Environ. Sci. Technol. 2008, 42, 9255–9261
Streamlined Combustion Gas Measurements for Improved National Dioxin Inventories ERIK SPINNEL,* JERKER FICK, PATRIK L. ANDERSSON, AND PETER HAGLUND Department of Chemistry, Umeå University, SE-901 87 Umeå, Sweden
Received May 7, 2008. Revised manuscript received October 5, 2008. Accepted October 8, 2008.
The analysis of PCDD/Fs requires advanced analytical instruments and is a complex, labor intensive process that consumes large quantities of high-purity solvents. It is therefore very expensive and thus problematic-or even impossible-for developing countries to afford to establish reliable PCDD/F source inventories in support of global and national emission reduction strategies. Low-cost reliable alternatives are needed to improve this situation. Therefore, a streamlined procedure for flue gas sampling and analysis has been developed and evaluated that utilizes a user-friendly polyurethane foam plug (PUFP) sampling technique. The collected samples are then shipped to a central laboratory for analysis where they are processed using a cost-efficient pressurized liquid extraction procedure with in-cell carbon cleanup (PLE-C) prior to analysis bygaschromatography-high-resolutionmassspectrometry(GCHRMS). The PLE-C technique has previously been validated for a range of matrices, and, in the present study, has been further improved by introducing an extraction cell coupling cartridge. The entire procedure was evaluated using three sets of samples: two from a laboratory-scale incinerator and one from a full-scale incinerator. The samples were expected to differ in PCDD/F levels and homologue patterns. Each sample was split and analyzed in parallel by both PLE-C and a reference procedure (Soxhlet extraction followed by a traditional multistep cleanup procedure, Sox-T). The results of analysis by PLE-C compared well with those from analysis by Sox-T. The difference between toxic equivalent (TEQ) values obtained using the PLE-C and reference technique for 11 PUFP samples ranged from -10% to +44% and the two techniques also yielded very similar PCDD/F homologue profiles. A principal component analysis (PCA) of the data showed that both methods were able to discriminate among the three sets of samples, thereby demonstrating that the between method variability was less than the between-sample variability. In summary, the results indicate that PUFP sampling followed by PLE-C extraction and cleanup provides a fast, relatively cheap, and reliable method for analysis of PCDD/Fs in flue gas.
Introduction Polychlorinated dibenzo-p-dioxins (PCDDs) and polychlorinated dibenzofurans (PCDFs) are two chemically similar * Corresponding author e-mail:
[email protected]; phone: +46-(0)90-786 9339. 10.1021/es801187n CCC: $40.75
Published on Web 11/16/2008
2008 American Chemical Society
groups of compounds, collectively referred to as “dioxins”. The PCDD/Fs include some of the most toxic compounds ever produced. These are chlorinated at each of the 2,3,7 and 8-positions and are known to cause adverse health effects such as cancer, liver damage, nausea, endocrine disruptions, and chloracne (1). In recent decades several significant sources of PCDD/Fs in the environment have been identified. These include pesticide impurities; chlorine production and chlorine bleaching at pulp and paper industries; hazardous waste and municipal solid waste (MSW) incineration; vehicle exhaust; metallurgical industries; and several other hightemperature processes (2-5). PCDD/Fs are regulated through the Stockholm Convention on Persistent Organic Pollutants (POPs) (6), which came into force in 2004 and was initially ratified by 128 parties. The Stockholm Convention is a global treaty designed to protect human health and the environment from POPs. The Convention promotes concerted actions to monitor and minimize the release of industrial byproduct POPs such as dioxins. Many developed countries have produced extensive inventories of PCDD/F sources. Much less is known about the most significant emission sources in emerging countries. However, there are indications that major PCDD/F sources in developing countries are different from those discussed above-for example back-yard burning, controlled burning to clear land or recycle nutrients, forest fires, and fires at dump sites (7). The lack of dioxin inventories in many countries and the high costs associated with establishing such inventories have been recognized within the United Nations Environment Program (UNEP). To assist countries in identifying sources and estimating releases of PCDD/Fs UNEP Chemicals has developed a “Standardized Toolkit for Identification and Quantification of Dioxin and Furan Releases” (http://www.chem.unep.ch/pops/pdf/toolkit/ toolkit.pdf). In addition, a series of workshops has been conducted to assist countries in taking measures to reduce and/or eliminate releases of PCDD/Fs into the environment. The toolkit is flexible and applicable to all countries. Even countries without any PCDD/F data may use the toolkit to screen industrial and other processes to make initial estimates of the scale of potential PCDD/F sources. However, it should be noted that such estimates have considerable uncertainties associated with them. To improve the estimates and the database that support the toolkit, there is a need for complementary measurements in the countries concerned. Thus user-friendly sampling techniques (especially for various combustion gases) and cost-efficient analytical procedures are urgently needed. Sampling of PCDD/Fs in such gases can be performed by various methods, many of which are costly and require trained personnel and complicated sampling equipment. However, attempts have been made to simplify the sampling while maintaining acceptable performance. One of the more interesting alternatives, “the PUFP sampler” utilizes a cartridge (in which are mounted a PUFP and a filter) connected to a pump and a simple flow measurement device (8). The PCDD/Fs are trapped by the PUFP and the filter during sampling, then the PUFP and filter are usually shipped to an environmental testing laboratory for extraction, cleanup, and analysis. The dioxins on the PUFP and filter are normally extracted and cleaned up using time-consuming and labor-intensive procedures prior to analysis using gas chromatography coupled to high-resolution mass spectrometry (GC-HRMS). This approach is expensive and hampered by low throughput. Alternative strategies have been recently developed including VOL. 42, NO. 24, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
9255
pressurized liquid extraction (PLE). This was first described by Richter et al. in 1996 (9) and has, for instance, been combined with an in-cell carbon trap (PLE-C) for combined extraction and cleanup. This method has been applied to diverse matrices including soil, sediment, fly ash, and fish oil and fish meal (10-12). It has also been shown that lowcost high-throughput techniques such as ELISA immunoassays could be used as a substitute for GC-HRMS or to complement it (11). There is great potential to further develop the PLE-C method and thus increase its utility and efficiency. In PLE-C, the PCDD/Fs are first extracted then trapped on active carbon (charcoal) and back-flushed from the carbon using toluene. In the current version of PLE-C the back-flush solvent passes through the sample matrix, which may result in the unwanted extraction of matrix constituents. This is clearly a limitation of the current method which could be avoided by suitable modification, such as removal of the sample prior to backflushing. Another drawback is that the volume of toluene required for efficient back-flushing exceeds the capacity of the receiving vial, which limits the number of samples that can be run in a single batch (11, 12). The aim of this study was to develop and validate a userfriendly PUFP sampling technique for combustion gases and a cost-efficient sample preparation procedure, utilizing novel extraction cell coupling assemblies for PLE-C (enable removal of the matrix before back-flushing the carbon trap). Once validated, this analytical strategy may be used to improve national and international dioxin inventories as well as the UNEP toolkit.
Experimental Section Chemicals. All solvents used for extraction and clean up (cyclohexane, acetone, toluene, n-hexane, n-heptane, dichloromethane, and methanol) were glass distilled and originated from Burdick & Jackson (Muskegon, MI). Tetradecane (olefinfree, pro analysis) and Celite were obtained from Fluka. Sulphuric acid, potassium hydroxide, silica gel (Kieselgel 60, 0.063-0.200 mm), and anhydrous sodium sulfate were purchased from Merck (Damstadt, Germany). SA 4PAH HF activated carbon used for PLE-C was kindly donated by Norit (Amersfoort, The Netherlands). AX21 used for traditional clean up is no longer commercially available but originated from the Anderson Development Company (Adrian, MI). Native and labeled 2,3,7,8-PCDD/Fs were obtained from Wellington Laboratories (Guelph, Ontario, Canada) and from Cambridge Isotope Laboratories (Andover, MA). An internal standard (IS) mixture containing all 17 13C12labeled 2,3,7,8-substituted PCDD/Fs (each at a nominal concentration 14 pg/µL) was prepared in toluene from individual solutions. The recovery standard (RS) mixture (with a nominal concentration of 14.5 pg/µL) was prepared from individual solutions of 13C12 -labeled 1,2,3,4-TCDD, 1,2,3,4,6PeCDF, 1,2,3,4,6,9-HxCDF, and 1,2,3,4,6,8,9-HpCDF. Quantification standards (QSs) for all native 2,3,7,8-substituted PCDD/Fs were prepared by adding IS and RS to 1 mL aliquots of a diluted native standard (Wellington Laboratories). Equipment. The PUFP sampling equipment has been described elsewhere (13). Detailed drawings are provided as Supporting Information to facilitate manufacture. Each sampler was fitted with a 50 mm × 60 mm diameter PUFP (Søm & Skumplast fabrikk AS, Sunde, Norway) and a 142 mm glass fiber filter (Type A/E; Gelman Laboratories, Ann Arbor; MI). For PUFP extraction, two different approaches were used: Soxhlet extraction and PLE. Soxhlet extraction was conducted using a 200 mL Soxhlet apparatus equipped with a Dean-Stark water trap. PLE was performed using an ASE200 system (Dionex). For the PLE experiments, custommade stainless steel guides (Unimeg, Umeå University, Sweden) with threads identical to the original ASE200 end 9256
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 24, 2008
caps were used to couple two 11 mL extraction cells together. Polyetheretherketone (PEEK) discs (3.0 mm thickness; Åkesson & Blomqvist AB, Sundsvall, Sweden) were used to create a seal between the two cells. The original PEEK discs provided by Dionex were replaced with thicker discs, because occasional leakage problems were experienced. The guide and cell assembly should not be tightened hard. It is when the cells are compressed in the ASE oven that the two cell segments are pressed into the PEEK disk and create a leak-tight seal. A complete drawing of the guides is provided as Supporting Information along with a photograph of the entire assembly. To separate the upper matrix compartment (which in this case contained the PUF plug and filter) from the lower adsorbent (activated carbon) compartment, a slightly modified stainless steel frit (approximately 1 mm smaller diameter than standard) was placed in the bottom of the upper cell, followed by a cellulose filter disk (Figure 1). Sampling. Three sets of flue gas samples were acquired: one from a full-scale plant and two from a laboratory-scale incineration reactor (collected at two different flue gas temperatures). These sets of samples were expected to differ in PCDD/F levels and homologue patterns and were used to test if both the reference method and the modified PLE-C analytical method were able to discriminate among different groups of samples, i.e., to check if the between-method variability was less than the between-sample variability. The sampling method has been described in detail by Marklund et.al. Sampling settings are given as Supporting Information. The full-scale plant (Umeå, Sweden) is a combined heat and power plant (built in 2000 by Von Roll Environmental Technology Ltd., Switzerland) with a thermal capacity of 66 MW (15 MW electric power and 57 MW hot water) (http:// www.aee-vonroll.com/downloads/umea.pdf). The plant has a four-pass steam generator, Von Roll ram feeder, pit cranes, ram expeller for slag, boiler ash removal system, and a selective noncatalytic NOx reduction system. Sorted municipal solid waste is the main fuel, which is incinerated on an aquaroll reciprocating grate (10 × 7.2 m) at 850-1000 °C, and the combustion capacity is 20 Mg/h (24 Mg/h at 10 MJ/ kg). The plant was fed waste fuel and no major deviation was reported in the incineration efficiency or fuel composition during the sampling event. Six flue gas samples (PUF 1-6) were sequentially collected, using the simplified cooled probe PUFP sampling technique for 15 min each at 200 °C. The remaining samples were collected from a 5 kW laboratory-scale fluidized-bed incinerator at Umeå University, recently described in detail by Wikstro¨m et al. (14). The reactor consists of a fluidized sand bed (primary-combustion zone), freeboard (secondary-combustion zone), convector (postcombustion zone), and air pollution control system. A standardized artificial municipal solid waste was used as fuel (Supporting Information, Table S1). Minor transient combustion was simulated, which was monitored via the CO level. Five flue gas samples, 45 min each, were collected from the laboratory-scale incinerator at two temperatures, approximately 300 °C (PUF 7-9) and 200 °C (PUF 10-11). No major temperature deviations or elevated CO events were observed in combustion during the sampling. Extraction and Cleanup. The weight of each PUFP and filter was measured and then split into four equal parts for which the weights also were measured. The total sampled volume was then divided into the four parts in accordance with the percentage of the total weight. The diagonally opposite quarters were combined to obtain representative duplicate samples for the PLE-C validation. A virgin PUFP and filter were analyzed with each set of samples as blank control. All PUFPs were spiked with IS and left to equilibrate for around 30 min prior to extraction.
FIGURE 1. Analysis of PCDD/Fs by PLE-C. Two 11 mL cells are coupled together with the coupling cartridge. The PUFP and filter paper are placed in the upper cell (r). A slightly modified stainless steel frit is placed in the bottom to separate the sample from the carbon trap (γ). The coupling cartridge (β) consists of two rings that join the two cells together with a PEEK seal. In the first step the cell assembly is extracted with n-heptane/acetone. The dioxins are retained in the carbon trap while interfering compounds are flushed through. Next, the coupling cartridge and the upper cell are removed (step 2). In the third and final step, a new end cap is attached then the cell is inverted and extracted with toluene. Reference data were obtained using Soxhlet extraction followed by a multistep cleanup (Sox-T). The extractions were performed with 300 mL of toluene for 24 h. Each resulting extract was evaporated to ca. 1 mL and cleaned up according to Liljelind et al. (15). Briefly, this involved open column chromatography on multilayer silica, highly activated alumina (Super Alox), and active carbon columns, followed by a miniaturized multilayer silica column. For PLE-C, two 11 mL cells were coupled together (the assembly is recognized as a 22 mL cell by the ASE200). The lower cell was packed with 2 g of 15% Norit carbon on Celite and was then washed and activated by passing toluene (10 mL), dichloromethane/methanol/toluene (10 mL; 15:4:1 v/v), dichloromethane/n-hexane (4 mL; 1:1 v/v), and, finally, n-hexane (10 mL) through the cell contents. The upper cell was attached and loaded with a stainless steel frit, the filter, and the PUFP (Figure 1). A cellulose filter disk was placed over the PUFP to prevent adsorbent material from being trapped by the end cap threading, which could cause leakage. The dioxins were then extracted from the PUFP and the filter using three cycles of n-heptane/acetone (static time 5 min; temperature 150 °C; flush volume 60%; pressure 2000 psi). The upper cell and the coupling assembly were removed and the cell with the carbon trap was sealed with a clean end cap. The carbon-containing cell was inverted and extracted with three cycles of toluene (static time 7 min; temperature 175 °C) to back-elute the PCDD/Fs. Tetradecane (50 µL) was added as a keeper and the toluene was evaporated. The residue was quantitatively transferred to a small multilayer silica column for purification. This was packed with 35% potassium hydroxide silica (0.2 g), activated silica (0.2 g), 40% sulfuric acid on silica (0.4 g), and sodium sulfate (∼0.1 g) and was eluted with 10 mL of n-hexane.
To assess the extraction efficiency of PLE-C, each sample was respiked with IS and processed by the Sox-T method. Finally, RS was added to all cleaned up samples extracts and the solvent was changed to tetradecane prior to analysis by GC-HRMS. Detection and Evaluation. The GC-HRMS analyses were performed using a HP6890 gas chromatograph equipped with a 60-m nonpolar column (DB5) and directly interfaced to a Micromass Ultima HRMS operating in selected ion recording mode with electron ionization at 35 eV. The system was tuned to a resolution of 10,000 or greater and the PCDD/F congeners were quantified using the isotope-dilution technique. The limit of detection (LOD) for each analyte was defined as the amount corresponding to a signal-to-noise ratio of three. The total sum of all isomers was calculated for each homologue to establish a homologue profile. Mono- to triCDD/Fs were not determined for the laboratory-scale incinerator samples. A toxic equivalent (TEQ) value was also calculated for each sample using individual toxic equivalency factors (TEFs) of the 2,3,7,8-substituted PCDD/F congeners (16). The TEQ calculation was performed for comparison reasons only and the derived TEQs do not necessarily reflect the potential risk associated with dioxins in the flue gas. Furthermore, since the analyses were performed using a single GC column there were some coelutions between 2,3,7,8- and non-2,3,7,8-substituted PCDD/Fs that caused a positive bias in TEQ values. However, the overall effect was the same for both analytical procedures and thus coelution did not influence the comparisons. Statistical Analysis. Principal component analysis (PCA), which can handle data matrices including correlated variables, was applied to study congener specific differences between the two methods (17). It can handle large data VOL. 42, NO. 24, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
9257
FIGURE 2. Compilation of TEQ values obtained from all the samples. PUF 1-6 were samples from the full-scale incinerator with 1 min intervals between the samplings. PUF 7-11 were sampled from the laboratory-scale fluidized bed reactor at a flue gas temperature of 300 °C (PUF 7-9) or 200 °C (PUF 10-11). The concentrations of analytes were normalized to 1 atm, 0 °C, 11% CO2. matrices including correlated variables. PCA extracts the major variation in the data and projects it into a few orthogonal variables called the principal components (PCs). These can be plotted as score plots which illustrate the major differences among the objects (here the 22 samples) and corresponding loading plots in which the parameters of greatest influence (in this case the measured PCDD/Fs) can be studied. PCA was conducted using Simca P+ 11.0 (Umetrics AB, Umeå, Sweden) and the significant principal components were selected to be those with eigenvalues >1.5.
Results The newly developed in-line extraction and cleanup methodology using PLE-C yieled very similar results (expressed on a TEQ basis) to the standard extraction and cleanup technique (Sox-T) for all flue gas samples, as clearly shown in Figure 2. The results from Sox-T analysis of samples from the full scale reactor ranged from 0.35 to 0.95 ng-TEQ/m3 while the results for replicate samples using PLE-C ranged 0.50 to 1.13 ng-TEQ/m3. There was also a clear covariation in the data produced over the sampling period (the six samples were collected sequentially), possibly reflecting variations in the PCDD/F concentrations in the flue gas due to fluctuations in the fuel composition (household waste) or incineration efficiency. For the laboratory-scale reactor the measured flue gas levels ranged from 0.21 to 0.38 ng-TEQ/ m3 using Sox-T and from 0.26 to 0.47 ng-TEQ/m3 using PLEC. The individual values from PLE-C ranged from -10% to +44% of those from Sox-T analysis. The correlation in a xy-plot was good with a R2 exceeding 0.90 for the TEQ values (SI Figure S2a). The correlation for the sum of each homologue was acceptable with a R2 ranging from 0.24 to 0.94 (SI Figure S4a-b). Raw data are given as Supporting Information (SI Figure S3a-d). The homologue sums obtained by the two techniques were also very similar, as demonstrated by the concentration ratios being close to one (Figure 3). The 95% confidence interval for mono- to tetra-CDF, tri-CDD, and octa-CDD/F included ratios of 1.0 and the data for these homologues were thus statistically equivalent for the two techniques (Student t-test). For the remaining homologues, PLE-C 9258
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 24, 2008
FIGURE 3. Comparisons of PLE-C vs Soxhlet derived concentration ratios for the various PCDD/F homologues. The mono- through tri-CDD/F homologue data are based on six replicates and the remainder are based on eleven replicates. Mono-CDDs were excluded because they were detected at significant concentrations in the Soxhlet blank. yielded statistically higher concentrations than Sox-T, probably due to a more efficient extraction. Highly chlorinated congeners that are particularly difficult to extract from particles have frequently been reported to be more efficiently recovered by PLE than by Soxhlet from a wide range of different matrices (18-20). PCA was applied to study congener-specific differences in extraction between the newly developed PLE-C methodology and the Sox-T reference method. The data examined in the statistical analysis included relative levels of 14 individual PCDD/Fs normalized to the sum of each class of compounds and the sum of each homologue group normalized to the total levels of PCDDs and PCDFs, respectively. In total, data for 22 samples and 22 variables were included in the PCA analysis which yielded four significant principal components where the first and second component explained 40 and 19% of the variation in the data, respectively (statistical information is given in SI). The PCA revealed clear differences between the samples from the full-scale reactor, seen
FIGURE 4. Principal component analysis (PCA) plots of the PLE-C and Sox-T data. Panel a shows a clear separation between the different incineration samples (1-6 and 7-11). PLE 2 appears to be an outlier due to high proportions of 123678-HxCDD, Sum HpCDF, and Sum HpCDD (Figure 4b), which were traced to cross-contamination between samples. Panels c and d show the plots obtained when PLE 2 and Sox 2 plus relative levels of 123678-HxCDD, Sum HpCDF, and Sum HpCDD were excluded. The resulting score plot (Figure 4c) still shows a clear separation between full-scale incinerator and laboratory-scale incinerator samples (1-6 vs 7-11), but also discriminates between laboratory scale samples collected at 300 °C (7-9) and those collected at 200 °C (10-11). Overall, small differences were observed between Sox-T and corresponding PLE-C samples, except for PLE 8 and Sox 8. clustered to the right in the score plot versus those from the laboratory-scale reactor (Figure 4a). The plots revealed that PLE2 clearly had a different congener pattern than the other samples from the full-scale reactor, notably higher 123678HxCDD, sum hepta-CDD, and sum hepta-CDF levels (Figure 4b) which were attributed to laboratory contamination of this specific sample. In support of this hypothesis, a pentachlorophenol-contaminated soil sample containing high concentrations of these specific congeners had previously been extracted using the same PLE equipment. Further PCA was therefore conducted, in which the three compromised data points were omitted (Figure 4c and 4d). The first two components of this analysis explained 44 and 14%, respectively, of the variation in relative levels. In the resulting plots, most duplicate samples (i.e., samples processed using PLE-C and corresponding samples processed using Sox-T) were found close to each other, indicating that similar congener patterns were obtained by both methods. The only pair of duplicates that differed significantly from one another was PLE 8 and Sox 8. In addition, all laboratoryscale reactor samples taken at 200 °C are separated from those sampled at 300 °C. This separation was correlated to the ratios of 1,2,3,7,8-penta-CDD and 1,2,3,4,7,8,9-HpCDF (Figure 4d). The main differences between laboratory-scale
and full-scale samples were a higher abundance of lightly chlorinated congeners (e.g., sum of tetra-CDF, sum of tetraCDD, and 2,3,4,7,8-penta-CDF) in the full-scale incinerator samples and a greater abundance of highly chlorinated congeners (e.g., sum of hexa-CDF and sum of hexa-CDD) in the laboratory-scale incinerator samples. In summary, the PCA analyses showed great similarities in the relative congener levels determined by the two methodologies, and the same grouping of incineration samples (Figure 4c) irrespective of the analytical method applied. The recoveries of the 13C12-labeled internal standards were also very similar (all recovery data were within acceptable limits, i.e., 47-113%). Furthermore, Soxhlet re-extraction of PUFs and filters (previously extracted by PLE-Cf) released very little additional PCDD/F and generally less than 5% of the amounts found in the PLE-C extracts. Overall these findings indicate that the PCDD/Fs were exhaustively extracted from the PUFP and filter, efficiently trapped in the carbon, and efficiently released from the carbon. Cost Saving Estimation. Sample preparation of PCDD/ Fs is often the bottleneck and hence the most expensive step in the total analysis process. A reduction in the cost of sample preparation also gives a reduction in the total analysis cost. VOL. 42, NO. 24, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
9259
TABLE 1. Estimated Amounts of Solvent and Labor and Associated Costs (In Euros) Required for Sample Preparation Using Sox-T and for PLE-C (Total Time to Process the Batch of Samples Is Also Indicated) technique
cost type
Sox-T
solvent volume
manual labor/total time and costs 1 sample
6 samples
24 samples
96 samples
8 h/25 h 600 35 40 675 2 h/3 h 150 5 10 165 4.1
16 h/33 h 200 35 40 275 9 h/15 h 113 5 10 128 2.2
64 h/100 h 200 35 40 275 12 h/36 h 38 5 10 53 5.2
256 h/400 h 200 35 40 275 40 h/136 h 31 5 10 46 5.9
700 mL labor solvent consumablesa total 100 mL
PLE-C labor solvent instrumentb total Sox-T/PLE-C a
Costs of carbon and silica-based adsorbents.
b
Depreciation costs based on 100 sample per month.
Table 1 summarizes the amounts of solvent required by the two methods, their manual labor requirements, and estimated total costs (assuming average costs of glass distilled solvents, technicians’ labor, and Sox-T consumables of 50 euros/liter, 75 euros/h, and 40 euros, respectively). Although the labor and solvent expenses vary among countries, the data indicate that significant potential savings can be made by changing from Sox-T to PLE-C. In cases where there are major differences in solvent and/or labor expenses, the overall costs can be readily adjusted using the data in Table 1. The calculations indicate that extraction and clean up of one sample using Sox-T would cost about 680 euros; 600 euros for labor, 40 euros for solvents, and 40 euros for other consumables (primarily prewashed carbon and silica-based adsorbents). The cost of PLE-C would be much lower at 165 euros: 150 euros for labor, 10 euros for instrument depreciation, and 5 euros for solvents. If the number of samples is increased, the cost per sample will be reduced. Soxhlet equipment generally consists of six units that operate in parallel. The total cost per sample for processing six samples by Sox-T and PLE-C would be 275 and 128 euros, respectively. The savings are particularly significant for the Sox-T method. However, since it is only possible to run six samples per batch a further increase in the number of samples would not further reduce the cost per sample. In contrast, for PLE-C, further increases in the number of samples to 24 or 96 would result in cost of 53 and 46 euros, respectively. This is much lower than the estimated cost of 275 euros for Sox-T, which incidentally is close to the average extraction and cleanup cost (225 euros) of four commercial laboratories that participated in the EU project DIFFERENCE (www.dioxins. nl). In summary, for large series of samples the use of PLE-C can result in a 5-fold reduction in both man-hours and total costs required for extraction and cleanup, compared to Soxhlet and multistage column cleanup procedures. Other expenses such as sampling, GC/HRMS analysis, data evaluation, and QA/QC are not included, but since sample preparation is often the rate-limiting step in PCDD/F analysis, any major reduction in sample preparation costs will lead to cheaper PCDD/F analysis.
Discussion This study has demonstrated that the PLE-C method for combined extraction and cleanup is reliable and produces PCDD/F results that are very similar to those from established Soxhlet/multi-column chromatography techniques. Combined with the PUFP-sampling technique it provides a convenient and cost-effective way to sample and analyze PCDD/Fs in flue gas streams. It enables sufficiently short sampling times (15 min) to allow meaningful studies of 9260
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 24, 2008
temporal variations studies (Figure 2, PUFs 1-6) and captures minor variations in congener and homologue patterns (Figure 4), which is very valuable in source characterization and source apportionment studies. The PUFP sampling equipment is simple in its construction and PUFP samplers can be made using limited resources, which is an important consideration for any sampling strategy intended for use in developing countries. The PUFP sampling equipment is also compact and easy to ship, so a laboratory can readily provide the person responsible for the sampling with precleaned sampling equipment and a calibrated flow measurement device. If not available locally, a suitable pump could be shipped together with the PUFP samplers. Sampling using the PUFP is simple and does not require any special training. However, it is of importance that the sampling personnel have prior experience in environmental sampling. When sampling has been completed, the equipment can be shipped back to the analytical laboratory. Laboratories that are suitably equipped and capable of measuring dioxins are available throughout the world and samples can readily be transferred to a competent laboratory in the region. Many such central laboratory facilities are already equipped with PLE equipment. If not, it would be better to invest in at least two systems because a back-up system is recommended and the equipment costs are acceptable in comparison to other instruments, such as a GC-HRMS. Due to its high capacity, low solvent consumption, and high degree of automation, the cost of such a purchase should be quickly recovered. The ASE200 system has the capacity to process 24 samples per batch. Since the extraction and cleanup are performed in one step (except for the small multilayer silica chromatography column that follows the PLE-C) it is possible to prepare a sample for GC-HRMS in approximately 2 h. Thus, the technique offers both short response times for priority samples and high throughput. The TEQ levels covered by the PUFP sampling coupled to PLE-C are in the range of 0.1-1 ng/m3. However, it is likely that a wider range of levels could be covered. Although, if extremely high levels are to be measured, confirmation with conventional Sox-T is recommended. To reduce the cost further, the PUFP sampling and PLE-C cleanup could be combined with a bioassay or an immunoassay system for PCDD/F detection and quantification. Since these techniques have been successfully applied to samples of complex matrices such as soil and sediment (11, 12), they are also likely to be applicable in analyses of combustion gas samples. In the cited studies, the PLE-C technique was used without the adaptor and the samples were loaded together with activated carbon in the same cells. The bioanalytical techniques have adequate selectivity to
discriminate against the coextracted material that may be present in the extracts. Laboratories that wish to perform PCDD/F analysis and do not have the cell coupling guides used in the present study may find testing the protocols discussed above helpful. It may be possible to streamline the present PLE-C method further by including the final polish step (using the miniature multilayer silica column) in the combined extraction and cleanup process. The adaptors could be used to couple three 11 mL cells (which the ASE200 system would recognize as a 33 mL cell) that could be loaded with (from bottom up) carbon, polar adsorbent, and sample. Semipolar and polar coextracted materials should be retained by the adsorbent while the PCDD/Fs should pass through and be adsorbed by the carbon. Prior to back-elution of the carbon, the sample and adsorbent modules would be removed. This approach is currently being investigated. In summary, the PLE-C method for sampling and analyzing PCDD/Fs in flue gas is simpler and more cost-effective than established methods. It is therefore attractive for largescale campaigns to improve national PCDD/F inventories, especially in developing countries, and to improve the emission factors used in the UNEP toolkit. In the coming months, the procedure will be field tested within the framework of a UNEP project on dioxin emissions from uncontrolled burning. Waste from Mexico and China will be burned under conditions simulating back-yard burning and city dump fires. The smoke will be collected using the PUFP technique and selected samples will be processed in parallel using the PLE-C procedure and the Soxhlet procedure to evaluate the performance of the novel procedure in measuring PCDD/Fs in combustion gases from uncontrolled fires.
Acknowledgments Lisa Lundin is acknowledged for excellent help during the PUFP sampling at the Dåva plant and Maria Calla is acknowledged for excellent help during PUFP sampling from the laboratory-scale reactor. Sigrid De Geyter is acknowledged for assistance and help during full-scale plant sampling.
Supporting Information Available This information is available free of charge via the Internet at http://pubs.acs.org.
Literature Cited (1) Schecter, A.; Birnbaum, L.; Ryan, J. J.; Constable, J. D. Dioxins: An overview. Environ. Res. 2006, 101, 419–428. (2) Rappe, C.; Glas, B.; Kjeller, L. O.; Kulp, S. E.; de Wit, C.; Melin, A. Levels of PCDDs and PCDFs in products and effluent from the Swedish pulp and paper industry and chloralkali process. Chemosphere 1990, 20, 1701–1706. (3) Kulkarni, P. S.; Crespo, J. G.; Afonso, C. A. M. Dioxins sources and current remediation technologies - A review. Environ. Int. In Press. (4) Olie, K.; Vermeulen, P.; Hutzinger, O. Chlorodibenzo-p-dioxins and chlorodibenzofurans are trace components of fly ash and flue gas of some municipal incinerators in the Netherlands. Chemosphere 1977, 8, 455–459.
(5) Rappe, C.; Kjeller, L. O.; Kulp, S. E.; de Wit, C.; Hasselsten, I.; Palm, O. Levels, profile and pattern of PCDDs and PCDFs in samples related to the production and use of chlorine. Chemosphere 1991, 23, 1629–1636. (6) Bouwman, H. South Africa and the Stockholm Convention on Persistent Organic Pollutants. S. Afr. J. Sci. 2004, 100, 323–328. (7) Gullett, B. K.; Lemieux, P. M.; Lutes, C. C.; Winterrowd, C. K.; Winters, D. L. Emissions of PCDD/Fs from uncontrolled, domestic waste burning. Chemosphere 2001, 43, 721–725. (8) Bruckmann, P.; Hackhe, K.; Konig, J.; Theisen, J.; Ball, M.; Pa¨pke, O.; Kirschmer, P.; Mulder, W.; Rappe, C.; Kjeller, L.-O. A comparative study for the determination of polychlorinated dibenzofurans and dibenzo(p)dioxins in ambient air. Chemosphere 1993, 27, 707–720. (9) Richter, B. E.; Jones, B. A.; Ezzell, J. L.; Porter, N. L.; Avdalovic, N.; Pohl, C. Accelerated Solvent Extraction: A Technique for Sample Preparation. Anal. Chem. 1996, 68, 1033–1039. (10) Haglund, P.; Sporring, S.; Wiberg, K.; Bjo¨rklund, E. Shapeselective extraction of PCBs and dioxins from fish and fish oil using in-cell carbon fraction pressurized liquid extraction. Anal. Chem. 2007, 79, 2945–2951. (11) Nording, M.; Nichkova, M.; Spinnel, E.; Persson, Y.; Gee, S. J.; Hammock, B. D.; Haglund, P. Rapid screening of dioxinscontaminated soil by accelerated solvent extraction/purification followed by immunochemical detection. Anal. Bioanal. Chem. 2006, 385, 357–366. (12) Spinnel, E.; Danielsson, C.; Haglund, P. Rapid and cost-effective analysis of polychlorinated dibenzo-p-dioxins and polychlorinated dibenzofurans in soil, fly ash and sediment certified reference materials using pressurized liquid extraction with an integrated carbon trap. Anal. Bioanal. Chem. 2008, 390, 411– 417. (13) Marklund, S.; So¨derstro¨m, G.; Ljung, K.; Rappe, C.; Kraft, M.; Hagenmaiert, H. Parallel sampling for dioxins using various sampling techniques at a Swedish municipal solid waste incinerator. Waste Manage. Res. 1992, 10, 21–36. (14) Wikstro¨m, E.; Andersson, P.; Marklund, S. Design of a laboratory scale fluidized bed reactor. Rev. Sci. Instrum. 1998, 69, 1850– 1859. (15) Liljelind, P.; So¨derstro¨m, G.; Hedman, B.; Karlsson, S.; Lundin, L.; Marklund, S. Method for multiresidue determination of halogenated aromatics and PAHs in combustion-related samples. Environ. Sci. Technol. 2003, 37, 3680–3686. (16) Van den Berg, M.; Birnbaum, L. S.; Denison, M.; De Vito, M.; Farland, W.; Feeley, M.; Fiedler, H.; Hakansson, H.; Hanberg, A.; Haws, L.; Rose, M.; Safe, S.; Schrenk, D.; Tohyama, C.; Tritscher, A.; Tuomisto, J.; Tysklind, M.; Walker, N.; Peterson, R. E. The 2005 World Health Organization Reevaluation of Human and Mammalian Toxic Equivalency Factors for Dioxins and Dioxin-like Compounds. Toxicol. Sci. 2006, 93, 223–241. (17) Jackson, J. A Users Guide to Principal Components; John Wiley & Sons: New York, 1991. (18) Preud’homme, H.; Potin-Gautier, M. Optimization of accelerated solvent extraction for polyhalogenated dibenzo-p-dioxins and benzo-p-furans in mineral and environmental matrixes using experimental designs. Anal. Chem. 2003, 75, 6109–6118. (19) Popp, P.; Keil, P.; Moder, M.; Paschke, A.; Thuss, U. Application of accelerated solvent extraction followed by gas chromatography, high-performance liquid chromatography and gas chromatography-mass spectrometry for the determination of polycyclic aromatic hydrocarbons, chlorinated pesticides and polychlorinated dibenzo-p-dioxins and dibenzofurans in solid wastes. J. Chromatogr. A 1997, 774, 203–211. (20) Kiguchi, O.; Saitoh, K.; Ogawa, N. Pressurized liquid extraction of polychlorinated dibenzo-p-dioxins, polychlorinated dibenzofurans and coplanar polychlorinated biphenyls from contaminated soil. J. Chromatogr. A 2007, 1144, 262–268.
ES801187N
VOL. 42, NO. 24, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
9261
