Environ. Sci. Technol. 2005, 39, 2349-2356
Activated Carbon Load Equalization of Discontinuously Generated Acetone and Toluene Mixtures Treated by Biofiltration CONGNA LI AND WILLIAM M. MOE* Department of Civil & Environmental Engineering, Louisiana State University, Baton Rouge, Louisiana 70803
Studies were conducted to evaluate use of granular activated carbon (GAC) as a passively operated loadequalization mechanism for biofilters treating gas streams with dynamically varying (intermittent) pollutant loading. In the initial stage of research, abiotic fixed-bed sorption experiments and numerical modeling were conducted to assess the degree of load-equalization achieved by GAC columns for air flows containing intermittent loading of acetone and toluene present as single-component contaminants and as a mixture. In the subsequent stage of research, an integrated system consisting of a GAC column in series before a biofilter was used to treat a gas stream containing a mixture of acetone and toluene at influent concentrations of 430 ppmv and 100 ppmv, respectively. To simulate loading conditions expected from an industrial process with intermittent operation, contaminated air was supplied 8 h/day and uncontaminated air was supplied 16 h/day. The system was operated with different empty bed contact times, as low as 2.5 s for the GAC column and 14.5 s for the biofilter. Performance of an additional, conventionally operated biofilter (i.e., without GAC load equalization system) was used as a basis of comparison. Data are presented which clearly demonstrate that passively operated GAC load-dampening systems installed in series before biofilters can lead to more uniform loading as a function of time and thereby improve biofilter treatment performance. Results also demonstrate that, because of competitive sorption, the degree of load equalization achieved for different constituents in multi-contaminant gas streams can vary markedly. A pore and surface diffusion model (PSDM) was able to accurately predict the degree of load-dampening achieved by GAC columns for single and multicomponent waste gases.
Introduction Because of their comparatively low cost, operational simplicity, and lack of secondary pollutant generation, biofilters are increasingly used to remove volatile organic compounds (VOCs) from contaminated air streams (1). In biofiltration, contaminants are transferred from an air stream into a biofilm immobilized on a solid support medium. Once in the biofilm, microorganisms biodegrade contaminants into environmentally acceptable end products including CO2, water, and additional biomass. This treatment technology is particularly * Corresponding author phone: (225)578-9174; fax: (225)5788652; e-mail:
[email protected]. 10.1021/es049152a CCC: $30.25 Published on Web 02/16/2005
2005 American Chemical Society
attractive for treating high volumetric flow airstreams containing low concentrations of biodegradable pollutants (1). Contaminant concentrations in most waste gas streams vary with time due to the unsteady-state nature of industrial processes that generate them (2). Biofiltration has proven successful at removing biodegradable VOCs from a wide variety of industrial and waste treatment operations under conditions of relatively steady loading; however, transient loading conditions pose challenges in biofilter design and operation. For example, short-term increases in VOC concentration can result in dynamic “shock” loads that exceed biological reaction capacities and result in contaminant emissions from biofilter systems (3-7). Intervals of low or no contaminant loading in biofilter systems (e.g., because of 8-h work days or process shutdowns) are also problematic because of starvation conditions imposed on the microbial populations. Although short-term interruptions in contaminant loading (on the order of minutes) are unlikely to cause problems (2), diminished contaminant removal for a period lasting several hours or even days following resumption of contaminant loading has been reported for cases where the duration of no loading was longer (on the order of several hours or days) because microbial populations can undergo sufficient endogenous decay or shifts in metabolism to result in diminished performance (3, 8-12). In general, longer duration shutdown periods require longer time for performance recovery. A potential method for achieving the goal of “buffering” dynamic loading (i.e., more evenly distributing contaminant loading over time) is to install an activated carbon column in series prior to the biofilter. The rationale for such a system is that, during periods of high contaminant loading, the adsorbent could temporarily accumulate contaminants and then subsequently desorb contaminant during periods when concentrations in the waste gas are low. In this manner, it could dampen fluctuations in organic loading to prevent shock loading of the biofilter and provide continuous feed to the biological system over periods when wastes are not generated. It may also allow smaller (and therefore less expensive) biofilters for treating discontinuously generated waste gases because the biofilter bed could be used more efficiently. Although this strategy, first proposed by Ottengraff (13), has been recommended by multiple authors (1, 14, 15), such a system has been experimentally tested only for a singlecomponent waste gas stream containing toluene (14). Little or no research has been conducted to assess performance of passively operated activated carbon load-dampening systems for treating multi-contaminant waste gas streams during unsteady-state loading conditions. When multiple contaminants are present, competitive sorption can occur (16-20). As a result, activated carbon may exhibit differing buffering capacities for different contaminants. Furthermore, an appropriate basis for design or analysis of buffering systems is currently lacking. In the first stage of research described herein, abiotic fixedbed adsorption/desorption experiments were conducted to evaluate the range of load-dampening expected for GAC columns subjected to various inlet concentrations of acetone and toluene as intermittently loaded single-component contaminants and as a two-component mixture. Intermittent loading conditions were intended to simulate a process where contaminants are generated 8 h/day. A pore and surface diffusion model (PSDM) was calibrated and validated for predicting dynamic behavior of the sorption systems. In the second stage of research, the impact of an activated carbon VOL. 39, NO. 7, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2349
FIGURE 1. Schematic diagram of the experimental apparatus used for the “buffered” biofilter. Components shown inside the dashed line were used for conducting abiotic sorption experiments. buffering system on biofilter performance was tested for treatment of a model gas stream containing a two-component mixture of acetone and toluene under intermittent loading conditions.
Materials and Methods Fixed-Bed Adsorption Experiments. A diagram of the experimental apparatus used in three separate fixed-bed adsorption tests is shown in Figure 1. Each column was constructed of PVC (7.62 cm i.d.) and contained a perforated plate that supported 6 cm depth glass beads (5 mm diameter), a thin layer of glass wool, 84.4 cm depth of GAC, another thin layer of glass wool, and another 6 cm of glass beads. The GAC, with mass of 1.69 kg, consisted of Calgon BPL 4 × 6 mesh (Calgon, Pittsburgh, PA), a material commonly used in adsorption of gas-phase contaminants. The GAC was washed with deionized water to remove fines and dried at 105 °C prior to use. Initial tests conducted prior to placement of activated carbon demonstrated that column components other than GAC had little or no adsorption capacity for acetone or toluene. Contaminant-free compressed air (average relative humidity 20%) flowed through a pressure regulator to reduce pressure and rotameter to control flow rate. The synthetic waste gas stream was generated by evaporating liquid VOCs (ACS reagent grade, Sigma, St. Louis, MO) into the air stream using syringe pumps (KD Scientific, Boston, MA). A microprocessor-based controller (Chron-Trol, San Diego, CA) turned syringe pumps on and off as necessary. The dynamic loading scenario investigated in this research consisted of 8-h contaminant loading followed by 16-h nonloading each day. During nonloading intervals, uncontaminated air flowed through the columns at the same rate as during contaminant loading intervals. Gas flow rates were 42.1, 38.4, and 38.4 L/min for GAC columns used to test adsorption of toluene, acetone, and their binary mixture, respectively. These correspond to empty bed contact times (EBCTs) in the GAC columns of 5.5, 6.1, and 6.1 s, respectively. For toluene adsorption tests, target influent contaminant concentrations during the daily 8-h loading cycle were 868 ppmv during the first 12 days, 488 ppmv for 7 days, and then 217 ppmv for 6 days (25 days total). For acetone adsorption tests, target influent concentrations during the daily 8-h loading cycle were 960 ppmv for the first 8 days, 545 ppmv for 7 days, and then 243 ppmv for 6 days (21 days total). 2350
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 7, 2005
During the mixture adsorption test, target influent concentrations for acetone and toluene during the 8-h loading cycle were 550 ppmv each (18 days total). All experiments were conducted at ambient laboratory temperature of 23 ( 2 °C. Modeling Adsorption/Desorption in the GAC Column. The pore and surface diffusion model (PSDM) described by Crittenden et al. (19) and Hand et al. (20) was used to predict the degree of load-dampening achieved by GAC columns under discontinuous loading conditions. The PSDM is a dynamic fixed-bed model that incorporates assumptions and governing equations described previously (19, 21). Simulations using the PSDM were performed using AdDesignS software (Michigan Technological University). In AdDesignS, an orthogonal collocation method is used to convert partial differential equations of the PSDM into a set of ordinary differential equations that are solved using a backward differential method (19, 21, 22). Ideal adsorbed solution theory (IAST) was used to predict competitive interactions between VOCs (18-20, 23). Dynamic adsorption calculations using this model require equilibrium parameters, kinetic parameters, physical properties of adsorbing compound(s) and adsorbent, fluid properties, influent concentrations, and column dimensions. A summary of parameter values and sources of data input to the model is presented in Table SI-1 (see Supporting Information). Influent contaminant concentrations and flow rates input to the model were identical to those experimentally tested for either the individual compounds or their mixture. The model was calibrated by adjusting the film masstransfer coefficients for toluene and acetone adsorption to achieve the best fit with experimental data obtained for the first loading condition of single-component adsorption experiments (i.e., the first 12 and 8 days of sorption data for toluene and acetone, respectively). In calibrating the model, primary consideration was given to matching the model prediction of the quasi-steady state maximum daily outlet VOC concentration with the measured data. Secondary consideration was given to matching model prediction of minimum daily outlet VOC concentration and shape of the quasi-steady state breakthrough curves with the measured data. Data from the last two loading conditions of singlecompound sorption experiments (lasting 7 and 6 days, respectively) as well as the entire duration of the binary mixture adsorption test were used for model validation. Biofilter Apparatus. In the second stage of experiments described herein, two laboratory-scale biofilter systems were experimentally tested to evaluate merits of activated carbon load-dampening systems for mixtures of acetone and toluene under intermittent loading conditions. One system (see Figure 1) consisted of an activated carbon column (to achieve loaddampening), packed column (to achieve humidification), and biofilter (to achieve contaminant biodegradation) installed in series. A second system, identical to the first except that the activated carbon column was omitted, was used for comparison purposes. These were designated as “buffered” and “unbuffered” biofilters, respectively. Gas supply and equipment for regulating air flow and introducing VOCs were identical to those used in abiotic sorption experiments. In the buffered system, GAC column components were identical to those used in abiotic sorption experiments except that mass of GAC (0.85 kg) and packed bed depth (42.2 cm) were reduced by one-half. Contaminated air was humidified to greater than 95% relative humidity using a packed column (constructed of PVC, 7.62 cm i.d., packed depth 100 cm) containing stainless steel 10 mm Interpack (Jaeger, Houston, TX). A peristaltic pump recirculated water at a rate of 200 mL/min from the bottom storage reservoir through a spray nozzle (Spray System, Wheaton, IL) located 10 cm above the packing. Makeup water was automatically added on an hourly basis
TABLE 1. Summary of Biofilter Loading Conditions buffered biofilter system
unbuffered biofilter system
parameter
phase 1
phase 2
phase 3
phase 1
period of operation (days) EBCT of biofilter (s) EBCT of activated carbon column (s) influent acetone concentration (ppmv) influent toluene concentration (ppmv) daily biofilter acetone loading (g m-3 d-1)a daily biofilter toluene loading (g m-3 d-1)a total daily biofilter VOC loading (g m-3 d-1)a contaminant loading interval (h/day)
0-40 58 10 430 100 506 185 691 8
41-60 29 5 430 100 1012 370 1382 8
61-71 14.5 2.5 430 100 2024 740 2764 8
0-40 58 430 100 506 185 691 8
phase 2 41-60 29 430 100 1012 370 1382 8
phase 3 61-71 17 418 96.6 1686 618 2304 8
a The daily biofilter contaminant loading is calculated as the mass of contaminants entering the system per day divided by the packed bed volume of the biofilter. Because contaminants were supplied during only one-third of the cycle length (i.e., 8 h/day), the instantaneous loading rate during the 8 h loading period was equal to 3 times the daily loading rate.
to compensate for evaporation loss and maintain approximately 1.3 L of water in the humidification system. Each biofilter, operated in an up-flow mode, consisted of a glass column with a bottom, top, and six sections (9.9 cm i.d.). A perforated stainless steel plate supported packing medium in each section. The column had total packed bed depth of 1.45 m and packed bed volume of 11.2 L. Glass marbles were placed in the bottom of the columns to evenly distribute airflow. The biofilter packing medium consisted of reticulated polyurethane foam cubes 1.2 cm per side containing approximately 7 pores per cm, bulk porosity of 97%, packed bed density of 30 kg/m3, and surface area (bulk) of approximately 210 m2/m3 (Honeywell-PAI, Lakewood, CO). Solenoid valves with stainless steel flow tubes (Asco Valve Inc., NJ) were installed prior to the GAC column inlet and after the GAC column, humidification column, and biofilter to allow automated sampling at different spatial locations. When a solenoid valve was turned on, a portion of the gas flow was diverted to the analytical instrument used to measure contaminant concentrations. Gas sampling lines were constructed using Teflon tubing. Biofilter Inoculation and Operation. Two separate sparged-gas reactors containing nutrient solution (6) were inoculated with activated sludge from a municipal wastewater treatment facility to enrich for toluene and acetone degrading cultures following a method described previously (7). Each biofilter was inoculated by temporarily submerging the packing medium in a mixture of the toluene and acetone degrading enrichment cultures as reported elsewhere (7). Following inoculation, both biofilters underwent three distinct periods of operation (arbitrarily designated phases 1, 2, and 3) that involved progressively higher contaminant loading rates as summarized in Table 1. During all loading conditions, contaminants were supplied in the influent gas stream for 8 h/day. During the remaining 16 h/day, contaminants were not supplied; however, uncontaminated air continued to flow through the system at the same rate as during the loading period. To allow direct comparison of the two systems based on EBCT of the biofilters, during phases 1 and 2, both biofilters received the same gas flow rate, and both systems received the same influent contaminant concentrations. To allow comparison of the two systems based on total system EBCT (rather than EBCT of only the biofilter portion of the complete system), during phase 3, gas flow rate to the unbuffered biofilter was lower than that of the buffered biofilter so that the unbuffered biofilter had the same EBCT (17 s) as the total EBCT of the activated carbon column (2.5 s) plus the biofilter (14.5 s) in the buffered system. Due to limited flow increments available using the syringe pumps, target influent contaminant concentrations entering the unbuffered biofilter system were slightly lower than those entering the buffered system during phase 3.
To avoid a lengthy start-up interval during which abiotic adsorption to GAC would account for a large percentage of contaminant removal in the buffered system, the activated carbon column was preloaded with contaminants prior to biofilter startup. In the preloading interval, the activated carbon column was operated with the same daily loading conditions as in phase 1 described above. Biofilter start-up did not occur until after the patterns of effluent acetone and toluene concentrations exiting the GAC column were similar in a day-to-day comparison. On day 9 and at 3-day intervals thereafter, nutrients were supplied to the biofilters by temporarily filling each biofilter column with freshly prepared nutrient solution and then draining by gravity for approximately 15 min before restoring normal operation (7). Analytical Techniques. Acetone, toluene, and CO2 concentrations were measured using a model 1312 photoacoustic multigas monitor (California Analytical, Orange, CA) equipped with four optical filters (UA# 0971, 0974, 0983, and SB0527). Relative humidity was measured using a traceable digital hygrometer/thermometer (Fisher Scientific, Pittsburgh, PA).
Results and Discussion Fixed-Bed GAC Adsorption/Desorption of Single Components. Experimentally measured GAC breakthrough curves for intermittent loading of toluene and acetone as singlecomponent contaminants are depicted in Figure 2a,c. As shown, under the discontinuous loading conditions, initial toluene and acetone breakthrough occurred after 6.3 and 2.4 days, respectively. Quasi-steady state conditions were attained rather quickly with both contaminants exhibiting a consistent pattern of attenuated effluent concentrations reoccurring on a daily basis. Daily peak effluent concentrations differed only slightly (within 5%), and mass balance calculations verified that contaminant mass entering and exiting GAC columns on a daily basis was essentially the same (within 3%) after quasi-steady state was reached at each loading condition. Model simulations of toluene and acetone as intermittently loaded single-component contaminants are shown in Figure 2b,d. When only a single parameter (film mass-transfer coefficient) was used for model calibration based on the first loading condition, the magnitude of load attenuation (i.e., the maximum and minimum outlet VOC concentrations) and general shape of the breakthrough curves predicted by the model had excellent agreement with measured breakthrough data. There is a small discrepancy, however, between the measured and modeled time to reach initial breakthrough, and the measured and modeled data were somewhat out of phase with respect to the timing of maximum daily breakthrough, particularly in the case of acetone. This resulted from the fact that, in model calibration, highest priority was VOL. 39, NO. 7, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2351
FIGURE 2. Experimental measurement (a) and model simulation (b) of toluene concentrations exiting the GAC column during intermittent loading (EBCT 5.5 s). Experimental measurement (c) and model simulation (d) of acetone concentrations exiting the GAC column during intermittent loading (EBCT 6.1 s): target influent (-); measured effluent (9); and modeled effluent (-). assigned to matching the magnitude of the maximum daily VOC concentration. It is anticipated that peak VOC concentration exiting a GAC column will be the most important metric in assessing effectiveness of load attenuation to improve biofilter performance; accurate model prediction of the timing of daily peak contaminant concentration is not expected to be a major concern for most passive load equalization applications. Thus, the model was successful in predicting what is expected to be the most important aspect of the GAC load attenuation process. The experimental data, supported by modeling results, demonstrate that a GAC column with relatively low EBCT (5.5 and 6.1 s for toluene and acetone, respectively) in comparison to EBCTs typical of biofilters can markedly dampen dynamic (intermittent) loading for both compounds. Peak contaminant concentrations exiting the GAC column were less than one-half of the influent concentration during each loading condition for both compounds. Furthermore, during the third loading conditions (influent concentration 217 ppmv toluene or 234 ppmv acetone), “ideal” buffering was achieved. Contaminants loaded to the GAC column during only one-third of the cycle duration (i.e., 8 h per 24 h period) exited the system evenly distributed in time at a level equal to one-third of the influent concentration. This nicely demonstrates that contaminant mass temporarily accumulated in the GAC column during intervals when influent contaminant concentrations are high can desorb within a sufficiently short time interval (i.e., during each loading cycle when influent contaminant concentrations are low) to be of practical benefit as a passively operated load equalization mechanism. The only driving force necessary for contaminant desorption was the naturally occurring decrease in influent contaminant concentration. Regeneration of the GAC column through other means (e.g., heating) was not necessary for successful load-dampening. For single-contaminant waste gas streams, load equalization such as that depicted in Figure 2 could provide an obvious practical benefit in design and operation of biofilters. Peak loading to the biofilter could be greatly reduced, and starvation during periods of low or no contaminant loading could be minimized or entirely avoided. If a biofilter’s design is to be based on peak loading as advocated by some authors (5), then the biofilter could be markedly reduced in size as compared to a system without buffering. Fixed-Bed GAC Adsorption/Desorption of Acetone and Toluene Mixture. Model predictions and experimentally measured effluent concentrations for the mixture of toluene and acetone (550 ppmv each) under discontinuous (8 h/day) 2352
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 7, 2005
FIGURE 3. Experimental measurement (a) and model simulation (b) of toluene and acetone concentrations exiting the GAC column during mixture adsorption experiments (EBCT 6.1 s). Both contaminants had an influent concentration of 550 ppmv during the 8 h/day loading interval. loading are depicted in Figure 3. Toluene was effectively buffered, and its breakthrough was similar to that observed for single-component adsorption (although with longer interval before initial breakthrough, consistent with that expected for lower influent concentration in comparison to data depicted in Figure 2). Breakthrough for acetone, however, was markedly different from that observed in the single-component adsorption process. Although initially wellbuffered (e.g., day 5), peak daily acetone concentrations exiting the GAC column progressively increased over time until toluene breakthrough reached quasi-steady state. On day 12, peak effluent acetone concentration was actually 16% higher than the influent. Mass balance calculations revealed that during the interval when acetone breakthrough was observed while toluene was not (days 5-12), mass of acetone exiting the GAC column was greater than mass entering on a daily basis. After both acetone and toluene reached quasisteady state (day 14), only a small fraction of acetone remained in the GAC bed. This phenomenon can be readily explained by competitive sorption effects caused by differences in physicochemical properties of the adsorbates (16, 17). VOC adsorption to GAC is governed by both activated carbon surface characteristics (e.g., pore volume, specific surface area) and VOC properties
(e.g., dipole moment, boiling point, and molecular weight) (24). Because the surface of activated carbon is nonpolar or slightly polar, it generally has higher affinity and higher adsorption capacity for nonpolar organics (e.g., toluene, dipole moment of 0.45 D) than it does for polar organics (e.g., acetone, dipole moment of 2.87 D) (25). Furthermore, toluene has a higher boiling point than acetone (111 vs 56.3 °C), and it also has a higher molecular weight (92 vs 58 g/mol), characteristics favoring adsorption. Thus, as the more strongly adsorbed toluene accumulated in the carbon bed, it displaced the more weakly adsorbed acetone. The combination of initially adsorbed acetone being displaced and emitted at the same time as additional acetone entering the system passed through unadsorbed resulted in effluent acetone concentration exceeding the influent for a period of time before quasi-steady state was reached. This effect (displacement of more weakly adsorbed component resulting in effluent concentrations for one or more compounds temporarily exceeding the influent) is consistent with patterns observed for competitive adsorption in processes treating multicomponent waste streams subjected to steady (rather than intermittent) loading conditions (16, 17, 20). The PSDM was successful in qualitatively and quantitatively predicting the pattern of VOC concentrations exiting the GAC column for the multicomponent gas stream. Both modeled and measured data demonstrate that for multicomponent waste gas streams containing constituents with different physicochemical properties (e.g., acetone and toluene), competitive adsorption can result in buffering of more weakly adsorbed components markedly different from that observed when contaminants are present as single components. The modeling approach described herein may prove useful for evaluating buffering expected in other systems where single or multiple contaminants are present. In the experiments described above, sorption was evaluated at a fixed EBCT. To experimentally evaluate the effect of EBCT on load attenuation and to evaluate whether activated carbon load-dampening can improve performance of biofilters treating multicomponent mixtures, additional experiments were conducted to compare performance of a buffered and unbuffered biofilter system. Numerous reports suggest that aromatic compounds (e.g., toluene) are more slowly degraded than ketones (e.g., acetone) in biofilters (7, 12, 26-28). Thus, decreasing the peak loading of toluene may be beneficial even if acetone is less attenuated or nonattenuated. Buffering during Biofilter Treatment. Typical contaminant concentrations measured over 48-h periods (i.e., two sequential loading cycles) at spatial locations prior to the biofilter inlets during the three different phases of biofilter operation are illustrated in Figures 4 and 5 for the unbuffered and buffered systems, respectively. Time zero in the figures corresponds to start of an 8-h loading period. As shown, measured influent acetone and toluene concentrations closely matched target concentrations (within 5%). The humidification column located in series prior to the unbuffered biofilter somewhat dampened loading for acetone but not toluene (see Figure 4). This is consistent with a previous report that scrubbing water in a packed bed humidification system can dampen concentrations of highly water-soluble compounds with low Henry’s Law constants (e.g., acetone) but is ineffective for less water-soluble compounds with higher Henry’s Law constants (e.g., toluene) (4). In this case, the volume of water recirculated in the humidification system was rather small (approximately 1.3 L); therefore, the loaddampening effect was exhausted rather quickly and contaminant concentrations entering the biofilter were essentially the same as those entering the system for a portion of the loading interval each day. As air flow rates and contaminant mass flow rates increased from phases 1 to 2 to 3,
FIGURE 4. Contaminant concentrations measured at spatial locations prior to the inlet of the unbuffered biofilter during phase 1 (data from days 37 and 38), phase 2 (data from days 55 and 56), and phase 3 (data from days 68 and 69): humidification inlet acetone (]); humidification inlet toluene (4); humidification outlet acetone ([); and humidification outlet toluene (2). the time for acetone to reach equilibrium in absorption and desorption in the water was consequently shortened. The daily mass of contaminants exiting the humidification column was essentially the same as that entering the column (within 1%), indicating excellent mass balance closure. As depicted in Figure 5, for the buffered biofilter system, the GAC column effectively buffered the dynamic loading of toluene to reasonably stable concentrations during all three phases, even with a GAC column EBCT as low as 2.5 s (Phase 3). Acetone was also dampened, but to a lesser extent than toluene, and the dampening decreased as EBCT decreased. Interestingly, acetone concentrations exiting the GAC column were out of phase with the inlet concentrations during phase 1. This was caused by a unique combination of column dimensions, gas flow rate, contaminant loading rate, and interval of contaminant loading, and would not be expected under all circumstances. As shown in the figure, the timing of when the peak daily acetone concentration exited the GAC column shifted as the gas flow rate increased in phases 2 and 3. Acetone concentrations exiting the GAC column were further attenuated after passing through the humidification column; however, because the volume of recirculated water was small, the dampening effect was rather small. Differences in degree of dampening observed for acetone depicted in Figure 5 in comparison to that shown in Figure 3 can be partially attributed to the lower feed concentration of toluene. At higher concentrations (Figure 3), the more strongly adsorbed component (toluene) has a greater competitive edge and consequently displaces more of the weakly adsorbed component (acetone) (29). Model simulations indicate that the magnitude of the initial “rollup” for acetone (before quasi-steady state of acetone adsorption is reached) increases in the presence of higher inlet toluene concentration (data not shown). Acetone was less attenuated at higher gas flow rates, consistent with the observation that sharper breakthrough curves occur at lower EBCTs in continuously loaded systems (29). Because of load equalization by the GAC column, peak contaminant loading to the buffered biofilter, especially for toluene, was substantially lower than peak loading to the unbuffered biofilter. Over the 71 days of operation, the difference between the total mass of each contaminant entering and exiting the GAC column was 0.7% and 5% for acetone and toluene, respectively. Considering analytical error associated with gas measurements and the integration VOL. 39, NO. 7, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2353
FIGURE 5. Acetone (top) and toluene (bottom) concentrations measured at spatial locations prior to the inlet of the buffered biofilter during phase 1 (left, data from days 28 and 29), phase 2 (middle, data from days 56 and 57), and phase 3 (right, data from days 69 and 70): GAC inlet acetone (O); GAC outlet acetone ([); humidification outlet acetone (]); GAC inlet toluene (b); GAC outlet toluene (2); and humidification outlet toluene (4).
FIGURE 6. Measured daily contaminant loading for the buffered biofilter (a) and the unbuffered biofilter (c). Removal efficiency in the buffered biofilter (b) and the unbuffered biofilter (d): acetone ([); toluene (4); and total of both VOCs (O). method employed in calculations, these data indicate a good closure on contaminant mass balance for the GAC column. Biofilter Performance. Target and measured biofilter daily contaminant mass loading rates are shown in Figure 6a,c. Data points depicted in the figure were calculated on a daily basis using concentrations measured at 1-h intervals. Lines represent target loading rates. As shown in the figure, measured loading rates were quite close to target levels. Figure 6b,d also depicts removal efficiency observed for buffered and unbuffered biofilters. Removal efficiencies were calculated on a daily basis using concentrations measured at biofilter inlets and outlets at 1-h intervals for each day data were collected. As shown, during the week following biofilter start-up, performance of both biofilters decreased over time. On day 9, a nutrient addition procedure was conducted, and performance of both biofilters began to improve. Thereafter, nutrients were added at approximately 3-day intervals. It is evident that a sufficient nutrient supply is vital for rapid biofilter start-up as reported previously (30). Performance of the buffered biofilter increased much more rapidly than that for the unbuffered system. By day 20, the buffered biofilter removed >99% of both contaminants and remained at that level until the end of phase 1. There was a small but noticeable decrease in performance of the buffered biofilter immediately after the contaminant loading rate was doubled (EBCT decreased to 29 s) at the start of phase 2. The system quickly acclimated to the new loading condition, and high removal efficiency (>98% for acetone and >99% for toluene) was observed until the end of phase 2. At the start of phase 3, when the contaminant loading rate was doubled again (EBCT decreased to 14.5 s), there was no 2354
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 7, 2005
decrease in removal efficiency for toluene. Acetone removal, however, decreased to 67% before gradually increasing to approximately 80% on day 71. Typical toluene, acetone, and CO2 concentrations exiting the buffered biofilter for each of the three phases of operation are depicted in Figure 7. CO2 concentrations depicted in the figure are concentrations above the inlet (i.e., CO2 production within the biofilter). During all three phases, toluene removal was essentially complete under the reasonably steady inlet concentrations achieved by GAC buffering. CO2 production increased as influent acetone concentration increased. During phases 1 and 2, although there was variation in acetone loading as a function of time during each daily loading cycle (see Figure 5), the buffered biofilter was able to exhibit a sufficiently high degradation rate that it could degrade contaminants at the rate they entered the biofilter. During phase 3, however, the acetone loading rate exceeded the system’s biodegradation capacity and acetone was transiently detected for approximately 6 h/day during the interval when acetone loading was highest (see Figure 7c). This difference in acetone removal is consistent with the degree of buffering observed in the system. During phases 1 and 2, the peak acetone concentration entering the biofilter was appreciably lower than that entering the GAC column (i.e., activated carbon attenuated peak acetone loading). During phase 3, the peak acetone concentration entering the biofilter was essentially the same as that entering the GAC column (see Figure 5c). In contrast to the relatively rapid startup and high contaminant removal efficiency observed in the buffered biofilter, the unbuffered biofilter exhibited lower treatment performance over the entire interval of study. As shown in
FIGURE 7. Typical effluent VOC and CO2 concentrations for the buffered biofilter during phase 1 (data from days 31 and 32), phase 2 (data from days 56 and 57), and phase 3 (data from days 68 and 69): effluent acetone ([); effluent toluene (4); and effluent CO2 (*). Biofilter VOC inlet concentrations correspond to the humidification outlet VOC concentrations as plotted in Figure 5.
FIGURE 8. Typical effluent VOC and CO2 concentrations for the unbuffered biofilter during phase 1 (data from day 38), phase 2 (data from day 56), and phase 3 (data from day 69): effluent acetone ([); effluent toluene (4); and effluent CO2 (*). Biofilter VOC inlet concentrations correspond to the humidification outlet VOC concentrations as plotted in Figure 4. Figure 6d, the unbuffered biofilter generally removed less than one-half as much toluene as the buffered biofilter and substantially less acetone as well. Even during phase 3, when daily contaminant loading to the unbuffered biofilter was 16.7% less than the buffered biofilter, performance was markedly lower. The lower performance, especially during startup, is consistent with the fact that intermittently loaded biofilters can take longer time to acclimate than biofilters receiving contaminant supply on a continuous basis (28). It is also consistent with the fact that contaminant concentrations entering the unbuffered biofilter were higher (i.e., the unbuffered system did not have the benefit of load attenuation provided by GAC in the buffered system). Typical toluene, acetone, and CO2 concentrations exiting the unbuffered biofilter during each phase of operation are shown in Figure 8. During all three phases, toluene breakthrough was observed shortly after start of contaminant loading each day. Acetone breakthrough, somewhat delayed by attenuation in the humidification system (see Figure 4),
was also consistently observed. An increase in CO2 production rate observed as acetone and toluene concentrations entering the biofilter increased indicates that the microbial population was able to rapidly increase its contaminant degradation rate; however, effluent contaminant concentration data demonstrate that the biofilter’s increased degradation rate was insufficient to remove all of the contaminants at the rate they entered the biofilter. During all three phases of operation, there was a small but noticeable decrease in effluent VOC concentrations over time during the loading period after inlet VOC concentrations stabilized at their peak influent concentrations. This suggests that the 16-h/day nonloading interval was a sufficiently long starvation period to result in decreased activity of toluene and acetone degrading microorganisms. Diminished performance following starvation conditions is consistent with previous reports on biofilter performance following nonloading periods (2, 3, 11). Such effects were not observed in the buffered biofilter where contaminants entered the biofilter more uniformly as a function of time due to buffering by the GAC column. At the end of experiments described above, an additional short-term test was conducted to further evaluate the effect of the GAC buffering system. The GAC column initially in series with the buffered biofilter was moved to a location in series prior to the humidification system in the previously unbuffered biofilter. Performance was evaluated over a 2-day period. Overall performance improved dramatically, with average toluene and acetone removal efficiency reaching 98% and 84%, respectively. This performance is comparable to that observed in the originally buffered system, and it provides further confirmation that load equalization provided by the GAC column was responsible for superior performance observed in the buffered system. A mass balance on carbon was conducted for each biofilter accounting for CO2, acetone, and toluene entering and exiting the systems in the gas phase during the entire period of operation (71 days), and biomass present at the end of the experiments (estimated from water displacement tests) (7). The difference between mass of carbon entering the system and that accounted for by mass of carbon exiting the system (as CO2, acetone, or toluene in the gas phase) or being present as biomass at the end of the experiment was 1.5% and 7.6% for buffered and unbuffered biofilters, respectively. Thus, there was reasonable closure on the carbon balance. More than 85% of carbon associated with acetone or toluene removed by the biofilters was transformed to CO2. Overall, results demonstrate that a passively operated activated carbon buffering system can achieve two advantages in operation of biofilters subjected to intermittent loading: (1) the magnitude of peak loading to the biofilter can be decreased, leading to more complete contaminant removal, and (2) because contaminant loading to the biofilter is more uniform as a function of time, diminished performance due to starvation conditions can be minimized. Because of competitive adsorption, however, a passively operated GAC column will not achieve an equal degree of load equalization for all compounds when multiple VOCs with dissimilar physicochemical properties are present.
Acknowledgments We gratefully acknowledge the Louisiana Board of Regents and BioReaction Industries for financial support, Calgon Carbon Corp. for providing activated carbon, and HoneywellPAI for providing biofilter packing medium.
Supporting Information Available Parameter values used in model simulations. This material is available free of charge via the Internet at http://pubs.acs.org. VOL. 39, NO. 7, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2355
Literature Cited (1) Devinny, J. S.; Deshusses, M. A.; Webster, T. S. Biofiltration for Air Pollution Control; CRC-Lewis Publishers: Boca Raton, FL, 1999. (2) Dirk-Faitakis, C.; Allen, D. G. Biofiltration of cyclic air emissions of R-pinene at low and high frequencies. J. Air Waste Manage. Assoc. 2003, 53, 1373-1383. (3) Martin, F. J.; Loehr, R. C. Effects of periods of nonuse on biofilter performance. J. Air Waste Manage. Assoc. 1996, 46, 539-546. (4) Al-Rayes, A. W.; Kinney, K. A.; Seibert, A. F.; Corsi, R. L. Load dampening system for vapor phase bioreactors. J. Environ. Eng. 2001, 127, 224-232. (5) Martin, R. W.; Li, H. B.; Mihelcic, J. R.; Crittenden, J. C.; Lueking, D. R.; Hatch, C. R.; Ball, P. Optimization of biofiltration for odor control: Model calibration, validation, and applications. Water Environ. Res. 2002, 74, 17-27. (6) Moe, W. M.; Li, C. Comparison of continuous and sequencing batch operated gas-phase biofilters for treatment of MEK. J. Environ. Eng. 2004, 130, 300-314. (7) Atoche, J. C.; Moe, W. M. Treatment of MEK and toluene mixtures in biofilters: Effect of operating strategy on performance during transient loading. Biotechnol. Bioeng. 2004, 86, 468-481. (8) Mohseni, M.; Allen, D. G.; Nichols, K. M. Biofiltration of alphapinene and its application to the treatment of pulp and paper air emissions. Tappi J. 1998, 81, 205-211. (9) Webster, T. S.; Cox, H. H. J.; Deshusses, M. A. Resolving operational and performance problems encountered in the use of a pilot/full-scale biotrickling filter reactor. Environ. Prog. 1999, 18, 162-172. (10) Park, J.; Kinney, K. A. Evaluation of Slip Feed System for Vaporphase Bioreactors. J. Environ. Eng. 2001, 127, 979-985. (11) Cox, H. H. J.; Deshusses, M. A. Effect of starvation on the performance and re-acclimation of biotrickling filters for air pollution control. Environ. Sci. Technol. 2002, 36, 3069-3073. (12) Moe, W. M.; Qi, B. Performance of a fungal biofilter treating gas-phase solvent mixtures during intermittent loading. Water Res. 2004, 38, 2258-2267. (13) Ottengraff, S. P. P. Exhaust gas purification. In Biotechnology; Rehm, H. J., Reed, G., Eds.; VCH Verlagsgesellschaft: Weinheim, Germany, 1986; Vol. 8, pp 425-452. (14) Weber, F. J.; Hartmans, S. Use of activated carbon as a buffer in biofiltration of waste gases with fluctuating concentrations of toluene. Appl. Microbiol. Biotechnol. 1995, 43, 365-369. (15) Swanson, W. J.; Loehr, R. C. Biofiltration: Fundamentals, design and operations principles, and applications. J. Environ. Eng. 1997, 123, 538-546. (16) Ruthven, D. M. Principles of Adsorption and Adsorption Processes; John Wiley & Sons: New York, 1984. (17) Yang, R. T. Gas Separation by adsorption processes; Butterworths: Boston, MA; Imperial College Press: London, UK, 1987.
2356
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 7, 2005
(18) Crittenden, J. C.; Luft, P.; Hand, D. W.; Oravitz, J. L.; Loper, S. W.; Ari, M. Prediction of multicomponent adsorption equilbria using ideal solution theory. Environ. Sci. Technol. 1985, 19, 1037-1043. (19) Crittenden, J. C.; Hutzler, N. J.; Geyer, D. G.; Oravitz, J. L.; Friedman, G. Transport of organic compounds with saturated groundwater flow: Model development and parameter sensitivity. Water Resour. Res. 1986, 22, 271-284. (20) Hand, D. W.; Crittenden, J. C.; Hokanson, D. R.; Bulloch, J. L. Predicting the performance of fixed-bed granular activated carbon adsorbers. Water Sci. Technol. 1997, 35, 235-241. (21) Mertz, K. A.; Gobin, F.; Hand, D. W.; Hokanson, D. R.; Crittenden, J. C. Manual: Adsorption Design Software for Windows (AdDesignS); Michigan Technological University, Houghton, MI, 1999. (22) Crittenden, J. C.; Wong, B. W. C.; Thacker, W. E.; Snoeyink, V. L.; Hinrichs, R. L. Mathematical model of sequential loading in fixed-bed adsorbers. J.-Water Pollut. Control Fed. 1980, 52, 2780-2795. (23) Crittenden, J. C.; Cortright, R. D.; Rick, B.; Tang, S. R.; Perram, D. Using GAC to remove VOCs from air stripper off-gas. J.-Am. Water Works Assoc. 1988, 80, 73-84. (24) Chiang, Y. C.; Chiang, P. C.; Chang, E. E. Effects of surface characteristics of activated carbons on VOC adsorption. J. Environ. Eng. 2001, 127, 54-62. (25) Yang, R. T. Adsorbents: Fundamentals and Applications; John Wiley & Sons: New York, 2003. (26) Webster, T. S.; Tongna, A. P.; Guarini, W. J.; Mcknight, L. Treatment of volatile organic compound emissions from a paint spray booth application using biological trickling filtration. In Proceedings of the 1998 USC-TRG Conf. on Biofiltration; Los Angeles, CA, Oct. 1998. (27) Song, J.; Kinney, K. A.; John, P. Influence of nitrogen supply and substrate interactions on the removal of paint VOC mixtures in a hybrid bioreactor. Environ. Prog. 2003, 22, 137-144. (28) Moe, W. M.; Qi, B. Biofilter treatment of VOC emissions from reformulated paint: Complex mixtures, intermittent operation, and startup. J. Air Waste Manage. Assoc., in press. (29) Mansour, A. R.; Shahalam, A. B.; Darwish, N. A comprehensive study of parameters influencing the performance of multicomponent adsorption in fixed-beds. Sep. Sci. Technol. 1985, 19, 1087-1111. (30) Moe, W. M.; Irvine, R. L. Effect of nitrogen limitation on performance of toluene degrading biofilters. Water Res. 2001, 35, 1407-1414.
Received for review June 6, 2004. Revised manuscript received December 28, 2004. Accepted January 5, 2005. ES049152A
