Dry Absorption of HCL and SO2 with Hydrated Lime from Humidified

The simultaneous absorption of HCl and SO2 by hydrated lime in a fixed-bed reactor has ... Because of the passage of the Clean Air Act Amendments in 1...
0 downloads 0 Views 162KB Size
4068

Ind. Eng. Chem. Res. 1999, 38, 4068-4080

Dry Absorption of HCL and SO2 with Hydrated Lime from Humidified Flue Gas Paul N. Chisholm and Gary T. Rochelle* Department of Chemical Engineering, The University of Texas at Austin, Austin, Texas 78712-1062

The simultaneous absorption of HCl and SO2 by hydrated lime in a fixed-bed reactor has been studied at conditions simulating humidified flue gas dry scrubbing. At 120 °C, an increase in relative humidity from 0% to 19% increased HCl removal and sorbent utilization. At 19% relative humidity (RH), the final loading of hydrated lime was 1.64 mol of HCl/mol of Ca2+. From 250 to 1000 ppm HCl, HCl removal was first order in HCl concentration. When SO2 was added to the feed gas, the total utilization of the sorbent by HCl and SO2 was not a function of gas concentration with 250-1000 ppm HCl and 0-2000 ppm SO2. However, the fraction of the hydrated lime converted by SO2 increased as the SO2/HCl feed ratio increased. With oxygen present in the feed, more SO2 was absorbed. Adding 150 ppm NO2 to the gas stream increased the final SO2 loading from 0.06 to 0.17 mol of SO2/mol of Ca2+. As the reactivity of SO2 and NO2 increased, the reactivity of HCl decreased slightly as a result of the competition for alkalinity with the other acid gases. The data were modeled using semiempirical flux equations based on a modified shrinking core model. Results from the parameter estimation were used to predict the absorption of HCl and SO2 on the surface of a bag filter. The predictions indicated that, with a humidified flue gas and 50% sorbent utilization, less than 20% HCl penetration is possible. However, even at the conditions with the greatest SO2 absorption (250 ppm HCl, 150 ppm NO2, and 2.5% O2), 90% SO2 penetration is predicted at 50% sorbent utilization. 1. Introduction Because of the passage of the Clean Air Act Amendments in 1990, many industrial processes must reduce acid gases emissions, including HCl and SO2. These two gases are emitted together, primarily in municipal waste combustors (MWC), hazardous waste incinerators (HWI), and coal-fired power plants. Published regulations (U.S. EPA, 1995) require new and large MWCs undergoing major modifications to remove 95% of the HCl and 75% of the SO2 emissions. Medical waste incinerators, a common HWI, must remove 99% of their HCl emissions and reduce the SO2 concentration to 55 ppm (U.S. EPA, 1997). New coal-fired power plants and those undergoing major modifications, besides controlling SO2 emissions, must also report HCl releases if they exceed 10 t/year. If the combustor has a fabric filter in place for particulate control, one strategy to meet these regulations is to inject a dry, alkaline sorbent into the flue gas after it has been humidified. The solids are added after the combustion chamber upstream of the fabric filter. Some acid gas absorption will take place in the duct. However, most will occur on the surface of the fabric filter, provided care is taken to ensure that injected solids do not fall out of the gas stream and are deposited on the filter bags. The competing technologies of dry sorbent injection are spray dryer absorption, wet scrubbers, and furnace sorbent injection. The advantages of dry sorbent injection include the absence of wastewater treatment and corrosion or steam plume problems (wet scrubbing), the absence of nozzle plugging (spray dryer absorption), greater sorbent utilization (furnace sorbent injection), its relative simplicity, and a small footprint allowing for easy retrofitting (Holzman and Atkins, 1988).

Other researchers (Weinell et al., 1992; Pakrasi, 1992; Jozewicz et al., 1990) have investigated the absorption of HCl by hydrated lime in humid gas with bench-scale apparatuses. They found that the absorption of HCl, like SO2, is enhanced at low temperature, primarily by an increase in the relative humidity (RH) of the gas. Pakrasi found that at 79 °C and 45% RH, complete utilization of hydrated lime by HCl is achieved. He concluded that incomplete utilization was due to the increasing barrier to diffusion presented by the calcium chloride product layer. Only limited studies of the simultaneous absorption of HCl and SO2 by hydrated lime have been performed (Matsukata et al., 1996; Jozewicz et al., 1990; Chang et al., 1989). Those researchers have found that increased HCl concentration increased SO2 absorption rates. Each of these studies have also shown that the presence of SO2 has little effect on HCl absorption. The objective of this study was to gain a comprehensive understanding of the absorption of HCl alone and of HCl and SO2 simultaneously by hydrated lime at lowtemperature flue gas conditions. Experiments were performed simulating the reaction conditions of acid gases with hydrated lime on a fabric filter for a MWC. The acid gases contacted hydrated lime in a fixed-bed reactor. The reactor outlet concentration of the acid gases was measured continuously with an FT-IR spectrophotometer. One advantage of this investigation over previous HCl-SO2-hydrated lime studies is that the rate of acid gas absorption was measured continuously through gas-phase analysis. Baseline studies were performed with the single-gas systemssHCl and SO2s and then with the combined gas system. The effects of oxygen, NO, and NO2 in the gas phase were also investigated. The bench-scale data were modeled using a modification to the shrinking core theory. Results from

10.1021/ie9806601 CCC: $18.00 © 1999 American Chemical Society Published on Web 09/14/1999

Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999 4069

the modeling were used to predict the absorption of HCl and SO2 by hydrated lime on the surface of a bag filter. 2. Reaction Chemistry At the temperature and most of the relative humidity ranges investigated, the most likely solid phases are calcium chloride dihydrate (Collins and Menzies, 1936) and calcium sulfite hemihydrate (Jones et al., 1976). The absorption of HCl, SO2, and NO2 can be achieved by irreversible acid-base reactions:

Ca(OH)2 + 2HCl f CaCl2‚2H2O (reaction 1) Ca(OH)2 + SO2 f CaSO3‚1/2H2O + 1/2H2O (reaction 2) Ca(OH)2 + 3NO2 f Ca(NO3)2 + NO + H2O (reaction 3) In this study, it was discovered that when CaSO3‚ is exposed to HCl vapor, SO2 is evolved and CaCl2‚2H2O is formed. Also, when CaCl2‚2H2O is exposed to SO2 in the absence of HCl, HCl is released by the solids and CaSO3‚1/2H2O is formed. This reversible reaction is given by Reaction 4:

1/ H O 2 2

CaSO3‚1/2H2O+ 2HCl + 1/2H2O T CaCl2‚2H2O + SO2 (reaction 4) When oxygen was present in the gas phase, S(IV) could be 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 this study was done at the lower border of the temperature range, it was assumed that S(VI) existed as gypsum:

CaSO3‚1/2H2O + 1/2O2 + 3/2H2O f CaSO4‚2H2O (reaction 5) NO2 is expected to act as a catalyst promoting S(IV) oxidation by reaction 5 (Shen, 1997). NO2 may also participate directly in S(IV) oxidation:

Ca(OH)2 + CaSO3‚1/2H2O + 2 NO2 + 1/2H2O f CaSO4‚2H2O + Ca(NO2)2 (6) 3. Experimental Section 3.1. Apparatus. The reactivity of hydrated lime with HCl and SO2 was studied in a fixed-bed reactor (see Figure 1). The solids were placed on an extra coarse, 0.2-cm thick glass frit at the bottom of the Pyrex reactor (1.9 cm in diameter and 13.4 cm in height). The pressure drop through the bed did not exceed 7 kPa. Cylinders of HCl, SO2, NO, and NO2, each 0.5% in nitrogen, were used to make the simulated flue gas. HCl was fed through a rotameter whose wetted surfaces were made exclusively of glass and Teflon. The SO2, NO, and NO2, along with filtered house air and nitrogen, were metered into the system using mass flow controllers. Liquid water, fed by a syringe pump, was vaporized in the furnace and blended with the rest of the simulated flue gas. The simulated flue gas, usually flowing at 1.5 SLPM (standard conditions: 1 atm, 21.1 °C), exited the furnace

Figure 1. Schematic of the fixed-bed reactor apparatus.

and was sent to a 0.8-L glass equilibration vessel that was submerged in the sand bath. The purpose of the vessel was to ensure the gas was at the bath temperature and to dampen out fluctuations caused by the evaporation of water in the furnace. After leaving the glass vessel, the gas was sent to the reactor where it contacted the hydrated lime (HL). HL was dispersed in quartz sand to minimize solids agglomeration and gas channeling. Following Karlsson et al. (1981), a sandto-sorbent mass ratio of at least 50 was used for all experiments. The reactor was immersed in a fluidized alumina sand bath to maintain a constant ((1 °C) temperature. After contacting the solids in the bed, the reactor outlet gas was mixed with 40-60 SLPM of filtered house air. Then, the diluted reactor effluent entered the gas cell operating with an FT-IR spectrophotometer. Dilution was necessary to prevent condensation in the lines downstream of the reactor and to reduce corrosion of the gas cell by the acid gases, particularly HCl. The gas cell (Infrared Analysis, Model G-5-22-V-AU) was 13 cm in diameter, was 56 cm long, and had an equivalent path length of 24 m. The FT-IR (Perkin-Elmer, System 2000) recorded infrared energy absorbance over time during an experiment. The tubing and fittings that contacted a high concentration (i.e., g20 ppm) of HCl were made of a perfluoroalkoxy polymer (PFA). The reactor was connected into the system with polysulfone quick-connect fittings. All the tubing and fittings from downstream of the furnace to just after the dilution air feed point were submerged in the sand bath. The sorbent, supplied by the Mississippi Lime Co. (MLC), was similar to the hydrated lime used by Nelli and Rochelle (1998). The specific surface area, as measured by nitrogen porosimetry (Micromeritics, Model ASAP 2000), was 21.1 ( 0.4 m2/g. Assuming the pores were cylindrical, the volume of the void space for pores between 24 and 1830 Å in diameter was 0.110 cm3/g. Approximately half of the pore volume consisted of pores less than 300 Å in diameter. According to MLC, the hydrated lime was composed of 97% calcium hydroxide with the balance being mostly calcium carbonate. A powder X-ray analysis exhibited calcium hydroxide peaks only. As determined by sieving, all the particles were smaller than 149 µm; 92% (by mass) were smaller than 45 µm. 3.2. Methods. For a typical experiment, the reactor was charged with HL and submerged in the sand bath. Nitrogen, with a known relative humidity, was passed through the reactor to humidify the solids. Flow was then directed around the reactor through a bypass line

4070

Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999

the two analyses confirmed the final loading of the hydrated lime by HCl generally within 5%. 4. Experimental Results and Discussion

Figure 2. Typical data that provided reactor outlet concentrations of HCl. At zero minutes, the flow was directed to the reactor from the bypass line. Experiment performed at 19% RH, 1000 ppm HCl, 120 °C, 18 mg of hydrated lime, and 1.5 SLPM. Experiment No. 5.

so that a simulated flue gas could be synthesized and allowed to reach steady state. The gas was then directed to the reactor where the acid gases contacted the sorbent. At a resolution of 8 cm-1, the FT-IR scans from 6500 to 500 cm-1 every 3 s. To reduce the noise from these scans, 5 consecutive spectra were averaged, giving an absorbance versus wavelength plot every 15 s. HCl, SO2, and NO2 each have a set of characteristic peaks that are unique. To determine a concentration versus time profile of an acid gas during an experiment, the area under a peak of the gas on the absorbance plot was determined for each time interval. The peak used for HCl was located from 2828 to 2814 cm-1; for SO2, from 1360 to 1340 cm-1; for NO2, 1655 to 1615 cm-1. The area next to each of these peaks was also calculated and subtracted from the area of the acid gas peaks to account for any baseline drift over the course of the experiment. Typical data for the HCl reaction with HL are shown in Figure 2. The wavenumber range used to analyze HCl was in the same vicinity as several minor water peaks. Consequently, the noise of the HCl signal was greater at higher relative humidity experiments where fluctuations in the water level were more pronounced. Another problem that was more evident during high relative humidity experiments was due to the adsorption of water on the walls of the gas cell. HCl absorbed into this water layer led to a damping of sharp changes in the HCl concentration. When flow was directed to the reactor from the bypass line, the HCl concentration at the outlet of the reactor dropped abruptly to almost zero as HCl reacted with the solids. With no HCl entering the gas cell, the HCl previously absorbed in the water layer was stripped into the gas phase. While this phenomenon does not affect final loading calculations, it does limit the accuracy of HCl removal data early in an experiment, particularly with those experiments that have high relative humidity and low HCl concentration. Analyses of the solids were periodically performed to confirm the gas-phase mass balance of HCl. The reacted solids were placed in a known volume of deionized water. The chloride concentration was determined using a chloride ion selective electrode (Fisher, Model 13-620519). An atomic absorption spectrophotometer (Varian, Model AA-1475) was employed to determine calcium concentration. The mole ratio of chloride to calcium from

Four sets of gas mixtures in nitrogen were studied: (1) HCl; (2) HCl and SO2; (3) HCl, SO2, and O2; (4) HCl, SO2, and O2 with NO or NO2. The experimental conditions, final loading of hydrated lime by HCl, SO2, and NO2, and total sorbent utilization are given in Table 1. 4.1. HCl Only. As expected, the absorption of HCl by hydrated lime was found to be sensitive to relative humidity. Figure 3 illustrates that HCl removal and loading increased dramatically with an increase in relative humidity. These data agree well with Pakrasi (1992). For example, he found that HCl reacted with HL to a maximum loading of 0.54 mol/mol of Ca2+ at 107 °C and 5.4% RH. Loading of an individual gas is defined as the total moles absorbed of that gas divided by the moles of calcium in the sorbent. For each 15-s time step, the moles of HCl absorbed in the bed was calculated by comparing the outlet HCl molar flow to the inlet molar flow. The difference was the amount of HCl that had reacted in the bed during that time step. The sum of the moles of HCl reacted through time τ divided by initial moles of HL is the average HCl loading of the sorbent throughout the bed at time τ. Increased reactivity of acid gases with increasing relative humidity has been seen throughout the literature (Pakrasi, 1992; Stouffer et al., 1989; Arthur,1998). The addition of deliquescent salts such as calcium chloride to calcium-based sorbents has also been shown to improve gas absorption (Ruiz-Alsop and Rochelle, 1985). For solutions in contact with CaCl2‚2H2O and saturated to calcium chloride at 120 °C, Collins and Menzies (1936) found that the vapor pressure of water over the solution was 35.3 kPa. Table 2 shows the number of monolayers that adsorbed as a function of the bulk-phase relative humidity, assuming the BET equation for the adsorption of water. Because calcium chloride reduced the water vapor pressure over the solids, a thicker layer of water adsorbed on the surface of the sorbent. In fact, an inlet gas relative humidity of 19% led to the formation of bulk water on the surface of the sorbent. As the product layer became thicker, the reactivity of HCl with the hydrated lime increased. Though Garea et al. (1997) have proposed an explanation of the role of relative humidity, their theory has not been conclusively proven. Close inspection of Figure 3 shows that, at 3.5% RH, HCl removal reached a maximum of 45% at the beginning of the experiment, decreased to about 10% removal at 0.1 mol of HCl/mol of Ca2+ loading, and then increased to about 40% removal before finally tapering off to 0% removal at a loading of 0.48 mol of HCl/mol of Ca2+. The same trend, though less pronounced, was evident in the experiment at 0.8% RH. The experiment at 3.5% RH was reproduced five times; the experiment at 0.8% RH was duplicated. Each of the seven experiments showed the same drop in reactivity at about 0.1 mol of HCl/mol of Ca2+ loading and then a resumption of reactivity. While it is not clear in Figure 3, there was a greater time lapse before the reaction restarted with the experiment at 0.8% RH than with the experiment at 3.5% RH. The deliquescent nature of calcium chloride may explain this unexpected phenomenon. The solids react to a loading of about 0.1 mol of HCl/mol of Ca2+,

Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999 4071 Table 1. Experimental Conditions and Results (All Experiments Performed at 120 °C and 1.5 SLPM, unless Indicated Otherwise) exp no.

rel. hum. (%)

mass sorbent (mg)

time (min)

HCl (ppm)

SO2 (ppm)

O2 (%)

NOx (ppm)

HCl load.a

SO2 load.a

NO2 load.a

total util.b (%)§

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

0 1 3.5 9 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 9 9

18 18 18 18 18 18 18 18 18 35 35 35 35 50 50 50 100 35 35 35 35 35 35 35 35 36 18

24 45 45 45 45 55 45 45 45 110 65 45 45 45 45 45 40 45 110 110 60 100 45 100 100 45 45

1000 1000 1000 1000 1000 250 500 2000 3500 250 500 1000 2000 1000 1000 1000 0 0 250 250 1000 250 0 250 250 1000 500

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

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5.5 1.2 5.5 5.5 0 0 0 2.5 0 0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 NO:150 NO2:150 NO2:150 NO2:150 0 0

0.101 0.173 0.488 1.07 1.61 1.66 1.66 1.43 1.27 1.58 1.47 1.52 1.48 1.60 1.54 1.37 0 0 1.48 1.52 1.50 1.52 0 1.44 1.33 1.122 1.020

0 0 0 0 0 0 0 0 0 0.0606 0.0748 0.0469 0 1.79e-3 5.02e-3 0.0704 0.108 0.204 0.0903 0.152 0.106 0.143 0.126 0.174 0.310 0 0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0.0534 0.123 0.100 0 0

5.05 8.65 24.4 53.5 80.5 83.0 83.0 71.5 63.5 85.6 81.0 80.7 74.0 80.2 77.5 75.5 10.8 20.4 83.0 91.2 85.6 90.3 15.3 95.6 102 56.1 51.0

a

Loading of acid gas A is given by mol of A/mol of Ca2+. b Total utilization takes into account stoichiometry: 1 /2nHCl + nSO2 + 1/2nNO2 total utilization ) × 100 nCa c Total flow rate: 0.69 SLPM.

Figure 3. Effect of relative humidity and loading on HCl removal. Experiments performed at 1000 ppm HCl, 120 °C, 18 mg of hydrated lime, and 1.5 SLPM. Experiments 1-5. Table 2. Thickness of Water Layer on Hydrated Lime at 120 °C bulk gas rel. hum. (%)

water partial pressure (kPa)

no. of water monolayers w/o CaCl2a

0 0.8 3.5 9 19

1.6 3.7 7.2 18 39

0 0.4 0.8 1.0 1.2

no. of water monolayers w/ CaCl2b 0 0.8 1.2 2.1 ∞ (bulk liquid)

a Assumes that no CaCl is present in the liquid layer. b As2 sumes that the liquid layer is saturated with CaCl2.

producing a layer of CaCl2‚2H2O on the HL particles per reaction 1. The calcium chloride then absorbed water from the vapor phase. The reaction started up again once there was a sufficient degree of surface water

Figure 4. Effect of HCl concentration and loading on HCl removal. Experiments performed at 19% RH, 120 °C, 18 mg of hydrated lime, and 1.5 SLPM (unless indicated otherwise). Experiments 5-9.

on the particles. The same phenomenon may have occurred with the higher relative humidity experiments, but because the water vapor rates were much greater, the rehydration of the particles took place too quickly to observe. While this behavior is interesting scientifically, further studies were not conducted to conclusively explain its cause. Over the typical HCl concentration range in flue gas for MWCs and coal-fired boilers, the HCl removal rate was first-order in the HCl concentration. Also, the final HCl loading was not a function of concentration. However, at an HCl concentration of 2000 and 3500 ppm, the increased HCl concentration decreased the final loading. This trend is evident in Figure 4. From 250 to 1000 ppm HCl, removal was equal within experimental error up to the final loading of 1.64 mol of HCl/mol of

4072

Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999

Ca2+ loading. At a concentration of 2000 and 3500 ppm, HCl removal remained high until a loading of approximately 0.6 mol of HCl/mol of Ca2+. However, the absorption shut off more quickly to final loadings of 1.44 mol of HCl/mol of Ca2+ at 2000 ppm HCl and 1.28 mol of HCl/mol of Ca2+ at 3500 ppm HCl. The experiments at 2000 and 3500 ppm were each reproduced three times. Experimental constraints prohibited experiments with a concentration greater than 3500 ppm. Previously, only Pakrasi (1992) had investigated the effect of increased HCl concentration on the final loading. The results of his experiments, which were conducted at conditions similar to those in Figure 4, did not reveal a decrease in HL utilization with the HCl concentration from 1000 to 5000 ppm. However, his analysis to determine HCl loading was done with a maximum sorbent exposure time of only 5 min. It is likely that, under highly reactive conditions, HL would have continued to react after the first 5 min of the experiment. Therefore, the loading at 5 min does not represent an accurate final loading because lower molar flows of HCl will take longer to completely react with HL. In addition, the average residence time of reactive solids is usually greater than the 5 min. Reduced final sorbent loading at higher HCl concentrations may have been due to an increase in the plugging of the pores of the sorbent. Simons’ (1992) analysis of the reaction of SO2 with calcined limestone showed that, under certain conditions, utilization is incomplete because pores became plugged near the external particle surface, leaving the interior of the sorbent unavailable for reaction. When the HCl concentration in the gas was high, the rate of reaction at the surface of the particle was greater. Therefore, reaction products accumulated at the opening of the pore, eventually leading to plugging. When the HCl concentration was low, the lower reaction rate at the surface of the solid allowed for increased diffusion of the gas into the interior of the particle. The results shown in Figure 4 imply that, at a concentration of 1000 ppm and below, the reaction rate at the surface of the particle was low enough to allow the HCl to diffuse into the pores and utilize more alkalinity. 4.2. HCl and SO2. As with HCl, the reactivity of SO2 with HL in the absence of HCl increased with increasing relative humidity. Unlike HCl, decreasing SO2 concentration increased the fraction of SO2 removed. These effects of relative humidity and SO2 concentration on SO2 removal (Chisholm, 1999) followed the same trends as those found by other researchers (Jorgensen et al., 1986; Klingspor et al., 1984; Ruiz-Alsop and Rochelle, 1985). Most of the experiments performed with HCl and SO2 were conducted at high relative humidity (19%). This was done because, at lower relative humidity, the reactivity of SO2 was too low to observe trends within acceptable precision. At early experimental times, the absorption of SO2 was limited by having to compete with HCl for alkalinity. Increased SO2 absorbance with a lower concentration of HCl is evident in Figure 5. Note that the total loading ((0.5 mol of HCl + mol of SO2)/ mol of Ca2+) is defined somewhat differently than the HCl loading (mol of HCl/mol of Ca2+). Later in the experiment, as the alkalinity of the sorbent became consumed, HCl reacted with CaSO3‚1/2H2O, releasing SO2 as indicated by reaction 4.

Figure 5. Effect of HCl concentration and total loading on SO2 removal. Experiments performed at 19% RH, 1000 ppm SO2, 35 mg of hydrated lime, 120 °C, and 1.5 SLPM. Experiments 10-13.

Figure 6. Effect of SO2 concentration and total loading on HCl removal. Experiments performed at 19% RH, 1000 ppm HCl, 35 mg of hydrated lime, 120 °C, and 1.5 SLPM. Experiments 12 and 14-16.

It was found that the concentration of SO2 in the gas phase had little effect on the reactivity of HCl. This is evident in Figure 6. The minor differences in the breakthrough curves can be attributed to differences in the quality of the distribution of HL in the reactor. If there was more sorbent on one side of the reactor, the curve would have broken early because some of the gas was bypassing the reactive solids on the side with less sorbent. The fact that all the experiments ended up at approximately the same final total loading suggested that there was, at most, only a minimal effect of SO2 concentration on HCl removal. In all of the HCl-SO2 experiments, the sorbent utilization achieved was the same as the utilization achieved with only HCl in the gas stream. At 19% RH and an HCl concentration of 1000 ppm and below, this final total loading was 0.82 (0.5 mol of HCl + mol of SO2)/mol of Ca2+. This fact implies that, with HCl and SO2, there seemed to be a maximum available alkalinity primarily defined by the relative humidity, regardless of the feed gas composition. When the ratio of SO2/HCl in the feed gas increased, slightly less CaCl2‚2H2O was formed in favor of a greater amount of CaSO3‚1/2H2O. Though with considerable scatter, this trend is evident in Figure 7. The fact that the final solids product composition is a function of the feed gas composition suggests that a pseudoequilibrium exists, probably in accordance with reaction 4. Experiments (Chisholm, 1999) where solids were exposed to just HCl or SO2 in sequence showed that not only would HCl react to drive off SO2 but also

Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999 4073

Figure 7. Effect of the SO2/HCl feed ratio on fraction of total loading by SO2. Experiments performed at 19% RH, 35 mg of hydrated lime, 120 °C, and 1.5 SLPM. Two data points had the same fraction of total loading by SO2 of 9.3% at an inlet SO2/HCl ratio of 2. Experiments 14-19.

Figure 8. Effect of O2 and HCl on the loading of SO2. Experiments performed at 19% RH, 1000 ppm SO2, 35 mg of hydrated lime, 120 °C, and 1.5 SLPM. Experiments 10, 17-21.

the reverse was true. The concentration of SO2 emitted when HCl was exposed to CaSO3‚1/2H2O increased as the HCl concentration increased. The same trend was observed when CaCl2‚2H2O was exposed to varying SO2 concentrations. Reagent calcium chloride also absorbed low levels of SO2 while simultaneously emitting HCl. One explanation for the pseudoequilibrium may be reaction kinetics. Another explanation may be that the chemistry dictates that HCl would react completely with CaCl2‚2H2O. However, CaSO3‚1/2H2O became trapped in the pores and was inaccessible to HCl. 4.3. HCl-SO2-O2 and Hydrated Lime System. Adding air to a dilute HCl gas stream did not affect HCl reactivity (experimental data not shown). With a similar sorbent, Chu and Rochelle (1989) have found that the presence of oxygen in the feed gas led to greater SO2 removal and final loading. With air present in the HClSO2 gas mixture, the oxidation of S(IV) to S(VI) by reaction 5 could take place. Figure 8 shows the effect of air when added to the HCl-SO2 mixture. Several conclusions can be drawn from the experiments shown in Figure 8. First, the presence of O2 increased the reactivity of SO2 with HL. This is clear, particularly in the experiments without HCl. When the experiments containing 5.5% O2 with and without 250 ppm HCl are compared, it is apparent that there was a slight increase in SO2 reactivity early in the experiment due to the presence of HCl. However, once the HCl moving front pushed to the end of the reactor and alkalinity started to become scarce, HCl reacted with

CaSO3‚1/2H2O, causing SO2 emissions. As the concentration of oxygen increased, the amount of sulfur that remained bound in the solids also increased. This suggests that, with a greater oxygen concentration, there was a greater amount of S(VI) formed through reaction 5. When there was only 250 ppm HCl in the gas stream, the SO2 did not have to aggressively compete with HCl; the SO2 absorption probably took place at the back of the reactor where HCl had not broken through. But when the concentration of HCl was increased to 1000 ppm, the rapid absorption of HCl and consumption of alkalinity led to reduced SO2 reactivity. This effect is seen clearly in the comparison of the experiments with 250 and 1000 ppm HCl with a constant O2 concentration of 5.5%. In summary, if SO2 did not have to compete with HCl for alkalinity, it reacted almost as if the HCl were not in the feed gas. Once SO2 reacted, the presence of oxygen in the gas stream allowed the irreversible oxidation of S(IV). But the oxidation did not take place quickly enough to force all the bound sulfur to S(VI). Consequently, SO2 was emitted as HCl began to break through to the back of the bed. At a high HCl concentration, SO2 had to compete with HCl for alkalinity. This competition caused a reduction in SO2 reactivity, even in the presence of oxygen. While the dynamics of SO2 absorption in the presence of HCl and O2 were somewhat complex, HCl reactivity was relatively uninteresting. For each of the experiments shown in Figure 8, the final loading of HCl was 1.5 mol of HCl/mol of Ca2+ within experimental error. Therefore, in the experiments where the SO2 loading was greater, the total loading was also greater. Without oxygen present, sorbent utilization with HCl-SO2 is about 82% at 19% RH. The experiment with 250 ppm HCl, 1000 ppm SO2, and 5.5% O2 at 19% RH yielded a final utilization of 91%s76% by HCl and 15% by SO2. 4.4. HCl-SO2-NOx-O2 and Hydrated Lime System. In general, NO accounts for approximately 90% of the NOx emitted in flue gas; the remaining 10% is NO2. Yuan and Fong (1992) studied the reactivity of the HCl-SO2-NO system with a calcium hydroxide slurry in a simulated spray-drying apparatus. Their results show that fractional removal of NO by the slurry is an order of magnitude less than those for HCl and SO2. They concluded that increasing NO concentration had a slight positive effect on SO2 absorption. In a study done without HCl, Nelli and Rochelle (1998) found that increasing NO2 concentration improved SO2 reactivity. Greater SO2 concentration also enhanced NO2 absorption. The experiments investigating the effect of NOx in the feed gas were performed at 250 ppm HCl. This was done so that the absorption of SO2 and NOx could be observed without those gases having to compete aggressively with HCl for alkalinity. A high concentration of SO2 and a low concentration of HCl most closely represent coalfired boiler flue gas. However, it is expected that the conclusions drawn below are qualitatively applicable to flue gases with other SO2/HCl ratios. Experiments performed in this study generally confirm the trends found by previous researchers. Figure 9 shows the effects of NO and NO2 on SO2 removal. In 50 min, the presence of 150 ppm NO decreased the SO2 loading by 0.05 mol of SO2/mol of Ca2+ when all other variables were held constant. This effect was probably due to the competition between SO2 and NO

4074

Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999

Figure 9. Effect of HCl, NO, and NO2 on SO2 loading. Experiments performed at 19% RH, 250 ppm HCl, 1000 ppm SO2, 35 mg of hydrated lime, 120 °C, and 1.5 SLPM. Experiments 10, 20, and 22-25.

for available alkalinity at the back of the sorbent bed. In the experiment without NO, by 100 min the HCl reactivity had forced the emission of SO2 by Reaction 4. Therefore, loading of SO2 in the experiments with and without NO ended up within 1 mol of SO2/mol of Ca2+ of each other. The loading of the solids by NO could not be accurately calculated because the NO peaks in the IR spectrum were washed out by water peaks. Though water was taken as part of the background, fluctuations in the syringe pump and the water content of the dilution air made the NO signal too noisy for a mass balance. Qualitatively, it was apparent that NO reacted at early experimental times but ceased absorbing (or desorbing) after about 3 min. One strategy for NO capture is to oxidize the relatively unreactive NO to NO2 by injecting a hydrocarbon such as methanol into the flue gas (Lyon et al., 1990). With this strategy in mind, experiments with NO2 added to the feed gas were conducted. The presence of NO2 clearly increased the reactivity of SO2 with HL. When 150 ppm NO2 was added to the gas stream in the absence of oxygen, the final loading of SO2 was 0.11 mol of SO2/mol of Ca2+ greater than that in the absence of NO2. With 150 ppm NO2 and 2.5% O2 in the feed gas, the final loading of SO2 reached 0.31 mol of SO2/mol of Ca2+. Even as the solids were approaching complete utilization, the absorption of SO2 continued. There are several potential causes of the continued SO2 reactivity. It is possible that NO2 was catalyzing the formation of sulfuric acid (Duecker and West, 1959). However, the analysis of reacted solids with a similar sorbent (Chisholm and Rochelle, 1999) did not support this conclusion. A free radical mechanism for the oxidation of S(IV) by NO2 has also been reported (Shen, 1997). But this mechanism was found to exist at high pH, a condition probably not present on the surface of highly converted solids. The improved SO2 absorption is probably due to enhanced sorbent utilization from the formation of sulfur-nitrogen compounds (Nelli and Rochelle, 1998). The loading of NO2 was calculated by means of the gas-phase mass balance. Figure 10 plots the loadings of NO2 for two of the same experiments shown in Figure 9. The addition of 250 ppm HCl to the system increased the NO2 loading. However, adding 2.5% O2 caused a reduction in the loading of NO2 by 0.02 mol of NO2 mol of Ca2+. This reduction in reactivity arose because with the oxygen in the feed stream, S(IV) could be oxidized

Figure 10. Effect of HCl and O2 on NO2 loading. Experiments performed at 19% RH, 250 ppm HCl, 1000 ppm SO2, 150 ppm NO2, 35 mg of hydrated lime, 120 °C, and 1.5 SLPM. Experiments 2325.

to S(VI). If S(IV) was oxidized by oxygen, it was not available to react with NO2 by reaction 6. As with the HCl-SO2-O2 system, the reactivity of HCl when NOx is added did not change considerably. Without NOx or oxygen in the gas, an experiment with 250 ppm HCl and 1000 ppm SO2 yielded a final loading of HCl of 1.58 mol of HCl/mol of Ca2+. The final loading of SO2 for the same experiment was 0.06 mol of SO2/ mole Ca2+. Under conditions with the most SO2 and NOx reactivity (150 ppm NO2 and 2.5% O2), the final loading of HCl dropped to 1.34 mol of HCl/mol of Ca2+ while the final loading of SO2 and NO2 were 0.31 mol of SO2/ mol of Ca2+ and 0.10 mol of NO2/mol of Ca2+, respectively. So even when HCl must compete the most for alkalinity, its reactivity still suffers only minimally. 5. Fixed-Bed Modeling and Bag Filter Predictions 5.1. Model Description. The primary objective of modeling the experimental results was to match the data with a set of equations that accurately described HCl and SO2 concentrations over the course of an experiment. The primary objective was not to prove conclusively any gas-solid reaction models. Results of the fixed-bed modeling were used to predict the absorption of HCl and SO2 by hydrated lime on the surface of a bag filter. 5.1.1. Fixed-Bed Mass Balance. The first step to modeling the gas system was to set up the appropriate mass-balance equations for the absorption of a gas by a solid in a fixed-bed reactor. Non-steady-state experiments were performed at integral conditions where there existed a concentration gradient through the bed in the same direction as the gas flow. The gas concentration as a function of time and distance is described by eq 1:

FA )

Q ∂CA 1 ∂CA a ∂t aAx ∂z

(1)

where FA is the flux of A to the sorbent in the fixed bed, CA is the concentration of species A, t is the experimental time, z is the distance in the bed, a is the sorbent surface area per unit reactor volume, Q is the gas flow rate, and Ax is the cross-sectional area of the reactor. In most cases, the time derivative is much less than the spatial derivative (von Rosenberg et al., 1977). Therefore, eq 1 can be reduced to

Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999 4075

FA ) -

Q dCA aAx dz

(2)

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, R is a stoichiometric factor (R ) 2 for HCl; R ) 1 for SO2), Sm is the molar surface area of the sorbent, and x is the conversion. Equations 2 and 3 were transformed to algebraic equations and Euler’s method was used to predict the gas concentration through the bed. Because Euler’s method was employed, care was taken to ensure that sufficiently small time and distance steps were taken during parameter estimation calculations. To estimate the parameters in a flux equation (discussed below), calculations began at the top of the reactor at time zero. A spatial mass balance in a finite element generated the inlet to the next finite element downstream. Another mass balance in the finite element determined the conversion for the next time step in the same element. Calculations were performed through the bed to generate an outlet concentration, CA,L|τ from the reactor at time τ. 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 caused confounding effects. Therefore, the gas cell surfaces were modeled as a well-mixed absorber/stripper where CHCl,L|τ was the inlet to the gas cell. The equation used to generate a modeled concentration, CHCl0|τ, to a concentration datum obtained at the same t, CHCl|τ, is given below:

CHCl0|τ ) CHCl,L|τ +

c (C | - CHCl0|τ-∆τ) (6) ∆τ HCl,L τ

where c is an empirical parameter estimated uniquely for each experiment. The driving force in eq 6 accounted for molecules of HCl stripping out of or absorbing into the water layer on the surfaces of the gas cell. Once the outlet concentrations were generated from the model at time τ, they were compared to the data taken for the same time. The same calculations through the bed were performed for each of the time intervals and compared to the experimental data. Once this was completed for an entire experiment, the parameters were adjusted to improve the fit and the calculations started again from the top of the bed at time zero. 5.1.2. Flux Equation. Challenges with modeling arose not with the mass balances discussed above but with the identification of an accurate equation describing the flux of the acid gases, FA. Many researchers (Gullett et al., 1992; Wang et al., 1996; Fonseca et al., 1998) have employed shrinking core models, either with particles or grains within particles. These models described the rate of absorption of an acid gas as being dictated by mass-transfer resistances in series. In most systems with small particles, the gas film and pore diffusion resistances do not limit the rate of gas absorption. At early times in an experiment before a product layer has deposited on the surface of a reactant particle, the rate of gas absorption was limited only by the kinetics of the surface reaction. As the acid gas contin-

ued to react with the sorbent particle, a product layer formed. This product layer then acted as a barrier between the gas and the reactive solid core of the particle, thereby limiting the rate at which the gas was absorbed. 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 1 δ + k s DA

(7)

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 developing eq 7, planar rather than spherical geometry was used. This was done primarily to simplify the flux expression. When the sorbents absorbed the acid gases studied, the volume of the particles expanded because the absorption products have greater molar volumes than hydrated lime. This effect was ignored again in the interest of simplicity. As shown in eq 7, the flux was a function of the product layer thickness. This thickness was not measured. Because conversion, x, was proportional to product layer thickness, δ was replaced by δox. Throughout the literature and in this study, it was apparent that most sorbents generally were not totally converted from the alkaline particle to the salt product. At some point during an experiment, the absorption rate declined to zero, despite the fact that there was still unreacted sorbent in the reactor. However, the shrinking core model predicts that the rate will not drop to zero until all the sorbent has been consumed. Incomplete utilization may be due to pore plugging (Simons and Garman, 1986) or reaction product expansion (Duo et al., 1994). Because of the discrepancy between the shrinking core model and the experimental results, a driving force based on conversion was added to eq 7:

FA )

CA(xT - x) 1 δ0x + k s DA

(8)

xT is defined as the conversion when reactivity of the sorbent terminated. The three regressed parameters in eq 8 are xT, ks, and DA/δ0. The value DA/δ0 will be referred to as DA,eff. Principles leading to eq 8 are used with both the single-gas system (HCl) and with the multi-gas system (HCl and SO2). However, another modification to the flux equation for SO2 was found to be necessary. This modification will be discussed later. This modification to the shrinking core equation and the other assumptions discussed above led to a semiempirical model for the flux. However, the primary objective of the modeling of experiments was to describe the data. The goal of modeling was not to determine conclusively an absorption mechanism or accurate physical constants. The parameter estimation package used was Generalized REGression (GREG) (Caracotsios, 1986; Stewart et al., 1992). GREG was employed in this work to estimate parameters with multivariable inputs but only one response vector. The modeled concentration of the

4076

Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999

outlet of the reactor normalized by its inlet concentration was compared to the same ratio determined experimentally. Code was written to link GREG with the fixed-bed integration routine. Parameters with confidence intervals were estimated and reported. 5.1.3. Bag Filter Performance Model. The objective of the modeling effort was to project how well hydrated lime and calcium silicate would perform absorbing HCl and SO2 on the surface of a bag filter. The bag filter system is physically different from the experimental system used in this study. In the fixedbed reactor, high concentration acid gas contacts mostly converted the sorbent in the front of the reactor. Conversely, in a bag filter the high-concentration gas will pass through fresh sorbent first and will contact the most highly converted solids just before it escapes the filter. In addition, the bag filter system has sorbent continually added to the filter cake, whereas the experimental system has a batch loading of sorbent. Because of these differences, predicting how well a sorbent would perform under given circumstances was not straightforward using only the experimental data. 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 bag house. Also, it was assumed that the sorbent contacts the surface of the bag to form a uniformly thick layer. No reaction took place in the duct between the introduction of the sorbent and when it reached the bag house. In addition, it was assumed that the presence of any other contaminates, gas or solid, would not influence the acid gas absorption. Changes in pressure drop with increased filter cake thickness were not calculated. One of the key operational conditions that impacted the performance projections 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% utilization 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 x (1 mol + 0.5 mol)) of sorbent was fed into the duct. 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, removal from that bag was lowest immediately after being cleaned and increased with increasing cycle time. The calculations for bag filter performance used the surface of the filter as the origin. As the cycle time continued, the filter cake became thicker. Over time, t*, the average penetration of gas A was defined as the time average of the instantaneous penetration of A from 0 to t*:

average penetration of A ) 1 - removal of A ) 1 t* CA(t) dt (9) t* 0 CA,in



where CA(t) is the instantaneous concentration of A leaving the bag house and CA,in is the inlet concentration of A to the bag house.

The quality of the bag filter performance projections were only as good as the model fit of the data. The experiments attempted to quantify the phenomena as accurately and precisely as possible. Imperfections with the data and the modeling, then, may have manifested in the performance projections. Consequently, caution should be used such that the precision of the performance predictions is not overestimated. 5.2. HCl Only. 5.2.1. Fixed-Bed Modeling. The focus of parameter estimation for the HCl-hydrated lime system was the experiments conducted at 9% and 19% humidity. As discussed earlier, the experiments at 0.8% and 3.5% RH exhibited unexpected, transient behavior. Because of the potential difficulty in modeling the data at these low relative humidities, the emphasis of modeling was placed on the high-humidity experiments. In addition, the high-concentration data (2000 and 3500 ppm HCl) were not modeled because those experiments exhibited final conversion different from those obtained at low and intermediate HCl concentrations. The data for the absorption of HCl by hydrated lime was first modeled with the flux described by eq 8. Using eq 8, the estimation routine systematically underpredicted the outlet concentration of HCl absorption in the first 5 min of each experiment. For each of the data sets, 100% removal of HCl was predicted for nearly all the experiments. For this reason, eq 8 was modified to a pseudo-second-order dependence on HCl concentration at early experimental times to better describe absorption. Beyond the time for initial reaction, HCl absorption was still described as first-order in HCl. This first-order dependence over the majority of the active absorption regime reflected the same concentration dependence seen in Figure 4. Equation 10 was used as the flux equation for the absorption of HCl by hydrated lime:

FA )

CHCl(xT - x) x 1 + ksCHCl Deff,H

(10)

The results of parameter estimation for experiments run at 9% and 19% at concentrations less than 2000 ppm are given in Tables 3 and 4, respectively. Note that, in Tables 3 and 4, the experimental data were modeled both individually and simultaneously with the other experiments performed at the same relative humidity. Different values for the gas cell correction constant, c, were estimated for each experiment and were all between 5.6 and 12.6 s. The values of ks were determined mostly during the first few minutes of the experiment before much of a product layer of calcium chloride had developed on the surface of the sorbent. Values of Deff,H were determined as HCl was breaking through the bed. The 95% confidence intervals of Deff,H for the two humidities almost overlapped. The values of xT were considerably different, as expected because of the different chloride capacities observed in the bench-scale studies. Figure 11 compares the model fit to the data for experiment 6, the experiment with the greatest error of those regressed. The fit shown represents a worst case indication of the accuracy of the model. 5.2.2. Bag Filter Predictions. The regressed parameters for the simultaneous regressions shown in Tables 3 and 4 were used to predict the absorption of HCl by hydrated lime on the surface of a bag filter. The plot showing the trade-off between sorbent utilization

Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999 4077 Table 3. Results of Parameter Estimation for the HCl-Hydrated Lime System at 9% RH HCl conc. (ppm)

exp. no.

4 26 27

sorb. load. (mg)

simult. model 2σ: 1000 2σ: 1000 2σ: 500 2σ:

18 36 18

mean model errora

Deff,H × 104 (m/s)

ks (m4/mol/s)

xT

3.20 0.36 3.03 0.55 3.90 0.97 4.33 0.93

0.613 0.185 0.352 0.230 0.172 0.063 0.872 0.568

0.497 0.006 0.465 0.011 0.508 0.008 0.465 0.010

simult. mean model errorb 4.22

3.09

4.13

3.48

4.20

4.29

4.76

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.

Table 4. Results of Parameter Estimation for the HCl-Hydrated Lime System at 19% RH HCl sorb. Deff,H exp. conc. load. × 104 ks no. (ppm) (mg) (m/s) (m4/mol/s) simult. model 2σ: 5 1000 18 2σ: 6 250 18 2σ: 7 500 18 2σ:

2.67 0.20 3.38 0.46 4.23 0.98 1.87 0.11

0.160 0.015 0.112 0.029 0.129 0.015 0.243 0.045

xT 0.795 0.008 0.748 0.012 0.802 0.012 0.794 0.010

mean simult. model mean model errora errorb 4.09 2.63

3.08

4.76

5.20

2.59

3.14

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.

Figure 12. HCl penetration as a function of hydrated lime utilization at varying cycle times. Performance projection with 19% RH and 1000 ppm HCl at 120 °C. The curves use the global parameters given in Table 3.

concentration and was similar in form to the flux equation for HCl:

FSO2 )

CSO2(xT - x) x 1 + ksSCHCl Deff,S

(11)

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: Figure 11. Comparison of model fit to the data for experiment 6: 19% RH, 250 ppm HCl, 18 mg of hydrated lime, 120 °C, and 1.5 SLPM.

and HCl penetration is shown in Figure 12. At low utilization (and a very high sorbent feed ratio), nearly 0% penetration is possible. As the sorbent feed ratio decreases, the utilization increases to almost xT. But when the sorbent is being utilized as much as possible, almost 100% HCl penetration is predicted. Figure 12 projects that to reduce HCl penetration to 10%, slightly more than half the sorbent will be utilized. 5.3. HCl and SO2. 5.3.1. Fixed-Bed Modeling. The addition of SO2 to the inlet gas stream led to an increase in the complexity of the modeling approach. 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 10. With respect to SO2 absorption, it was shown above that the presence of O2 and NO2 as oxidizing agents of S(IV) had a positive influence on SO2 absorption. The flux equation for SO2 absorption did not explicitly account for O2 or NO2

1

xτ )

/2nHCl|τ + nSO2|τ nCa

(12)

At early experimental time (x f 0), eq 11 described the SO2 flux as being first-order in both SO2 and HCl concentrations. The same effect of HCl concentration on SO2 absorption was seen with calcium silicate as the sorbent (Chisholm and Rochelle, 1999). 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). Other experiments (Chisholm, 1999) demonstrated that an increase in relative humidity over hydrated lime increased SO2 absorption. As the relative humidity increased, the thickness of the layer of liquid water adsorbed on the sorbent increased. If the solids had surface solution that was saturated to CaCl2, the vapor pressure of water over the solids would have been reduced. This reduction may have led to a thicker adsorbed water layer and increased SO2 absorption.

4078

Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999

Table 5. Results of Parameter Estimation for the HCl-SO2 System exp. no.

HCl conc. (ppm)

SO2 conc. (ppm)

NO2 conc. (ppm)

O2 conc. (%)

Deff,H × 104 (m4/mol/s)

ks,H (m/s)

25

250

150

2.5

24

250

150

0

20

250

21

1000

16

1000

12

1000

1000 2σ: 1000 2σ: 1000 2σ: 1000 2σ: 2000 2σ: 1000 2σ:

2.33 0.23 3.87 1.05 2.08 0.18 0.443 0.025 1.36 0.06 0.987 0.054

0.185 0.029 0.0660 0.0075 0.0728 0.0084 0.0735 0.0062 0.146 0.023 0.108 0.016

0

5.5

0

5.5

0

0

0

0

xT

Deff,S × 107 (m/s)

ks,S × 102 (m4/m/s)

0.951 0.028 0.939 0.031 0.974 0.029 0.802 0.018 0.794 0.011 0.861 0.015

9.80 2.10 5.35 1.63 2.92 0.86 1.52 ∞b 0.405 0.235 2.97 1.14

5.87 2.90 4.47 7.22 15.4 ∞b 1.43 0.42 2.40 0.11 1.93 0.29

mean model errora 4.08 4.91 5.03 3.38 2.53 2.93

a The error reported is the square root of the average model error squared multiplied by 100. b ∞ means the parameter was determined to be very large or had a very large confidence interval.

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 and SO2 in the inlet gas stream, eqs 11 (HCl) and 12 (SO2) indeed described the data effectively. The results of the parameter estimations for six experiments from the HCl-SO2hydrated lime system are given in Table 5. For each of the experiments modeled, a unique set of regressed parameters were determined. 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 S(IV) oxidation. There were large confidence intervals for some SO2 flux parameters due to the fact that only a small amount of SO2 absorption took place over a rather narrow range of experimental time. Consequently, there were not enough data in some cases to accurately estimate parameters. Values for Deff,H and ks,H were approximately equal to the same values estimated in the absence of SO2. This was expected because the addition of SO2 to the gas did not have a strong influence on HCl absorption. The values for Deff,S and ks,S were determined to be at least 2 orders of magnitude less than Deff,H and ks,H. Because HCl reacted considerably more than SO2, it was expected that the SO2 parameters would be less than the HCl parameters. Deff,S, like SO2 loading, generally decreased in the absence of NO2 or O2 or in the presence of excess HCl. Figure 13 shows a typical fit of the HCl and SO2 data. 5.3.2. Bag Filter Predictions. The bag filter performance calculations predicted what was qualitatively seen in the fixed-bed experiments: HCl penetration was low while SO2 penetration was high. Projections of HCl and SO2 penetration under two sets of conditions that allowed for improved SO2 absorption are illustrated in Figure 14. On one hand, high levels of HCl removal were achieved, regardless of NO2 and O2 concentration. On the other hand, almost 70% of the inlet SO2 penetrated through the bag filter at 60-min cycle time, even in the presence of 150 ppm NO2 and 2.5% O2. HCl and SO2 penetration as a function of sorbent utilization is given in Figure 15. To reduce HCl penetration to 10%, hydrated lime utilization was approximately 40%. At this utilization, 85% of the SO2 fed penetrated through the bag filter. The conditions for the

Figure 13. Comparison of model fit to the data for experiment 24: 19% RH, 250 ppm HCl, 1000 ppm SO2, 150 ppm NO2, 35 mg of hydrated lime, 120 °C, and 1.5 SLPM.

Figure 14. Penetration of HCl and SO2 with varying NO2 and O2 concentrations in a bag filter system. Performance predictions with a sorbent feed ratio of 1.5 at 19% RH and 120 °C with 250 ppm HCl and 1000 ppm SO2. Parameters used are taken from Table 5 for experiments 20 and 24.

prediction depicted in Figure 15 delivered the greatest SO2 loading in the fixed-bed experiments. So the projection given in the figure below represents a best case scenario for SO2 absorption with hydrated lime. The bag filter model assumed that the sorbent undergoes only one pass. It is possible that an analysis that includes sorbent recycle may make the use of hydrated lime more attractive for both HCl and SO2. In the absence of any sorbent recycle, the small measure of SO2 reactivity does not make hydrated lime an attractive option for acid gas emissions control from municipal waste combustors or coal-fired boilers.

Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999 4079

Literature Cited

Figure 15. HCl and SO2 penetration as a function of cycle time in a bag filter system. Performance prediction at 19% RH, 120 °C, 250 ppm HCl, 1000 ppm SO2, 150 ppm NO2, and 2.5% O2. Parameters used are taken from Table 5 for experiment 24.

Conclusions In the single-gas system, the reactivity of HCl with hydrated lime (HL) increased considerably with increasing relative humidity. At 1000 ppm HCl and below, the rate of absorption was first-order with respect to HCl concentration and the final loading was constant. Greater than 1000 ppm HCl, increasing concentration decreased the final loading. The decrease in final loading may have been due to pore plugging near the external particle surface. When HCl and SO2 were simultaneously exposed to HL, the final total loading of the solids did not depend on the SO2/HCl feed ratio. However, as the SO2/HCl ratio increased, the fraction of the converted solids that consisted of CaSO3‚1/2H2O increased while the fraction of CaCl2‚ 2H2O decreased slightly. This phenomenon suggested that there was a pseudoequilibrium between the reaction products and the acid gases. The addition of oxygen to the HCl-SO2 system enhanced SO2 absorption but did not affect HCl reactivity. On one hand, the presence of NO in the HCl-SO2O2 system did not affect the final loading by HCl or SO2. On the other hand, the presence of NO2 enhanced SO2 absorption. While the presence of O2 improved SO2 reactivity, the presence of O2 caused a decrease in NO2 loading. Increasing SO2 and NO2 reactivity had a minor negative effect on the absorption of HCl. The conditions that maximize sorbent utilization are high relative humidity, SO2 concentration, NO2 concentration, and oxygen concentration and low HCl concentration. A model was developed to describe the experimental data. Semiempirical flux equations for HCl and SO2 were developed that fit the data well. Parameters defined by the flux equations were estimated. An expression to correct for transient effects due to absorption of HCl in the gas cell was included. Regressed parameters were used to predict the penetration of HCl and SO2 in a bag filter system. Injection of hydrated lime was projected to be an effective sorbent for HCl absorption only at high relative humidity. Almost 10% HCl penetration was predicted for a sorbent utilization of 50% at 19% RH. Low levels of HCl and high levels of SO2 penetration were predicted for HCl and SO2 absorption by hydrated lime. At conditions that optimize SO2 reactivity but do not necessarily reflect real flue gas (19% RH and 150 ppm NO2), predictions for penetration at 50% sorbent utilization were 20% and 95% for HCl and SO2, respectively.

Arthur, L. F. Silicate Sorbents for Flue Gas Cleaning. Ph.D. Dissertation, The University of Texas at Austin, Austin, TX, 1998. Caracotsios, M. Model Parametric Sensitivity Analysis and Nonlinear Parameter Estimation. Theory and Applications. Ph.D. Dissertation, The University of Wisconsin-Madison, Madison, WI, 1986. Chang, J. C. S.; Brna, T. G.; Sedman, C. B. Pilot Evaluation of Sorbents for Simultaneous Removal of HCl and SO2 from MSW Incinerator Flue Gas by Dry Injection Process. In Proceedings of the International Conference on Municipal Waste Combustion, Hollywood, CA, 1989. Chisholm, P. N. The Absorption of Hydrogen Chloride and Sulfur Dioxide by 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 Calcium Silicate from Humidified Flue Gas, Submitted to J. Eng. Chem. Res. 1999. 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 (2), 175. Collins, E. M.; Menzies, A. W. C. A Comparative Method for Measuring Aqueous Vapor and Dissociation Pressures, with Some of Its Applications. J. Phys. Chem. 1936, 40, 379. Duecker, W. W.; West, J. R. The Manufacture of Sulfuric Acid; Reinhold Publishing Corp.: New York, 1959. Duo, W.; Seville, J. P. K.; Kirby, N. F.; Clift, R. Formation of Product Layers in Solid-Gas Reactions for Removal of Acid Gases. Chem. Eng. Sci. 1994, 49 (24A), 4429. 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. Garea, A.; Viguri, J. R.; Irabien, A. Kinetics of Flue Gas Desulphurization at Low Temperatures: Fly Ash/Calcium (3/1) Sorbent Behavior. Chem. Eng. Sci. 1997, 52 (5), 715. Gullett, B. K.; Jozewicz, W.; Stefanski, L. A. Reaction Kinetics of Ca-Based Sorbents with HCl. Ind. Eng. Chem. Res. 1992, 31, 2437. Holzman, M. I.; Atkins, R. S. Retrofitting Acid Gas Controls: A Comparison of Technologies. Solid Waste Power 1988, 2 (5), 28. Jones, B. F.; Lowell, P. S.; Meserole, F. B. Experimental and Theoretical Studies of Solid Solution Formation in Lime and Limestone SO2 Scrubbers; Environmental Protection Technology Series, EPA-600/2-76-273a; Environmental Protection Agency: Washington, DC, 1976. Jorgensen, C.; Chang, J. C. S.; Brna, T. G. Evaluation of Sorbents and Additives for Dry SO2 Removal. Presented at the Spring National AIChE Meeting, New Orleans, LA, 1986. 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. Karlsson, H. T.; Klingspor, J.; Bjerle, I. Adsorption of Hydrochloric Acid on Solid Slaked Lime for Flue Gas Clean Up. J. Air Pollut. Control Assoc. 1981, 31 (11), 1177. Klingspor, J.; Stromberg, A.; Karlsson, H. T.; Bjerle, I. Similarities between Lime and Limestone in Wet-Dry Scrubbing. Chem. Eng. Process. 1984, 18 (5), 239. Lyon, R. K.; Cole, J. A.; Kramlich, J. C.; Chen, S. L. The Selective Reduction of SO2 to SO3 and the Oxidation of NO to NO2 by Methanol. Combust. Flame. 1990, 81, 30. 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.; Rochelle, G. T. Simultaneous Sulfur Dioxide and Nitrogen Dioxide Removal by Calcium Hydroxide and Calcium Silicate Solids. 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. Ruiz-Alsop, R. N.; Rochelle, G. T. Effect of Deliquescent Salt Additives on the Reaction of SO2 with Ca(OH)2. In Proceedings of the 189th Meeting of the American Chemical Society, Miami, FL, 1985.

4080

Ind. Eng. Chem. Res., Vol. 38, No. 10, 1999

Shen, C. H. Nitrogen Dioxide Absorption in Aqueous Sodium Sulfite. Ph.D. Dissertation, The University of Texas at Austin, Austin, TX, 1997. Simons, G. A. Predictions of CMA Utilization for In-situ SO2 Removal in Utility Boilers. Resour., Conserv., Recycl. 1992, 7, 161. Simons, G. A.; Garman, A. R. Small Pore Closure and the Deactivation of the Limestone Sulfation Reaction. AIChE J. 1986, 32 (9), 1491. Stewart, W. E.; Caracotsios, M.; Sørensen, J. P. Parameter Estimation from Multiresponse Data. AIChE J. 1992, 38 (5), 641. Stouffer, M. R.; Yoon, H.; Burke, F. P. An Investigation of the Mechanisms of Flue Gas Desulfurization by In-Duct Dry Sorbent Injection. Ind. Eng. Chem. Res. 1989, 28, 20. U.S. EPA. Standards of Performance for New Stationary Sources and Emission Guidelines for Existing Sources: Hospital/ Medical/Infectious Waste Incinerators; 40 CFR 48348; Environmental Protection Agency: Washington, DC, Sept 15, 1997. U.S. EPA. Standards of Performance for New Stationary Sources and Emission Guidelines for Existing Sources: Municipal Waste

Combustors; 40 CFR 65387; Environmental Protection Agency: Washington, DC, Dec 19, 1995. von Rosenberg, D. U.; Chambers, R. P.; Swan, G. A. Numerical Solution of Surface Controlled Fixed-Bed Adsorption. Ind. Eng. Chem. 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. Weinell, C. E.; Jensen, P. I.; Dam-Johansen, K.; Livbjerg, H. Hydrogen Chloride Reaction with Lime and Limestone: Kinetics and Sorption Capacity. Ind. Eng. Chem. Res. 1992, 31, 164. Yuan, C. S.; Fong, Z. T. Experimental Study on Simultaneous SO2, HCl, and NO Removal via Spray Drying Technology. Presented at the 85th Annual Meeting & Exhibition of the Air & Waste Management Association, Kansas City, MO, 1992.

Received for review October 16, 1998 Revised manuscript received July 7, 1999 Accepted July 21, 1999 IE9806601