Understanding Mercury Transformations in Coal-Fired Power Plants

Mercury capture in coal-fired power plants depends greatly upon its speciation in the post-combustion flue gases. Whereas elemental mercury (Hg0) is ...
1 downloads 0 Views 199KB Size
Download PDF
Environ. Sci. Technol. 2007, 41, 7870-7875

Understanding Mercury Transformations in Coal-Fired Power Plants: Evaluation of Homogeneous Hg Oxidation Mechanisms BALAJI KRISHNAKUMAR† AND J O S E P H J . H E L B L E * ,‡ Department of Chemical Engineering, University of Connecticut, Storrs, Connecticut 06269-3222, and Thayer School of Engineering, Dartmouth College, Hanover, New Hampshire 03755-8000

Homogeneous mercury oxidation mechanisms described by Niksa and Qiu, and three theoretical mercury oxidation reaction rate constants developed by Wilcox were evaluated for their predictions of the extent of mercury oxidation under coal combustion conditions. Predictions were compared to results from bench-scale experiments to determine whether such models were suitable for predicting measured levels of homogeneous mercury oxidation. Experiments considered different flue gas compositions (O2, Cl, NO, and SO2) and quench rates to provide a broad range of conditions for analysis. Regardless of the mechanism chosen, most mercury oxidation was predicted to occur at temperatures below 900 K. The Niksa mechanism predicted Hg oxidation to occur only in systems that were close to isothermal at temperatures above 900 K followed by a rapid gas quench. This mechanism provided the best agreement with the experimental data of Sliger. The Qiu mechanism predicted Hg oxidation in several experimental systems and conditions fairly accurately although it did not provide the best agreement in all cases. Qiu mechanism predictions for the experimental system at the University of Connecticut operating at an equivalence ratio of 0.9 in the presence of HCl/Cl2 and also SO2 were within the bounds of experimental uncertainty. Additionally, for an experimental dataset obtained from the University of Utah for quench rates of 210 and 440 K/s in the presence of HCl, the Qiu model predicted the experimental observations with a high degree of accuracy. The effects of flue gas composition and quench on Hg oxidation are qualitatively represented by the Qiu mechanism suggesting a relative robustness of the model, although there is still need to refine rate constants to achieve greater accuracy. The Wilcox rate constants when substituted in the Qiu mechanism predicted near-complete oxidation of Hg irrespective of HCl concentrations in systems that involve flue gas quench below temperatures of 450 K.

Introduction On March 15, 2005, the U.S. Environmental Protection Agency (EPA) issued the Clean Air Mercury Rule establishing a target * Corresponding author e-mail: [email protected]. † University of Connecticut. ‡ Dartmouth College. 7870

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 22, 2007

of 70% reduction in mercury (Hg) emissions from coal-fired power plants by the year 2018 (1). Air emissions of mercury are a concern because of their eventual deposition, transformation to methyl mercury, and biomagnification in fish. The U.S. EPA estimates that each year approximately 300,000 newborns in the United States face an elevated risk of developing disabilities due to methyl mercury exposure arising out of consumption of contaminated fish (2). Mercury capture in coal-fired power plants depends greatly upon its speciation in the post-combustion flue gases. Whereas elemental mercury (Hg0) is water-insoluble, the oxidized form (Hg2+) is water-soluble and readily captured by scrubbers. Due to the complexity of factors affecting Hg oxidation, statistical models based on regression analysis have not been successful in predicting Hg oxidation under the necessary range of operating environments (3). It is, therefore, important to develop mechanistic models capable of elucidating Hg transformation in coal-fired utilities and provide guidance both for prediction of coal quality impacts on Hg emissions as well as for facilitation of the development of Hg control technology. Mercury oxidation in combustion systems is believed to proceed through both homogeneous and heterogeneous pathways, generally mediated by chlorine species. Hg oxidation in these systems is kinetically limited (4, 5) with extents between 0 and 99% observed in practical combustion systems (6, 7). Several kinetic reaction mechanisms have been proposed to elucidate homogeneous Hg transformations, with the majority treating oxidation as proceeding through an intermediate species believed to be HgCl (8-11). Researchers have also explored Hg/O interactions, but it was concluded that HgO (if present) does not affect the Hg chemistry (12, 13). The chemical kinetic mechanisms reported in the literature have been used to model bench-scale experimental data and in most cases, the rate constants were empirically fit to predict a single dataset. In cases where the mechanisms were evaluated against a broad range of datasets, arbitrary quench profiles were imposed due to the nonavailability of such data for the systems modeled. This paper addresses these shortcomings by analyzing bench-scale experiments using recently reported homogeneous Hg oxidation mechanisms and theoretical rate constants, and evaluating their ability to predict Hg speciation in well-controlled experiments designed to address homogeneous chemistry alone. Experimental Systems and Mechanisms. Data from homogeneous mercury oxidation experiments by Sliger et al. (9), Sterling et al. (14) (referred to as UConn), and Fry et al. (15) (referred to as UUtah) were considered in examining the models. All were flame-based systems and sufficient information regarding operating conditions such as flue gas composition and temperature profile was available in the literature. Systems using bottled gases were not evaluated because they were not considered representative of the radical-rich flame conditions of utility boilers. All kinetic models were solved using the SENKIN routine of CHEMKINIII software. The Hg oxidation mechanisms evaluated in this study were Niksa et al. (3, 10) (referred to as Niksa), and Qiu et al. (11) (referred to as Qiu). Both the Niksa and Qiu Hg mechanisms use the framework proposed by Widmer et al. (8), consisting of an eight step elementary reaction sequence for Hg oxidation and presented in Table 1. Recently, Wilcox et al. (16-18) have reported rate constants for three of these eight Hg reactions, which were calculated using theoretical methods involving quantum chemistry and transition state 10.1021/es071087s CCC: $37.00

 2007 American Chemical Society Published on Web 10/10/2007

TABLE 1. Widmer’s Eight Step Elementary Hg Oxidation Mechanism with Three Theoretically Calculated Rate Constants by Wilcoxa reactions R1 R2 R3 R4 R5 R6 R7 R8 a

Hg + Cl + M ) HgCl + M Hg + Cl2 ) HgCl + Cl Hg + HCl ) HgCl + H Hg + HOCl ) HgCl + OH HgCl + Cl + M ) HgCl2 + M HgCl + Cl2 ) HgCl2 + Cl HgCl + HCl ) HgCl2 + H HgCl + HOCl ) HgCl2 + OH Units, cm-mol-s-K.

b

Arrhenius rate constantsb from Wilcox (16-18)

temperature range (K)

4.25 × 1013 exp(-8588/T)

393-1500

2.09 × 1017 exp(-37952/T)

298-2000

4.50 × 1013 exp(-3049/T)

298-2000

Rate constants are for the reverse reaction.

theory. The three reactions and the corresponding rate constants calculated by Wilcox et al. are also presented in Table 1. In this study, the three rate constants of Wilcox et al. were substituted separately into the Niksa and Qiu mechanisms to evaluate the effect of these theoretical reaction rate constants on Hg oxidation. To calculate the extent of Hg oxidation in combustion systems, a complete kinetic model must include a submechanism containing reactions involving C-H-N-O-SCl species in addition to the Hg reaction mechanism. Another important requirement is thermodynamic data used to solve the energy balance of the system and to provide equilibrium constants needed for calculation of the reverse reaction rate constants. Complete submechanisms have been reported for both the Niksa and Qiu mechanisms and the thermodynamic data used by the respective authors are also available. In the calculations considered in this paper, initial values for species composition were determined by assuming equilibrium at a sufficiently high postflame temperature in each case. The initial composition for each mechanistic study was determined separately for the different conditions using the corresponding thermodynamic data. Note that the Wilcox rate constant for reaction R1 in Table 1 is defined for the temperature range 393-1500 K, whereas both the Sliger and UConn experiments involve lower temperatures. For these lower temperatures, a linear extrapolation of the Wilcox R1 rate constant was used. The reaction submechanism assembled by Qiu et al. (11) was modified slightly by removing two reactions involving SO2 species originally taken from the Leeds database (19), consistent with the revised version presented in the same. This modification ultimately did not have a bearing on overall Hg oxidation. In addition, thermodynamic parameters of some species including Hg0 were modified by including additional data (20, 21) to broaden the temperature range of applicability. The reported mechanism and thermodynamic data were otherwise taken “as-is” for each model, with the Niksa and Qiu submechanisms consisting of 160 and 158 elementary reactions, respectively. Beyond these steps, the development of the various kinetic mechanisms was not reviewed. The emphasis here was on identifying the models that best predicted homogeneous Hg oxidation under wellcontrolled experimental conditions. Overview of Model Prediction. Sliger et al. Dataset. Sliger et al. (9) studied homogeneous mercury oxidation in a downfired natural gas furnace having a multiple layer refractory design. The postflame gas at a temperature of 1195 K had a

FIGURE 1. Comparison of Hg oxidation in Sliger et al. experiments (9: I, b: II, 2: III) with the predictions of Niksa, Qiu and Wilcox/ Niksa rate constants (light solid, dotted, and dark solid curves for I, II, and III respectively). I, II, and III represent Hg levels of 53, 590, and 1390 µg/Nm3 respectively.

baseline composition of 6.2% CO2, 7.4% O2, 12.3% H2O, and 25 ppm NO. Mercury was injected into the burner as an acetate solution and three Hg concentrations of 53, 590, and 1390 µg/Nm3 (designated here as case I, II, and III respectively) were examined. HCl gas was introduced downstream of the burner and was varied in the range of 55-545 ppm. Residence time in the furnace was 1.4 s during which time the gases cooled by 55 K prior to collection through a quartz probe with a quench rate of approximately 5400 K/s. In modeling this system, the initial species composition was determined from equilibrium at 1195 K, and the concentration of NO was fixed at the measured furnace exhaust value of 25 ppm. Model predictions of Hg oxidation under the conditions of Sliger’s experiments using the Qiu and Niksa mechanisms are shown in Figure 1. The Niksa predictions presented here are largely identical to those reported by Niksa et al. in their original evaluation and deviate marginally for some cases (less than 4% error) possibly because of minor differences in the quench profiles. These predictions therefore serve as a benchmark for validating the simulation procedure and implementation of the Niksa mechanism in this paper. From Figure 1, it is seen that the Sliger et al. experiments are better matched by the predictions of the Niksa mechanism than the Qiu mechanism, particularly at HCl concentrations greater than 170 ppm. The Qiu mechanism consistently predicts lower extents of Hg oxidation than were observed in the experiments for all cases with HCl greater than 280 ppm. At lower HCl levels however it closely predicts the experimental observations at the lower two Hg concentrations. Both the Niksa and Qiu mechanisms appear to capture the overall profile of Hg oxidation observed in the Sliger experiments. To evaluate the Wilcox rate constants, the three Hg reactions specified in Table 1 (reactions R1, R5, and R7) were first combined with the five remaining Hg reactions in the eight step Hg scheme and the C-H-O-N-S-Cl submechanism of Niksa et al. (“Wilcox/Niksa”). A similar combination was also developed with the Hg reactions and submechanism reported by Qiu et al. (“Wilcox/Qiu”). The predictions of the Wilcox/Niksa combination are compared with predictions of the Niksa mechanism for the Sliger experiments in Figure 1. The predictions of Wilcox/Qiu are not presented as this combination of rate constants predicted complete oxidation under all conditions except for case III and 55 ppm HCl for which 83% oxidation was predicted. As shown in Figure 1, for conditions corresponding to case I, VOL. 41, NO. 22, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

7871

the lowest Hg concentration examined by Sliger et al., the Wilcox/Niksa combination predicts near-complete oxidation at HCl concentrations above 170 ppm, whereas the maximum measured extent of Hg oxidation was 41% at 280 ppm HCl. For case II, this combination overpredicts Hg oxidation by a factor of 2 at all HCl levels except the lowest concentration of 55 ppm, where it underpredicts the measurement (8%). At higher HCl levels (>280 ppm), predictions range from 90 to 99% oxidation. For case III, predictions are within 10% of measurements for HCl levels of 280-545 ppm, but significantly underpredict the measured extent of oxidation at 55 and 170 ppm HCl. All the mechanisms predict that Hg oxidation occurs in the quench stage alone in this system, where super-equilibrium concentrations of Cl atoms exist due to the high quench rate. The three cases examined in the Sliger dataset differ only in their Hg concentration, and measurements indicate that Hg oxidation increases with increasing Hg concentration. The Niksa and Qiu mechanisms, however, predict an increase in oxidation only as Hg concentration increases from case I to II, whereas lower oxidation is predicted for case III for the same HCl concentration. It is, therefore, inferred that the model predictions for Hg oxidation, though weakly dependent on Hg concentration, nevertheless exhibit a maximum at Hg concentrations intermediate to those of cases II and III. In contrast, an inverse dependence of the extent of oxidation on Hg concentration is apparent in the predictions using the Wilcox/Niksa set of rate constants for all three cases as the level of oxidation progressively drops with increasing Hg concentrations. UConn Dataset. The UConn experimental system (14, 22) consists of a quartz tube reactor with postcombustion cooling rates ranging from 200 to 400 K/s, a range representative of utility boilers. Flame conditions were generated by burning CH4 in O2 and adding N2 as a diluent at equivalence ratios (φ) of 0.9 and 0.98. Both Hg (from a permeation device) and Cl (as HCl or Cl2) were added to the system by injection into a mixing chamber located immediately downstream of the flame zone. An average gas concentration of 50 µg/Nm3 Hg was maintained, whereas HCl concentrations were varied from 100 to 300 ppm. Effects of NO and SO2 in the presence of Cl2 and HCl were also studied. The temperature profile of the system was measured for φ ) 0.9 as a part of this study, whereas for φ ) 0.98, an estimated profile previously reported by Qiu et al. (11) was used. Equilibrium compositions calculated at 1136 and 1398 K were used as initial species compositions for φ ) 0.9 and 0.98 respectively. Baseline values of NO at these equivalence ratios were fixed at the measured exhaust levels of 130 and 160 ppm. For the UConn experiments, the Niksa mechanism predicted negligible Hg oxidation under all HCl concentrations and flame equivalence ratios. Oxidation levels predicted by the Qiu mechanism were close to experimental values, particularly at an equivalence ratio of 0.9 as shown in Figure 2. For case II (φ ) 0.98), the Qiu predictions show a greater divergence from experiment than at other conditions. Potential inaccuracy in the estimated temperature profile may be contributing, although this factor alone is insufficient to explain the disparity between the model predictions and experiments. When the Wilcox Hg reactions are substituted into the Niksa and Qiu mechanisms following the procedure described above, varying results are observed in the model predictions. Whereas the Wilcox/Niksa combination predicts negligible oxidation for all conditions, the Wilcox/Qiu combination predicts near-complete oxidation. This is in contrast to predictions for UUtah dataset (discussed later) where the Wilcox/Qiu combination consistently predicted lower levels of oxidation in comparison to the Qiu mechanism 7872

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 22, 2007

FIGURE 2. Comparison of Hg oxidation in the presence of HCl observed in UConn experiments ((: I, 9: II) with the predictions of Qiu (solid curve: I, dotted curve: II) mechanism. I and II represent O ) 0.9 and 0.98, respectively.

FIGURE 3. Hg oxidation profile using the Wilcox reaction rate constants substituted in the Qiu mechanism compared against the experimental value (b) and the Qiu mechanism predictions for UConn experimental conditions of O ) 0.9 and 150 ppm HCl. for the same HCl level. To investigate this further, and to evaluate the effect of the individual Wilcox rate constants, a test case from the UConn dataset corresponding to case I (φ ) 0.9) with 150 ppm HCl was chosen. The Wilcox reactions R1, R5, and R7 (presented in Table 1) were both individually and as a combination (R1-R5, R1-R7, R5-R7) replaced in the Qiu mechanism and calculations then conducted for each of the resulting six cases. The Hg oxidation profile for these test runs, those of Qiu, and those of Wilcox/Qiu are presented in Figure 3. The cases where Wilcox R7 was substituted are not presented as this rate constant did not alter the predictions when substituted in the Qiu mechanism either alone or with R1 and/or R5. Wilcox’s rate constants for R1 and R5 caused an increase in the final oxidation by 45 and 25 (absolute) percent, respectively, compared to the baseline Qiu mechanism. Complete oxidation was predicted with both R1 and R5 replaced, and as expected, the Hg oxidation profile for R1/R5 in Figure 3 was identical to that of the Wilcox/Qiu combination (R1, R5, and R7 replaced). It should be noted from the oxidation profile presented in Figure 3 that whereas the final extent of oxidation was indeed high for the Wilcox rate constants, the extent of oxidation was lower than or comparable to the Qiu mechanism predictions for temperatures greater than 450 K. At lower temperatures, a sudden increase in oxidation is observed when the calculations are conducted with the Wilcox/Qiu rate constants. At a tem-

FIGURE 4. Effect of Cl2 and NO/SO2 on Hg oxidation in UConn experiments (hollow bars) compared with the predictions using the Qiu mechanism (solid bars). perature of approximately 450 K, Wilcox/Qiu predicts 10% Hg oxidation, which is identical to the measured value. At the lower temperature limit (393 K) prescribed by Wilcox for R1, Wilcox/Qiu predicts 90% oxidation, which is still significantly higher than both the experiments and the predictions of the Qiu mechanism. In addition to evaluating the effects of HCl levels and flame equivalence ratios on Hg oxidation, the effects of NO (30 and 100 ppm added to baseline NO of 130 ppm) and SO2 (100 and 400 ppm) in the presence of HCl and/or Cl2 were also studied at φ ) 0.9 using the two mechanisms as well as with the substituted Wilcox constants. Note that experimental data were not available under all conditions and that the cases with SO2 addition in the presence of HCl and NO are parametric test cases used to study model predictions for a broader range of conditions. Examining this broader range of conditions, the Niksa mechanism, both as is and with the Wilcox reaction rate constants substituted again predicted negligible oxidation. Complete Hg oxidation was, however, predicted by the Wilcox-substituted Qiu mechanism under all HCl concentrations with added NO and SO2. The UConn experimental data indicated that at 100 ppm HCl, addition of 30 and 105 ppm NO to the baseline NO resulted in negligible change in Hg oxidation, while at 300 ppm HCl, addition of 30 and 105 ppm NO caused 3 and 15% inhibition, respectively. Calculations of Hg oxidation using the Qiu mechanism did not deviate appreciably from baseline predictions for either case implying that the marginal inhibition observed experimentally is not predicted by these models. In contrast, greater inhibition due to SO2 addition was both experimentally observed and predicted. Measurements of Hg oxidation with SO2 (100, 400 ppm) and NO (30, 105 ppm) addition in the presence of Cl2 (250, 500 ppm) at φ ) 0.9 are presented in Figure 4. The results of calculations using the Qiu mechanism are also shown. Note that the baseline level of NO under all these conditions was 130 ppm. Addition of SO2 causes suppression of Hg oxidation at both 250 and 500 ppm Cl2 concentrations but the inhibition is lower at the higher concentrations of Cl2. Although the model predictions deviate from the measured extent of Hg oxidation under baseline Cl2 concentrations (Figure 4), the predictions for conditions with SO2 addition correlate well with observed values. For SO2 addition at the higher Cl2 concentration (500 ppm), the Qiu

mechanism predictions deviate by less than 4% from experimentally measured values. Similar predictions were obtained for the parametric test cases involving the addition of NO and SO2 in the presence of HCl (100, 300 ppm) at φ ) 0.9. Addition of 100 ppm NO to the baseline value of 130 ppm at both HCl concentrations resulted in negligible change in the predicted extents of Hg oxidation by the Qiu mechanism. The addition of SO2 resulted in 15-20% (absolute) inhibition under all cases, with the extent of inhibition less at the higher HCl concentration. Increasing the baseline NO concentration to 230 ppm for fixed concentrations of HCl (100, 300 ppm) and SO2 (100, 400 ppm) did not have any noticeable effect on the predictions. A sensitivity analysis was performed for the Qiu mechanism using the UConn experimental conditions of φ ) 0.9, 150 ppm HCl, and baseline NO, both in the presence and absence of 100 ppm SO2 to understand the inhibitory effect of SO2 addition on Hg oxidation. It was found that the three reactions involving S/Cl/O species that were part of the flue gas sub-mechanism did not alter the Cl species pool. The reaction HCl + OHdCl + H2O was found to be the primary pathway for releasing Cl atoms into the flue gas. Addition of SO2 resulted in the consumption of OH radicals through the reactions HOSO2 + O2dHO2 + SO3 and HO2 + OHdH2O + O2. Although the reverse reaction of NO2 + OHdHO2 + NO regenerated a fraction of OH, the destruction reactions were found to proceed at a faster rate resulting in a net reduction in the OH concentration. Consequently, lower amounts of Cl atoms were liberated in the presence of SO2 resulting in reductions in the level of Hg oxidation. Comparing the Clatom profile for the cases with and without SO2 addition, it was found that addition of 100 ppm SO2 reduced the peak Cl levels by a factor of 2 for the UConn experiments at φ ) 0.9, 150 ppm HCl, baseline NO. It is possible that reactions other than the ones stated above are important under different temperatures and concentrations of HCl, O2, SO2, and NO, and hence, the results of the above sensitivity analysis must be treated as “case-specific”. It is, however, suggested that the S/O/Cl reactions do not directly suppress Hg oxidation. Reductions in Cl levels with SO2 addition are achieved by altering concentrations of other flue gas species such as OH, which are responsible for releasing Cl atoms into the system. UUtah Dataset (15). Experiments at the University of Utah were conducted in a 1000 Btu/h down-fired natural gas furnace, which employed a premixed quartz glass burner. A quartz tube reactor of 4.7 cm internal diameter was enclosed in an electrical resistance heater up to a length of 132 cm. The tube extended 79 cm below the heater to the sampling point and was wrapped with heat tape and insulation, thus allowing control of the quench rate in this section. The mercury concentration in all experiments was 25 µg/dryNm3, and Cl (injected as Cl2) was varied between 0 and 600 ppm. Under these conditions, the baseline exhaust gas (wet basis) consisted of 18.1% H2O, 8.6% CO2, 0.72% O2, 27 ppmv NO, and 7.4 ppmv CO. The experimental conditions in this study are represented as (quench rate, residence time): I: 210 K/s, 6s; and II: 440 K/s, 7s. The flame equilibrium composition at 1100 K, a temperature close to the initial postflame temperature reported for the two cases, was used as the initial value in all calculations. This is not expected to affect predictions because in all cases, comparable maximum temperatures of approximately 1350 K were attained prior to quench. Apart from the minor change in initial temperature, the remainder of the profile was included as reported (15). The concentration of NO was fixed at the reported exhaust value of 27 ppm. Model evaluation of the UUtah data showed that the Niksa mechanism both alone and with the substitution of the Wilcox VOL. 41, NO. 22, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

7873

FIGURE 5. Evaluation of Hg oxidation in UUtah experiments (9: I, (: II) using Qiu mechanism (dark dotted and solid curves for I and II respectively) and Wilcox/Qiu (light dotted and solid curves for I and II, respectively) combination. rate constants predicted negligible Hg oxidation under all conditions considered. Recall that a similar result was obtained with these two sets of rate constants for the UConn dataset. In contrast, levels of oxidation predicted by the Qiu mechanism showed very good agreement with the UUtah data especially for case II (440 K/s). The model predictions of the Qiu and the Wilcox/Qiu mechanisms are compared with the experimental values in Figure 5. For the lower quench rate (case I), the Qiu mechanism overpredicts oxidation by 11 and 17% (absolute) at 200 and 300 ppm HCl, respectively, and underpredicts oxidation by an average of 7% (absolute) at higher HCl concentrations (400-600 ppm). The modified mechanism using the substituted Wilcox rate constants (Wilcox/Qiu) consistently predicted lower extents of oxidation than the Qiu mechanism alone. The levels of oxidation predicted by the former compare well with experiments at low HCl levels (100-300 ppm) for case I (210 K/s) and at high HCl levels (400-600 ppm) for case II (440 K/s). The divergence was large (greater than 30% absolute difference) for case I at HCl levels greater than 300 ppm, whereas for case II, the error at lower HCl levels up to 300 ppm was approximately 14% (absolute). Recall that for UConn dataset evaluations, complete oxidation was predicted using the Wilcox/Qiu combined rate constants under all conditions due to a substantial increase in oxidation at temperatures below 450 K. The Wilcox/Qiu predictions for the UUtah dataset do not show such high levels of oxidation, likely because the lowest temperature attained in the system was approximately 570 K. Based on results obtained in the UConn dataset evaluations, it is expected that Wilcox/Qiu combination would predict complete oxidation for the UUtah dataset as well if the flue gas was quenched further to temperatures lower than 450 K. To examine this, a modified quench was imposed on case II at the lowest Cl concentration (100 ppm HCl) such that the flue gas was allowed to cool from 570 to 300 K at 75 K/s. As expected, Hg oxidation predicted using Wilcox/Qiu was 99.9% at the end of the imposed simulation period, whereas the Qiu mechanism predicted approximately 63% oxidation under similar conditions. A sharp increase in oxidation below 450 K was observed for Wilcox/Qiu as the Hg oxidation increased from 50 to 90% during the flue gas quench from 450 to 420 K in 0.4s.

Discussion The Niksa mechanism has been shown to predict the extent of Hg oxidation fairly accurately for one experimental system and less well for others. This difference is attributed to 7874

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 22, 2007

differences in the predicted Cl-species profile at temperatures below 900 K. Most Hg oxidation is expected to occur at temperatures below 900 K. For the Sliger experiments, which include an initial near-isothermal high-temperature region followed by a very high quench (5400 K/s), the Niksa mechanism predicts a gradual decrease in Cl concentration prior to the onset of the rapid quench, during which time superequilibrium concentrations of Cl radicals exist and initiate and sustain Hg oxidation. For the other experimental studies, where quench rates are comparatively low, almost all Cl radicals are quenched by 900 K resulting in prediction of negligible Hg oxidation due to the nonavailability of chlorine in the appropriate temperature window. For homogeneous Hg oxidation to be predicted by this mechanism, it is likely that flue gases would need to be subjected to hightemperature isothermal conditions prior to a rapid quench to sustain a sufficient Cl atom pool. To explore this further, a parametric study of the Niksa mechanism was conducted for conditions relevant to the Sliger dataset. Case I at 170 ppm HCl was chosen, and a nominal quench rate of 200 K/s was used instead of the original value of 5400 K/s. As expected, a very low level of Hg oxidation (∼2%) was predicted at 200 K/s quench. In contrast, the Qiu mechanism generally predicts a maximum in the Cl atom concentration before the initiation of recombination and therefore sustains the Cl radical pool in the required temperature window for all experiments. In summary, the Qiu homogeneous Hg oxidation mechanism was found to provide quantitative agreement with the broadest set of experimental data. Although the Niksa mechanism predictions were closest to the one dataset generated under conditions of extremely rapid quench initiating at high temperatures, it did not demonstrate good agreement with the two other experimental systems considered. Parametric tests suggest that for oxidation to be predicted accurately by the Niksa mechanism, very high quench rates are required. Incorporation of the Wilcox reaction rate constants, which were developed using theoretical methods, resulted in poorer agreement with experimental data especially for systems involving final temperatures lower than 450 K. The sudden and rapid increase in oxidation below 450 K predicted by the Wilcox/Niksa combination is primarily attributed to Wilcox rate constant R1 and to a lesser extent R5. Although the value of the Wilcox rate constant for R1 was linearly extrapolated to temperatures below the lowest limit prescribed by Wilcox (393 K), even at 393 K, the predicted extents of oxidation were substantially higher than the measured values or those predicted using the Qiu mechanism. It should be considered that the theoretical rate constants calculated by Wilcox were incorporated into a mechanism with rate constants involving empirical approximations. These empirical rate constants also contribute to the predictions of Hg oxidation when using the Wilcox substituted rate constants. It is, therefore, possible that a complete set of theoretical rate constants building on the initial three provided by Wilcox might provide better agreement with experimental data. To ascertain the utility of transition state theory rate constants, the rate constants for all Hg oxidation reactions must be calculated using such methods for the entire temperature range of application and must be validated using experimental data.

Acknowledgments This study was sponsored by the U.S. Department of Energy under the contract no. DE-PS26-03NT41634-9. We also thank Professor Jianrong Qiu of Huazhong University of Science and Technology and Dr. Stephen Niksa of NEA for providing their mechanisms and for helpful discussions.

Literature Cited (1) U.S. Environmental Protection Agency. Mercury; http:// www.epa.gov/mercury. (2) U.S. Environmental Protection Agency. Mercury; http://www.epa.gov/mercury/exposure.htm. (3) Niksa, S.; Fujiwara, N. Predicting extents of mercury oxidation in coal-derived flue gases. J. Air Waste Manage. 2005, 55, 930939. (4) Frandsen, F.; Dam-Johansen, K.; Rasmussen, P. Trace elements from combustion and gasification of coal- An equilibrium approach. Prog. Energy Combust. Sci. 1994, 20, 115-138. (5) Senior, C.L.; Sarofim, A.F.; Zeng, T.; Helble, J.J.; Mamani-Paco, R. Gas-phase transformtions of mercury in coal-fired power plants. Fuel Process. Technol. 2000, 63, 197-213. (6) Chu, P.; Porcella, D.B. Mercury stack emissions from U.S. utility power plants. Water, Air, Soil Pollut. 1995, 80, 135-144. (7) U.S. Environmental Protection Agency. Electric Steam Generating Units; http://www.epa.gov/ttn/atw/combust/utiltox/ utoxpg.html. (8) Widmer, N.C.; West, J.; Cole, J.A. Proceedings of the Air and Waste Management Association Annual Conference, Salt Lake City; AWMA: Pittsburgh, PA, 2000. (9) Sliger, R.N.; Kramlich, J.C.; Marinov, N.M. Towards the development of a chemical kinetic model for the homogeneous oxidation of mercury by chlorine species. Fuel Process. Technol. 2000, 65-66, 423-438. (10) Niksa, S.; Helble, J.J.; Fujiwara, N. Kinetic modeling of homogeneous mercury oxidation: The importance of NO and H2O in predicting oxidation in coal-derived systems. Environ. Sci. Technol. 2001, 35, 3701-3706. (11) Qiu, J.; Helble, J.J.; Sterling, R. Proceedings of the 12th International Conference on Coal Science, Cairns, Australia, 2003. (12) Edwards, J.R.; Srivastava, R.K.; Kilgroe, J.D. A study of gas-phase mercury speciation using detailed chemical kinetics. J. Air Waste Manage. 2001, 51, 869-877.

(13) Xu, M.; Qiao, Y.; Zheng, C.; Li, L.; Liu, J. Modeling of homogeneous mercury speciation using detailed chemical kinetics. Combust. Flame 2003, 132, 208-218. (14) Sterling, R.O.; Qiu, J.; Helble, J.J. Proceedings of the 227th ACS Spring National Meeting, Anaheim, CA, 2004; American Chemical Society: Washington, DC, 2004. (15) Fry, A.; Cauch, B.; Lighty, J.S.; Silcox, G.D.; Senior, C.L. Proceedings of the 31st International Symposium on Combustion, Heidelberg, Germany, 2006. (16) Wilcox, J; Robles, J.; Marsden, D; Blowers, P. Theoretically predicted rate constants for mercury oxidation by hydrogen chloride in coal combustion flue gases. Environ. Sci. Technol. 2003, 37, 4199-4204. (17) Wilcox, J.; Marsden, D.; Blowers, P. Evaluation of basis sets and theoretical methods for estimating rate constants of mercury oxidation reactions involving chlorine. Fuel Process. Technol. 2004, 85, 391-400. (18) Wilcox, J.; Blowers, P. Decomposition of mercuric chloride and application to combustion flue gases. Environ. Chem. 2004, 1, 166-171. (19) The University of Leeds and E¨ otv¨os University Combustion Simulations. Sulfur Mechanism Extension; http:// garfield.chem.elte.hu/Combustion/sox.htm. (20) http://webbook.nist.gov/chemistry/. (21) http://cea.grc.nasa.gov. (22) Mamani-Paco, R.; Helble, J.J. Proceedings of the Air and Waste Management Association Annual Conference, Salt Lake City; AWMA: Pittsburgh, PA, 2000.

Received for review May 9, 2007. Accepted August 20, 2007. ES071087S

VOL. 41, NO. 22, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

7875