Environ. Sci. Technol. 2002, 36, 5199-5204
Uncertainties in Charring Correction in the Analysis of Elemental and Organic Carbon in Atmospheric Particles by Thermal/Optical Methods
TABLE 1. Experimental Parameters of the Two Thermal Methods Used in This Study analysis method
HONG YANG AND JIAN ZHEN YU* Department of Chemistry, Hong Kong University of Science & Technology, Clear Water Bay, Kowloon, Hong Kong a
Thermal/optical methods are widely used in the determination of aerosol organic carbon (OC) and elemental carbon (EC) collected on quartz filters. A fraction of OC undergoes charring to form pyrolytically generated EC (PEC) during thermal analysis. The correct speciation of OC and EC in thermal/optical methods depends on one of the following two assumptions: (1) PEC evolves before native EC evolves in the analysis or (2) PEC and native EC have the same apparent light absorption coefficient (σ) at the monitoring light wavelength. Neither of these assumptions has actually ever been checked or tested. The first assumption is invalidated by the observation that the combustion of PEC overlaps that of native EC despite multiple stepwise combustion at temperatures ranging from 575 to 910 °C. An examination of σ versus EC evolution indicates that the σ values of PEC and EC are not the same in most cases and the σ value of PEC is not constant during a single thermal analysis. The second assumption is thus invalid as well. The measured EC concentrations can either overestimate or underestimate the true native EC concentrations depending on the relative magnitude of the σ values of the PEC and native EC at the point where the instrument sets the EC/OC split line. Both over- and underestimation have been observed in real aerosol samples. The unequal σ values of PEC and EC also explain that different temperature programs, when employed to analyze the same filter samples, systematically yield different EC and OC concentrations. Our findings imply that minimizing charring improves the accuracy of the EC/OC split in thermal/ optical methods.
Introduction Thermal/optical methods are widely used in the determination and speciation of aerosol organic carbon (OC) and elemental carbon (EC) (1-5). In these methods, carbon on filter substrates is made to evolve through programmed progressive heating in a controlled atmosphere. It has long been recognized that a fraction of OC is pyrolytically converted to EC during analysis (2, 6-8). This conversion leads to difficulty in accurate speciation of OC and EC. The use of an optical system for continuous monitoring of filter transmittance or reflectance has been used to correct for pyrolytically generated EC (PEC). The common practice is * Corresponding author phone: (852)2358-7389; fax: (852)23581594; e-mail:
[email protected]. 10.1021/es025672z CCC: $22.00 Published on Web 10/24/2002
2002 American Chemical Society
carrier gas
Ace-Asia
UST-3
He-1 He-2 He-3 He-4 He/O2-1a He/O2-2 He/O2-3 He/O2-4 He/O2-5
340 °C, 60s 500 °C, 60s 615 °C, 60s 870 °C, 90s 575 °C, 45s 650 °C, 45s 725 °C, 45s 800 °C, 45s 910 °C, 100s
250 °C, 150s 500 °C, 150s 650 °C, 150s 850 °C, 150s 650 °C, 150s 750 °C, 150s 850 °C, 150s 890 °C, 150s
A mix of 1% oxygen in UHP helium.
to consider the amount of C oxidation necessary to return the filter transmittance/reflectance to the prepyrolytic conversion value as the PEC. The point at which the filter transmittance/reflectance reaches its initial value is defined as the split between OC and EC. Carbon that evolves prior to the split is operationally defined as OC and carbon that evolves after the split is EC. For the convenience of later discussion, we define apparent EC (AEC) as the EC amount as determined by the instrument using the above correction scheme; and the true native EC (NEC) is the EC present in the aerosols before the thermal alteration. Such a correction scheme is valid only if one of the following two assumptions holds: (1) PEC evolves before native EC evolves in the analysis or (2) PEC and native EC have the same absorption coefficient (σ) at the monitoring light wavelength. The validity of these assumptions has not been established. Johnson et al. (1) was the first to note the necessity of the second assumption; however, they did not offer any evidence to support their statement in their paper. In this work, we have designed experiments to show that neither of the two assumptions holds. We also have examined the consequences of the erroneous assumptions on the accuracy of EC measurements.
Experimental Section Aerosol Materials. Aerosol samples collected in five locations in Asia were used in this study to capture large variations in aerosol chemical composition. Quartz fiber filters (PallGelman, Ann Arbor, MI) were the collection substrates for all the samples. Total suspended particulate (TSP) samples were collected onto 20 × 25 cm quartz fiber filters in Hong Kong, Guangzhou, China, and on the South China Sea during a cruise over periods of 4-72 h at flow rates of 0.61-1.13 m3/min using a high volume sampler (Andersen Instruments, Smyrna, GA). Aerosols of less than 2.5 µm were collected in Nanjing and Hong Kong onto 20 × 25 cm quartz fiber filters over 12-h periods at a flow rate of 1.13 m3/min using a high volume sampler fitted with an impactor inlet of 2.5 µm cut size (Andersen Instruments, Smyrna, GA). Aerosols of less than 5 µm were collected onto 20 cm filters in Cheju Island, South Korea at a flow rate of 0.50 m3/min using a highvolume particle trap impactor/denuder sampler designed by CalTech (9). Our previous work has shown that the water-soluble organic carbon (WSOC) fraction of ambient aerosols tends to char and it accounts for a significant fraction of overall PEC (7). This feature of WSOC makes it possible to generate PEC, free of native EC, from real aerosol materials. The WSOC fraction was isolated from the rest of aerosol materials using VOL. 36, NO. 23, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5199
FIGURE 1. Thermograms of (a) an untreated aerosol sample (PM2.5_UST_3-7-01) and its water extract and (b) charcoal. water extraction of the filters. The aerosol water extracts were obtained by mixing filter pieces with water in a sonication bath. The water extracts were then filtered and concentrated with a rotary evaporator to suitable volumes for subsequent use. The aerosol water extracts were deposited onto a filter before the thermal analysis. Deposition was accomplished through either liquid doping with a syringe or nebulization of the solution followed by collection of the resulting WSOC aerosol onto the filter. A CaSO4 drying column (20 cm long, 1 cm i.d.) was placed between the nebulizer and the filter to remove water. For the examination of the thermal characteristics of pure PEC, a prebaked clean quartz filter was used to receive the water extracts. For the study of the effect of additional PEC on the EC measurement, a piece of filter preloaded with aerosol particles was removed from its 20 × 25 or 20 cm parent filter to receive the aerosol water extracts. Charcoal aerosols were generated in our lab by passing charcoal powder (-20+50 mesh) suspended in a methanol solution through the nebulizer followed by a drying column. The charcoal particles were then collected onto a 47 mm filter. Carbon Analysis. Carbon analysis of the aerosol samples was carried out by placing a filter strip of 1× 1.45 cm in a 5200
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 23, 2002
thermal/optical aerosol carbon analyzer (Sunset Laboratory, Forest Grove, OR). The carbon analyzer employs a temperature- and atmosphere-controlled oven to evolve carbon species and a laser at a wavelength of 680 nm to monitor the filter transmittance (4). Table 1 lists the two thermal methods used in this study, Ace-Asia and UST-3. The first thermal method was used in our lab to analyze aerosol samples collected during the Aerosol Characterization Experiment in Asia (Ace-Asia).
Results and Discussion Generation of PEC from Aerosol Water Extracts or Sucrose. Deposition of pyrolizable OC (aerosol water extracts or sucrose solution) onto either blank filters or aerosol-laden filters was carried out through liquid doping or nebulization application. When direct liquid doping was used, the resulting aqueous OC deposition was not evenly distributed on the filter due to wicking effect of the quartz fibers. Filters that were doped with aqueous OC and were retrieved after thermal analysis in the helium stage showed darker coloration at the filter edges. This indicated that the aqueous OC tended to migrate away from the filter center and concentrate more on the filter edges during drying. It was also noted that the
FIGURE 2. Hypothetic thermograms of pure PEC, pure NEC, and a mixture of the two in a HeO2 atmosphere. pyrolizable OC could penetrate deep in the filter fibers when applied through liquid doping, whereas realistic aerosol deposition was mainly on the filter surface. Despite the two above-mentioned deviations of doped filters from realistic aerosol deposition, we have reasons to believe that the PEC obtained this way provides a close simulation of the PEC that is formed on a realistic aerosolladen filter. We have observed that blackness due to pyrolysis extended completely through the filter if the analysis was stopped after the helium stage and the filter was retrieved from the carbon analyzer. That is, the filter was black only at the aerosol-laden side before C analysis; however, after the filter went through the He stage of the analysis, both sides of the filter were black. Huntzicker et al. (2) also reported this phenomenon and suggested that pyrolysis at least partially occurred as a result of the interaction between organic vapors moving through the filter fibers. We have additional evidence to support this hypothesis. For filters loaded with WSOC, either through nebulization or liquid doping, the instrument could always correctly set the EC/OC split line, i.e., at the end of each analysis when all the PEC was oxidized and the filter transmittance (T) returned to the initial T value. Had the PEC been formed where the pyrolizable OC was, the filter T in the center, which the laser monitored, would return to the initial T earlier than the filter T at the edges due to less pyrolizable OC in the filter center caused by the wicking effect. This would lead to an EC/OC split line set at a point before all the PEC was oxidized, which was incorrect since there was no EC in the aqueous OC doped filter. In addition, we have carried out analysis of filters doped with different loadings of a same WSOC sample. When we plotted the maximum light absorbance versus the PEC amount, a linear curve was obtained. This again suggests that the resulting PEC from the WSOC doped was more likely distributed evenly on the filter. In summary, existing evidence supports the hypothesis that the PEC formation involves organic vapors moving through the filter fibers. Such a vapor deposition process has the effect of smoothing an initially uneven deposition and redistributes pyrolizable OC on the surface deep into the filter fibers. Consequently, this makes it not important whether the aqueous OC is initially evenly distributed through the filter and whether the WSOC is initially located at the surface or deep in the fibers.
FIGURE 3. Filter absorbance as a function of unburned EC in thermal analysis for aerosol samples collected in five different locations. Assumption 1. Figure 1a shows the thermograms of an aerosol filter strip and its WSOC fraction. In the thermal analysis of the aerosol sample, the transmittance decreases and reaches its minimum during the first stage of the analysis in helium, signaling formation of PEC. The second stage of the analysis is carried out in a mixture of helium and oxygen; as a result, the transmittance increases and eventually returns to the initial value due to combustion of the light-absorbing PEC and EC. The transmittance reaches its maximum value at the end of the analysis when all the PEC and EC are burned. The first assumption, i.e., PEC evolves before native EC evolves in the analysis, is invalidated on the basis of three observations. First, PEC and NEC are not resolved; instead they appear as a broad peak in the thermogram of the untreated filter (Figure 1a). Second, when the water-soluble portion of OC is subjected to the same thermal analysis, the evolution of WSOC continues into the He/O2 stage and goes beyond the EC/OC split line (Figure 1a). Third, a comparison of the combustion of charcoal and PEC generated from WSOC shows that their evolution times overlap (Figure 1a,b). In conclusion, Figure 1 demonstrates that the evolution of PEC VOL. 36, NO. 23, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5201
TABLE 2. Effect of Additional PEC on the Apparent EC Concentrations EC (µgC/cm2)b samplea
analysis method
untreated filter
filter deposited with sucrosec
charcoal PM2.5-HK-28-Jun-01 PM2.5-HK-3-Jul-01 PM2.5-HK-3-Jul-01 PM2.5-HK-3-Jul-01 PM2.5-NJ-4-Feb-01 TSP-Sea-summer TSP-GZ-summer
Ace-Asia Ace-Asia Ace-Asia UST3 Ace-Asia Ace-Asia Ace-Asia Ace-Asia
2.60 ( 0.33 0.39 ( 0.01 1.87 ( 0.03 2.32 ( 0.07 1.87 ( 0.03 8.65 ( 0.32 1.53 ( 0.05 0.76 ( 0.02
1.96 ( 0.30 0.46 ( 0.01 2.67 ( 0.21 2.48 ( 0.01
filter doped with WSOCd
∆EC/EC
2.25 ( 0.26 9.55 ( 0.13 1.84 ( 0.06 0.95 ( 0.10
-25 ( 17% 18 ( 4% 43 ( 11% 7 ( 3% 20 ( 14% 10 ( 4% 20 ( 5% 25 ( 13%
a The sample description contains information on aerosol type, sample location, and sampling date in sequence. Sampling locations: HK stands for Hong Kong, NJ for Nanjing, GZ for Guangzhou, and Sea for South China Sea. b The EC concentrations are expressed as average ( standard deviation, which were obtained from multiple analyses. c Sucrose was deposited to the filters through nebulization. d The WSOC was added to the filters through liquid spiking using a syringe.
and NEC overlaps in the thermograms. In short, the first assumption does not hold. Assumption 2. With the first assumption proved to be invalid, the correct speciation of EC and OC relies on the second assumption, i.e., that PEC and NEC have the same absorption coefficients at the monitoring light wavelength. The derivation of this assumption is as follows. The filter absorbance is determined by the instrument on the basis of eq 1
A ) ln
T To
(1)
where T and To are, respectively, the filter transmittance at a given time and the transmittance through a blank filter (10, 11). The filter transmittance at the end of each analysis, when all the carbon on the filter has been removed, is taken as To. The absorption coefficient, σ, in the unit of m2/g is
σ ) 100 ×
A UEC
(2)
where UEC is the amount of unburned EC in the unit of µg/cm2 on the filter. The initial light absorbance of the filter is attributed to NEC, therefore
100 × Ai ) σNEC[NEC]
(3)
For the purpose of understanding the overlapping features of PEC and NEC, hypothetic evolution thermograms are constructed for pure PEC, pure NEC, and a mixture of the two in Figure 2. With the instrument-defined EC/OC split line as the division line, PEC1 and PEC2 are defined as the portion of PEC that is oxidized before and after the split line, respectively. Similarly, NEC1 and NEC2 are defined as the fraction of NEC that is oxidized before and after the split line, respectively. The measured EC concentration (AEC) is the sum of NEC2 and PEC2, whereas the true native EC concentration is the sum of NEC1 and NEC2. At the split line, the light absorbance, As, is attributed to NEC2 and PEC2:
100 × As ) σNEC[NEC2] + σPEC2[PEC2]
(4)
The instrument’s EC/OC split definition dictates that Ai and As are equal, which leads to
σNEC[NEC1] ) σPEC2[PEC2]
(5)
Consequently, AEC can be expressed as
[AEC] ) 5202
9
σNEC [NEC1] + [NEC2] σPEC2
(6)
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 23, 2002
FIGURE 4. Light absorption coefficient of unburned EC during the thermal analysis for (a) an PM2.5 aerosol collected in Hong Kong and (b) a laboratory prepared charcoal filter. It is clear from eq 6 that AEC is equal to NEC only when σNEC and σPEC2 have the same values. Aerosol samples collected in the five different locations were examined for their absorbance variation with the amount of unburned EC on the filter, which includes both native EC and PEC formed during the first stage of the analysis (Figure 3). The slope of the absorbance versus UEC determines σ. A linear curve would indicate a constant σ value. The curves in Figure 3 are linear at the lower end of the unburned EC. The portions in the vicinity of the instrumental
FIGURE 5. Comparison of PEC formation in the two thermal analysis methods, Ace-Asia and UST-3. EC/OC split lines are curved, indicating that σ changes as EC is burned. This, in turn, suggests that the unburned EC is heterogeneous and its components (i.e., PEC and NEC) are oxidized in a nonproportional fashion. Figure 3, thus, invalidates the second assumption. Implication for Carbon Measurements by Thermal Methods. It is apparent that a deviation from the second assumption would lead to incorrect determination of the EC/OC split line and consequently to uncertainties in the NEC concentration measurement. The amount of deviation is
∆EC ) [AEC] - [NEC] )
(
)
σNEC - 1 [NEC1] σPEC2
(7)
One can see from eq 7 that AEC underestimates the NEC concentration when σNEC < σPEC2, whereas AEC overestimates the NEC concentration when σNEC > σPEC2. The magnitude of deviation of AEC from NEC is proportional to the amount of NEC that is oxidized before the split line and the ratio between the σ values of NEC and PEC2.
The above logic is tested by comparing the AEC concentrations in aerosol samples of the same NEC loadings but of different PEC loadings. This is achieved by depositing sucrose or aerosol water extracts onto aerosol-laden filters through nebulization or direct liquid spiking. Table 2 lists the AEC values of the filters before and after deposition of the additional PEC. The AEC concentrations in the ambient aerosol samples increase with the additional PEC from either sucrose or WSOC, with a percentage change ranging from 7% to 43%. After consideration of the measurement uncertainties, these values are statistically different from zero by student t-test at 95% confidence level. The AEC concentration in the charcoal sample, however, decreases with the introduction of the additional PEC. An examination of the σ profile reveals that whether AEC increases or decreases with the additional PEC is explained by the relative magnitude of σPEC2 and σNEC. Figure 4 shows two examples, an aerosol sample taken in Hong Kong and the charcoal filter sample. The additional PEC for the aerosol sample is generated from its WSOC fraction. The σ values of the additional PEC are obtained by analyzing the same amount of WSOC through liquid spiking. PEC from the aerosol water extract shows a decreasing σ as its oxidation progresses (Figure 4a). The σ value drops from an initial value of 103 to 8 m2/g at the end of its evolution. In the portion of PEC that evolves after the split line in the analysis of the doped filter sample, the σ values are lower than the σ values of the unaltered filter. Consequently, the split line is earlier than it should be, leading to a higher AEC value. The same logic explains that the AEC value is lower after addition of sucrose to the charcoal filter. In the case of the charcoal sample, the σ values for charcoal EC are lower than those for the mixture of charcoal EC and PEC from sucrose (Figure 4b). Equation 7 indicates that the degree of PEC overcorrection or undercorrection depends on the degree of NEC and PEC overlap and also on the difference of the σ values between the two forms of EC. The degree of overlap is greatly influenced by the thermal programs. As a result, different thermal programs would lead to different AEC concentrations. The UST-3 and Ace-Asia methods employ steps with similar temperatures but differ in the residence time at each temperature step. The UST-3 method uses longer residence time at each temperature step, which allows a more complete evolution of carbon at each temperature. This, in turn, leads to less OC available for charring. As a result, the charring amount is reduced when using the UST-3 method. Figure 5 compares the amount of carbon that remains after the first stage in helium in the two methods for 12 aerosol samples
TABLE 3. Comparison of EC Measurements by the Two Thermal Methods, Ace-Asia and UST-3 EC/TCb
PEC+NEC (µgC/cm2)
samplea
TCb (µgC/cm2)
Ace-Asia
UST-3
PM5-Cheju-24-Mar-01 PM5-Cheju-29-Mar-01 PM5-Cheju-10-Apr-01 PM5-Cheju-14-Apr-01 PM5-Cheju-19-Apr-01 TSP-Sea-summer TSP-GZ-summer PM2.5-HK-03-Jul-01 TSP-HK-27-Jul-01(day) TSP-HK-4-Jul-01(night) PM2.5-NJ-11-Feb-01 PM2.5-NJ-04-Feb-01
9.57 14.64 12.70 5.25 3.37 2.03 5.15 7.30 38.33 45.79 11.14 31.14
0.34 0.19 0.12 0.18 0.35 0.38 0.30 0.26 0.36 0.22 0.26 0.29
0.36 0.20 0.13 0.22 0.41 0.43 0.35 0.30 0.38 0.23 0.24 0.27
∆ECb
(µgC/cm2)
-0.18 -0.25 -0.06 -0.21 -0.19 -0.11 -0.28 -0.29 -1.03 -0.33 0.26 0.54
∆EC/ECc
Ace-Asia
UST-3
-5% (10%) -8% (5%) -4% (8%) -18% (10%) -14% (9%) -13% (8%) -15% (9%) -13% (9%) -7% (4%) -3% (5%) 10% (3%) 6% (5%)
4.77 7.36 4.98 2.19 1.37 0.76 1.89 2.48 19.39 21.85 5.45 16.73
3.16 5.47 4.75 1.62 1.37 0.87 1.47 1.89 15.35 19.69 4.20 12.78
a The sample description contains information on aerosol type, sample location, and sampling date in sequence. Sampling locations: HK stands for Hong Kong, NJ for Nanjing, GZ for Guangzhou, and Sea for South China Sea. b The TC and EC/TC values are the average values of 2-3 measurements. ∆EC)EC(Ace-Asia)-EC(UST-3). c The EC values in ∆EC/EC are based on measurements using the UST-3 method. The numbers in the parentheses are one standard deviation.
VOL. 36, NO. 23, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5203
collected from the five different locations in Asia. The amount of unburned EC before the introduction of O2 with the UST-3 method is approximately 82% of that with the Ace-Asia method. With the NEC concentrations remaining the same regardless of the differences in the temperature programs, there is consequently less PEC formed in the UST-3 method. The percentage PEC decreases range from 15% to almost 100%. As a result of less interference from PEC, AEC values given by the UST-3 method are closer to the true NEC values in each sample. Table 3 lists the AEC concentrations for the set of aerosol samples using the two temperature programs. If the AEC concentrations given by the UST-3 method were taken as the “true” NEC, the Ace-Asia method would underestimate the NEC concentrations for three of the five Cheju samples, two of the three Hong Kong samples, the South China Sea sample, and the Guangzhou sample. The underestimation ranged from 7% to 18%. For the other two Cheju samples and one Hong Kong sample, the underestimation was not statistically different from zero when the measurement uncertainty was considered. On the other hand, the AceAsia method would overestimate the NEC concentrations by 7-10% for the samples collected from Nanjing. In the Nanjing samples, the σ value of the unburned EC at the instrumentset EC/OC split was found to be lower with the Ace-Asia method than that with the UST-3 method. This indicates that the portion of PEC remaining after the split has a lower σ value than that of NEC in the Nanjing samples. In aerosol samples taken in the other four locations, the reverse relationship between the σ values of PEC and NEC was observed. Consequently, eq 7 explains both the positive bias of the EC measurement in the Nanjing samples and the negative bias in the samples from the other locations when the samples are analyzed using the Ace-Asia method. We have demonstrated that the true EC concentrations can be either overestimated or underestimated depending on whether the fraction of PEC that is burned after the EC/OC split line has a higher or a lower σ value than that of NEC. The σ value of PEC formed from aerosol water extracts has been shown to vary with its evolution during a single thermal analysis. It is plausible to expect that the σ value of PEC is affected by the composition of its organic precursors, which vary greatly among aerosols collected from different
5204
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 23, 2002
locations and at different times. As a result, the magnitude of the uncertainty arising from the incorrect EC/OC split is expected to vary from one aerosol sample to another. Before we are able to devise a means to set a correct EC/OC split line, it is important to employ thermal conditions that minimize charring formation in order to improve the accuracy of the EC and OC determination. Further work is ongoing to seek an optimal thermal method that produces a minimum PEC.
Acknowledgments This work was supported by the Research Grants Council of Hong Kong, China (HKUST6185/00P and DAG00/01.SC17).
Literature Cited (1) Johnson, R. L.; Shah, J. J.; Cary, R. A.; Huntzicker, J. J. An Automated Thermal-Optical Method for the Analysis of Carbonaceous Aerosol. In Atmospheric Aerosol: Source/Air Quality Relationships; ACS Symposium Series 167, Macias, E. S., Hopke, P. K., Eds.; American Chemical Society: Washington, DC, 1981; pp 223-233. (2) Huntzicker, J. J.; Johnson, R. L.; Shah, J. J.; Cary, R. A. Analysis of Organic and Elemental Carbon in Ambient Aerosols by a Thermal-Optical Method. In Particulate Carbon: Atmospheric Life Cycle; Wolff, G. T., Klimisch, R. L., Eds.; Plenum Press: New York, 1982; pp 79-85. (3) Chow, J. C.; Watson, J. C.; Pritchett, L. C.; Pierson, W. R.; Frazier, C. A.; Purcell, R. G. Atmos. Environ. 1993, 27A, 1185-201. (4) Birch, M. E.; Cary, R. A. Aerosol Sci. Technol. 1996, 25, 221-241. (5) Chow, J. C.; Watson, J. C.; Crow, D.; Lowenthal, D. H.; Merrifield, T. Aerosol Sci. Technol. 2001, 34, 23-24. (6) Cadle, S. H.; Groblicki, P. J.; Stroup, D. P. Anal. Chem. 1980, 52, 2201-2206. (7) Yu, J. Z.; Xu, X.; Yang, H. Environ. Sci. Technol. 2002, 36, 754761. (8) Huebert, B. J.; Charlson, R. J. Tellus B 2000, 52, 1249-1255. (9) Mader, B. T.; Flagan, R. C.; Seinfeld, J. H. Environ. Sci. Technol. 2001, 35, 4857-4867. (10) Hansen, A. D. A.; Rosen, H.; Novakov, T. Sci. Total Environ. 1984, 36, 191-196. (11) Gundel, L. A.; Dod, R. L.; Rosen, H.; Novakov, T. Sci. Total Environ. 1984, 36, 197-202.
Received for review March 25, 2002. Revised manuscript received July 22, 2002. Accepted September 17, 2002. ES025672Z
