Environ. Sci. Technol. 2003, 37, 3995-4000
Chlorinated Aromatics from Combustion: Influence of Chlorine, Combustion Conditions, and Catalytic Activity T. O ¨ B E R G * ,† A N D T . O ¨ HRSTRO ¨ M‡ Department of Biology and Environmental Science, University of Kalmar, SE-391 82 Kalmar, Sweden, and Bergstro¨m & O ¨ hrstro¨m, P.O. Box 3028, SE-611 03 Nyko¨ping, Sweden
Research on the formation of chlorinated aromatics in combustion processes has mainly taken place in the laboratory. Previous attempts to correlate observation data from commercial plants have been inconclusive. This study reports on the outcome of an industrial experiment in a full-scale afterburner. The influence of chlorine input, combustion temperature, and catalytic activity was investigated in a factorial design with two blocks. Polychlorinated benzenes, dibenzo-p-dioxins, and dibenzofurans were formed both at combustion temperatures and below 400 °C. The results show that all three experimental factors have statistically significant impact on the formation and release of these toxic byproducts. The quantitative dependence between chlorine input and the occurrence of chlorinated aromatics is of particular interest due to previous controversy. The purpose with this study was to ensure that the installation of a boiler for energy recovery would not cause elevated emissions of chlorinated aromatics. The experiment demonstrated that this risk is probably low, since the presence of catalytic material or an increase in chlorine input is required for this to happen. A general conclusion was that industrial experimentation employing the principles of statistical design could improve the validity in recommendations regarding commercial plant operation.
Introduction Polychlorinated benzenes (CBz), dibenzo-p-dioxins (PCDD), dibenzofurans (PCDF) were detected in fly ash from municipal solid waste incinerators (MSWI) more than 25 years ago (1-3). De novo formation of chlorinated aromatics was discussed by Olie et al. (2), but a more comprehensive treatment was given by Bumb et al. (4). The quantitative relationships between process factors and chlorinated aromatics in flue gas emissions were identified as an important research topic from the beginning (5). Early full-scale experimental investigations in combustion plants highlighted the temperature dependence of both the formation and the decomposition mechanisms for chlorinated aromatics (6, 7). Subsequent laboratory studies indicated that the time/temperature relationship for the thermal decomposition in an afterburner environment could be described with a pseudo-first-order expression (8, 9). Data reported in a recent full-scale study from an industrial * Corresponding author e-mail:
[email protected]. † University of Kalmar. ‡ Bergstro ¨m & O ¨ hrstro¨m. 10.1021/es034056f CCC: $25.00 Published on Web 07/26/2003
2003 American Chemical Society
afterburner also showed this exponential relationship between decomposition of chlorinated benzenes and the combustion temperature (10). Other early full-scale experimental investigations identified statistically significant quantitative relationships between the chlorine content of the fuel and the emissions of various groups of chlorinated aromatics (11-13). Many attempts were later made to correlate observation data and establish links between plant operation, fuel composition, and formation and release of chlorinated aromatics (14-17). Conclusive cause-effect relationships are, however, difficult to establish without experimentation (18). It is therefore not surprising that results from experiments, both in the laboratory and in full-scale processes, show a better agreement than results from pure observation studies. Laboratory investigations also confirmed the earlier reported results on the importance of both the combustion conditions and the chlorine input (19-28). Low-temperature (300-400 °C) catalytic formation is a third important factor in the overall assessment of emissions of chlorinated aromatics from combustion plants. Interestingly, this formation pathway was simultaneously indicated both by measurement results from full-scale combustion plants and in laboratory experiments (29-32). Copper is a known active oxychlorination catalyst in the Deacon process and linked to the low-temperature formation of chlorinated aromatics through numerous laboratory experiments (3339). Investigators have also verified the catalytic effect of copper in pilot-scale combustion systems and reactors, where copper chloride or other copper compounds have been added to the fuel, have been injected into the combustion zone, or have let reactants pass copper-containing material (40-47). Rapid gas cooling has sometimes been suggested to counteract the copper-catalyzed formation of chlorinated aromatics (48-51), but this would be in conflict with other environmental goals such as efficient energy recovery. The scaling-up of results from a laboratory environment to a full-scale technological process is not a trivial problem, but systematic experimental studies on the formation of chlorinated aromatics in full-scale combustion plants are seldom reported. Statistical experimental design allows the simultaneous assessment of both main effects and interactions of treatment factors and has been used with great success in the chemical industry since the 1950s (52). Unfortunately, even today many experimental studies still employ the traditional “one-factor-at-a-time” approach and thereby neglect the importance of interactions between factors. The present investigation was undertaken to assess the consequences of the installation of a boiler for energy recovery in an industrial plant. A review of the literature did not provide a sufficient basis to assess the possible effects and interactions of changes in combustion temperature, chlorine input, and catalytic effects. Data were available from laboratory- and pilot-scale investigations, but results from industrial plants were generally missing and interactions were not reported. A possible approach could have been to simply evaluate the installation of a test boiler, without changing the normal operating conditions. However, such a study would neither provide information about the possible constraints for pollutant formation in this particular plant nor add the extra confidence provided to the evaluation by “positive controls”. This extra validation was needed, and the purpose of this study was thus expanded, not only to investigate the effect of a boiler installation but also to evaluate how changes in chlorine input, combustion temperature, and catalytic VOL. 37, NO. 17, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3995
FIGURE 1. Experimental setup (acronyms LPG and ID denote liquid petroleum gas and induced draft). material influence the formation and release of chlorinated aromatics from this industrial afterburner. A coherent experimental treatment may also be of interest to help resolve some of the controversy that still exists on the role of chlorine in pollutant formation.
Experimental Section The experimental investigation was carried out in one of the afterburners in the steelworks of SSAB Tunnplåt AB (a subsidiary of SSAB Svenskt Stål AB), in the middle of Sweden. The afterburner is an air pollution control device and reduces the emission of organics from a paint-drying oven, by combustion of the solvent-rich off-gases. Liquid petroleum gas (LPG) is used as the supplementary fuel to obtain suitable combustion conditions in the afterburner. During the experiment, a paint system based on poly(vinyl chloride) was in use. Volatile components used in this paint system include dialkyl (C9-C11) phthalates, hydrotreated light petroleum distillates, and butyl diglycol acetate. Experimental Setup. A used 30-kW boiler for domestic heating was refurbished for the purpose of the experiment, Figure 1. The boiler inlet was attached to the gas duct directly after the outlet from the combustion chamber and recuperator (a heat exchanger). The flue gas flow from the afterburner is ∼17 000 m3/h (dry gas at 101.3 kPa and 0 °C). The flue gases from the main duct were passed at 100-130 m3/h through the boiler, and the gas flow was controlled with a small induced draft (ID) fan. Inside the boiler, the flue gases first passed through the original furnace tube and then, after the rear turning box, into the second pass convection tubes. The temperature upon entering the boiler was 410-430 °C, in the turning box 305350 °C, and at the outlet 100-125 °C. The water temperature in the boiler was controlled to 80 °C with a programmable shunt valve. Measurements. All measurement results have been normalized to dry gas, a pressure of 101.3 kPa (1 atm), a temperature of 0 °C, and an oxygen concentration of 16% (v/v). Carbon monoxide, total organic carbon (TOC), and oxygen were measured after the boiler with continuous instruments. Hydrogen chloride and chlorine gas were determined using U.S. EPA method 26 (53). Organic components were sampled both before and after the boiler, using an all-glass sampling train according to the European standard 1948 (54), Figure 2. An accredited laboratory (ALcontrol AB) carried out the analyses of PCDD and PCDF according to the European 3996
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 17, 2003
standard protocol (54). The same laboratory analyzed chlorinated benzenes by high-resolution gas chromatography and selected ion monitoring mass spectrometry. The analytical procedure was described in an earlier paper (55). Analytical results for PCDD and PCDF were reported as the sum of each congener group (tetra-octa) and for the 17 specific congeners used to calculate the international toxic equivalent quantity (I-TEQ) (56). The detection limits for specific congeners of PCDD and PCDF were in the range 0.002-0.03 ng/sample. Analytical results for chlorinated benzenes were reported for 11 congeners (di-hexa) with detection limits in the range 0.001-0.01 µg/sample. Congeners below the limit of quantification were treated as zero when sums of congeners or I-TEQ were calculated. The gas volumes of samples varied between 5 and 14 m3 dry gas at 101.3 kPa, 0 °C, and 16% O2. Experimental Design. The experiment was set up as a factorial design with eight individual runs in two blocks (57). Each test run was 7 h, with 3 h for sampling and 4 h for preconditioning. Three experimental factors were investigated. The chlorine input was changed by pumping 30 L/h of 4 M HCl (technical grade) directly into afterburner through the LPG burner flame. The normal postflame combustion temperature is 705 °C, but during these tests, the temperature was varied up to 750 °C. Low-temperature catalytic activity was induced by mounting and removing a copper sheet in the boiler (cylinder shaped and lining the walls of the furnace tube). The experimental conditions are easier to replicate with this arrangement, compared to manipulating the element composition of particles in the flue gas or deposited onto the boiler surfaces. Whichever method chosen, the catalyst effect must still be regarded as a qualitative factor. Splitting the experiment into two blocks with one month between provided an opportunity to assess the influence of the operation time for the boiler. This allowed time for particle buildup on boiler surfaces and, since hydrogen chloride was present, the formation of an initial equilibrium concentration of metal chlorides and oxychlorides. The boiler was in operation during the whole test period, and the copper sheet was mounted and removed on a weekly basis to allow conditioning with the normal flue gas. This blocking arrangement also reduces the within-treatment variability and improves precision in the treatment comparisons. The run order within each block was randomized in order to guarantee inferential validity (57). Data Analysis. The outcome of the experiment was evaluated using standard statistical methodology. Simple linear models were fitted to the data using multiple linear regression, and the effects to include were selected from normal probability plots. Analysis of variance (ANOVA) was used to assess statistical significance of the models and the various model parameters. Lack of fit cannot be separated from pure error since no replicates were run, and significance is thus estimated directly from the residuals. An adequate model requires the errors to be normally and independently distributed with constant variance. To fulfill this requirement, a variance stabilizing transformation of the dependent Y-variable (response) may be needed. Suitable power transformations (YR) were found empirically (57). The data analysis was carried out using the software Unscrambler v7.6 (CAMO ASA, Oslo, Norway) and Statistica v6.1 (StatSoft Inc., Tulsa, OK). The computations were run on a PC.
Results and Discussion Experimental settings and results are summarized in Table 1. The combustion quality, expressed as the concentrations of CO and TOC, is directly linked to the combustion
FIGURE 2. All-glass sampling train.
TABLE 1. Experimental Settings and Resultsa run order
experimental factors: addition of chlorine (HCl) combustion temp °C copper lining sheet block factor: conditioning time (weeks) measurements, boiler inlet: Σchlorinated benzenes, µg/m3 sdg 16% O2 PCDD/PCDF, ng of I-TEQ/m3 sdg 16% O2 measurements, boiler outlet: CO mg/m3 sdg 16% O2 TOC mg C/m3 sdg 16% O2 HCl mg/m3 sdg 16% O2 Cl2 mg/m3 sdg 16% O2 Σchlorinated benzenes, µg/m3 sdg 16% O2 PCDD/PCDF, ng of I-TEQ/m3 sdg 16% O2 a
1
2
3
4
5
yes 705 yes
no 705 no
yes 750 no
no 750 yes
no 750 no
0
0
0
0
2.2
0.43
1.2
na
na
1460 16 259
