Environ. Sci. Technol. 2006, 40, 2714-2720
Development of Cost-Effective Noncarbon Sorbents for Hg0 Removal from Coal-Fired Power Plants JOO-YOUP LEE,† YUHONG JU,‡ T I M C . K E E N E R , * ,† A N D RAJENDER S. VARMA‡ Department of Civil & Environmental Engineering, University of Cincinnati, Cincinnati, Ohio 45221-0071, Clean Processes Branch, Sustainable Technology Division, National Risk Management Research Laboratory, U. S. Environmental Protection Agency, MS 443, 26 W. Martin Luther King Drive, Cincinnati, Ohio 45268
Noncarbonaceous materials or mineral oxides (silica gel, alumina, molecular sieves, zeolites, and montmorillonite) were modified with various functional groups such as amine, amide, thiol, urea, and active additives such as elemental sulfur, sodium sulfide, and sodium polysulfide to examine their potential as sorbents for the removal of elemental mercury (Hg0) vapor at coal-fired utility power plants. A number of sorbent candidates such as amine- silica gel, urea- silica gel, thiol- silica gel, amide-silica gel, sulfuralumina, sulfur-molecular sieve, sulfur-montmorillonite, sodium sulfide-montmorillonite, and sodium polysulfidemontmorillonite, were synthesized and tested in a lab-scale fixed-bed system under an argon flow for screening purposes at 70 °C and/or 140 °C. Several functionalized silica materials reported in previous studies to effectively control heavy metals in the aqueous phase showed insignificant adsorption capacities for Hg0 control in the gas phase, suggesting that mercury removal mechanisms in both phases are different. Among elemental sulfur-, sodium sulfide-, and sodium polysulfide-impregnated inorganic samples, sodium polysulfide-impregnated montmorillonite K 10 showed a moderate adsorption capacity at 70 °C, which can be used for sorbent injection prior to the wet FGD system.
Introduction According to the analyses conducted by the U.S. EPA (1) and the United Nations (2), recent estimates of annual total global mercury emissions from natural and anthropogenic sources range from approximately 4400 to 7500 tons per year. Anthropogenic U.S. mercury emissions are estimated to account for approximately 3% of the total global emissions (132-225 tons/yr), and coal-fired power plants in the U.S. are estimated to contribute to approximately 1% of the total global mercury emissions, which accounts for 33% of anthropogenic U.S. emissions (44-75 tons/yr). Three forms of mercury are released from coal-fired boilers: elemental mercury (Hg0), oxidized mercury (Hg+ or Hg2+), and par* Corresponding author e-mail:
[email protected]. ‡ National Risk Management Research Laboratory, U. S. Environmental Protection Agency. † University of Cincinnati. 2714
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 8, 2006
ticulate mercury (1, 2). Among these forms, elemental mercury can be transported thousands of miles before depositing to land and water (1, 2). On March 15, 2005, the U.S. EPA announced the Clean Air Mercury Rule (3) to permanently limit mercury emissions from coal-fired power plants. The first-phase cap is 38 tons annually beginning in 2010, with a final cap set at 15 tons starting in 2018, resulting in nearly 70% reductions of 1999 emission levels. Sorbent injection is one of the most promising technologies for application to the utility industry as virtually all coalfired boilers are equipped with either an electrostatic precipitator (ESP) or a baghouse (4). Among various sorbents tested under the Department of Energy’s (DOE) field testing program (4), the most widely tested and promising sorbent is found to be raw activated carbon, which has displayed the capability of capturing both elemental and oxidized mercury from flue gas streams. However, activated carbon still has the following limitations (4, 5): (1) raw activated carbon is still expensive (e.g., Norit DARCO FGD activated carbon, DOE’s benchmark sorbent, costs $0.42/lb); (2) it requires a very high carbon-to-mercury mass ratio (3000∼100 000), especially in flue gases with low HCl content such as subbituminous and lignite coals; (3) it degrades the quality of captured fly ash because it is a general sorbent which also absorbs oxygen from the air. Thus, it adversely impacts its sales as a coal combustion byproduct for Portland cement in concrete. The DOE’s early estimate cost for disposal of fly ash containing spent activated carbon is approximately $3 billion per year, which is approximately three times as much as the utility industry makes a year by selling their fly ash as a coal combustion byproduct (6). Consequently, there is a strong desire to develop efficient and cost-effective noncarbon-based sorbents which may not have an adverse impact on fly ash sales and lead to increased cost for disposal. The primary objective of this study was to develop and test novel cost-effective noncarbonaceous solid sorbent materials for sorbent injection suitable for removal of mercury from power plant emissions, preferably as a discrete waste to minimize formation of toxic waste generation. The target cost of the developed materials was aimed to be comparable to or less than the current market prices of raw activated carbon. It is worthwhile to reiterate that noncarbonaceous sorbents would be an economically viable option, although their prices are comparable to raw activated carbon, as long as they do not negatively impact on fly ash sales and do not require landfill disposal. An array of sorbent materials were synthesized and evaluated in a lab-scale fixed-bed system, and their relative effectiveness in mercury capture forms the subject matter of this paper.
Experimental Section Mercury Measurement. An elemental mercury permeation tube (VICI Metronics, Inc., Poulsbo, WA) was used for steady Hg0 vapor generation. An ultrahigh purity argon (>99.999%) carrier gas flow rate was set to 100 mL/min to steadily inject Hg0 vapor into the fixed-bed system, and was maintained at all times with a mass flow controller (MFC, Model GFC 171, Aalborg Instruments and Controls, Inc., Orangeburg, NY). The 3-cm long Hg0 permeation tube immersed in a water bath was set to release Hg0 vapor at a rate of 91 ng/min, which gave 11 ppbv of an inlet Hg0 concentration. The water bath was successfully operated at 55.5 °C within (0.2 °C to meet the specified inlet Hg0 concentration. The influent Hg0 vapor concentration was repeatedly measured with 4% (w/v) KMnO4/10% (v/v) H2SO4 impinger solutions used in the Ontario Hydro Method (7), and the mercury concen10.1021/es051951l CCC: $33.50
2006 American Chemical Society Published on Web 03/15/2006
FIGURE 1. Schematic diagram of a lab-scale fixed-bed system for Hg0 adsorption test. tration in the acidified KMnO4 solution was determined using a cold vapor atomic absorption spectrophotometer (CVAAS, Model 400A, Buck Scientific Inc., East Norwalk, CT). The variations in mercury concentration were confirmed within (0.5 ppbv. An online Hg analyzer (UV-1201S with mercury analysis kit, Shimadzu Corp., Columbia, MD), calibrated using the calibrated Hg0 permeation tube, was also used to observe breakthrough curves, and to study the dynamic adsorption capacity of sorbents. The gas-phase mercury detection limit for this online instrument was 4 µg/Nm (3). A main argon flow rate (900 mL/min) controlled by a MFC was also supplied and mixed with the Hg0-laden stream. Then, the total stream entered the on-line mercury analyzer and its effluent gas stream was captured by the acidified KMnO4 solution to analyze the mercury content by the CVAAS. A blank test was carried out in order to examine the adsorption of mercury vapor on the tubing, reactor, and blank glass fiber filter prior to the main experimental study on mercury uptake by sorbents. The system was cleaned with 10%(v/v) nitric acid and deionized water before each experiment to remove residual mercury in the system as described in the Ontario Hydro Method. The amount of residual mercury in the tubing and the reactor wall turned out to be negligible in comparison with the amount of mercury recovered at the outlet of the system for a 10-min testing period. Fixed-Bed Adsorption Tests. A lab-scale fixed-bed apparatus was constructed, as shown in Figure 1, to explore the uptake capacity of Hg0 vapor by sorbents. The fixed-bed reactor was constructed to allow for a total flow of 1 L/min gas throughput at 23 °C. The 1.27-cm i.d. reactor made of borosilicate material was determined to meet a superficial velocity of 13 cm/s at 23 °C in the empty bed reactor. The superficial velocity of the simulated flue gas was chosen to simulate a flow pattern among the ductwork of coal-fired
utility (∼50 ft/s ) ∼1500 cm/s), an electrostatic precipitator (ESP) (∼5 ft/s ) ∼152 cm/s), and a fabric filter (∼3 ft/min ) ∼2 cm/s). The sorbent sample was mixed with silica (SiO2, Fisher Scientific, fine granules, particle size: 149-420 µm) as a diluent prior to being packed in the reactor. About 2030 mg of each sorbent in 6 g of silica was used, and the bed material was supported by a fritted quartz disk with a Teflon O-ring and a glass fiber filter with a nominal 1 µm pore diameter in order to minimize channeling, and to prevent the escape of sorbent through the bed. An additional filter system with a glass fiber filter with a nominal 0.7 µm pore diameter was used at the outlet of the reactor to capture sorbent particles potentially escaping from the bed. The test conditions are summarized in Table 1. During each test, the mercury-laden inlet gas bypassed the sorbent bed and passed to the analytical system until the desired inlet mercury concentration was established. Then, the adsorption test was initiated by diverting the gas flow through the sorbent column in downflow mode to minimize the potential for fluidization of the bed. All of the tubing and valves in contact with Hg0 were constructed from Teflon, which has been demonstrated to have good chemical resistance and inertness toward elemental mercury. The sorbent bed and filter system was placed in a temperaturecontrollable convection oven (Stabil-Therm electric utility oven, model OV-500C-2, Blue M Electric Company, Blue Island, IL), which could maintain the system temperature within (0.5 °C. A Teflon-coated thermocouple was installed inside the fixed-bed reactor to monitor and manually control the gas temperature at the inlet of the sorbent bed as needed. Tests were carried out in the reactor that was maintained at 70 and 140 °C under Hg0-laden argon flow for screening purpose. Since the adiabatic saturation temperature of a wet scrubber with approximately 13% H2O vapor in a typical flue gas is around 52 °C, 70 °C was selected to implement this mercury control method across a wet scrubber slightly above VOL. 40, NO. 8, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2715
TABLE 1. Summary of Test Conditions item
conditions 1/
2-in. (1.27 cm) i.d. borosilicate 23, 70, 140 1,000 @ 23 °C, 1,159 @ 70 °C, 1,395 @ 140 °C flow mode downflow superficial velocity in 13 @ 23 °C, 15 @ 70 °C, an empty reactor (cm/s) 18 @ 140 °C residence time in 0.23 @ 23 °C, 0.20 @ 70 °C, an empty reactor (s) 0.17 @ 140 °C Reynolds number 97 @ 70 °C sorbent 20-30 mg in 6 g of a sand bed gas argon inlet Hg0 concentration 91 ng/min ) 11 ppbv ) 91 µg/m3 @ 23 °C adsorption capacity Up to 83% total breakthrough; determination spent sorbent analysis; Ontario Hydro impinger solution analysis
reactor temperature (°C) flow rate (cm3/min)
the adiabatic saturation conditions. A higher temperature, 140 °C, was also chosen to simulate similar conditions between an air preheater and a particulate control device for sorbent injection (8). If an experiment for each sorbent, which was first conducted at 70 °C, was successful, then another test was then carried out at 140 °C. The total gas flow rate was monitored at the outlet of the impinger system using a bubble flow meter. It was confirmed that consistently reproducible results were obtained with this experimental setup. Data Analysis. The mercury uptake capacity (or sorbent adsorption capacity) of a known weight of a sorbent for the on-line Hg analyzer tests in argon flow was calculated in terms of µg Hg adsorbed/g sorbent material from the breakthrough curve (total effluent mercury concentration versus time) for the sorbent at each set of test conditions. The area under the inlet Hg0 concentration line and a breakthrough curve was used to determine how much mercury was adsorbed by the sorbent. By material balance, the area between the curve and the line provides the information on the total Hg0 solute adsorbed onto a sorbent if the entire bed reaches equilibrium with Hg0 vapor. The area up to a break-point time, which is often taken as a relative concentration (a ratio of effluent to influent solute concentration) of 0.05 or 0.1, represents the actual amount of the solute adsorbed. The area under the breakthrough curve was constructed by interpolating the sampling data points of the online mercury analyzer readings. The shape of the breakthrough curve is a good indicator of mass-transfer resistance. If the width of the mass-transfer zone is narrow relative to the bed length, the breakthrough curve will be rather steep, and most of the sorbent capacity will be utilized at the break point, which is desirable for efficient use of the sorbent. Since the empty-bed contact time is less than 0.2 s at 70 and 140 °C in the fixed bed, Hg0 adsorption for a successful sorbent will be a mass-transferlimited process. A next-step research effort would be to build
FIGURE 2. Proposed structures of (a) amine, (b) urea, (c) thiol, (d) amide functionalized silica surface. a larger pilot-scale entrained-flow reactor for the same particle size and superficial velocity so that further experiments can be carried out in order to obtain mass-transfer correlations for the estimation of an injection rate for a specific sorbent. The adsorption capacities of the tested sorbents were usually measured up to 83% of the inlet mercury concentration and compared with those of the Norit FGD activated carbon. The adsorption capacities obtained using the online mercury analyzer were also compared with mercury content captured in the KMnO4 impinger solution using the CVAAS. The discrepancy between mercury measurements using two different instruments (online analyzer and CVAAS) was usually in the range of 3 and 15%, and impinger analysis data was taken as the uptake capacities of the sorbents. These uptake capacity values, together with the optimal temperature windows, were used to characterize sorbents. Desirable characteristics in a novel sorbent include a high uptake capacity for Hg0 across the broadest optimal temperature window independent of flue gas compositions. Synthesis of Sorbent. Activated Carbon. A commercially available activated carbon, DARCO FGD (DOE’s benchmark sorbent, Norit America Inc., Marshall, TX) was selected as a benchmark sorbent. It is a lignite-coal-based powdered activated carbon, and its physical and chemical properties have been well documented (9). Functionalized Silica. The surface of silica gel was modified by various functional groups such as amine, urea, thiol, amide, etc., as illustrated in Figure 2, and the synthesized sorbents were summarized in Table 2. Amino-functionalized silica, effective in capturing heavy metals such as Co2+, Fe3+, Cu2+, and Zn2+ ions in the aqueous phase (10, 11), was synthesized using 3-aminopropyltrimethoxysilane and a silica precursor, tetraethyl orthosilicate (all purchased from Aldrich Chemical Corp., St. Louis, MO) in the aqueous phase. A typical synthesis procedure involved mixing a surfactant, cetryltrimethylammonium bromide (CTAB), a silica precursor, tetraethylothosilicate (TEOS), an amine precursor, 3-aminopropyltrimethoxysilane, distilled water, and sodium hydroxide (NaOH) in a molar ratio of 0.12:1.0:0.15:130:0.7. The
TABLE 2. Functionalized Silica Prepared for Hg0 Sorption Tests functional group
precursor
remarks
amine urea thiol amide
3-aminopropyltrimethoxysilane 1-[3-(trimethoxysilylpropyl)] urea 3-mercaptopropyl-trimethoxysilane 3-aminopropyl)triethoxysilane and chloroacetamide
used for heavy metal treatment in the aqueous phase10,11 co-benefit with SCR13 proven to be efficient for Hg treatment in water14 used for heavy metal treatment15
2716
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 8, 2006
TABLE 3. Comparison of Hg0 Uptake Capacities Reported in Previous Studies and Current Work inlet Hg0 concentration (µg/Nm3) 83
T (°C)
gas
linear velocity (cm/s)
online Hg analyzer
avg. particle size (µm)
surface area (m2/g)
Ar
13 @ 23 °C
AASa
∼15 (FGD)
547
Ar ∼4 simulated 18 @ 135 °C gas N2 ∼10 @ standard temp. (N/A)
AFSb AAS
N/Ac ∼15 (FGD)
AAS
∼10 @ standard temp. (N/A) 344-497 @ 36 °C 6 @ room temp.
4860 53-61
23 70 140 138 135
249
23
8
140 140
N2
6420 110
36 25, 140
Air N2
a
AAS: atomic absorption spectrophotometer.
b
Hg0 uptake capacity (µg/g)
ref current work
650 (AC-1) 600
123 81 ∼0 370 ∼0 w/o HCl
∼15 (FGD)
547
∼120
17
AAS
∼15 (FGD)
547
AAS AAS
104-147 ∼75 (BPL, Calgon carbon)
1250 1026
0-25 269 (6 h test results) 1420 ∼10
8 9
18 19 20
AFS: atomic fluorescence spectrophotometer. c N/A: not available.
FIGURE 3. Breakthrough curves for Norit FGD activated carbon with 11 ppbv inlet Hg0 concentration in argon at 23, 70, and 140 °C. mixture was then heated with continuous stirring at 100 °C for 24 h. The solid white product was then recovered by filtration and was refluxed in ethanol to remove the surfactant, CTAB. The final material was washed with copious amounts of distilled water and ethanol, and dried under vacuum at 100 °C for 24 h. A urea-functionalized silica sample was also prepared from 1-[3-(trimethoxysilylpropyl)] urea and mesoporous silica gel (nominal pore diameter: 100 Å, Aldrich Chemical Corp., St. Louis, MO) as described in the literature (12). Urea injection has been evaluated for mercury control wherein urea is an ammonium source (13). The urea-functionalized silica was prepared by suspending 10.0 g of silica gel in 100 mL of dry toluene with an excess amount (1.23 g) of 1-[3(trimethoxysilylpropyl)] urea. Thiol-functionalized sol-gel material and polymers, reported to be very effective in mercury adsorption in the liquid phase and as liquid crystal materials (14), were synthesized in an analogous manner as amino-functionalized silica described earlier. In view of the reported utility of polystyrene sulfone amide-based resin to selectively adsorb mercuric ions over many metallic ions in the aqueous phase (15), an amide functionalized silica was prepared by a simple nucleophilic substitution reaction of suspension of amino-silica (2.00 g) with excess chloroacetamide (0.25 g) in dry toluene.
FIGURE 4. Breakthrough curves for functionalized silica sorbents with 11 ppbv inlet Hg0 concentration in argon at 70 °C. Sulfur Impregnated Inorganic Sorbents. Elemental sulfur was precipitated by dissolving sodium thiosulfate (1.50 g, Na2S2O3, Aldrich Chemical Corp., St. Louis, MO) with 2 M hydrochloric acid (5.0 mL) in distilled water (10 mL) as shown in reaction 1 (16). Montmorillonite K 10 (MK10, 5 g, Aldrich Chemical Corp., St. Louis, MO) was then added to the suspended solution, and the mixture was agitated at room temperature for 30 min. The resulting solid was filtered out and washed with distilled water three times (5 mL × 3), and dried under vacuum at 50 °C for 16 h to afford sulfurimpregnated montmorillonite clay (6.05 g) as a pale yellow solid. Likewise, sulfur-impregnated molecular sieves (13X, Acros Organics USA, Morris Plains, NJ) and alumina (Al2O3, type: neutral Brockmann activity I, ACROS Organics, Morris Plains, NJ) were also prepared. These two materials (molecular sieves and alumina) were used as high surface area sorbents.
Na2S2O3 + HCl f Na2SO3 + SV + H2O
(1)
Sodium Sulfide- and Sodium Polysulfide-Impregnated Montmorillonite Clay. Sodium sulfide (Na2S, 0.60 g, Aldrich Chemical Co) was slowly added to a suspended solution of montmorillonite K 10 (2 g) in 10 mL distilled water with vigorous agitation, and the mixture was kept at 45 ( 2 °C for VOL. 40, NO. 8, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2717
FIGURE 5. Breakthrough curves for various sulfur-, sodium (poly)sulfide-promoted sorbents with 11 ppbv inlet Hg0 concentration in argon at 70 and 140 °C. 30 min. The liquid was then removed by filtration, and the resulting dark colored solid was dried under vacuum for 18 h below 50 °C to afford sodium sulfide-impregnated montmorillonite clay (2.45 g) as a black solid. Sodium polysulfide (e.g., sodium tetrasulfide, Na2S4, and/ or sodium pentasulfide, Na2S5) was prepared as shown in Reaction 2 (16) using sodium sulfide (Na2S, 1.50 g, Aldrich Chemical Corp., St. Louis, MO) and ground sulfur fine powder (0.65 g, Aldrich Chemical Corp., St. Louis, MO), which were added under vigorous agitation to the suspended solution of montmorillonite K 10 clay (2.00 g) and 2 N aqueous NaOH solution (0.50 g) in 40 mL acetone at room temperature. The mixture was then kept at 65 ( 2 °C under the reflux condition for 6 h. The solvent was then removed from the resulting brown-colored suspension below 60 °C using a rotary evaporator (BU ¨ CHI Rotavapor R-200, BU ¨ CHI Laboratory Equipment, Switzerland) to give sodium polysulfide-impregnated montmorillonite clay (Na2Sn-clay, 3.65 g) as a dark brown solid. H2O ∆
Na2S + S 9 8 Na - S(n) - Na Clay
(2)
Results and Discussion Norit FGD Activated Carbon. The performance of the Norit America’s raw FGD activated carbon was evaluated prior to evaluation of candidate sorbents in the fixed-bed column with 11 ppbv inlet Hg0 concentration in argon flow at three
different temperatures of 23, 70, and 140 °C in duplicate as shown in Figure 3. It showed average Hg0 adsorption capacities of 123 (121 and 124) µg Hg0/g FGD at 23 °C, 81 (76 and 86) µg Hg0/g FGD at 70 °C, and ∼0 at 140 °C when the online analyzer readings were recorded up to 83% breakthrough points. The data showed that the experimental results (i.e., adsorption capacities obtained from breakthrough curves) were consistently reproducible under the same experimental conditions, and the capacity of Hg0 adsorption onto activated carbon increases as adsorption temperature decreases, suggesting a physical adsorption mechanism. It was also noted that the ability of FGD activated carbon to remove Hg0 was almost negligible in the absence of HCl gas at 140 °C, as corroborated in a previous study conducted with 53-61 µg Hg/Nm3 inlet Hg0 concentration at 135 °C (9). However, it is worth mentioning that it is almost impossible to simply compare the uptake capacities of Hg0 obtained from this work with various values reported in the literature since they were acquired under different experimental conditions (e.g., Hg0 inlet concentration, flow pattern, residence time, simulated flue gas compositions, and temperature) and using a variety of activated carbons (e.g., surface area, pore size distribution, chemical compositions, particle size, etc.). Thus, the equilibrium adsorption capacities reported in the literature, as shown in Table 3, are for comparison purposes in terms of experimental conditions and different types of activated carbons. Although other results obtained under similar experimental conditions in previous studies (9, 17, 18) are not always consistent, it is noteworthy that the Hg0 uptake capacities of raw FGD activated carbon obtained at 23 and 140 °C from this work are quite comparable to the data reported in two previous studies (9, 17). Functionalized Silica. The functionalized silica samples with amine, urea, thiol, and amide functionalities were tested under argon flow at 70 °C. Although these sorbents are reported to show fairly good uptake capacities of mercuric ions in the aqueous phase (10, 11, 13-15), they demonstrated negligible gas-phase adsorption capacities toward Hg0 as shown in Figure 4. These results indicate that the adsorption mechanism(s) of Hg0 in the gas phase is completely different from that(those) of mercuric ions (Hg+/Hg2+) in the liquid phase. Sulfur, Sodium Sulfide-, Sodium Polysulfide-Impregnated Inorganic Sorbents. Clay minerals exist abundantly in nature, and their high surface area, adsorptive and ionexchange properties, and ease in dispersing in liquids render them ideal candidates for catalytic applications (21-23) and effective in the remediation of various heavy metals including Ni2+, Cu2+, Mn2+, Fe3+, Cr3+, Cd2+, Pb2+, Hg2+, and methylmercury (CH3Hg+) in soil and the aqueous solutions (2428). Clay minerals are made up of layered aluminosilicates, and montmorillonite is one of the major components of bentonite and fuller’s earth with a general formula, Al2O3‚
TABLE 4. Summary of Sulfur- and Sodium (poly)Sulfide-doped Sorbents in Figure 5 sorbent
descriptions
raw MK10
raw montmorillonite K-10
S-Al S-MS S-MK10 Na2S-MK10-1
sulfur-impregnated alumina (Al2O3) sulfur-impregnated molecular sieve sulfur-impregnated MK10 Na2S-impregnated MK10
Na2S-MK10-2
Na2S-impregnated MK10
Na2Sn-MK10
Sodium polysulfide-impregnated MK10
2718
9
remarks reported geometric mean particle size: ∼2 µm Average pore size:
