Environ. Sci. Technol. 2004, 38, 4694-4700
PCDD/F TEQ Indicators and Their Mechanistic Implications J E O N G - E U N O H , †,| ABDERRAHMANE TOUATI,‡ B R I A N K . G U L L E T T , * ,† A N D JAMES A. MULHOLLAND§ Office of Research & Development, National Risk Management Research Laboratory, U.S. Environmental Protection Agency (E305-01), Research Triangle Park, North Carolina 27711, ARCADIS G & M, P. O. Box 13109, Research Triangle Park, North Carolina 27709, School of Civil and Environmental Engineering, Georgia Institute of Technology, Atlanta, Georgia 30332-0512, and Oak Ridge Institute for Science and Education Postdoctoral Program, Oak Ridge, Tennessee 37831
Stack gas samples from two incinerator facilities with different operating conditions were investigated to understand how indicators of toxic equivalency (TEQ) from among the 210 polychlorinated dibenzo-p-dioxin/ furan (PCDD/F) isomers varied. This effort was motivated by the need to find more easily monitored indicator compound(s) of TEQ and to reconcile the varying indicator compounds reported in the literature. The measured isomer patterns were compared with those expected from known formation mechanisms to identify the dominant mechanism(s) and explain why certain compounds are relevant TEQ indicators. Despite differences in the facility types and operating conditions, a common pattern was found for the highly chlorinated (4Cl and higher) PCDDs/Fs. A combination of chlorination/dechlorination reactions as the dominant formation mechanism for PCDF was consistent with the observed isomer patterns, whereas condensation reactions of phenolic precursors appeared to be responsible for PCDD formation. PCDF isomers, rather than the PCDD isomers, were more closely related to the TEQ measure, likely because the chlorination mechanism favors 2,3,7,8-Clsubstitution more than the phenol condensation mechanism. Unlike highly chlorinated PCDD/F isomer patterns, less chlorinated PCDD/F patterns (especially, mono- and diCDF) were sensitive to operating conditions and facility type. Competing formation mechanisms were inferred from the variation of observed isomer distribution patterns; this sensitivity resulted in relatively low correlations of these isomers with PCDD/F TEQ values. This suggests that any use of the low-chlorinated compounds as TEQ indicators for online monitoring processes are likely best suited for plantspecific, rather than universal, applications. In addition to many of the highly chlorinated (penta-CDF, hexa-, and heptaCDD/F) isomers being identified as strong TEQ indicators, 1 of 12 (8%), 5 of 17 (29%), and 5 of 28 (18%) of the separable * Corresponding author phone: (919)541-1534; fax: (919)541-0554; e-mail:
[email protected]. † U.S. Environmental Protection Agency. ‡ ARCADIS G & M. § Georgia Institute of Technology. | Oak Ridge Institute for Science and Education Postdoctoral Program. 4694
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 17, 2004
tri-CDD, tri-CDF, and tetra-CDF isomers, respectively, were identified as strong (R2 > 0.7) TEQ indicators in both incinerators.
Introduction Polychlorinated dibenzo-p-dioxins and polychlorinated dibenzofurans (PCDD/F) are hazardous pollutants formed during combustion processes. Regular PCDD/F emission monitoring from incinerators would assist in ensuring minimal environmental releases, but extractive sampling and analysis is very expensive and time-consuming. Numerous research studies have been conducted to establish indicator compounds for fast and less costly prediction of PCDD/F toxic equivalent (TEQ) concentrations. A common approach is to find a single congener which sufficiently represents toxicity. 2,3,4,7,8-Pentachlorodibenzofuran (PeCDF) is regarded as the most promising TEQ indicator due to its high toxic equivalent factor (TEF) and its commonly high concentration relative to the other 16 toxic congeners (1). Besides measuring solely the 2,3,7,8-Cl-substituted isomers comprising the TEQ measure, other PCDD/F isomers, as well as semivolatile and volatile organic halogen compounds such chlorophenols and chlorobenzenes, show promise as indicator compounds for the emissions of PCDD/F (2-4). Despite many studies that have found potential indicator compounds, there is no agreement on use of a single, “universal” TEQ indicator. For example, Tuppurainen et al. (2) reported that almost all the chlorophenol (ClPh) isomers were correlated strongly with gas-phase PCDD/F TEQ, especially PCDFs, but Blumenstock et al. (3) found that the ClPhs were related with PCDD TEQ better than PCDF TEQ. Recently, studies making use of monoto tri-CDD/F as indicator compounds were performed and good correlation results were reported with TEQ values (57). The interest in use of mono- to tri-chlorinated isomers as possible indicator compounds for PCDD/F TEQ is due to the similarities of these compounds with TEQ-based congeners and the emergence of fast, in-situ, on-line instruments adapted to PCDD/F measurement, such as resonanceenhanced multiphoton ionization (REMPI) with time-of-flight mass spectrometry (TOFMS) (7). Oser et al. (7) reported that 2-monochlorodibenzodioxin/dibenzofuran (2-MCDD/F) and 2,4-dichlorodibenzofuran (2,4-DiCDF) showed a positive correlation with TEQ, whereas Blumenstock et al. (3) did not find a significant relationship with these isomers. One possible reason for the published discrepancies in selection of TEQ indicators may be due to the variation of PCDD/F formation mechanisms between plants and under different operating conditions. The correlative relationships found between indicators and TEQ values have not yet been explained through specific formation mechanisms nor has it been explained how these relationships varied with operating condition and facility type. PCDD/F isomer patterns detected in incinerator emissions can provide insight into the chemical mechanism of their formation due to the specific isomer patterns expected from each mechanism (8). Recently, it was suggested that similar isomer patterns in combustion residues are found in each group of homologues even though the homologue patterns varied with the combustion conditions (9). Many studies have been performed to understand the mechanisms that control isomer pattern similarity, and several PCDD/F isomer prediction models have been developed based on different formation mechanisms (8, 10-15). Wehrmeier et al. (13) suggested that PCDD/F distributions are controlled by the thermodynamic 10.1021/es034997s CCC: $27.50
2004 American Chemical Society Published on Web 08/03/2004
stability of the congeners such that a consistent isomer pattern would be observed regardless of various combustion conditions. But Addink et al. (14) reported that their experimental results could not confirm thermodynamicallycontrolled isomer distributions except for heptachlorinated dibenzofurans (HpCDFs). Iino et al. (15) presented a PCDD/F isomer distribution prediction model based on a dechlorination mechanism which was in good agreement with PCDFs but not PCDDs, suggesting that different mechanisms could apply to PCDD and to PCDF. They assumed that the chromatograph peak ratios relative to the total hepta-CDD/F isomers were determined by position-specific dechlorination kinetics from the respective octa-CDD/F and that these kinetic probabilities could subsequently predict the relative concentrations of lower-chlorinated isomers. Ryu et al. (1011) developed a chlorination-based PCDD/F prediction model based on statistical factors and partitioning between monochlorinated products. Ryu et al. (12) also developed a PCDF isomer distribution prediction model based on phenol condensation reactions, using relative rates of product formation from a known distribution of chlorinated phenols. This model can predict PCDF isomer distributions with reasonable accuracy from measurement of the distributions of phenol and 19 chlorinated phenols (12). Wikstro¨m et al. (16) reported that the PCDDs are mainly formed by chlorophenol condensation, while the PCDFs are formed from a non- or a low-chlorinated precursor followed by further chlorination reactions. These studies suggest that the dominant formation mechanism for PCDD is phenol condensation and for PCDF is chlorination/dechlorination. This work examines PCDD/F indicator/TEQ relationships and how selection of these indicators can be understood in terms of the dominant formation mechanisms. The potential variation of PCDD/F isomer distributions due to different formation mechanisms and, hence, selection of specific indicators for TEQ prediction, are examined. Two dissimilar facilities operating under a range of conditions were studied to understand the role of the dominant formation mechanisms in determining which PCDD or PCDF homologues have the greatest potential to provide TEQ indicators.
Experimental Section Two data sets were used for the present study: 11 samples were from a municipal solid waste incinerator (stoker-fired) near Norfolk, Virginia (6), hereafter termed, “Norfolk” and 8 samples were from a waste- and #2 fuel-oil-fired North American Package Boiler (“Boiler”) facility in Research Triangle Park, North Carolina (17). These distinctive facility and fuel types are in common use throughout the United States. In the Boiler work, samples were taken at different combustion conditions such as fuel type, combustion efficiency, and input Cl concentration. The Norfolk facility consists of four 450-metric-ton-per-day refuse-derived fuel (RDF) boilers. The Norfolk incinerator fired RDF only or cofired RDF with two types of coals to see the effect of coal cofiring on PCDD/F formationsone was a high-sulfur (3.40%) coal and the other was low-sulfur (0.70%). The Boiler is a 732-kW, 3-pass, firetube, marine utility boiler capable of firing natural gas or a variety of fuel oils. Stack gas samples were collected on the boiler while co-firing fuel oil and a simulated hazardous waste, consisting of dichlorobenzene and copper naphthenate, under a variety of different combustion conditions. Detailed information on sampling and operating conditions has been described elsewhere (6, 17). All 210 PCDD/F isomers were analyzed in all of the samples by highresolution gas chromatography (HRGC)/low resolution mass spectrometry (LRMS) (Hewlett-Packard 5890/ 5971) with a DB-Dioxin column. Statistical Analysis. Correlation coefficients (R2) were calculated for the relationship between PCDD/F isomer
concentrations and the TEQ values. To evaluate similarities and differences in the PCDD/F isomer distributions among the samples and see the relationship between homologue/ isomer concentration and TEQ values, principal component analysis (PCA) was used. PCA is a multivariate statistical analysis that allows evaluation of the absolute and relative importance of variables as well as graphical representation of the same (18). PCA was performed using software SIMCA-P 7.01 (Umetrics, Sweden). Prior to analysis, the PCDD/Fs homologue and isomer data were normalized to their respective total and total homologue concentrations, as described below.
Results and Discussion Relationship between Homologue Concentration and TEQ. The relationship between homologue concentrations and TEQ values for these two data sets (ranging from ∼1 to ∼70 ng TEQ/m3, 7% O2) was investigated using PCA. For each of the 19 (11 Norfolk and 8 Boiler) runs, the total homologue concentration was normalized by the concentration sum of the total 210 PCDD/F isomers. The PCA “objects” were each sample run and the “variables” were the normalized homologue concentrations and TEQ values (ng TEQWHO98/m3). While the score plot (not shown) of the homologue profiles of these 19 data runs showed some distinction according to incinerator types, the loading plot (Figure 1) allows general cross-facility relationships between homologue and TEQ concentrations to be examined. The hexa-CDF and heptaCDD homologue concentrations are extremely close to the TEQ values. The mono- and di-CDD/F and tri-CDD homologue concentrations are not proximal to the TEQ (located in the lower left corner), indicating that their concentrations did not relate well with the other PCDD/F homologues and the TEQ values. It was reported that the less-chlorinated, mono- to tri-CDD/F homologues were more affected by operating conditions than the more highly chlorinated (4Cl and higher) PCDD/F homologues (16). This result suggests that the formation mechanism of the less chlorinated DD/F homologues is somewhat different from that of the higher chlorinated PCDD/F and, ultimately, these low-chlorinated DD/F may be more difficult to relate to the TEQ value. Figure 1 also shows that the tetra- and penta-CDD homologues did not correlate well with other highly chlorinated homologues, nor with TEQ. Relationship between Isomer Concentration and TEQ. The relationship between PCDD/F isomer concentrations and TEQ values in the separate facility and combined facility data sets was determined by a correlation analysis. The number of isomers within each homologue which showed a positive, moderate (R2 > 0.5) correlation with TEQ divided by the number of chromatographically separable isomers within the homologue for the separate and combined data sets are shown in Table 1. The “combined” column indicates the fraction of isomers from the two facilities’ combined and separate data which showed a positive, moderate (R2 > 0.5) correlation with TEQ, allowing for the possibility of finding more general indicators from a wider range of data. The “common” column indicates the fraction of the same isomers from both facilities (the intersection) that match the R2 criteria. The last column of Table 1 shows the fraction of each homologue’s isomers that satisfies the more rigorous criteria of a positive, strong (R2 > 0.7) correlation with TEQ for the each data set and the combined data set. These isomers, representing the selected set of strongly correlating isomers from the combined facility data as well as their pooled data, may be considered “strong” indicators for use as universal indicators across facilities. The values of 0.5 and 0.7 were arbitrarily chosen to reflect initial approximations of “good” and “improved” choice of indicator compounds that could explain the variance of the PCDD/F data. Selection VOL. 38, NO. 17, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4695
FIGURE 1. PCA results for the relationship between homologue concentrations and TEQ.
TABLE 1. Number of Isomersa Which Show Positive, Moderate-to-Strong Correlation with TEQ in Each Separate and Combined Data Sets separate (R 2 > 0.5)
c
Norfolk
Boiler
common (R 2 > 0.5) NF, B
combined (R 2 > 0.5) NF, B, NF + B
strong combined indicator isomers (R 2 > 0.5) NF, B, NF + B (isomer identificationb)
MCDF DiCDF TriCDF TeCDF PeCDF
0/4 0/14 16/17 18/28 14/19
0/4 6/14 13/17 17/28 19/19
0/4 0/14 13/17 13/28 14/19
0/4 0/14 11/17 16/28 18/19
0/4 0/14 5/17 (148/127, 123/247/178/146, 237, 238, 267) 5/28 (1278/2368, 1267/1469/2467, 2347, 2378,c 3467) 10/19 (13468, 12467, 12479/14678, 12478, 13469,12378, 12678/23468, 12369/12489/12679, 12349/23478) 8/13 (134678, 124678, 134679, 124679,123478, 123678) 123467/123479, 234678) 1/4 (1234678) 0/1 0/2 0/7 1/12 (237) 0/16 0/11 4/7 (123469/123478, 123678, 123789, 123467) 2/2 (1234679,1234678) 0/1
HxCDF
13/13
13/13
13/13
12/13
HpCDF OCDF MCDD DiCDD TriCDD TeCDD PeCDD HxCDD HpCDD OCDD
4/4 1/1 0/2 1/7 9/12 2/16 6/11 7/7 2/2 0/1
4/4 1/1 0/2 0/7 3/12 0/16 3/11 6/7 2/2 1/1
4/4 1/1 0/2 0/7 2/12 0/16 3/11 6/7 2/2 0/1
3/4 1/1 0/2 0/7 3/12 7/16 51/1 7/7 2/2 1/1
a x/y ) Number of isomer peaks which showed positive correlation with TEQ/number of total separable isomer peaks. b (a/b) ) coeluted isomers. Boldface ) TEF isomers.
and use of specific isomers as TEQ indicators for on-line monitoring processes can only be established upon field validation of R2 values and consideration of other data quality indicators, such as recovery of dynamically spiked standards and detection limits. The tri-CDD, tri-CDF, and tetra-CDF homologues, respectively, contain 1 of 12 (8%), 5 of 17 (29%), and 5 of 28 (18%) of their isomers as potential universal 4696
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 17, 2004
indicators, along with most of the highly chlorinated DD/DF isomers. This result agrees well with the relationship between the homologues and TEQ values (Figure 1), with one exception (tri-CDD). Eleven of the seventeen isomers that are assigned TEF values and, hence, comprise the TEQ measure, were found to be good TEQ predictors. The remaining six TEF isomers also showed a positive correlation
TABLE 2. R2 Value of PC1 in the Separate and Combined Data Sets (Unit, %) separate
MCDF DiCDF TriCDF TeCDF PeCDF HxCDF HpCDF DiCDD TriCDD TeCDD PeCDD HxCDD
combined
NF
Boiler
NF + B
92.3 91.1 95.1 89.6 90.8 98.5 100 87.0 77.1 87.9 90.7 97.6
83.0 88.2 80.1 60.7 73.3 76.4 99.3 79.0 96.2 66.0 81.6 77.7
63.2 76.5 85.9 67.8 77.5 86.0 99.9 74.8 75.8 71.3 84.7 90.6
with TEQ but none had a strong (>0.7) R2 value in all the cases. A considerable number of non-TEF isomers, including Tri-CDD/Fs, were also found to be strong TEQ predictors, supporting the case for on-line, correlative monitoring of PCDD/F emissions by measuring candidate TEQ indicator isomers such as 2,3,7- or 2,3,8-TriCDF. Mechanistic Implications. The determination of select isomers as TEQ indicators raises questions about their universal applicability for different combustion sources. Different chemical mechanisms of PCDD/F formation are known to result in distinctive homologue, isomer, and TEQ interrelationships under a limited range of operating conditions. Identification of the dominant formation mechanism at a source may assist in explaining what TEQ/indicator relationships can be expected. Mechanism identification may also offer answers to, for example, why this study finds that tetra- and penta-CDD are not related with TEQ and why some non-TEF-based isomers are related with TEQ. As described before, similar isomer distribution patterns were reported in combustion samples from multiple sources (13) and specific PCDD/F isomer distribution patterns were developed for different PCDD/F formation mechanisms (1012, 13, 15). Similarities between the observed isomer distribution patterns and the isomer patterns expected from known mechanisms will suggest the dominant formation mechanism and possibly explain selection or exclusion of potential indicators. The similarity in isomer distribution patterns can explain why many other isomers besides TEF isomers have a correlative relationship with TEQ. For example, if the isomer pattern is similar and if a certain isomer, such as 2,3,4,7,8PeCDF, is related to TEQ, then all the other isomers in the PeCDF homologue must have a relationship with TEQ because of the transitive property. In our study, good agreement was observed in the case of highly chlorinated PCDD/F isomers (almost 100% of the isomers were related with TEQ), but the less chlorinated DD/F isomers were not as correlated with TEQ (∼50% of the isomers were related with TEQ) (Table 1). To further understand isomer/TEQ relationships, PCA was used with relative isomer concentrations, first, to examine isomer pattern similarities within each PCDD/F homologue across facility type and operating conditions (including mono- through tri-CDD/F), and second, to determine the homologue-specific dominant formation mechanisms in each facility. The first PCA compared relative isomer patterns between both facilities using the isomer concentration normalized by the sum of the total isomer concentration within each homologue, for all 19 runs. The R2 value of the PC1, representing the data variance which can be described by the first principle component, is presented in Table 2. DiCDD
and HpCDD were not applicable to PCA since these homologues have only two isomers each. PC1 explained 60.7100% of the variance in the PCDD/F isomer distributions, depending upon the homologue, in each facility, suggesting the possibility of similar characteristic isomer distribution patterns within entire homologue groups. However, the R2 values were not as high in the combined data sets, especially for the less chlorinated DD/F, suggesting some differences in formation mechanisms across facilities. For example, the R2 value for MCDF at Norfolk was 92.3% and that of the Boiler was 83% but the common value was down to 63.2% when the data sets were combined (Table 2). This result indicates different isomer distribution patterns between facility types, especially for the less chlorinated DD/F. PCA loading plots of the isomer concentrations within homologues of low chlorine content (Figure 2a) and high chlorine content (Figure 2b) show, by the separate grouping of the facility data (circles), that the isomer patterns are facility-specific. Further, inter-facility variation, although run B4 is exceptional, in mono- and di-CDFs (Figure 2a) was much larger than that of the hexa- and hepta-CDFs (Figure 2b). The same trend was observed in PCDDs (not shown). This suggests that different isomer distribution patterns and formation mechanisms, especially in less chlorinated DD/F, may exist in a combustion facility during variation of combustion operating conditions such as feed type and rate. The sensitivity of the mechanisms that form the less chlorinated congeners to changes in operating conditions may have been the cause of the relatively low correlation of these compounds with PCDD/F TEQ values (Table 1). For the second PCA analysis, three predicted PCDD/F isomer distribution models based on different formation mechanisms were examined with the Norfolk and Boiler data to identify potential dominant formation mechanisms in this study. First, Iino and co-workers’ (15) dechlorination-based PCDD/F isomer distribution prediction model was used; second, Ryu et al.’s (10-11) chlorination-based PCDD/F prediction model was used; third, the phenol-condensationbased PCDF isomer distribution prediction model of Ryu et al. (12) was used based on a typical distribution of chlorinated phenols, coupled with PCDD isomer distributions based on phenol condensation reactions of a mixture of the 10 chlorinated phenols found in greatest abundance in incinerator flue gas [as per Ballschmiter et al. (19)]. To examine isomer distributions and these mechanistic models, the isomers within each homologue were taken as the PCA objects. The PCA variables were the relative isomer concentrations, defined again as the isomer concentration divided by its total homologue concentration (as in Figure 2), from the Norfolk and Boiler samples and as predicted by the chlorination, dechlorination, and phenol condensation models. The facility-averaged isomer patterns from the 11 Norfolk and the 8 Boiler data sets were used, although the same result was obtained when every single run’s isomer pattern was separately tested. The loading plot (Figure 3) shows how the variables (three different models and the two incinerators’ data sets) are related to each othersvariables which have similar isomer distribution patterns are closely located. The loading plots for each PCDF homologue are presented in Figure 3. The Boiler data are located near the phenol condensation model for mono- and di-CDF but the Norfolk data are located near the chlorination model. This indicates that the mono- and di-CDF distribution patterns from the Boiler data are similar to those produced by phenol condensation, whereas the Norfolk data patterns are similar to those produced by a chlorination mechanism (Figure 3a and b). Formation mechanisms observed in the mono- to di-CDFs seem to be facility-specific (Figure 3a and b), although there is the possibility that there is not one dominant mechanism (i.e., VOL. 38, NO. 17, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4697
FIGURE 2. PCA results for isomer similarity with facility types: (a) mono- and Di-CDFs, (b) hexa- and hepta-CDF (circles were drawn to show the data groups by facility type).
FIGURE 3. Validation of PCDF formation mechanism in Boiler and Norfolk data with PCA: (a) mono-CDF, (b) di-CDF, (c) tri-CDF, (d) tetra-CDF, (e) penta-CDF, (f) hexa-CDF, (g) hepta-CDF. Based on phenol condensation model (12), chlorination model (10, 11), dechlorination model (15). more than one mechanism contributes significantly to an isomer set). Dechlorination best accounts for the remaining chlorinated DFs (tri- to hepta-chlorinated) regardless of facility type (Figure 3c-h). For tri- and tetra-CDFs, the Boiler data are located near the dechlorination model prediction, while the Norfolk data are located between the dechlorination and chlorination predictions (Figure 3c and d). For pentato hepta-CDFs, both data sets are closely located with dechlorination model (Figure 3e-g). These results suggest that dechlorination is the dominant PCDF formation mechanism regardless of facility type and operating condition, as proposed by other researchers (8, 16). As an illustrative example, Figure 4 shows the di-CDF isomer distribution pattern, presented as the fraction of the isomer’s concentration to its total homologue concentration. The distinctions in the patterns confirm that different dominant formation mechanisms exist between the Norfolk and Boiler data for this illustrated homologue. The predominant 2,4-DiCDF in 4698
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 17, 2004
the Boiler data, which was mainly formed by phenol condensation (12), provides the main distinction. For less chlorinated DFs (mono- and di-CDF), other PCDF formation mechanisms might be contributing according to the facility type and operating condition. This would result in different isomer distribution patterns and relatively low correlations between less chlorinated DF isomers and PCDD/F TEQ values. The fact that the sample variation between the Boiler and Norfolk data decreased with increasing dibenzofuran chlorine number (from Figure 3a-g) indicates the sensitivity of the formation mechanism in low-chlorinated furans to facility type. For the PCDDs, multiple formation mechanisms were inferred in the less chlorinated DDs but, mainly, the phenol condensation mechanism was dominant, confirming previous studies (15, 16). The PCDD isomer distribution patterns and, hence, the formation mechanism of PCDDs are more consistent than those of PCDFs.
FIGURE 4. Comparison of di-CDF isomer distribution pattern between incinerators and models.
TABLE 3. R2 Value in Three Categorized HxCDF Isomer Groups sample type
chlorination
dechlorination
mix (chlorination/dechlorination)
all isomers
NF boiler avg (NF/Boiler)
0.08 0.05 0.09
0.51 0.47 0.49
0.49 0.42 0.51
w/o toxic isomers
NF boiler avg (NF/Boiler)
0.23 0.16 0.29
0.97 0.91 0.95
0.89 0.80 0.91
only toxic isomers
NF boiler avg (NF/Boiler)
0.35 0.37 0.33
0.13 0.12 0.14
0.70 0.69 0.69
target isomers
Relationship between Chlorination and Toxic Isomers. Although dechlorination appears to be the dominant formation mechanism for highly chlorinated furans (g4 Cl, see Figure 3), dechlorination and chlorination are likely to occur simultaneously on particle surfaces. The 2,3,7,8-Cl substituted isomers might be expected to respond differently to chlorination or dechlorination mechanisms than the non-2,3,7,8Cl substituted isomers, due to position-specific effects (e.g., electronic and steric hindrance effects) caused by the location of the chlorines. Indeed, previous studies already showed that the chlorination of PCDD/F is favored at the 2,3,7,8positions (10). To evaluate the possibility of simultaneous occurrence of both chlorination and dechlorination mechanisms and their effect on isomer distribution patterns, the average of the isomer distribution patterns predicted by both the chlorination and dechlorination models were determined. In this manner, half of the isomers were produced by DD/F chlorination and half by dechlorination from OCDD/F. This averaged chlorination/dechlorination distribution was separately compared with the observed patterns of toxic isomers (2,3,7,8-Cl-substituted), nontoxic isomers, and the combined isomer sets for the Boiler, Norfolk, and averaged Boiler + Norfolk data sets. The HxCDF homologue was selected for this evaluation because it contains the largest number of 2,3,7,8-Cl-substituted isomers, hence the largest number of isomers for comparison with the predictions. Correlation factors (R2) on the average isomer patterns were obtained (Table 3) to see the relationship between three different models (chlorination, dechlorination, mixed) and the three HxCDF isomer groups. (Note that the effect of using the average isomer pattern was minimal; the relative standard deviation of the R2 values for the fit of the optimal models
to the individual isomer patterns, reflecting run to run variation, did not exceed 13%). The first group, containing all of the HxCDF isomers in the Norfolk, Boiler, and averaged Norfolk and Boiler samples, showed a relatively high R2 value (∼0.51) with the dechlorination model compared to chlorination model (R2 ∼ 0.08), which is the same result as in Figure 3f. The second group, which excluded the toxic isomers, similarly was aligned with the dechlorination model but to an even greater extent (R2 ≈ 0.95). Unlike the first and second HxCDF isomer groups, the R2 value of the toxic isomers in each data set showed a closer relationship with the chlorination model (R2 ≈ 0.33) than the dechlorination model (R2 ≈ 0.14) while showing an even stronger relationship with the mixed mechanism model (R2 ≈ 0.69). This mixed model also related very well with the first and second HxCDF isomer groups (Table 3) due to the weighting of the dechlorination model. Thus, while the nontoxic isomers are clearly more allied with a dechlorination mechanism, the toxic isomers show a greater influence of a chlorination mechanism, as evidenced by the relatively higher R2 value of the chlorination model. Nonetheless, improvements in the Pearson coefficients for the mixed chlorination/dechlorination model show that the patterns of the toxic isomers are affected by both mechanisms. Other facts in this current paper verify the close relationship between chlorination and TEQ. One potential strong indicator isomer in the tri-CDD homologue is 2,3,7-triCDD which is the predominant isomer in the DD chlorination mechanism model (10). These results offer an explanation for why PCDFs have more strong TEQ indicators than PCDDs and why TeCDD and PeCDD do not have any strong TEQ indicators. However, unlike the other PCDD isomers, most of the HxCDD and HpCDD isomers have the potential to be used as TEQ VOL. 38, NO. 17, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
4699
indicators because most of the isomers in these homologues are toxicity-weighted and have a more consistent distribution pattern than that of low-chlorinated dioxins. Besides the toxicity-weighted isomers, many other PCDD/F isomers appear useful as common TEQ indicators for the two different facilities despite their operation under significantly different conditions. However, application of these potential indicator isomers is premature, pending additional data and study. Compared to PCDDs, PCDFs have more potential robust TEQ indicator isomers because of the close relationship between chlorination/dechlorination formation mechanisms for PCDFs and the TEQ-related compounds. For on-line monitoring, a few tri-CDD/F isomers showed potential to be used as robust TEQ indicators for these two facilities, unlike the mono- and di-CDD/F isomers.
Acknowledgments J.-E.O.’s work was supported in part by the Postdoctoral Fellowship Program of Korea Science & Engineering Foundation (KOSEF) and the Postdoctoral Research Program at the EPA’s National Risk Management Research Laboratory, administered by the Oak Ridge Institute for Science and Education (ORISE) through interagency agreement DW89938167 between the U.S. Department of Energy and the U.S. EPA.
Literature Cited (1) Iino, F.; Takasuga, T.; Touati, A..; Gullett, B. K. Waste Manage. 2003, 23, 729-736. (2) Tuppurainen, K. A.; Ruokoja¨rvi, P. H.; Asikainen, A. H.; Aatamila, M.; Ruuskanen, J. Environ. Sci. Technol. 2000, 34, 49584962. (3) Blumenstock, M.; Zimmermann, R.; Schramm, K. W.; Kettrup, A. Chemosphere 2000, 40, 987-993.
4700
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 17, 2004
(4) Kaune, A.; Lenoir, D.; Schramm, K. W.; Zimmermann, R.; Kettrup, A.; Jaeger, K.; Ru ¨ ckel, H. G.; Frank, F. Environ. Eng. Sci. 1998, 15 (1), 85-95. (5) Blumenstock, M.; Zimmermann, R.; Schramm, K. W.; Kettrup, A. Chemosphere 2001, 42, 507-518. (6) Gullett, B. K.; Wikstro¨m, E. Chemosphere 2000, 40, 1015-1019. (7) Oser, H.; Thanner, R.; Grotheer, H. H.; Gullett, B. K.; Natschke, D.; Raghunathan, K. Combust. Sci. Technol. 1998, 134, 201-220. (8) Iino, F.; Tsuchiya, K.; Imagawa, T.; Gullett B. K. Environ. Sci. Technol. 2001, 35, 3175-3181. (9) Schramm, K. W.; Wehrmeier, A.; Lenoir, D.; Henkelmann, B.; Hahn, K.; Zimmermann, R.; Kettrup, A. Organohalogen Compd. 1996, 27, 196-200. (10) Ryu, J. Y.; Mulholland, J. A.; Byoung, C. Chemosphere 2003, 51, 1031-1039. (11) Ryu, J. Y.; Mulholland, J. A.; Dunn, J. E. Organohalogen Compd. 2003, 63, 49-52. (12) Ryu, J. Y.; Mulholland, J. A.; Oh, J. E.; Nakahata, D. T.; Kim, D. H. Chemosphere 2004, 55, 1447-1455. (13) Wehrmeier, A.; Lenoir, D.; Schramm, K. W.; Zimmermann, R.; Hahn, K.; Henkelmann, B.; Kettrup, A. Chemosphere 1998, 36, 2775-2801. (14) Addink, R.; Govers, H. A. J.; Olie, K. Environ. Sci. Technol. 1998, 32, 1888-1893. (15) Iino, F.; Imagawa, T.; Gullett, B. K. Environ. Sci. Technol. 2000, 34, 3143-3147. (16) Wikstro¨m, E.; Tysklind, M.; Marklund, S. Environ. Sci. Technol. 1999, 33, 4263-4269. (17) Gullett, B. K.; Touati, A; Lee, C. W. Environ. Sci. Technol. 2000, 34, 2069-2074. (18) Jackson, J. E. A User’s Guide to Principal Components; Wiley: New York, 1991. (19) Ballschmiter, K.; Braunmiller, I.; Niemczyk, R.; Swerev, M. Chemosphere 1988, 17, 995-1005.
Received for review September 11, 2003. Revised manuscript received June 9, 2004. Accepted June 9, 2004. ES034997S