOM Competitive Mechanisms Included figure number and NOM competitive mechanisms modeled
FIGURE 5. HSDM kinetic parameters as a function of NOM surface loading (PB fraction) External Surface Pore-Blockage Effect Parameter Determination: kxf,min and r. Once the initial kinetic parameters for all 3 components were defined, the next step was to quantify the surface-blockage effect of the PB compound on kf values for the other components. To do this, the data from preloaded SBA tests presented in Figure 1 of Part I were analyzed as follows. First, the surface loading of total NOM was calculated on the basis of the DOC removed during preloading. Second, the surface loading of the SC fraction was estimated using a conventional HSDM for batch reactors and the SC parameters determined earlier. Third, the PB fraction surface loading, qPB, was calculated by subtraction using eq 2. Finally, the kf values were determined on the basis of the fitting of the initial atrazine breakthrough data with a conventional HSDM, and these values were normalized by dividing by kf,0. The left curve of Figure 5 shows the relationship between qPB and kf/kf,0, which was fitted using the empirical equation proposed in accompanying manuscript (eq 5 in Part I) to determine the two surface blockage parameters, R and kxf,min, listed in Table 1. Intraparticle Pore-Blockage Parameter Determination: qcr and β. To quantify the intraparticle pore-blockage effect, batch kinetic tests with preloaded PGAC were performed following the procedure described by Ding et al. (10). PGAC was used to achieve high-NOM surface loadings in short preloading times because past studies showed that Ds did not reach a plateau value even at high-NOM surface loadings (10, 20). Figure 6 presents the atrazine adsorption kinetic curves obtained for various doses of PGAC preloaded with NCEL water and one kinetic curve with non-preloaded carbon. The conventional HSDM model fit of each kinetic curve revealed that NOM preloading decreased Ds for atrazine. The NOM surface loading for each kinetic test was obtained by measuring DOC removal during preloading, and
Figure 7 direct site competition only Figure 8 direct site competition and intraparticle pore blockage Figure 9 direct site competition, intraparticle pore blockage, and external surface pore blockage Figure 10 direct site competition, intraparticle pore blockage, and external surface pore blockage (moving-bed configuration)
0.5 min 1.0 min 1.5 min 2.0 min EBCT EBCT EBCT EBCT 0.89
0.79
-0.04 -1.09
0.85
0.83
0.76 -0.45
0.87
0.87
0.95
0.96
0.88
0.87
0.90
0.89
the SC fraction surface loading was calculated assuming that equilibrium was achieved during preloading. Then qPB was calculated by subtraction using eq 2. The relationship between qPB and the normalized surface diffusion coefficient, Ds/Ds,0, is also plotted in Figure 5. These data were fit using the empirical equation presented in the accompanying manuscript (eq 6 in Part I) to obtain the intraparticle poreblockage parameters, qcr and β, listed in Table 1. A comparison of the surface and intraparticle pore-blockage effects shown in Figure 5 reveals that intraparticle pore blockage develops slower initially, but it eventually becomes dominant at higher PB surface loading. Previous research showed that predominantly microporous adsorbents, like F-400, are more susceptible to intraparticle pore-blockage effects than those with a greater proportion of mesopores (20). Model Verification: Fixed-Bed Reactor. Atrazine breakthrough data at four different EBCTs (0.5-2.0 min) were obtained for a fixed-bed GAC column configuration operated for 1400 h. These data sets were compared to model predictions in Figures 7-9 for verification of the modeling/ parameterization approach and to compare the different competitive effects of NOM. The model was first run by only incorporating the competitive effect of direct competition for adsorption sites. The resulting predictions are compared to the experimental breakthrough data in Figure 7, and this revealed that ignoring the intraparticle and external poreblockage effects significantly underestimates the breakthrough at early run times for all EBCTs and that the degree of underestimation becomes more pronounced as EBCT increases. At an EBCT of 2 min, the maximum predicted breakthrough was 10 times lower than that of the experimental data. This trend of declining predictive ability with increasing EBCT is consistent with deeper bed sections being VOL. 40, NO. 21, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
6815
FIGURE 7. Fixed-bed GAC adsorber breakthrough data vs model predictions with direct site competition effect only
FIGURE 8. Fixed-bed GAC adsorber breakthrough data vs model predictions with direct site competition and intraparticle poreblockage effects
FIGURE 9. Fixed-bed GAC adsorber breakthrough data vs predictions with direct site competition, intraparticle pore-blockage effects, and external surface pore-blockage effects preloaded for a longer time before exposure to the trace compound. The R2 values, comparing the observed data and predicted breakthrough curves, are provided in Table 2, and they confirmed that the predictive ability declined sharply at higher EBCTs. Note that negative R2 values indicate that the predictive ability is worse than merely modeling the data as the mean of all samples. Predictions by a second model simulation taking into account both direct competition for sites and intraparticle pore-blockage effects, but still ignoring external surface poreblockage, are compared to experimental results in Figure 8. Accounting for intraparticle pore blockage effects increased the breakthrough predictions at relatively long run times, but early run time predictions were unchanged and still significantly below the measured data for all EBCTs. Break6816
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 21, 2006
FIGURE 10. Moving-bed GAC adsorber breakthrough data vs predictions with direct site competition, intraparticle pore-blockage effects, and external surface pore-blockage effects through profiles at 1.5 and 2.0 min EBCT were underpredicted at all times, and a slight overprediction of breakthrough occurred for the two shorter EBCTs at later run times. This suggests that GAC exposed to trace compound at the outset might experience slightly less-potent NOM preloading effects. The R2 values provided in Table 2 confirmed that including the intraparticle pore-blockage effects generally improved model agreement with experimental data, but it was still inadequate at larger EBCTs with the 2.0 min EBCT still having a negative R2 value. The model was subsequently modified to simulate all three of the proposed competitive mechanisms: direct site competition, intraparticle pore blockage, and external surface pore blockage. As shown in Figure 9 and Table 2, including these three NOM competitive effects resulted in adequate predictions for all EBCTs with R2 values ranging from 0.87 to 0.96. The addition of the external surface pore blockage was particularly important for predicting early breakthrough levels. Notice that the relatively modest overprediction at higher breakthrough levels for relatively short EBCTs was not exacerbated by incorporating surface-blockage effects because film resistance becomes negligible at higher surface loadings. Recognizing the high degree of simplification embedded in the 3-component GAC model and its parameters to characterize the complex competitive effects of heterogeneous NOM, the accuracy of the predictions in Figure 9 is quite remarkable. Model Verification: Moving-Bed Reactor with CounterCurrent Adsorbent Flow. The model with all three competitive mechanisms was also applied to verify the ability to simulate counter-current adsorbent flow. Figure 10 presents the model predictions in comparison to experimental data at four EBCTs during counter-current adsorbent flow. The sharp drops in the breakthrough profiles at approximately 1400 and 2100 h correspond to replacing half of the GAC bed, and the sharp drops at approximately 2800 and 3150 h correspond to replacing one-quarter of the GAC bed. Once again, the predictions agree well with the experimental data and accurately reflect the complex features of the breakthrough curves, which was confirmed by the R2 values in Table 2 that range from 0.87 to 0.90. Figures 9 and 10 demonstrate that the 3-component COMPSORB-GAC modeling approach and parameterization procedure developed in this study provides an accurate tool to support rational design and optimization efforts for a variety of GAC adsorber configurations. Summary and Future Research Recommendations. When the three mechanisms of direct competition for sites were considered, intraparticle pore blockage and external surface pore blockage were sufficient to accurately model the complex competitive effects of NOM in a GAC system.
The surface-blockage effect, ignored in most previous GAC kinetic models, should be considered when the maximum breakthrough goal is low because this effect is most pronounced at those times. Confidence in the model is bolstered by the fact that all model parameters were determined in relatively short-term, independent tests and did not need to be calibrated with long-term GAC column data. This is possible because NOM competitive effects were related to NOM surface loading rather than preloading time as used in several previous GAC modeling studies. In addition, the fact that NOM preloading and kinetic parameters can vary axially is a distinct advantage of the COMPSORB-GAC approach that increases model generality for predicting untested reactor conditions. Overall, the proposed 3-component COMPSORBGAC model and parameter determination approach provide a reasonable balance between the need for independent, accurate performance prediction and the need for model simplicity that can facilitate practical applications. Potential improvements in the 3-component GAC modeling approach might include the following: a more independent parameter determination for the SC fraction of NOM, simplifications to reduce the number of parameterization experiments, verification with different combinations of adsorbate and adsorbent, and development of a method to estimate 3-component model parameters for different adsorbates on the basis of reference compounds, as identified in previous studies (21, 22). In addition, a better fundamental understanding of the effects and interactions of trace compound size, NOM molecular size distribution, carbon pore size distribution, and the chemical characteristics of these substances is expected to yield more efficient adsorption processes and better prediction of complex competitive adsorption behavior.
Acknowledgments The authors thank S. Qi, L. Ding, and Q. Li for many fruitful discussions. The research presented in Parts I and II of this study was supported by the American Water Works Association Abel Wolman Fellowship, Suez Environment, the USEPA Science to Achieve Results Fellowship, and the WaterCAMPWS, a Science and Technology Center of Advanced Materials for the Purification of Water with Systems under National Science Foundation agreement number CTS0120978. The opinions in this paper do not necessarily reflect those of the sponsors.
(6)
(7)
(8)
(9)
(10)
(11) (12)
(13)
(14)
(15)
(16)
(17)
(18)
(19)
(20)
Literature Cited (1) Li, Q. L.; Marinas, B. J.; Snoeyink, V. L.; Campos, C. ThreeComponent Competitive Adsorption Model for Flow-through PAC Systems. 1. Model Development and Verification with a PAC/Membrane System. Environ. Sci. Technol. 2003, 37 (13), 2997-3004. (2) Speth, T. F., Evaluating Capacities of GAC Preloaded with Natural Water. J. Environ. Eng. 1991, 117 (1), 66-79. (3) Carter, M. C.; Weber, W. J. Modeling Adsorption of TCE by Activated Carbon Preloaded by Background Organic Matter. Environ. Sci. Technol. 1994, 28 (4), 614-623. (4) Hand, D. W.; Crittenden, J. C.; Arora, H.; Miller, J. M.; Lykins, B. W. Designing Fixed-Bed Adsorbers to Remove Mixtures of Organics. J. Am. Water Works Assoc. 1989, 81 (1), 67-77. (5) Summers, R. S.; Haist, B.; Koehler, J.; Ritz, J.; Zimmer, G.; Sontheimer, H. The Influence of Background Organic Matter
(21)
(22)
on GAC Adsorption. J. Am. Water Works Assoc. 1989, 81 (5), 66-74. Sontheimer, H.; Crittenden, J. C.; Summers, R. S. Activated Carbon for Water Treatment, 2nd ed.; DVGW-Forschungsstelle: Karlsruhe, Germany, 1988. Jarvie, M. E.; Hand, D. W.; Bhuvendralingam, S.; Crittenden, J. C.; Hokanson, D. R. Simulating the Performance of Fixed-Bed Granular Activated Carbon Adsorbers: Removal of Synthetic Organic Chemicals in the Presence of Background Organic Matter. Water Res. 2005, 39 (11), 2407-2421. Zimmer, G.; Haist, B.; Sontheimer, H. The Influence of Preadsorption of Organic Matter on the Adsorption Behavior of Chlorinated Hydrocarbons. In Proceedings of the 1987 AWWA Annual Conference; American Water Works Association: Denver, CO, 1987. Li, Q. L.; Marinas, B. J.; Snoeyink, V. L.; Campos, C. ThreeComponent Competitive Adsorption Model for Flow-through PAC Systems. 1. Model Application to a PAC/Membrane System. Environ. Sci. Technol. 2003, 37 (13), 3005-3011. Ding, L.; Marinas, B. J.; Schideman, L. C.; Snoeyink, V. L.; Li, Q. L. Competitive Effects of Natural Organic Matter: Parameterization and Verification of the 3-Component Adsorption Model COMPSORB. Environ. Sci. Technol. 2006, 40 (1), 350-356. Randtke, S. J.; Snoeyink, V. L. Evaluating GAC Adsorptive Capacity. J. Am. Water Works Assoc. 1983, 75 (8), 406-413. Crittenden, J. C.; Luft, P.; Hand, D. W. Prediction of Multicomponent Adsorption Equilibria in Background Mixtures of Unknown Composition. Water Res. 1985, 19 (12), 1537-1548. Najm, I. N.; Snoeyink, V. L.; Richard, Y. Effect of Initial Concentration of a SOC in Natural Water on Its Adsorption by Activated Carbon. J. Am. Water Works Assoc. 1991, 83 (8), 5763. Losovik, P. A. A Calculation Method to Estimate the Concentration and Elemental Composition of Organic Substances in Natural Waters. J. Anal. Chem. 2006, 61 (6), 544-549. Knappe, D. R. U.; Snoeyink, V. L.; Roche, P.; Prados, M. J.; Bourbigot, M. M. Atrazine Removal by Preloaded GAC. J. Am. Water Works Assoc. 1999, 91 (10), 97-109. Smith, E. H.; Weber, W. J. Evaluation of Mass-Transfer Parameters for Adsorption of Organic Compounds from Complex Organic Matrixes. Environ. Sci. Technol. 1989, 23 (6), 713-722. Weber, W. J.; Liu, K. T. Determination of Mass Transport Parameters for Fixed Bed Adsorbers. Chem. Eng. Commun. 1980, 6, 49-60. Peel, R. G.; Benedek, A.; Crowe, C. M. A Branched Pore KineticModel for Activated Carbon Adsorption. AIChE J. 1981, 27 (1), 26-32. Li, Q. L.; Snoeyink, V. L.; Marinas, B. J.; Campos, C. Elucidating Competitive Adsorption Mechanisms of Atrazine and NOM Using Model Compounds. Water Res. 2003, 37 (4), 773-784. Li, Q. L.; Snoeyink, V. L.; Marinas, B. J.; Campos, C. Pore Blockage Effect of NOM on Atrazine Adsorption Kinetics of PAC: The Roles of PAC Pore Size Distribution and NOM Molecular Weight. Water Res. 2003, 37 (20), 4863-4872. Crittenden, J. C.; Sanongraj, S.; Bulloch, J. L.; Hand, D. W.; Rogers, T. N.; Speth, T. F.; Ulmer, M. Correlation of Aqueous-Phase Adsorption Isotherms. Environ. Sci. Technol. 1999, 33 (17), 29262933. Magnuson, M. L.; Speth, T. F. Quantitative Structure-Property Relationships for Enhancing Predictions of Synthetic Organic Chemical Removal from Drinking Water by Granular Activated Carbon. Environ. Sci. Technol. 2005, 39 (19), 7706-7711.
Received for review March 14, 2006. Revised manuscript received July 28, 2006. Accepted August 2, 2006. ES060603W
VOL. 40, NO. 21, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
6817
