Article pubs.acs.org/est
Near-Real-Time Combustion Monitoring for PCDD/PCDF Indicators by GC-REMPI-TOFMS Brian K. Gullett,*,† Lukas Oudejans,† Dennis Tabor,† Abderrahmane Touati,‡ and Shawn Ryan† †
U.S. Environmental Protection Agency, Office of Research and Development (E305-01) Research Triangle Park, North Carolina 27711, United States ‡ ARCADIS U.S., Inc., 4915 Prospectus Drive, Suite F, Durham, North Carolina 27713, United States ABSTRACT: The boiler exit flue gas of a municipal waste combustor was sampled to evaluate an online monitoring system for chlorobenzene congeners as indicators of polychlorinated dibenzodioxin and dibenzofuran (PCDD/PCDF) concentrations. Continuous measurements of chlorobenzene congeners using gas chromatography coupled to a resonance-enhanced multiphoton ionization - time-of-flight mass spectrometry (GCREMPI-TOFMS) system were compared over 5-min periods with conventional sampling methods for PCDD/PCDF. Three pairs of values were taken every hour over a period of three days to characterize the combustor’s response to transient operating conditions (shutdowns and startups). Isolation of specific chlorobenzene congeners from other samemass compounds was accomplished by using a GC column separator ahead of the REMPITOFMS. The 50-fold variation of PCDD/PCDF concentration was paralleled by similar changes in monitored compounds of 1,4-dichlorobenzene, 1,2,4-trichlorobenzene, 1,2,3trichlorobenzene, and 1,2,4,5-tetrachlorobenzene. A correlation of R = 0.85 and 0.89 was established between 40 pairs of simultaneous 5-min GC-REMPI-TOFMS measurements of 1,2,4-trichlorobenzene and 5 min conventional sampling and analysis for the TEQ and Total measures of PCDD/PCDF, respectively. The GC-REMPI-TOFMS system can be used to provide frequent measures of correlative PCDD/PCDF concentration thereby allowing for an understanding of measures to minimize PCDD/PCDF formation and develop operational feedback to limit emissions.
■
INTRODUCTION Analyses for polychlorinated dibenzodioxin and dibenzofuran (PCDD/PCDF) emissions from industrial plants typically require a multihour extractive sample taken on an annual or less frequent basis, followed by several weeks of analytical work to determine concentrations. From a regulatory or public interest viewpoint, this results in an infrequent and potentially minimally representative monitoring scheme. From a control perspective, this time lag prevents the plant from linking emissions to plant operation or waste characteristics. Contemporary methods for continual sampling of emissions make possible approximately 2−4 week, cumulative measurements.1,2 These methods are useful for delayed assurance of compliance and provide eventual insight into past, overall plant performance. However, both the 4 h annual sample and the cumulative sampling methods require laboratory analysis, resulting in a typical delay on the time scale of weeks between sampling and receipt of emissions results. Recent work3−5 has demonstrated 7-fold increases in PCDD/PCDF emissions during 1 h combustor transients including shutdowns and startups, consistent with others work6,7 in which PCDD/PCDF raw gas levels increased by one to 2 orders of magnitude during transient combustion conditions. The extent to which these transient emissions may affect short- and long-term stack emission values and, hence, compliance issues is undetermined. The rapid variation of PCDD/PCDF, as well as other copollutants, due to transients, © 2011 American Chemical Society
fuel changes, and operating variations suggest that fast online monitoring is necessary in order to effect changes in operating conditions that will reduce or prevent conditions favorable to PCDD/PCDF formation. PCDD/PCDF monitoring methods6−14 are more recent developments and could offer benefits of continual facility compliance characterization, feedback to the plant operators for pollutant formation minimization, assurance to the public of compliance, and minimization of overdesign of gas cleaning systems. These methods often rely on correlations with similar, coformed pollutants or PCDD/PCDF precursors themselves that are present in much higher concentrations than the target PCDD/PCDF or measurement of a subset of the 17 toxic PCDD/PCDF congeners. This correlation with indicator compounds is necessary due to lack of sensitivity (relevant PCDD/PCDF concentrations are typically on the order of parts per quadrillion, ppq) as well as an inability to measure all of the 17 toxic equivalency factor (TEF)-weighted congeners that comprise the PCDD/PCDF toxic equivalency (TEQ) measure. The indicator/TEQ relationships have typically been established by conventional, multihour sampling and analysis (gas chromatography/mass spectrometry, GC/MS). Common Received: Revised: Accepted: Published: 923
August 5, 2011 November 14, 2011 December 5, 2011 December 5, 2011 dx.doi.org/10.1021/es2027339 | Environ. Sci. Technol. 2012, 46, 923−928
Environmental Science & Technology
Article
180.15 These compounds were evident since the observed intensity ratio for m/z 180:182:184 did not follow the fixed ratios 1.00:0.97:0.32 anticipated for isotopic ratios for trichlorobenzenes. In addition, no wavelength dependence was observed in the recorded REMPI signal when the first laser wavelength was changed. Attempts to find a wavelength region where 1,2,4-trichorobenzene could be detected with no interferences or large background signals present were unsuccessful. While in this instance the m/z 180 signals were found to correlate reasonably well (R > 0.86) with PCDD/ PCDF TEQ,15 the presence of multiple compounds with an unknown relationship with halogenated organics argue for unambiguous identification of 1,2,4-TrClBz. The ability to measure facility transients for indicator concentrations with REMPI-TOFMS and to correlate their levels with PCDD/PCDF TEQ and Total (tetra- to octachlorinated congeners) values has yet to be established. This paper reports on the first REMPI-TOFMS measurements for specific PCDD/PCDF indicator compounds along with simultaneous conventional sampling for PCDD/PCDF TEQ values.
indicator compounds cited have included lowly chlorinated (less than 4 Cl’s) PCDD/PCDF congeners15 monochlorobenzenes;16 specific compounds such as 1,4-dichlorobenzene, 2,4,6-trichlorodibenzofuran, and 2,4-dichlorophenol;17 monochlorobenzene;18 and 1,2,4-trichlorobenzene.15 The latter work determined a correlation coefficient of R = 0.97 and 0.98 for 60 and 5 min duration samples, respectively, between PCDD/ PCDF TEQ and 1,2,4-TrClBz for a municipal waste combustor.15 Chlorobenzenes are not believed to be direct precursors to PCDD/PCDF but rather are formed in parallel reactions with PCDD/PCDF.19 Still others have found chlorobenzene correlations with PCDD/PCDF to be sitespecific,13 indicating instead that low volatility organohalogen compounds are better universal indicators. However, this latter conclusion was based on fitting a linear model to two data sets of nonoverlapping values, such that combination of the data sets improved each individual plant’s model. To address needs for fast, PCDD/PCDF-predictive measurements, the method of resonance-enhanced multiphoton ionization (REMPI) coupled to time-of-flight mass spectrometry (REMPI-TOFMS) has seen considerable development. REMPI-TOFMS is a laser-based, soft ionization process which combines enhanced sensitivity and selectivity with real time response. The ionization schemes include a two step process using two different photons (1 + 1′) in which the first laser wavelength can be tuned to resonantly enhance the excitation/ ionization process;20 the second fixed wavelength (213 nm) generates the molecular ion which is subsequently extracted into the TOFMS. The current mobile system (120 cm ×150 cm ×60 cm) is isomer selective and reaches low ppb-level detection limits for many common aromatic-structured combustion pollutants. Other near real time measurement methods include vacuum ultraviolet (VUV) single photon ionization (SPI) ion trap (IT) time-of-flight mass spectrometry (TOFMS) which has been successfully employed to measure trichlorobenzenes9 and pentachlorodibenzofurans (PeCDFs)12 at a waste incinerator. The former showed excellent detection limits and fast (18 s) response time but did not demonstrate correlations with PCDD/PCDF. The latter reference found strong correlations of the measured PeCDF homologue with PCDD/PCDF TEQ (two of the 28 PeCDF compounds are TEF-weighted), but its measurement cycle of 2−6 h limits its responsiveness. In order to enhance sensitivity, multiphoton ionization studies have also been performed in which the traditional nanosecond laser system was replaced by a femtosecond laser system.21 A modest improvement in sensitivity up to a factor ten has been shown, depending on the specific PCDD/PCDF isomer. The practicality of operating such femtosecond laser system (cost, size, and robustness) at a municipal waste combustor site makes this option less attractive at this time. Earlier work at an industrial hazardous waste incinerator showed that REMPI-TOFMS could monitor concentration changes on the order of minutes by successfully monitoring transient aromatic concentrations including monochlorobenzene during the introduction of barrels of liquid hazardous waste.22 REMPI, like all of the online monitors, requires a correlation with more readily measurable indicator compound(s) to indicate either total PCDD/PCDF or the PCDD/PCDF toxic equivalency measure, TEQ. When REMPI-TOFMS was applied to detection of 1,2,4-TrClBz in the boiler exit flue gas of this municipal waste combustor, discrete identification of the target signal was confounded by unforeseen compounds at m/z
■
EXPERIMENTAL SECTION The flue gas from a municipal waste combustor (MWC) with multiple refuse-derived-fuel (RDF) fired boilers (>400 Mg/ day) was sampled after the boiler chamber and prior to the air cleaning devices. Simultaneous, 5-min sampling for PCDD/ PCDF by standard extractive methods23 and select chlorobenzene congeners by GC-REMPI-TOFMS were conducted. The boiler was operated under conditions of shutdowns and startups. Boiler shutdowns would normally occur with operating problems such as jams in the fuel feeding system or for routine maintenance but were initiated for these tests in an attempt to create a large dynamic range of emissions during transient operating conditions. The possible presence of duct concentration and flow gradients were evaluated for their potential detrimental effects on establishing PCDD/PCDF concentration correlations with a single sample probe location. Three simultaneous 30-min measurements were taken through the 10 cm ports at 35%, 45%, and 55% of the cross-duct diameter (∼ 2 m) to check for stratification of emissions and flow. The flow velocities showed only a 1.8% relative standard deviation. Concentrations of PCDD/PCDF and 1,2,4-trichlorobenzene (1,2,4-TrClBz) determined by conventional sampling and analysis had relative standard deviations of only 9.8% and 13%, respectively. These results suggest that the duct flow and concentrations are sufficiently homogeneous that a single-location measurement is representative and allows for determination of correlations. Thirty-seven PCDD/PCDF samples were taken before the spray dryer using U.S. EPA Method 2323 over a three day period during shutdowns (SD), startups (SU), and “late” startup (LSU) conditions. Shutdown conditions typically lasted about 60 min, followed by an approximate 60 min startup period. The late startup condition was defined as the period after the startup period when the system had not yet reached steady state. Samples defined as LSU were taken from about 60 min to up to 180 min after startup. Three additional samples were taken during a preshutdown phase (PSD) in which the plant’s stack CO was rising, necessitating a subsequent boiler shutdown and remedial action. It should be noted that the three enforced shutdowns during a three day plant are likely to have prevented the plant from a return to steady state 924
dx.doi.org/10.1021/es2027339 | Environ. Sci. Technol. 2012, 46, 923−928
Environmental Science & Technology
Article
sampled onto the first trap held at 50 °C, to avoid breakthrough of target compounds. After 500 mL of flue gas volume was sampled (at ∼100 mL/min flow rate), the first trap was flash heated to 190 °C, and its contents were focused onto the second trap in series (held at 50 °C) into a total volume of 25 mL. The 50 °C temperature set point was previously determined to avoid both breakthrough of target compounds and an unacceptably long cool-down period after flash heating. Breakthrough was determined by dynamically spiking a chlorobenzene mixture onto Tenax held at a range of temperatures and monitoring downstream using REMPITOFMS. All gases are transferred through the preconcentrator using silica lined, heated stainless tubing to prevent adsorption or condensation of the sample. Between tests, the concentrator traps underwent a bake-out procedure (at 200 °C) as part of the operational cleaning procedures. Preliminary real time REMPI-TOFMS analysis for 1,2,4TrClBz in the flue gas of the same MWC found that multiple compounds were ionized at the 284/213 nm wavelengths with m/z values 180, 182, and 184 amu as the major identifying (chlorine) isotopes, complicating congener isolation and correlation with PCDD/PCDF.15 The m/z 180 interferents were tentatively identified as stilbene and methyl 9H-fluoren-9one. In order to isolate the trichlorobenzene isomers from other m/z 180 compounds, the samples were thermally released from the second trap onto a chromatographic column separator (60 m length, DB5, Agilent Technologies Inc.) prior to the REMPI-TOFMS inlet. This second trap was initially heated to 160 °C before bake-out of both traps at 200 °C for 5 min. A shorter retention time was obtained by using a higher flow rate of argon through the GC column while maintaining sufficient time resolution to resolve individual trichlorobenzene isomers. Distinction of chlorobenzene isomers was made possible by switching the first laser wavelength to each congener’s or isomer’s specific absorption wavelength. The 12−15 min output trace from the GC column allowed a concentration to be determined for multiple chlorobenzene isomers which were determined by integration of the REMPI trace, normalization to sample volume, and comparison to the REMPI response from the TO-14 calibrated gas mixture containing all three dichlorobenzene isomers and 1,2,4-TrClBz. These quality checks were conducted at the start and at the end of each sampling day. Quantification of target analytes that were not present in the calibrated gas mixture was accomplished prior to this study through measurement of their response with respect to 1,2,4-TrClBz and equal amounts of chlorobenzenes from a mono- to hexachlorobenzene standard (o2si, Inc., Charleston, SC) injected onto the GC column. REMPI was operated in a 2-color, 2-photon mode for chlorobenzene detection. During each GC column run, wavelengths of the first laser photon (OPOTEK Vibrant, Carlsbad, CA) were set at appropriate (elution) times in resonance with the electronic S1←S0 transition of the target chlorobenzene congener. Four isomer specific wavelengths in the range of 280−293 nm were identified28 for sequential detection of 1,4-dichlorobenzene (DiClBz), 1,2,4-TrClBz, 1,2,3-TrClBz, and 1,2,4,5-tetrachlorobenzene (TeClBz) as the targets eluted from the GC column. A fixed 213 nm wavelength photon completed the ionization process for all target analytes. REMPI-TOFMS measurements were made with a two second integration time (average of 20 mass spectra) during the elution
conditions. This is particularly true for PCDD/PCDF concentrations where lagging, or “memory”, effects are likely to have been expected.24 PCDD/PCDF were sampled isokinetically through a glass fiber filter and XAD resin using glass-lined, stainless steel sampling probes coupled to heated (T = 170 °C) Teflon lines. Three 5-min duration samples were collected over each 1-h period, resulting in approximately 15 min between samples. Analysis of the samples followed procedures in EPA Method 23.23 Presampling, pre-extraction, and recovery standards were used throughout the process to determine sampling and analytical efficiency. The combined XAD sorbent, filter, and glassware rinses were extracted with toluene overnight with a Dean−Stark apparatus to remove moisture and isolate the target analytes. The cooled extracts were concentrated to 50 mL with a three-ball Snyder column to prevent loss of target compounds. A Zymark turbovap II was used to further concentrate to 1 mL, from which a portion of the extract was concentrated to 100 μL and then diluted to 12 mL in hexane. The solution was cleaned by sequential solid phase cleanup and extraction with various solvents and an acidic silica gel, basic silica gel, neutral silica gel, basic alumina, and a Celite carbon mix using an automated system. The resulting extract in toluene was concentrated to 100 μL, and recovery standards were added prior to analysis. The samples were analyzed by high resolution gas chromatography/high resolution mass spectrometry (HRGC/HRMS) for the 17 TEF-weighted PCDDs/Fs and tetra- to hepta-CDD/CDF total homologue amounts under procedures recorded elsewhere.25 2005 TEFs26 were used to calculate the TEQ value. The flue gas stream was simultaneously sampled for GCREMPI-TOFMS analysis with a second, colocated probe, approximately 30 cm from the Method 2323 probe. The sampling duration was 5 min, and the frequency was 3 h−1. The flow rate for the REMPI probe was 100 mL/min, and the probe was oriented away from the main flow direction to minimize particle collection. A heated (T = 150 °C) glass/Teflon microfiber filter (Unique Product International, FLT-1584A) and silicon steel transfer line prior to the REMPI inlet prevented any trace particulate matter from clogging the pulsed inlet valve. The filter was changed daily after sampling approximately 7.5 L to minimize phase bias through potential adsorption or desorption of target analytes. Previous work27 has shown over 90% downstream recovery of 2-MCDF through a fly ash-loaded (1 mg/cm2) filter indicating a lack of significant phase bias through the heated filter. This minimal partitioning bias is expected to be even less significant for the more volatile chlorobenzene target compounds in this work. The GC-REMPI-TOFMS system was calibrated daily with N2 gas; a 14-compound, 100 ppb (each) calibration gas mixture (EPA TO-14 Aromatics Subset Mix, Sigma Aldrich); and subsequent N2 gas to ensure no carryover from the TO-14 mixture. In between collected samples, 1,4-dichlorobenzene and 1,2,4-TrClBz were monitored in real time through sampling from the same calibrated gas mixture to ensure consistent performance of the REMPI-TOFMS system. More details on the REMPI-TOFMS operations can be found elsewhere.27 In order to increase the concentration of the target analytes prior to GC-REMPI-TOFMS analysis, the 5 min flue gas sample was introduced into a concentrator (Entech 7100A, Entech Instruments, Inc., Simi Valley, CA). This concentrator utilized two (Tenax) trapping stages in series. The flue gas was 925
dx.doi.org/10.1021/es2027339 | Environ. Sci. Technol. 2012, 46, 923−928
Environmental Science & Technology
Article
Table 1. Boiler Exit PCDD/PCDF Concentrations by Method 2323 Sampling and HRGC/HRMS Analysisa boiler condition
average
relative std dev
maximum
minimum
TEQ (ng TEQ/m3) shutdown (SD) startup (SU) late startup (LSU) preshutdown (PSD) Totalb (ng/m3) shutdown (SD) startup (SU) late startup (LSU) preshutdown (PSD)
73.0 21.0 89.5 96.9 80.4 3304 1170 4040 4120 3510
0.89 0.66 0.88 0.65 0.14 0.89 0.62 0.91 0.67 0.13
256 40.0 256 198 92 12600 2320 12600 9020 4020
4.9 4.9 8.5 5.9 69 256 301 366 256 3140
no. of condition series
no. of samples 40 10 15 12 3 40 10 15 12 3
4 5 4 1 4 5 4 1
a
Quality criteria were acceptable: only two of 40 samples had recoveries of a single presampling standard (1,2,3,4,7,8,9-HpClDF) above the 130% criterion, at 133% and 141%. bSum of tetra- to octa-chlorinated congeners.
of the GC column. Mass gates were set in the TOFMS to avoid detector overload on molecular masses outside the mass range of interest (m/z 140−220). More details on the REMPITOFMS instrument are available elsewhere.20
■
RESULTS AND DISCUSSION The 40 pre-APCS PCDD/PCDF TEQ concentrations ranged from 5 to 256 ng TEQ/m3 over five series of startup and shutdown conditions plus the one preshutdown condition (see Table 1). The shutdown periods (SD) show, on average, the lowest concentrations (average 21 ng TEQ/m3), while the startup (SU) and late start up (LSU) periods averaged 90 and 97 ng TEQ/m3, respectively. The highest concentrations were observed during the start-up conditions. The relative standard deviations ranged from 65 to 89% (with the exception of the three PSD samples). Similar trends are observed for the PCDD/PCDF Total, sum of tetra- to octa-chlorinated congeners. This large range of concentrations allows for a robust correlation to be established for a wide range of fuel types and operating conditions. GC-REMPI-TOFMS measurements of the as-sampled concentrations of four chlorobenzenes, potential predictors of PCDD/PCDF TEQ, are shown in Table 2. Four potential
Figure 1. GC-REMPI-TOFMS chromatogram at the 1,2,4-trichlorobenzene wavelength with inset showing associated mass spectrum.
The wide range of PCDD/PCDF concentrations observed (50fold, Table 1) are reflected in the 25- to 100-fold range of these select ClBzs. The lowest chlorobenzene concentrations were about five times above their method detection limits, suggesting that for this plant, sampling times even shorter than five minutes would result in measurable indicator concentrations and, hence, PCDD/PCDF correlations. Unequal numbers of samples reflect the operator’s choice to focus on specific congeners. The three plant conditions provide a wide range of paired concentration values from which to establish a robust correlation of indicator compounds and PCDD/PCDF. Table 3 illustrates the ability of the ClBz indicator compounds to predict PCDD/PCDF TEQ and Total over
Table 2. Boiler Exit Concentrations of Indicators Measured by GC-REMPI-TOFMS (μg/m3) indicator
average
relative std dev
maximum
minimum
no. of values
1,4-DiClBz 1,2,4-TrClBz 1,2,3-TrClBz 1,2,4,5TeClBz
10.31 10.39 14.29 1.42
1.52 0.99 0.98 0.98
86.99 55.61 71.10 5.63
0.84 0.93 3.04 0.13
52 52 34 31
PCDD/PCDF indicators, 1,4-diClBz, 1,2,4-TrClBz, 1,2,3TrClBz, and 1,2,4,5-TeClBz, were targeted. The chlorobenzene indicators were isolated from the other interferents with the aid of the chromatographic column separator prior to the REMPITOFMS inlet. Figure 1 displays the GC-REMPI-TOFMS chromatogram at the 1,2,4-trichlorobenzene wavelength. Minimal loss of sensitivity with degree of chlorination was observed, contrary to suggestions that multiphoton ionization schemes are suspect for detection of aromatics with increasing degree of ionization.9 Laboratory tests with an injected mixture of equal concentrations of 1,4-DiClBz, 1,2,4-TrClBz, and 1,2,4,5:TeClBz into the GC-REMPI-TOFMS found relatively equivalent peak area responses of 1.00:0.66:0.77, respectively.
Table 3. Correlation Coefficient, R, of PCDD/PCDF with Indicators indicator
TEQ
Total
no. of data pairs
1,4-DiClBz 1,2,4-TrClBz 1,2,3-TrClBz 1,2,4,5-TeClBz
0.256 0.852 0.627 0.783
0.322 0.887 0.691 0.819
40 40 27 24
the three day test period. The principal target indicator, 1,2,4TrClBz, predicted TEQ at R = 0.85 suggesting that REMPITOFMS monitoring of this compound can provide an excellent 926
dx.doi.org/10.1021/es2027339 | Environ. Sci. Technol. 2012, 46, 923−928
Environmental Science & Technology
Article
eleven PCDD/PCDF TEQ values. This suggests that the current correlative model for PCDD/PCDF TEQ was consistent despite three years’ time and fuel, boiler unit, and operational changes. While the universality of these indicator compounds and their predictor models across facility types remains to be further verified, these results are encouraging for intrafacility comparison. Derived models will likely vary somewhat depending on the dominant mechanism of pollutant formation which may be a function of plant type, waste/fuel types, and operating/combustion conditions. The models may also be multi-indicator and are not necessarily linear with concentration. Multi-indicator models have been shown to yield improved PCDD/PCDF predictions.29 Figure 3 shows the three day, temporal trend of PCDD/ PCDF TEQ concentration through the multiple shutdowns and
indication of PCDD/PCDF concentration trends. Of particular interest is the ability of the correlation to hold during each phase of plant operation. During the shutdown, startup, and late startup phases, the correlation coefficient, R, of 1,2,4TrClBz with PCDD/PCDF TEQ was 0.61, 0.89, and 0.81, respectively (data not tabulated). A plot of the 5-min, conventionally sampled PCDD/PCDF concentrations versus the GC-REMPI-TOFMS-sampled 1,2,4-TrClBz concentrations is shown in Figure 2. This demonstrates that regardless of the
Figure 2. Correlation of PCDD/PCDF concentration from 5-min EPA Method 23 sampling with GC-REMPI-TOFMS measurements of 1,2,4-trichlorobenzene concentration.
plant operating phase, following the concentration of 1,2,4trClBz allows prediction of PCDD/PCDF levels. This is of particular note for the late start up phase, where PCDD/PCDF levels would be expected to be dropping and more consistent than during upset periods. Similar results are observed for PCDD/PCDF Total where shutdown, startup, and late startup phases had correlation coefficients of 0.72, 0.91, and 0.84 with 1,2,4-TrClBz. Use of a simple linear model to predict PCDD/PCDF TEQ with REMPI-TOFMS measurements of 1,2,4-TrClBz resulted in PCDD/PCDF (ng TEQ/m3) = 18.8 + 4.9 * [1,2,4-TrClBz (μg/m3)] with parameter statistics in Table 4.
Figure 3. Comparison of PCDD/PCDF concentration from 5-min EPA Method 23 measurements and model predictions using GCREMPI-TOFMS measurements of 1,2,4-trichlorobenzene concentration in the linear model.
start ups, along with the GC-REMPI-TOFMS-measured concentration of 1,2,4-TrClBz. The 1,2,4-TrClBz concentration follows the M23 PCDD/PCDF concentration well over the three days’ worth of boiler shutdowns and startups. Peak PCDD/PCDF concentrations during startups through late startups are mimicked by the concentration of 1,2,4-TrClBz. The increase of PCDD/PCDF and 1,2,4-TrClBz levels over the three days of transient operations are likely due to cumulative degradation of combustion efficiency and resultant formation of carbonaceous structures.18
Table 4. Parameters for Linear Model of 1,2,4-TrClBz and PCDD/PCDF TEQ (F Ratio = 100.37) term
estimate
std error
t ratio
prob > |t|
intercept 1,2,4-TrClBz
18.8 4.9
7.7 0.5
2.4 10.0
0.0194
