Absorption of HCl and SO2 from Humidified Flue Gas with Calcium

1048

Ind. Eng. Chem. Res. 2000, 39, 1048-1060

Absorption of HCl and SO2 from Humidified Flue Gas with Calcium Silicate Solids Paul N. Chisholm and Gary T. Rochelle* Department of Chemical Engineering, The University of Texas at Austin, Austin, Texas 78712-1062

The absorption of HCl and SO2 with calcium silicate was studied in a bench-scale, fixed-bed reactor at 120 °C. From 0 to 3.5% relative humidity (RH), an increase in relative humidity increased sorbent utilization by reaction with HCl. From 3.5 to 19% RH, final sorbent loading by HCl was constant at 1.4 mol/mol of Ca2+. The absorption rate of HCl was first-order in HCl concentration from 250 to 3250 ppm. When calcium silicate was exposed to HCl and SO2 simultaneously in the absence of O2 or NO2, no SO2 remained loaded with the solids at the end of an experiment. Any SO2 that was absorbed was eventually emitted from the solids in favor of increased HCl absorption. The addition of O2 to the simulated flue gas caused improved SO2 absorption but had little effect on HCl absorption. A dramatic increase in final SO2 loading and a decrease in final HCl loading was observed when NO2 was added to the gas stream. Adding 50 ppm NO2 increased SO2 loading from 0 to 0.73 mol/mol of Ca2+ and decreased HCl loading from 1.4 to 0.61 mol/mol of Ca2+. In the presence of NO2, increasing the SO2/HCl inlet ratio increased final SO2 loading and decreased final HCl loading. In experiments without HCl from 90 to 150 °C, it was found that a low concentration of NO2 increased final SO2 loading more at higher temperatures. The experimental data from the fixed bed were modeled using a modification of the shrinking core model. Flux equations and estimated parameters were then used to predict the performance of HCl and SO2 absorption by calcium silicate on the surface of a bag filter. The predictions suggested that, at reasonable gas conditions in the absence of SO2, HCl penetration through the bag filter can be reduced below 20%. With simulated municipal waste combustion flue gas (low SO2/HCl ratio) with 50 ppm NO2, HCl and SO2 penetration can be reduced to less than 5%. At coal-fired boiler conditions (high SO2/HCl ratio) with 50 ppm NO2, HCl penetration can be reduced to 2% while SO2 penetration was predicted to be 40%. 1. Introduction Historically, SO2 emissions have received more attention than other acid gases because of the presence of high levels of sulfur in fuels, particularly coal. As a result of the Clean Air Act Amendments passed in 1990, emissions of other acid gases such as HCl are being more closely regulated. High levels of gaseous HCl and SO2 can be found in the waste gases from municipal waste combustion (Donnelly, 1991; Holzman and Atkins, 1988), hazardous waste incineration, and coal-fired boilers. The U.S. EPA has promulgated regulations governing the emissions of HCl and SO2 from municipal waste combustors (U.S. EPA, 1995) and hospital waste incinerators (U.S. EPA, 1997). To meet acid gas control regulations, a variety of control strategies are employed. One strategy is to remove the acid gas by injecting dry, alkaline solids into the flue duct. The gas will react with the sorbent in the duct and on the surface of the existing particulate control device. The advantages of dry sorbent injection over competing technologies are its operational simplicity, reduced capital cost, and small footprint (White and Vancil, 1989). The major disadvantage of dry sorbent injection is that it requires a high sorbent feed rate to compensate for low sorbent utilization. The most commonly used sorbent is hydrated lime. A high surface area sorbent made by slurrying hydrated lime with gypsum and a silica source has been * To whom correspondence should be addressed.

developed (Arthur and Rochelle, 1998) to increase sorbent utilization. The calcium silicate solids have been shown to be more reactive with SO2 and NO2 than hydrated lime (Nelli and Rochelle, 1998; Jozewicz et al., 1988; Chu and Rochelle, 1989). Less detailed studies of the reaction of HCl with calcium silicate solids have also been performed (Jozewicz et al., 1990; Cain, 1993). The objective of this work is obtain an understanding of the absorption of HCl and SO2 with calcium silicate solids. An emphasis will be placed on understanding the interactions of other gases usually present in flue gass NO, NO2, O2, and water vaporson HCl and SO2 reactivity.

2. Sorbent Preparation and Acid Gas Reaction Chemistry 2.1. Sorbent Preparation. The calcium silicate solids used for all the experiments discussed in this study were prepared by slurrying hydrated lime, pulverized, post-consumer glass, and gypsum in a 1:1:0.5 mass ratio. The slurry was 20 wt % solids and was maintained at 92 °C. After 50 h, the slurry was filtered, then vacuum-dried at 90 °C and >30 in. Hg vaccuum, and sieved through an 80-mesh screen to break up the powder. A thorough treatment of the effect of slurrying conditions is given in Arthur (1998). The production of amorphous calcium silicate hydrates took place by the

10.1021/ie990493k CCC: $19.00 © 2000 American Chemical Society Published on Web 03/14/2000

Ind. Eng. Chem. Res., Vol. 39, No. 4, 2000 1049

following mechanism:

Ca(OH)2 T Ca2+ + 2OH(SiO2)x + 2H2O + OH- T (SiO2)x-1 + Si(OH)5-

(1) (2)

Ca2+ + ySi(OH)5- + (2 - y)OH- + (z - 2y -1)H2O T (CaO)(SiO2)y(H2O)z (3) The formation of calcium silicate solids began with the dissolution of hydrated lime. Exposed to a high-pH solution, the silica framework broke down and silica was released. The silica then reacted with calcium and precipitated as a calcium silica hydrate. The value of y for the solids used in this study is 0.67 (Arthur, 1998). Gypsum was added to promote high surface area formation by increasing the concentration of calcium ions in the slurry (Kind et al., 1994). 2.2. Acid Gas Reaction Chemistry. The absorption of HCl, SO2, and NO2 can be achieved by irreversible acid-base reactions:

participate directly in S(IV) oxidation:

(CaO)(SiO2)y(H2O)z + 2HCl f CaCl2‚2H2O + yH2SiO3 + (z - y - 1)H2O (4)

3. Sorbent Characterization and Experimental Methods

(CaO)(SiO2)y(H2O)z + SO2 f CaSO3‚1/2H2O + yH2SiO3 + (z - y - 1/2)H2O (5) (CaO)(SiO2)y(H2O)z + 3NO2 f Ca(NO3)2 + NO + (z - y)H2O + yH2SiO3 (6) Calculations based on thermodynamic data from Sinke et al. (1985) and Garvin et al. (1987) suggested that calcium chloride undergoes a phase transition at 120 °C when the relative humidity (RH) is 2.2%. Above 2.2% RH, the dihydrate is stable; below 2.2% RH, the monohydrate is the stable calcium chloride salt. A more detailed discussion of the phase transition calculation is given later. At the temperature and relative humidity range investigated, the most likely solid phase for calcium sulfite was the hemihydrate (Jones et al., 1976). In this study, it was discovered that when CaSO3‚ 1/ H O is exposed to HCl vapor, SO is evolved and 2 2 2 CaCl2‚2H2O is formed. This interaction is given by reaction (7):

CaSO3‚1/2H2O+ 2HCl + 1/2H2O T CaCl2‚2H2O + SO2 (7) When oxygen is present in the gas phase, S(IV) is oxidized to S(VI). According to Jones et al. (1976), the dihydrate of calcium sulfate (gypsum) converts to a hemihydrate between 120 and 140 °C. Because most of this study was performed at the lower border of the temperature range, it was assumed that S(VI) exists as gypsum:

CaSO3‚1/2H2O+ 1/2O2 + 3/2H2O f CaSO4‚2H2O

(8)

NO2 is expected to act as a catalyst promoting S(IV) oxidation by reaction (8) (Shen, 1997). NO2 may also

Figure 1. Pore volume distribution of calcium silicate solids and hydrated lime as determined by nitrogen porosimetry.

(CaO)(SiO2)y(H2O)z + CaSO3‚1/2H2O + 2NO2 f CaSO4‚2H2O + Ca(NO2)2 + yH2SiO3 + (z - 3/2 - y)H2O (9)

3.1. Sorbent Characterization. As determined by the BET method, the surface area of the calcium silicate sorbent was found to be 87.5 ((0.7) m2/g. The pore volume distribution of the calcium silicate solids as determined by nitrogen porosimetry is shown in Figure 1. Hydrated lime (Mississippi Lime Co.) is shown for comparison. X-ray diffraction analysis of the calcium silicate showed that the solids were almost completely amorphous (Arthur and Rochelle, 1998). To determine the capacity of the sorbent to absorb acid gases (defined herein as alkalinity), the solids were dissolved in an HCl solution at a pH of ≈3. After the solution was titrated to a phenolphthalein end point with NaOH, the alkalinity of the calcium silicate was found to be 6.49 mmol of CaO/g of sorbent. The mass of calcium and sodium was found to be 6.20 and 0.894 mmol/g of sorbent, respectively, when atomic absorption (AA) spectroscopy was used. Note that, throughout this text, alkalinity from both Ca2+ and Na+ will be lumped together and referred to as moles of Ca2+. 3.2. Experimental Apparatus and Methods. This discussion of the experimental system will be a brief overview. A more detailed treatment of the methods and apparatus is given in Chisholm and Rochelle (1999). The contacting of acid gases with the sorbent took place in a Pyrex glass reactor that was 2.8 cm in diameter and 16.9 cm in length. The sorbent, dispersed in an excess of quartz sand, rested at the bottom of the reactor on a 0.4-cm-thick extra-course frit. Pressure drop through the bed did not exceed 7 kPa. The simulated flue gas was blended and sent to the reactor which rested submerged in a fluidized sand bath. After the acid gases reacted with the sorbent, the reactor outlet was diluted heavily with house air and sent to an on-line FT-IR system. The FT-IR system consisted of a gas cell with an equivalent length of 24 m as well as the IR source and detector. Water which adsorbed on the surfaces of the gas cell led to a time lag in the analytical system, especially with HCl. The FT-IR generated a full spectrum every 15 s. HCl, SO2, and NO2 all absorbed at least moderately well

1050

Ind. Eng. Chem. Res., Vol. 39, No. 4, 2000

Table 1. Experimental Conditions and Final Loading Results (Flow Rate for All the Experiments Was 1.5 SLPM and the Calcium Silicate Loading Was 36 mg) exp. no.

RH (%)

T (°C)

time (min)

HCl (ppm)

SO2 (ppm)

O2 (%)

1 2 3 4 5 6 7 8 9 10c 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27d 28

0 0.8 1.8 3.5 9 19 3.5 3.5 3.5 3.5 19 19 19 19 19 19 19 19 0.8 9 19 1.5 1.5 1.5 10 10 3.5 19

120 120 120 120 120 120 120 120 120 120 120 120 120 120 120 120 120 120 120 120 120 150 150 150 90 90 120 120

10 45 45 45 45 45 60 45 45 45 100 80 50 90 90 80 75 80 50 75 75 45 45 45 45 45 50 60

1000 1000 1000 1000 1000 1000 250 500 2000 3250 250 250 0 250 250 250 250 250 250 250 1000 0 0 0 0 0 1000 1000

0 0 0 0 0 0 0 0 0 0 1000 1000 1000 1000 1000 1000 1000 1000 1000 1000 250 1000 1000 1000 1000 1000 0 250

0 0 0 0 0 0 0 0 0 0 0 5.5 2.5 2.5 0 2.5 0 2.5 2.5 2.5 2.5 2.5 2.5 2.5 2.5 2.5 0 2.5

29 30

9 19

120 120

45 45

0 0

1000 1000

2.5 2.5

a

HCl ldga

SO2 ldga

NO2 ldga

util.b(%)

0.33 0.746 0.862 1.421 1.342 1.412 1.374 1.283 1.499 1.397 1.440 1.332 0 0.612 0.853 0.503 1.427 1.116 0.344 0.745 1.244 0 0 0 0 0 1.468 1.324

0 0 0 0 0 0 0 0 0 0 0 0.153 0.350 0.732 0.901 1.040 0.009 0.331 0.135 0.716 0.197 0.091 0.324 0.322 0.306 0.542 0 0.145

0 0 0 0 0 0 0 0 0 0 0 0 0 0.259 0.690 0.411 0 0 0.051 0.174 0.067 0 0.022 0.060 0 0.183 0 0

16.5 37.3 43.1 71.1 67.1 70.6 68.7 64.2 75.0 69.9 72.0 81.9 35.0 120 167 150 72.3 88.9 30.7 118 85.3 9.10 33.5 35.2 30.6 63.4 73.4 80.7

0 0

0.761 0.915

0.208 0.115

86.5 97.2

0 0 0 0 0 0 0 0 0 0 0 0 0 NO2:50 NO2:150 NO2:150 NO:150 NO:150 NO2:50 NO2:50 NO2:50 0 NO2:25 NO2:50 0 NO2:150 0 NO: 140 NO2: 10 NO2:150 NO2:150

Loading of acid gas A is given in mol of A/mol of Ca2+. b Total utilization takes into account acid-base reaction stoichiometry: 1

total utilization ) c

NOx (ppm)

Total flow rate: 0.75 SLPM.

d

/2nHCl + nSO2 + 1/2nNO2

Calcium silicate loading: 72 mg.

nCa

× 100

in the IR band at specific wavenumbers. The extent to which one of these molecules absorbed was directly correlated with concentration. With the knowledge of what the concentrations are at both the inlet and the outlet of the reactor, the rates of the reactions of HCl, SO2, and NO2 could be determined by a gas-phase mass balance. 4. Results and Discussion The experimental results that were obtained can be grouped in three sets of gas systems: (a) HCl; (b) HCl, SO2, NOx, and O2; (c) SO2, NOx, and O2. The first two sets, all at 120 °C, were the original scope of this study. However, experimental data from these first two sets led to additional experiments that focused on gaining an understanding of the interactions of SO2 and NO2 as a function of temperature and relative humidity. The experimental conditions and final loading of HCl, SO2, and NO2 are shown in Table 1. 4.1. HCl. The literature has shown that, at low temperature (25-175 °C), the reactivity of acid gases with calcium-based sorbents increases with increasing relative humidity. Generally, increasing relative humidity up to about 90% increases the utilization of the solids monotonically. As seen in Figure 2, HCl loading at the end of the experiments increased from 0.32 to 1.42 mol/ mol of Ca2+ when the relative humidity was increased from 0 to 3.5%. However, at 3.5, 9, and 19% RH, HCl loading throughout the experiments was not a strong function of relative humidity. In each of these three experiments, the HCl removal approached 100% until

Figure 2. Effect of relative humidity on HCl removal and loading. Experiments performed at 1000 ppm HCl, 120 °C, 36 mg of calcium silicate, and 1.5 SLPM. Experiments 1-6 (Table 1).

about 0.8 mol/mole Ca2+ and then fell off to zero HCl removal at approximately 1.4 mol/mol of Ca2+. Because these results were unexpected, each of the 3.5, 9, and 19% RH experiments were reproduced two more times. Typical results are shown in Figure 2 when multiple experiments were performed at the same conditions. It appears that the maximum calcium silicate loading is 1.4 mol of HCl/mol of Ca2+ or 70% utilization. A calculation using the densities of calcium silicate (Kind et al., 1994), calcium chloride dihydrate, and silicic acid (Perry et al., 1984) and the void volume of calcium silicate indicated that the pores would become plugged at an HCl loading of 1.58 mol/mol of Ca2+. It is possible that an increase in relative humidity above 3.5% did


not increase the final HCl loading because the pores became plugged. Sinke et al. (1985) reported that the triple point for CaCl2‚H2O, CaCl2‚2H2O, and saturated calcium chloride existed at 176 °C. With use of this information and the heats of formation from Garvin et al. (1987), it was calculated that, at 120 °C and a relative humidity less than 2.2%, the monohydrate was the stable solid phase. At a relative humidity greater than 2.2%, the dihydrate salt was the stable solid phase. It is possible that increasing the relative humidity above 2.2% did not increase the HCl loading because the larger dihydrate salt blocked the pores. Because the pores were blocked, HCl could not diffuse into the center of the particle and react with the 30% of the calcium that remained unutilized. In an attempt to confirm the existence of two different levels of hydration, two experiments, one at 1.8% RH and the other at 3.5% RH, were performed where calcium silicate was dispersed in glass wool instead of quartz sand. The conditions were the same as experiments 3 and 4 (Table 1). X-ray powder diffraction analyses were performed on samples of both reacted sorbents in an attempt to understand the discontinuity in the effect of relative humidity. The X-ray analyses (Chisholm, 1999) show that both samples contained the dihydrate of calcium chloride based on a standard from Hanawalt et al. (1938). Clearly, these results did not confirm the calculation regarding the hydrate phase transition discussed above. It is possible that, during the time interval (∼ 30 min) between extracting the solids from the reactor and loading them into the X-ray diffractometer, the monohydrate may have converted to the dihydrate salt. The results of the study by Sinke et al. (1985) may support this conclusion as they found the monohydrate to be unstable. Scanning electron microscopy (SEM) pictures were taken on the same reaction products analyzed with X-ray powder diffraction. The objective of the SEM studies was to determine if there was any visible difference in the reaction products of the experiments conducted at 1.8 and 3.5% RH. Both samples looked nearly identical under the electron microscope and very similar to the fresh sorbent (Chisholm, 1999). Consequently, the SEM analyses, like the X-ray powder diffraction analyses, did not confirm different levels of hydration of the reaction products. An analysis of the reactivity of HCl as a function of concentration is more straightforward than that of the reactivity as a function of relative humidity. It is shown in Figure 3 that, at concentrations ranging from 250 to 3250 ppm, HCl removal as a function of HCl loading is zero-order with respect to concentration within experimental error. This indicates that the reaction rate is first-order with respect to HCl concentration at a fixed HCl loading. The modeling work discussion in a later section addresses the dependence of HCl concentration on absorption in more detail. 4.2. HCl, SO2, NOx, and O2. 4.2.1. Effect of O2. Most of the experiments which included SO2 were performed at 19% RH and 120 °C. This high humidity was chosen because HCl, SO2, and NO2 reactivity generally increased with increasing relative humidity. Conducting experiments at 19% RH allowed for better quantification of acid gas interactions. It was found that when both HCl and SO2 were present in the absence of an S(IV) oxidizing agent (O2 or NO2), the absorption of HCl would

Figure 3. Effect of HCl concentration on removal. Experiments performed at 3.5% RH, 120 °C, 36 mg of calcium silicate, and 1.5 SLPM. Experiments 4 and 7-10 (Table 1).

Figure 4. Effect of time and O2 and HCl concentrations on SO2 loading. Experiments performed at 19% RH, 1000 ppm SO2, 120 °C, 36 mg of calcium silicate, and 1.5 SLPM. Experiments 11-13 (Table 1).

lead to the emission of SO2 from the reacted solids. This phenomenon is shown by the lower curve in Figure 4. When both HCl and SO2 were present in the gas phase, each gas had a moving reactivity zone in the bed. In these zones, HCl reacted to form CaCl2‚2H2O while SO2 formed CaSO3‚1/2H2O. When the calcium silicate solids that were available for reaction were mostly consumed, HCl reacted with any available calcium sulfite per reaction (7). Consequently, SO2 was emitted from the solids in favor of increased HCl reactivity. As shown in Figure 4, the fixed bed continued to remove SO2 through about 10 min. Then, the reaction of calcium sulfite with HCl caused SO2 to be released, decreasing the SO2 loading. The curve showing HCl loading as a function of time was essentially the same with 1000 ppm SO2 as without any SO2. This fact suggested that the absorption rates of HCl with both calcium silicate and CaSO3‚1/2H2O are approximately the same. When 5.5% O2 was added to the gas stream, SO2 would react with the solids and maintain the SO2 loading after about 10 min from the start of the test run. The presence of O2 allowed for the oxidation of S(IV) to S(VI) per reaction (8). Unlike calcium sulfite hemihydrate, when gypsum was exposed to HCl, it was stable and would not emit SO2. The HCl loading at the end of the experiment was ≈0.07 mol/mol of Ca2+ less when SO2 and O2 were present than when they were absent from the gas mixture. The total loading with SO2 and O2 present was 1.49 mol/mol of Ca2+; the total loading in the HCl-only system was 1.42 mol/mol of Ca2+.

1052


Figure 5. Effect of time and NO2 on SO2 absorption. The experiment was performed at 19% RH, 250 ppm HCl, 1000 ppm SO2, 120 °C, 36 mg of calcium silicate, and 1.5 SLPM. Experiments 12 and 14-16 (Table 1).

Figure 6. Effect of time and NO2 on HCl loading. The experiment was performed at 19% RH, 250 ppm HCl, 1000 ppm SO2, 120 °C, 36 mg of calcium silicate, and 1.5 SLPM. Experiments 12 and 1416 (Table 1).

4.2.2. Effect of NO2. The addition of NO2 to the simulated flue gas greatly enhanced the final SO2 absorption. This effect is shown in Figure 5. Assuming there was no effect of varying oxygen concentration from 2.5 to 5.5%, the addition of 50 ppm NO2 increased the SO2 loading to 0.73 from 0.15 mol/mol of Ca2+. Tripling the NO2 concentration to 150 ppm NO2 led to a 42% increase in SO2 loading. The increase in SO2 reactivity with calcium silicate by the addition of NO2 and O2 was generally at the expense of reduced reactivity of HCl. This reduction is apparent in Figure 6. In the presence of varying levels of NO2, the nature of the competition can change from a system dominated by HCl reactivity to one dominated by SO2. With the addition of 150 ppm NO2, the final SO2 loading increased by a factor of about 7. Conversely, adding 150 ppm NO2 decreased the final HCl loading by a factor of 2.6. If it were assumed that the only reaction products of HCl, SO2, and NO2 with calcium silicate were the chloride, sulfite (or sulfate), and nitrate (or nitrite) salts of calcium, there would be a limit to the absorption capacity of the sorbent. On the basis of the stoichiometry of these salts, utilization can be defined in terms of the moles of HCl, SO2, and NO2 loaded onto the solids: 1

utilization )

/2nHCl + nSO2 + 1/2nNO2 nCa

× 100

(1)

If only calcium salts were the reaction products, the

maximum utilization would have been 100%. However, in the experiments with NO2 and O2, utilization was calculated to be as high as 167% with SO2 absorption accounting for 104% utilization. This fact suggested that there may be other reaction products. Others (Nelli, 1997; Jarvis et al., 1985) have found sulfur-nitrogen species such as hydroxyaminedisulfonate (HADS) and aminedisulfonate (ADS) when reaction systems have included SO2 and NO2. With respect to stoichiometry, neither HADS nor ADS consume any of the available calcium or sodium. Therefore, any sulfur-nitrogen compound formation would be like “free” reactivity because the SO2 or NO2 absorption would not be at the expense of alkalinity. An ion chromatography (Varian 2000i/SP with a Dionex AS4A column) method similar to Nelli’s (1997) was developed to analyze for HADS and ADS. Reaction products were recovered from the reactor at the end of the experiments with the conditions of experiments 16 and 26 and analyzed to determine the reaction products. The conditions of the experiment were set to maximize SO2 absorption: high humidity (19% RH), low HCl concentration (0 and 250 ppm), high SO2 concentration (1000 ppm), and high NO2 concentration (150 ppm) with O2 added (2.5%). Only chloride, sulfate, nitrate, and low levels of nitrite and sulfite were present in the reaction products solution. No HADS, ADS, or any other unidentified peaks were seen. With the conditions of experiment 26, the nitrogen balance of the reaction products closed to within 11% of the gas-phase balance. Though calculations for chlorine and nitrogen balances were less precise with experiment 16 because of overlapping chromatograph (IC) peaks, the mass balances for chlorine and nitrogen closed within 20% and 5%, respectively. The inconclusive results of the IC analysis made it difficult to confidently ascertain the reaction products. It is possible that the true reaction products changed when the solids were dissolved in water at room temperature. It was also suspected that the production of H2SO4 in the layer of surface water on the solids may have accounted for the SO2 absorption in excess of 100% utilization. When the reacted solids were removed and dissolved in 100 mL of water, the pH of the resulting solution was measured to be 4.3. If it were assumed that all the acid gas loading over 100% utilization was due to H2SO4 formation, the pH of the solution with the reacted solids would have to have been ≈2.9. A potentiometric titration of fresh calcium silicate solids was performed to check for the existence of any buffers such as bicarbonate that may have led to utilization in excess of 100%. No buffers were found. In summary, the results of the IC analysis, the pH calculation of the reacted solid solution, and the potentiometric titration did not lend any clear insight into the mechanism of the “superabsorption” of SO2. 4.2.3. Effect of NO. The simultaneous addition of NO and O2 together to the simulated flue gas increased SO2 absorption. This is evident in the top curve shown in Figure 7. The addition of just NO did not increase SO2 absorption. As was mentioned above, adding 5.5% O2 prevented the desorption of SO2 by reaction (7). When both NO and O2 were present, low levels of NO2 were seen in the IR spectrum. It is likely that the reaction

NO + 1/2O2 f NO2

(10)

took place in the experimental equipment, perhaps in


Figure 7. Effect of time, NO, and O2 on SO2 loading. The experiments were performed at 19% RH, 250 ppm HCl, 1000 ppm SO2, 120 °C, 36 mg of calcium silicate, and 1.5 SLPM. Experiments 11, 12, 17, and 18 (Table 1).

Figure 8. Effect of time and relative humidity on SO2 and NO2 loading. Experiments were performed at 250 ppm HCl, 1000 ppm SO2, 50 ppm NO2, 120 °C, 36 mg of calcium silicate, and 1.5 SLPM. Experiments 14, 19, and 20 (Table 1).

the water vaporization furnace. This low level of NO2 in the reactor inlet then led to slightly increased levels of SO2 absorption. The elevated final SO2 loading of 0.33 mol/mol of Ca2+ with 150 ppm NO and 2.5% O2 present was much less than the final loading when the gas inlet included 50 ppm NO2 (Figure 5). 4.2.4. Effect of Varying Relative Humidity. All of the experiments discussed above were performed at 19% RH. It was shown earlier that increasing the relative humidity above 3.5% had no positive effect on HCl reactivity. However, it was expected that high humidity levels provided for increased reactivity of SO2 (Chisholm, 1999). The 19% RH at 120 °C corresponds to 35 vol % water vapor, a concentration greater than what is generally found in most municipal waste combustors and coal-fired boiler flue gases. Therefore, two additional experiments, both at 120 °C, were conducted at 9% RH (16 vol % water) and 0.8% RH (1.5 vol % water). The comparison of the results from these experiments with one conducted at 19% RH is shown in Figure 8. The final SO2 loading for the 9 and 19% RH experiments shown in Figure 8 were essentially the same at 0.73 and 0.72 mol/mol of Ca2+, respectively. These data show that, within experimental error, HCl and SO2 react the same under these conditions at 9 and 19% RH. As shown in Figure 2, it was not expected that HCl absorption would change significantly between the 9 and 19% RH. However, other experiments (Chisholm, 1999) involving just SO2 showed increased SO2 loading

Figure 9. Effect of time and SO2/HCl inlet ratio on HCl and SO2 loading. Experiments performed at 19% RH, 120 °C, 50 ppm NO2, 2.5% O2, 36 mg of calcium silicate, and 1.5 SLPM. Experiments 14 and 21 (Table 1).

when the humidity was increased from 9 to 19%. Because the HCl loading profile for the two highhumidity experiments were the same, the SO2 that passed through the sorbent bed was exposed to the same mass of the chloride salt in both experiments. It appears that the deliquescent nature of calcium chloride caused a fixed level of surface moisture to be associated with the solids in both high-humidity experiments. Therefore, the important parameter that dictated SO2 absorption was surface moisture, not vapor-phase water concentration. Provided there was enough NO2 and O2 available to prevent the emission of SO2, the calcium chloride bound to the solids eliminated relative humidity as the important variable. Experiments were performed without HCl at 9 and 19% RH and 120 °C with 1000 ppm SO2, 150 ppm NO2, and 2.5% O2. The experiment at 9% RH showed a final SO2 loading of 0.76 mol/mol of Ca2+ while the experiment at 19% RH had a final SO2 loading of 0.92 mol/ mol of Ca2+. These results along with those presented in Figure 8 support the idea that the presence of calcium chloride on the surface of the sorbent compensated for the differences in humidity. When the relative humidity was reduced to 0.8% RH, the HCl loading at the end of the experiment decreased to 0.34 mol/mol of Ca2+. Without sufficient calcium chloride associated with the solids or enough humidity, the reactivity of SO2 dropped considerably. Though there was sufficient NO2 and O2 to prevent emission of SO2, there was insufficient moisture to induce the absorption of SO2. 4.2.5. Effect of Varying HCl and SO2 Concentrations. Most of the experiments discussed thus far were conducted with 250 ppm HCl and 1000 ppm SO2. When these concentrations were reversed, the reactivity of HCl increased considerably at the expense of SO2 (Figure 9). Both experiments show that, in the presence of 50 ppm NO2 and 2.5% O2, both HCl and SO2 loadings increase dramatically with concentration. 4.3. SO2, NOx, and O2. As shown in Figure 5, the final SO2 loading in the presence of HCl increased from 0.15 to 1.04 mol/mol of Ca2+ when 150 ppm NO2 was added to the gas. In the absence of HCl, Nelli (1997) noticed an increase in final SO2 loading from 0.71 to 0.91 mol/mol of Ca2+ at 70 °C and 60% RH when the NO2 concentration increased from 0 to 223 ppm. To better understand the role of temperature in the SO2NO2 interaction, several experiments were conducted

1054


temperature has just been cooled 60 °C without the injection of water. SO2 absorption is greater at the lower temperature and at greater relative humidity. Even if a high level of NO2 ()25 ppm NO2) is present in the high-temperature flue gas, SO2 loading and sorbent utilization will not increase beyond the low-temperature utilization. 5. Fixed-Bed Modeling and Bag Filter Predictions

Figure 10. Effect of time and NO2 on SO2 loading at 150 °C. The experiments were conducted at 1.5% RH, 0 ppm HCl, 1000 ppm SO2, 36 mg of calcium silicate, and 1.5 SLPM. Experiments 2224 (Table 1).

5.1. Model Description. The primary objective of modeling was to identify flux expressions and regressed parameters describing the absorption of HCl and SO2 by calcium silicate solids. These equations and parameters were then used to predict the removal of HCl and SO2 from flue gas on the surface of a bag filter. 5.1.1. Fixed-Bed Mass Balance. The modeling approach was similar to the methods used to analyze the reaction of acid gases with hydrated lime (Chisholm and Rochelle, 1999). The first step in modeling the gas system was to set up the appropriate mass-balance equations for the absorption of a gas by a solid in a fixedbed reactor. Unsteady-state experiments were performed at integral conditions where a concentration gradient through the bed existed in the same direction as the gas flow. Assuming that the spatial gradient is greater than the time gradient (von Rosenberg et al., 1977), the gas concentration is described by eq 2:

FA ) -

Figure 11. Effect of time and NO2 on SO2 loading at 90 °C. Experiments conducted at 10% RH, 1000 ppm SO2, 2.5% O2, 36 mg of calcium silicate, and 1.5 SLPM. Experiments 25 and 26 (Table 1).

at varying temperatures in the absence of HCl. Three experiments performed at 150 °C are shown in Figure 10. This set of experiments suggested that there existed a ceiling concentration of NO2 above which additional NO2 did not assist SO2 absorption. The maximum benefit of NO2 addition at 150 °C was seen at some concentration less than or equal to 25 ppm NO2. Assuming that the presence of HCl does not strongly impact the SO2-NO2 interaction, Figure 5 shows that changing the NO2 concentration from 50 to 150 ppm had a definite positive influence on SO2 absorption. It may be concluded then that, at 120 °C, the ceiling level of NO2-assisted SO2 absorption was above 150 ppm NO2. As seen in Figure 10, the addition of 25 ppm NO2 to 1000 ppm SO2 at 1.5% RH and 150 °C increased the final SO2 loading from 0.090 to 0.32 mol/mol of Ca2+ or by 256%. Experiments were also conducted at 10% RH and 90 °C. These data are shown in Figure 11. Increasing the NO2 concentration from 0 to 150 ppm increased the final SO2 loading from 0.31 to 0.54 mol/mol of Ca2+ or by 74%. It is clear that low levels of NO2 can have a much greater positive impact on SO2 absorption at higher temperature. The experiments conducted at 1.5% RH and 150 °C had the same water concentration as those at 10% RH and 90 °C. In a real system, a flue gas at the lower

Q dCA aAx dz

(2)

where FA is the flux of A to the sorbent in the fixed bed, CA is the concentration of species A, z is the axial distance in the bed, Q is the gas flow rate, a is the sorbent surface area per unit volume, and Ax is the cross-sectional area of the reactor. The solid-phase reactivity is described by eq 3:

FA ) -

R dx R dCS ) a dt Sm dt

(3)

where CS is the concentration of the sorbent, t is the experimental time, R is a stoichiometric factor, Sm is the molar surface area of the sorbent, and x is the conversion. An equation describing the localized flux of gas A was assumed (discussed below). The experimental time and reactor distance from eqs 1 and 2 were discretized into finite elements and Euler’s method was employed. Calculations were performed through the bed to generate an outlet concentration, CA,L|τ from the reactor at each time, τ. During an estimation, time and distance intervals were reduced until a decrease in an interval led to a zero decrease in the sum of squares to three significant figures. For SO2, CSO2,L|τ was compared directly to the datum taken at time τ, CSO2|τ. However, with the more soluble HCl, water absorption on the surfaces of the gas cell led to confounding effects. Therefore, a time lag correction was applied to HCl data. Details of this correction are given in Chisholm and Rochelle (1999). 5.1.2. Development of Flux Equation. Many researchers have employed shrinking core models, either with particles or grains within particles (Chisholm and Rochelle, 1999; Wang et al., 1996; Pakrasi, 1992; Fon-

Ind. Eng. Chem. Res., Vol. 39, No. 4, 2000 1055 Table 2. Results of Parameter Estimation for the HCl-Calcium Silicate System exp. no.

RH (%)

2

0.8

3

1.8

4 7 8 9 27 5 6

HCl conc. (ppm)

sorbent loading (mg)

1000 2σ: 1000 2σ:

36 36

Deff × 105 (m/s)

ks × 104 (m/s)

xT

mean model errora (%)

0.178 0.014 0.275 0.032

13.6 2.2 21.3 0.12

0.37c

4.05

0.43c

2.61

simultaneous mean model errorb (%)

This Set of Parameters Is from the Simultaneous Modeling of Experiments at 3.5, 9, and 19% RH experiments 4-9 and 31 2.23 6.22 0.644 2σ: 0.18 3.25 0.006 3.5 1000 36 1.53 6.22d 0.641 3.86 2σ: 0.13 0.017 3.5 250 36 2.80 4.88 0.670 4.12 2σ: 0.35 2.83 0.009 3.5 500 36 2.35 9.43 0.603 3.77 2σ: 0.37 13.30 0.011 3.5 2000 36 1.93 12.0 0.642 3.46 2σ: 0.48 10.7 0.024 3.5 1000 72 2.58 15.6 0.617 2.43 2σ: 0.14 6.8 0.013 9 1000 36 1.17 6.22d 0.680 4.05 2σ: 0.069 0.011 19 1000 36 2.76 2.67 0.626 2.61 2σ: 0.58 1.37 0.012

4.56 4.65 5.72 4.48 3.49 6.45 2.73 2.67

a The error reported is the square root of the average model error squared multiplied by 100 when the experiment was modeled individually. b The error reported is the square root of the average model error squared multiplied by 100 when the experiments were modeled simultaneously. c The xT value was fixed at the value determined by the gas-phase mass balance to allow for independent estimation of Deff and ks. d Value was fixed at the value determined by the simultaneous regression value to allow for independent estimation of Deff and xT.

seca et al., 1998). Assuming first-order kinetics and diffusion through the product layer were the limiting rates in gas absorption, the following equation describes the flux of gas A,

FA )

CA δ 1 + k s DA

(4)

where ks is the first-order surface rate constant, δ is the product layer thickness, and DA is the diffusion coefficient of the diffusing species through the product layer. In the development of eq 4, planar rather than spherical geometry was used. This was done primarily to simplify the flux expression. In addition, when the sorbents absorbed the acid gases studied, the volume of the particles expanded. This effect was ignored also in the interest of simplicity. As shown in eq 3, the flux was a function of the product layer thickness. This thickness was not measured at any time. Because conversion was proportional to product layer thickness, δ was replaced by δ0x. The value DA/δ0 will be referred to as Deff,A. The shrinking core model predicts that the rate will not drop to zero until all the sorbent has been consumed. However, it has been seen in this study and throughout the literature that absorption rates often drop to zero before the sorbent is fully converted. Because of the discrepancy between the shrinking core model and the experimental results, a driving force based on conversion was superimposed on eq 4:

FA )

CA(xT - x) l x + ks Deff,A

(5)

xT is defined as the conversion when the reactivity of the sorbent terminated. The three regressed parameters

in eq 4 are xT, ks, and Deff,A. The parameters were regressed using an estimation package (Caracotsios, 1986). Principles leading to eq 4 were used with both the single-gas system (HCl) and with the multigas system (HCl and SO2). However, another modification to the flux equation for SO2 was found to be necessary. This modification will be discussed later. 5.1.3. Bag Filter Performance Model. The objective of the modeling effort was to predict how well hydrated lime and calcium silicate would perform absorbing HCl and SO2 on the surface of a bag filter. The bag filter performance model assumed that the flux expressions that govern the absorption of HCl and SO2 in the experimental system also described the flux expressions for the baghouse. One of the key operational conditions that impacted the performance predictions was the sorbent feed ratio. This ratio was the number of moles of sorbent fed to the system divided by the number of moles needed to remove all the acid gas assuming 100% conversion of the solids by simple acid-base reactions. For example, if 1 mol each of SO2 and HCl entered a flue duct, a sorbent feed ratio of 2 meant 3 mol (2 × (1 mol + 0.5 mol)) of sorbent was fed into the duct. This study did not include the flow rate of NO2 when the sorbent feed ratio was calculated. The second important operational variable in a bag filter system was the cycle timesthe period between cleanings of the bag. Bag filters are generally cleaned by a reversed air stream, by a pulse jet, or by simply shaking it. After a bag was cleaned, it was assumed that there was no cake on the surface of the filter. Consequently, acid gas penetration from that bag was greatest immediately after being cleaned and decreased with increasing cycle time. The penetration of HCl or SO2 was calculated by averaging, for a given cycle time, the instantaneous penetration over the entire cycle. 5.2. HCl Only. 5.2.1. Parameter Estimation. The results of the parameter estimations for the system with calcium silicate absorbing HCl are given in Table 2. For

1056


each of the experiments listed in the table, a separate regression was performed. For the experiments at 3.5, 9, and 19% RH, no effect on HCl absorption was seen when the relative humidity, HCl concentration, or sorbent loading was changed. For this reason, the last seven experiments shown in Table 2 were regressed simultaneously to produce global absorption parameters. Regressed values of Deff and xT were found to be constant at each concentration, loading, and relative humidity at 3.5, 9, 19%. This was expected as the experiments showed no effect of these variables over the ranges mentioned. The value of ks, though with considerable scatter, was found to be constant for all the experiments. There were two minor problems with the parameter estimation results shown in Table 2. For the experiments at 0.8 and 1.8% RH, the value of xT was set at the value determined experimentally. Without setting xT for these two regressions, the estimation routine had difficulty converging on a reasonable solution. The second problem was that the values of ks were found to have wide confidence intervals. In two of the regressions, the value of ks was set to the value determined by the global estimation. Problems with xT at low humidity and ks arose because of a moderately high (≈0.8) normalized covariance between Deff and ks. The large covariance was probably due to the fact that after a finite element had undergone a few time steps of contact with HCl, x changed at a very slow rate. Consequently, the two terms in the denominator of eq 4 were of similar formsa constant. Therefore, the value of ks was determined over only a few time steps before considerable conversion takes place in a finite element. The difficulties in obtaining accurate Deff and ks values with small confidence intervals persisted throughout the modeling effort. The mean model error for the set of parameters was an indicator of how well the model fit the data. For the set of data modeled simultaneously, the error for an experiment with the global set of parameters was greater than the error for the individually regressed set of parameters. This was due to the fact that, with an experiment being regressed individually, the parameters can adjust to any specific trend or error of that experiment. Constrained by a global averaging from the six other experiments, experimental errors are “averaged out” and the estimation routine can more accurately determine parameters. A plot comparing the model fit to the data for a typical experiment (experiment 8) based on the results from the simultaneous regression is shown in Figure 12. The ratio of CHCl/CHCl,0 is the reactor outlet concentration normalized by the inlet concentration. 5.2.2. Bag Filter Performance Projections. The parameters estimated from the fixed-bed reactor were used to estimate the absorption dynamics of HCl by calcium silicate in a bag filter system. The same flux equation was used for the performance projections as that used in the fixed-bed modeling. Figure 13 shows the effect on HCl penetration of varying HCl concentrations while holding the sorbent feed ratio constant. As the HCl concentration increases, the mass of the sorbent fed to the system also increased. Therefore, there is more sorbent available to react with greater concentrations of HCl. The fraction of HCl absorbs immediately

Figure 12. Typical fit using the parameters from the simultaneous regression of all HCl experiments greater than 1.8% RH. Experiments performed at 3.5% RH, 500 ppm HCl, 36 mg of calcium silicate, 120 °C, and 1.5 SLPM. Experiment 8 (Table 1).

Figure 13. Penetration of HCl as a function of HCl concentration in a bag filter system. Performance projection with a sorbent feed ratio of 1.5, 3.5% RH, and 120 °C. The curves use the global parameters given in Table 2.

once it contacts fresh sorbent; this was the same for all concentrations per unit bed thickness. But because the bed is thicker with more inlet HCl, a greater fraction of HCl is removed by solids at high (but less than xT) conversion. The curve labeled “constant cake depth” in Figure 13 plots the penetration of HCl at 250, 500, and 1000 if the bag filter was cleaned when sorbent accumulation on the filter reached a given depth. This curve assumed that the only contribution to the filter cake was sorbent addition. Because the sorbent feed rate (not ratio) is 4 times greater with 1000 ppm HCl versus 250 ppm, the time required to accumulate the same cake thickness for the 1000 ppm case was a quarter of the time for the 250 ppm case. When this thickness was reached, the penetration at 250 ppm was essentially the same as that at 1000 ppm. The constant cake depth curve was added to point out that when taking pressure drop into account, increased HCl concentration did not significantly decrease HCl penetration. Perhaps the most useful depiction of the results of the performance projection is the curve shown in Figure 14. The utilization, defined as the average through the filter cake of the moles of HCl absorbed adjusted for reaction stoichiometry, is an indication of how much of the sorbent is actually used. Ideally, low HCl penetration would have been possible at high sorbent utiliza-


III. For x g xT, FSO2 )

k0 x

(6c)

where x at time τ is the sum of the moles of HCl, nHCl, and SO2, nSO2, that have been absorbed since time zero, adjusted for reaction stoichiometry and divided by the initial moles of alkalinity: 1

xτ )

Figure 14. HCl penetration as a function of utilization at varying cycle times. Performance projection with 3.5% RH and 1000 ppm HCl at 120 °C. The curves use the global parameters given in Table 2.

tion. The shape of the curves shown in Figure 14 illustrates the tradeoff between increasing sorbent utilization while at the same time increasing HCl penetration. For a cycling time of 20 min, this figure indicates that to remove greater than 90% of the inlet gas at 1000 ppm HCl and 3.5% RH, less than half of the sorbent will actually be involved in HCl absorption. A utilization of 50% with a 40-min cycle time translates to a sorbent feed ratio of 2 and an HCl removal of about 95%. 5.3. HCl-SO2. 5.3.1. Parameter Estimation. The addition of SO2 to the inlet gas stream led to an increase in the complexity of the modeling approach, especially with the concurrent addition of NO2 and O2 with SO2. Though analysis of the gas system became more complex, the flux equation for HCl was the same as the one used for the HCl-only system given by eq 4. With respect to SO2 absorption, it was shown in Figures 4 and 5 that the presence of O2 and NO2 as oxidizing agents of S(IV) had a positive influence on SO2 absorption. The positive impact of oxidizing agents on SO2 absorption was particularly clear over long experimental times when SO2 absorption was reduced to a slow but steady rate. At a high inlet SO2/HCl ratio, the presence of NO2 led to SO2 absorption rates never quite dropping to zero. The HCl absorption rate, on the contrary, did drop to zero when the conversion of the solids reached xT. While the flux equation for SO2 absorption was similar in form to the flux equation for HCl, the flux equation for SO2 took into account long-term, nonzero absorption rates. The equations describing SO2 absorption took the following form:

I. For x e 0.95xT, FSO2 )

CSO2(xT - x)

(6a)

1 x + ks,SCHCl Deff,S

II. For 0.95xT e x e xT, FSO2 )

CSO2(xT - x) 1 x + ks,SCHCl Deff,S

+

k0 x

(6b)

/2nHCl|τ + nSO2|τ nCa

(7)

Definitions for ks,S and Deff,S are analogous to those for HCl presented in eq 5. k0 is an empirical constant with units of flux. The SO2 flux was described by two mechanisms: (1) reaction with the sorbent and (2) longterm absorption, perhaps due to sulfur-nitrogen compound formation. It was necessary to break down the flux equation into three regimes because once conversion became greater than the termination conversion, xT, the term (xT - x) became negative. With this term negative, the model predicted the desorption of both HCl and SO2. The transition regime between these two mechanisms is described by eq 6b. The lower boundary, 0.95xT, was set somewhat arbitrarily. Neither estimated parameters nor the sum of squares was sensitive to the transition regime lower border from 0.75xT to 0.97xT. It was necessary to include a transition zone to allow for a smooth transition from the regime governed by sorbent absorption to the regime with long-term, slow SO2 absorption. The expression for the long-term SO2 absorption included a 1/x term because it was noted that the absorption rate, though fairly steady, did slow slightly as utilization values increased in excess of 100%. At early experimental times (x f 0) eqs 6a and 6b described the SO2 flux as being first-order in both SO2 and HCl concentrations. It is believed that the deliquescent nature of CaCl2 salts improved SO2 absorption. This same effect was seen by other researchers (Matsukata et al., 1996; Jozewicz et al., 1990). It was noted earlier that an increase in relative humidity over calcium silicate increased SO2 absorption. If the solids had a surface solution that was saturated to CaCl2, the partial pressure of water over the solids would have increased. This increase in water vapor partial pressure, then, would have increased SO2 absorption. Several other forms of eq 6a were regressed, including equations where the order on SO2 concentration was varied widely. However, only eq 6a estimated the same approximate value of ks with experiments ranging from 250 ppm HCl/ 1000 ppm SO2 to 1000 ppm HCl/250 ppm SO2. While the flux equation employed for SO2 absorption clearly did not evolve from first principles, it is important to recall that the primary objective was to describe the experimental data. With HCl andSO2 in the inlet gas stream, eqs 4 (HCl) and 5 (SO2) indeed described the data effectively. Parameters estimations for seven experiments from the HCl-SO2 are given in Table 3. Note that Deff,H and ks,H are regressed constants analogous to those presented in eq 4. The HCl and SO2 flux equations do not explicitly account for the effects of O2 or NO2. Instead, the values for the regressed parameters were allowed to vary at different O2 and NO2 concentrations. Consequently, the parameters themselves account for different degrees of

1058


Table 3. Results of Parameter Estimation for the HCl-SO2 System exp. no.

HCl conc. (ppm)

SO2 conc. (ppm)

NO2 conc. (ppm)

O2 conc. (%)

Deff,H × 105 (m/s)

ks,H × 105 (m/s)

16

250

150

2.5

17

250

150

0

14

250

50

2.5

20b

250

50

2.5

12

250

0

5.5

21

1000

50

2.5

28

1000

1000 2σ: 1000 2σ: 1000 2σ: 1000 2σ: 1000 2σ: 250 2σ: 250

NO:140 NO2:10

2.5

11.3 8.0 5.43 1.20 6.38 1.93 d d 56.2 207 2.33 0.45 2.22

5.73 2.12 7.68 1.13 6.62 1.95 4.88 1.95 2.93 0.025 15.0 6.2 3.22

0.48

2.18

2σ:

Deff,S × 106 (m/s)

ks,S × 102 (m4/m/s)

k0 × 109 (mol/ m2/s)

errora (%)

0.935 0.037 0.942 0.040 0.943 0.048 0.843 0.074 0.843 0.013 0.827 0.019 0.766

4.37 0.43 1.30 0.11 2.08 0.20 2.87 1.39 0.083 c 7.82 0.85 5.33

2.35 1.54 4.45 1.09 2.78 1.88 0.503 0.370 15.4 d 2.17 1.15 5.15

8.0 c 8.0 c 3.8 c 5.2 c 0 c 0 c 0

4.08

0.023

0.79

4.63

c

xT

4.10 4.75 6.82 3.81 3.32 4.10

a

The error reported is the square root of the average model error squared multiplied by 100 when the experiment was modeled individually. b Experiment performed at 9% RH. c Parameter was determined manually to two significant figures and does not have a 2σ estimate. d Parameter was determined to be very large or had a very large confidence interval.

Figure 15. Typical model fit for HCl-SO2. Experiment performed at 19% RH, 250 ppm HCl, 1000 ppm SO2, 50 ppm NO2, and 2.5% O2 with 36 mg of calcium silicate at 120 °C and 1.5 SLPM. Experiment 14 (Table 1).

S(IV) oxidation. The value of k0 was determined by manually adjusting it to match the average slope of conversion versus time after xT was surpassed. The values for k0 given in Table 3 are the best values within two significant figures. Generally, the estimation routine would estimate k0 to be zero with an infinite 2σ value because the concentration was not sufficiently sensitive to k0. However, with k0 set to zero, the outlet concentration from the reactor was systematically overpredicted. This problem led to an underprediction of sorbent conversion at values greater than xT. To capture the “superabsorption” of SO2 by calcium silicate well in excess of 100% utilization, k0 was estimated manually. The comparison between the data and the modeled concentration is shown in Figure 15 for both HCl and SO2. The ratio of CA/CA,0 is the reactor outlet concentration of A normalized by the inlet concentration. Similar to the regression results for HCl absorption, problems arose due to large normalized covariance ()0.8) between Deff,H, ks,H, and ks,S. With large covariance, some of the parameters had rather wide confidence intervals. Wide confidence intervals for a parameter indicated that the experimental data were not particularly sensitive to that parameter. However, when the parameters were used to predict bag filter performance, the bag filter model was perhaps more more sensitive to some terms than the fixed-bed model. Figure 16 shows the effect of changing a parameter from its

Figure 16. Effect on penetration of varying ks,S from the center to the top and bottom of the confidence interval. Performance projection with a calcium silicate feed ratio of 1.5 at 19% RH, 250 ppm, 1000 ppm SO2, 150 ppm NO2, 2.5% O2, and 120 °C. Experiment 16.

regressed value to the top and bottom of the 2σ confidence interval. In general, ks,S tended to give the widest confidence intervals so it was used as the example for the figure. The curves shown in Figure 16 provide a qualitative measurement of the limits on the precision of the bag filter performance projections due to large confidence intervals. On the basis of the example shown in Figure 16, the effect of varying a parameter through its 2σ confidence interval seemed to have less of an effect on HCl penetration than on SO2 penetration. There was a considerable effect on SO2 penetration when ks,S was decreased to the bottom of the confidence interval. SO2 penetration increased from 33 to 51% when ks,S was decreased from 2.35 to 0.81 m4/m/s. Experiment 16 was chosen as the example for discussion because the results of the parameter estimation produced relatively large confidence intervals. With this in mind, the results from experiment 16 may represent a worst case scenario. 5.3.2. Bag Filter Performance Projections. Plots of HCl and SO2 penetration as a function of sorbent utilization are shown in Figures 17 and 18, respectively, where the solid data points refer to coal-fired flue gas and open data points to municipal waste combustion flue gas. The model predicted similar penetrations of HCl with a municipal waste combustion and a coal-fired


Figure 17. HCl penetration as a function of utilization at varying inlet SO2/HCl concentration ratios in a bag filter system. Performance projection at a 40-min cycle time with 19% RH, 50 ppm NO2, and 2.5% O2 at 120 °C. The curves were generated from parameters listed in Table 3 for experiments 14 and 21.

Figure 19. Comparison of the performance of calcium silicate and hydrated lime. Performance projection at a 40-min cycle time with 19% RH, 250 ppm HCl, 1000 ppm SO2, and 2.5% O2 at 120 °C. The calcium silicate curve was generated from parameters listed in Table 3 for experiment 14. The hydrated lime curve was taken from Chisholm and Rochelle (1999).

drated lime. A previous study (Chisholm and Rochelle, 1999) generated data and a model similar to the one presented in this study but with hydrated lime employed as the sorbent. In general, the absorption of HCl alone at high humidity was similar with both sorbents. At 3.5% RH, the maximum loading of HCl (xT) was much greater with calcium silicate compared to that with hydrated lime. When HCl and SO2 were absorbed simultaneously, the addition of NO2 enhanced SO2 absorption much more with calcium silicate as the sorbent than with hydrated lime as the sorbent. The comparison of the performance of calcium silicate and hydrated lime in Figure 19 is shown as an example. The dissertation of Chisholm (1999) presents a more thorough comparison of the two sorbents. 7. Conclusions

Figure 18. SO2 penetration as a function of utilization at varying inlet SO2/HCl concentration ratios in a bag filter system. Performance projection at a 40-min cycle time with 19% RH, 50 ppm NO2, and 2.5% O2 at 120 °C. The curves were generated from parameters listed in Table 3 for experiments 14 and 21.

boiler flue gas. The only difference in the two curves was at high utilization; penetration of HCl with a SO2/ HCl ratio of 4.0 is less than that with a ratio of 0.25. This difference arose because the sorbent exhibited a greater capacity (xT) with a high SO2/HCl feed ratio. As seen in Figure 18, differences in SO2 penetration were considerable as the SO2/HCl ratio varied. At the conditions of municipal waste combustion, SO2 penetration decreased rapidly from 100% to almost 5% when the utilization decreased from 80 to 60%. At coal-fired boiler conditions, SO2 penetration decreased much less as the utilization was reduced from 85 to 40%. Below 40% utilization, adding additional sorbent (and reducing utilization) caused a sharp reduction in SO2 penetration. The results from 40 to 85% utilization at a high SO2/ HCl ratio suggested that as more sorbent is added, it preferentially reacts with HCl. When only 1% of the HCl penetrated (a utilization ) 40%), the slope on the SO2 penetration curve increased, indicating that the additional sorbent was reacting with SO2. 6. Comparison of Hydrated Lime with Calcium Silicate One of the broad objectives of this work was to compare the performance of calcium silicate with hy-

With HCl as the only acid gas, utilization and removal increased at 3.5% compared to 0% RH. No increase in HCl reactivity was seen with increased relative humidity at 19% compared with that at 3.5% RH. HCl absorption was determined to be first-order in a HCl concentration from 250 to 3250 ppm. In the absence of any oxidizing species (O2 or NO2), feeding HCl and SO2 simultaneously led to HCl reacting completely with any deposited S(IV). Therefore, the solids at the end of the experiment had a loading of HCl but no loading of SO2. With the addition of O2, loaded S(IV) was oxidized to S(VI). This oxidation led to a nonzero loading of SO2 at the end of the experiments. At low HCl and high SO2 concentrations, adding 150 ppm NO2 helped the calcium silicate solids to absorb almost 7 times the SO2 absorbed in the absence of NO2. In gas systems with a high HCl concentration and low SO2 concentrations, the loading of HCl increased at the expense of SO2 loading. The reverse was also true with gas systems where the SO2 concentration was high and HCl concentration was low. In studies with SO2, NO2, and O2 in the absence of HCl, low concentrations of NO2 were found to facilitate greater SO2 absorption at higher temperatures. Experimental evidence suggested that at a given temperature there existed an NO2 concentration above which no added SO2 loading took place. This NO2 concentration ceiling decreased with increasing temperature. Semiempirical flux equations for HCl and SO2 that fit the data well were developed. A parameter estima-

1060


tion routine was employed to estimate parameters in the flux equations. Regressed parameters were used to predict the removal of HCl and SO2 in a bag filter system. On the basis of bag filter predictions, calcium silicate is an effective sorbent for HCl control at moderate-tohigh relative humidity. At a relative humidity of 3.5% or greater and reasonable operating conditions, bag filter performance projections for HCl absorption by calcium silicate showed that penetration can be dropped to almost 10%. At reasonable operating conditions, less than 20% penetration of HCl was possible with a sorbent utilization of 50%. In the presence of NO2, calcium silicate solids were predicted to be effective for simultaneous HCl and SO2 control, especially at a low SO2/HCl ratio. Penetration of HCl was predicted to be low at nearly all the conditions studied. SO2 penetration decreased with an increasing SO2/HCl ratio. When HCl and SO2 are simultaneously absorbed by calcium silicate, an inlet SO2/HCl ratio of 0.25 led to less than 5% penetration of both HCl and SO2 in the presence of 50 ppm NO2. As the SO2/HCl ratio increased, the reactivity of HCl increased and SO2 reactivity decreased. At reasonable bag filter conditions for a coal-fired boiler flue gas, 2% HCl and 40% SO2 penetration was predicted with a sorbent utilization of 50%. Literature Cited Arthur, L. F. Silicate Sorbents for Flue Gas Cleaning. Ph.D. Dissertation, University of Texas at Austin, Austin, TX, 1998. Arthur, L. F.; Rochelle, G. T. Preparation of Calcium Silicate Absorbent from Recycled Glass. Environ. Prog. 1998, 17 (2), 86. Cain, P. M. Removal of Hydrogen Chloride from Flue Gas by Dry Sorbents Prepared from the Hydrothermal Reaction of Silica and Calcium Hydroxide. Ph.D. Dissertation, The University of Tennessee, Knoxville, TN, 1993. Caracotsios, M. Model Parametric Sensitivity Analysis and Nonlinear Parameter Estimation. Theory and Applications. Ph.D. Dissertation, The University of Wisconsin-Madison, Madison, WI, 1986. Chisholm, P. N. Absorption of Hydrogen Chloride and Sulfur Dioxide with Calcium-Based Sorbents from Humidified Flue Gas. Ph.D. Dissertation, The University of Texas at Austin, Austin, TX, 1999. Chisholm, P. N.; Rochelle, G. T. Dry Absorption of HCl and SO2 with Hydrated Lime from Humidified Flue Gas. Ind. Eng. Chem. Res. 1999, 38, 4068. Chu, P.; Rochelle, G. T. Removal of SO2 and NOx from Stack Gas by Reaction with Calcium Hydroxide Solids. J. Air Pollut. Control Assoc. 1989, 39, 175. Donnelly, J. R. Overview of Air Pollution Controls for Municipal Waste Combustors. In Proceedings, Second International Conference on Municipal Waste Combustion, Tampa, FL, 1991; U.S. Environmental Protection Agency: Washington, DC, 1991. Fonseca, A. M.; O Ä rfa˜o, J. J.; Salcedo, R. L. Kinetic Modeling of the Reaction of HCl and Solid Lime at Low Temperatures. Ind. Eng. Chem. Res. 1998, 37, 4570. Garvin, D.; Parker, V. B.; White, H. J., Jr. CODTA Thermodynamic Tables; Hemisphere: New York, 1987. Hanawalt, J.; Rinn, H.; Frevel, L. Chemical Analysis by X-ray Diffraction. Classification and Use of X-Ray Diffraction Patterns. Anal. Chem. 1938, 10, 457.

Holzman, M. I.; Atkins, R. S. Retrofitting Acid Gas Controls: A Comparison of Technologies. Solid Waste Power 1988, 2 (5), 28. Jarvis, J. B.; Nassos, P. A.; Stewart, D. A. A Study of SulfurNitrogen Compounds in Wet Lime/Limestone FGD Systems. In Proceedings, EPA/EPRI Symposium on Flue Gas Desulfurization, Cincinnatti, OH, 1985; Electric Power Research Institute: Palo Alto, CA, 1985. Jones, B. F.; Lowell, P. S.; Meserole, F. B. Experimental and Theoretical Studies of Solid Solution Formation in Lime and Limestone SO2 Scrubbers; EPA-600/2-76-273a; Environmental Protection Agency: Washington, DC, 1976. Jozewicz, W.; Jorgensen, C.; Chang, J. C. S.; Sedman, C. B.; Brna, T. G. Development and Pilot Plant Evaluation of SilicaEnhanced Lime Sorbents for Dry Flue Gas Desulfurization. J. Air Pollut. Control Assoc. 1988, 38, 796. Jozewicz, W.; Chang, J. C. S.; Sedman, C. B. Bench-Scale Evaluation of Calcium Sorbents for Acid Gas Emission Control. Environ. Prog. 1990, 9 (3), 137. Kind, K. K.; Wasserman, P. D.; Rochelle, G. T. Effects of Salts on Preparation and Use of Calcium Silicates for Flue Gas Desulfurization. Environ. Sci. Technol. 1994, 28, 277. Matsukata, M.; Takeda, K.; Miyatani, T.; Ueyama, K. Simultaneous Chlorination and Sulphation of Calcined Limestone.Chem. Eng. Sci. 1996, 51 (11), 2529. Nelli, C. H. Nitrogen Dioxide Removal by Calcium Silicate Solids. Ph.D. Dissertation, University of Texas at Austin, Austin, TX, 1997. Nelli, C. H.; Rochelle, G. T. Simultaneous Sulfur Dioxide and Nitrogen Dioxide Removal by Calcium Hydroxide and Calcium Silicate Solids. J. Air Waste Manage. Assoc. 1998, 48, 174. Pakrasi, A. Kinetic Studies on the Removal of Hydrogen Chloride from Flue Gas by Hydrated Lime Powders in a Bench Scale Fixed Bed Reactor. Ph.D. Dissertation, The University of Tennessee, Knoxville, TN, 1992. Perry, R. H.; Green, D. W.; Maloney, J. O. Perry’s Chemical Engineers’ Handbook, 6th ed.; McGraw-Hill: New York, 1984. Shen, C. H. Nitrogen Dioxide Absorption in Aqueous Sodium Sulfite. Ph.D. Dissertation, The Department of Chemical Engineering, The University of Texas at Austin, Austin, TX, 1997. Sinke, G. C.; Mossner, E. H.; Curnutt, J. L. Enthalpies of Solution and Solubilities of Calcium Chloride and Its Lower Hydrates. J. Chem. Thermodyn. 1985, 17, 893. U.S. EPA. Standards of Performance for New Stationary Sources and Emission Guidelines for Existing Sources: Municipal Waste Combustors; 40 CFR 65387; U.S. Environmental Protection Agency: Washington, DC, Dec 19, 1995. U.S. EPA. Standards of Performance for New Stationary Sources and Emission Guidelines for Existing Sources: Hospital/ Medical/Infectious Waste Incinerators; 40 CFR 48348; U.S. Environmental Protection Agency: Washington, DC, Sept 15, 1997. von Rosenberg, D. U.; Chambers, R. P.; Swan, G. A. Numerical Solution of Surface Controlled Fixed-Bed Adsorption. Ind. Eng. Chem. Res. Fundam. 1977, 16 (1), 154. Wang, W.; Ye, Z.; Bjerle, I. The Kinetics of the Reaction of Hydrogen Chloride with Fresh and Spent Ca-Based Desulfurization Sorbents. Fuel 1996, 75 (2), 207. White, D. M.; Vancil, M. A. Review of Dry Injection Technology for Reducing Emissions from Municipal Waste Combustors. In Proceedings, First International Conference on Municipal Waste Combustion; Hollywood, CA, 1989; U.S. Environmental Protection Agency: Washington, DC, 1989.

Received for review July 12, 1999 Revised manuscript received January 19, 2000 Accepted February 7, 2000 IE990493K

Absorption of HCl and SO2 from Humidified Flue Gas with Calcium

Recommend Documents