Relationship between Particle Mass and Mobility for Diesel Exhaust Particles

Transcription

1 Environ. Sci. Technol. 2003, 37, Relationship between Particle Mass and Mobility for Diesel Exhaust Particles KIHONG PARK, FENG CAO, DAVID B. KITTELSON, AND PETER H. MCMURRY* Department of Mechanical Engineering, University of Minnesota, 111 Church Street SE, Minneapolis, Minnesota We used the aerosol particle mass analyzer (APM) to measure the mass of mobility-classified diesel exhaust particles. This information enabled us to determine the effective density and fractal dimension of diesel particles as a function of engine load. We found that the effective density decreases as particle size increases. TEM images showed that this occurs because particles become more highly agglomerated as size increases. Effective density and fractal dimension increased somewhat as engine load decreased. TEM images suggest that this occurs because these particles contain more condensed fuel and/ or lubricating oil. Also, we observed higher effective densities when high-sulfur EPA fuel ( 360 ppm S) was used than for Fischer-Tropsch fuel ( 0 ppm S). In addition, the effective density provides the relationship between mobility and aerodynamic equivalent diameters. The relationship between these diameters enables us to intercompare, in terms of a common measure of size, mass distributions measured with the scanning mobility particle sizer (SMPS) and a MOUDI impactor without making any assumptions about particle shape or density. We show that mass distributions of diesel particles measured with the SMPS-APM are in good agreement with distributions measured with a MOUDI and a nano-moudi for particles larger than 60 nm. However, significantly more mass and greater variation were observed by the nano- MOUDI for particles smaller than 40 nm than by the SMPS- APM. Introduction Fine and ultra-fine particles in the ambient atmosphere are of current interest because of their adverse health effects and their impacts on the earth s radiation balance, visibility impairment, and atmospheric chemistry. The inability of conventional techniques to measure certain properties of fine and ultra-fine particles, especially nonspherical particles that consist of complex multicomponent mixtures including semivolatile compounds, limits the accuracy with which such particles can be characterized. Diesel particulate emissions are a major source of fine and ultra-fine particles to the atmosphere. In this study, we focus on the relationship between mobility size and mass of diesel particles in the nm mobility equivalent diameter range. This relationship enables us to determine their effective density, their * Corresponding author phone: (612) ; fax: (612) ; mcmurry@me.umn.edu. fractal dimension, and the relationship between mobility size and aerodynamic size. Several previous studies have reported measurements of the effective density of diesel exhaust particles. The effective density is important because it determines particle transport properties and establishes the relationship between mobility size and aerodynamic size. All of this work has involved the use of differential mobility analyzers (DMA) (2, 3), which classify particles according to mobility size, and impactors, which classify particles according to aerodynamic size (4-7). The following relationship is then used to calculate the effective density and to establish the relationship between mobility size and aerodynamic size (8): F eff d me 2 C me ) 6m πd me C me )F o d ae 2 C ae (1) where F o is the unit density (1 g/cm 3 ), F eff is the effective density, d ae is the aerodynamic equivalent diameter, d me is the mobility equivalent diameter, C is the slip correction factor (9), and m is particle mass. Ahlvik et al. (10) used a DMA and an electrical low pressure impactor (ELPI) to measure the effective density of diesel exhaust particles. Effective density was obtained by varying its value iteratively until the number of particles in each ELPI channel equaled the number of particles in the corresponding size range in the differential mobility particle sizer (DMPS) distribution. Also, they determined the effective density by following the methodology described by Kelly and McMurry (11) whereby a DMA is used to select particles of known mobility followed by an impactor to determine their aerodynamic size. Maricq et al. (12) also used a DMA and an ELPI and determined the effective density by measuring the distribution of aerodynamic diameters for a selected mobility diameter. They reported that diffusion and electrostatic losses of small particles through the upper stages of the ELPI distorted the aerodynamic size distributions and that the effective density varied by (20% because of these losses. They observed somewhat higher than expected particle concentrations in the lowest stage (37 nm cut size) because of particle bounce from the upper stages. Virtanen et al. (13) determined the effective density by simultaneously measuring the mobility and aerodynamic size distributions with a scanning mobility particle sizer (SMPS) and an ELPI, respectively. In their method, size-dependent effective density was varied to find the best fit of the ELPI and SMPS distributions. In this study, we used the DMA-APM technique to determine the relationship between mobility size and mass for diesel particles (14). With this technique, the DMA selects particles of known mobility, and the aerosol particle mass analyzer (APM) (1) measures their mass (or distribution of masses, if particles of a given mobility contain several types of particles with different masses). These measurements enable us to determine the effective density and fractal dimension. Because measurements are carried out on gasborne particles, our measurements are less susceptible to artifacts because of volatilization or adsorption that occur, for example, with filters or impactors. We have supplemented the mass and mobility measurements with TEM images of mobility-classified particles. The TEM images provide valuable qualitative insights into the relationship between particle morphology and particle properties. Previous studies (15-20) have shown that the fractal dimension is a useful mathematical construct for determining certain properties of agglomerate particles. The following /es025960v CCC: $ American Chemical Society VOL. 37, NO. 3, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY Published on Web 12/19/2002

2 TABLE 1. Summary of Diesel Engine Operating Conditions with Results of Effective Densities of Diesel Exhaust Particles a engine type DR ratio ) DR1 DR2 fuel type speed (rpm) load (Nm) mobility size (nm) effective density (g/cm 3 ) For Measurements of Mass-Mobility Relationship John Deere 660 EPA ( 360 ppm S) (10%) John Deere 660 EPA ( 360 ppm S) (50%) John Deere 660 EPA ( 360 ppm S) (50%) Volkswagen 204 EPA ( 360 ppm S) (0%) Volkswagen 213 Fisher-Tropsh ( 0 ppm S) (0%) For SMPS and MOUDI Comparison John Deere 360 EPA ( 360 ppm S) (50%) na na a DR, dilution ratio; DR1, first dilution ratio; DR2, second dilution ratio; na, not applicable. relationship is used to define the mass fractal dimension, D f (15): N ) C( R g a ) Df where N is the number of primary particles in the aggregate, R g is the radius of gyration, a is the radius of the primary particles, and C is a constant. If the primary particle size is constant, then N varies in proportion to particle mass, m. Furthermore, Schmidt-Ott et al. (16) have shown that the radius of gyration is linearly proportional to the mobility diameter in the continuum regime and in the free molecular regime if the fractal dimension is greater than 2. Our measurements were made in a regime where this should apply. It follows, therefore, that In principle, the effective density defined by eq 1 is related to the fractal dimension in eq 3 as follows: Since diesel particles are not spherical (D f < 3) in the size range that we studied, the effective density should decrease as mobility size increases. When we analyzed our data, we obtained a single overall value of D f for all particle sizes studied under given engine operating conditions using eq 3. Rather than using eq 4 to calculate effective densities with this value of D f, we independently obtained the effective density for each particle size (the method to determine the effective density is described in the following section). Another significant aspect of this work is that it enables us to rigorously intercompare size distributions measured with the SMPS and an impactor without the need to make any assumptions about particle shape, density, or composition. This was done by using APM measurements of particle mass to establish the relationship between number and mass (2) m ) C D me D f (3) F eff ) C D me D f -3 (4) distributions and by using the relationship between mobility and aerodynamic sizes measured with the DMA-APM system. Experimental Section Sampling and Dilution System for Diesel Exhaust. The diesel engines used in this study are a Volkswagen (Volkswagen TDI, 4 cylinder, 1.9 L, 66 kw) and a John Deere (John Deere 4045, 4 cylinder, 4.5 L, 75 kw). The matrix of measurements and engine operating conditions for our study are shown in Table 1. A schematic of the sampling and dilution system is shown in Figure 1. To realistically simulate the mixing of diesel exhaust into ambient air, the Variable Residence Time Dilution System (VRTDS) (21) was employed. The VRTDS consists of two sequential dilution stages. The residence time in the dilution system can be adjusted by controlling the retractable tube length. The hot diesel exhaust particles are sampled through an insulated and heated stainless steel tube to the primary dilution stage where they are diluted in an air ejector with clean cooled air from a compressed air line. During the first dilution stage, the diesel exhaust emissions undergo nucleation, adsorption, condensation, and coagulation (22). The exhaust is further diluted by clean dry air at the secondary dilution stage to obtain an appropriate concentration of the diesel exhaust particles; no additional particle formation processes occur during the secondary dilution process. Kittelson et al. (22) have shown that when diesel particles are emitted into the atmosphere from a moving vehicle, they undergo nucleation and condensation processes that are similar to those observed in this dilution system. Therefore, we believe that the particle properties observed in this study should be similar to those for diesel particles emitted to the atmosphere. The dilution ratio is monitored by measuring the ratio of NO x concentrations in the diluted exhaust to those in the undiluted exhaust. The dilution ratios for each operating condition are included in Table 1. APM. The APM consists of two cylindrical electrodes that rotate at the same speed. Charged particles are introduced ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 3, 2003

3 FIGURE 1. Schematic of sampling and dilution system for diesel exhaust particles. FIGURE 2. Schematic of the DMA-APM system. axially through the small annular gap between the two cylindrical electrodes and rotate at the same angular speed as the electrodes. A voltage is applied to the inner APM electrode with the outer cylindrical electrode grounded. Thus, particles experience centrifugal and electrostatic forces, which act in opposite directions. When these forces are balanced, particles penetrate through the APM and reach the downstream particle detector. Further details on the APM are described in previous papers (1, 14) DMA-APM System. A schematic of the DMA-APM system is shown in Figure 2. The diesel exhaust particles are first introduced to a Po-210 neutralizer to obtain a Boltzmann equilibrium charge distribution before they enter the DMA. A Collison atomizer is used to generate polystyrene (PSL) spheres, which are used as density standards. The DMA was operated at an aerosol flow rate of 1.49 L/min and a sheath flow rate of 10.3 L/min. For particles larger than 100 nm, the mobility-classified particles include a significant fraction of multiply charged particles. Our previous paper (14) showed that singly and doubly charged particles of a given electrical mobility are cleanly separated by the APM. For example, the APM voltage required to classify doubly charged particles is 2.4 times the value required for singly charged particles of the same density and mobility (300 nm mobility size). If the sampled aerosols were to contain an external mixture of particles with densities that differed by a factor of about 2.4, however, the doubly charged more massive particles would be classified by the APM at the same voltage as the singly charged less massive particles. In this case, distinguishing the distribution of masses would require a more comprehensive inversion of the DMA-APM data or the use of a charger that minimizes the production of multiply charged particles (23). By scanning the APM voltage, the mass distribution for the mobility-classified particles can be measured. For each APM voltage, their number concentration is measured by the CNC (TSI 3760), and the peak APM voltage corresponding to the peak mass for the mobility-selected particles is determined. From the measured peak APM voltage, particle mass is calculated with the following equation: qv APM m ) π 6 dve3 F true ) π 6 dme3 F eff ) r 2 ω 2 ln(r 2 /r 1 ) where m is the particle mass, r is the radial distance to the annual gap from the axis of rotation, ω is the APM rotational speed, r 1 is the inner radius, r 2 is the outer radius, q is the particle charge, and V APM is the APM voltage applied to inner electrode. The mobility equivalent diameter (d me) equals the (5) VOL. 37, NO. 3, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 579

4 FIGURE 3. Normalized concentration downstream of the APM vs the APM classifying voltage for PSL spheres and diesel particles. Measurements were done using particles with a mobility size of 107 nm, VW engine, 0% load, and 1250 rpm. volume equivalent diameter (d ve) for spherical nonporous particles and is greater than the volume equivalent diameter for agglomerate particles such as diesel soot (8). Therefore, the effective density (F eff) equals the inherent density (F true) for spherical particles and is less than the inherent density for diesel particles. Following the methodology described by McMurry et al. (14), the effective density of diesel particles was obtained by carrying out sequential measurements on diesel particles and polystyrene (PSL) spheres of the same mobility size. This methodology can reduce measurement uncertainties that would occur if the DMA flow rates and classifying voltage were used to calculate the mobility diameter. An example from such measurements is shown in Figure 3. The effective density is F eff,diesel )F true,diesel d ve,diesel 3 V APM,diesel d )F 3 true,psl (6) V me,diesel APM,PSL where F true,psl is 1.05 g/cm 3, and V APM,diesel and V APM,PSL are the APM classifying voltages corresponding to the peak-normalized concentration for PSL and diesel particles downstream of the APM. The effective density determined in this way for the diesel particle data shown in Figure 3 is 0.80 ( 0.02 g/cm 3. In obtaining densities using eq 6, the APM voltage at the peak of the concentration versus voltage curve is used. These N(V) relationships are highly repeatable; we have previously shown that this peak location can be used to measure densities within 4% (14). However, the peak widths are quite broad and are determined by the probability that particles are transferred through both the DMA and the APM for given DMA and APM operating conditions. We refer to this probability, which is determined by a convolution of the DMA (3) and APM transfer functions (1), as the DMA-APM transfer function. Ehara et al. (1) discussed the N(V) relationship for APM-classified PSL spheres. Transmission Electron Microscope (TEM). We took TEM images for particles of mobility sizes between 50 and 220 nm and analyzed them with an image analysis program (Digital Micrograph 3, Gatan Inc.). A low-pressure impactor (LPI) (24), which consists of 8 stages and has a minimum cut size of 50 nm, was employed to collect the DMA selected particles, which were collected on TEM grids (see Figure 2). Equation 1 was used to identify the impactor stage with the appropriate aerodynamic size cut. A JEOL 1210 TEM (accelerating voltage: kv, magnification: ) was used FIGURE 4. Mass-mobility diameter relationship of diesel particles for two different engine loads. The fractal dimension is obtained using eq 3 (John Deere engine, 10 and 75% loads, 1400 rpm, 360 ppm EPA fuel). to obtain the TEM images. Copper 200 mesh grids coated with carbon and silicon monoxide films were used as TEM substrates. Micro-Orifice Uniform Deposit Impactor (MOUDI). A MSP MOUDI (25) and a MSP nano-moudi were used to measure mass size distributions of diesel exhaust particles as a function of aerodynamic diameter. The nano-moudi was connected downstream of the regular MOUDI to measure particles smaller than 56 nm, the lowest cut size of the regular MOUDI. The nano-moudi has stages with aerodynamic cut points of 30, 20, and 10 nm. The total flow rates for the regular MOUDI and the nano-moudi are 30 and 10 L/min, respectively. The MOUDI measurements were done under the same engine operating conditions (John Deere, 50% load, 1000 rpm) as the SMPS measurements but were made at different times. Previous results have shown that aerosols produced under same operating conditions with this engine are reproducible. The average of three MOUDI distributions obtained under identical engine operating conditions was used to compare the SMPS measurements. To minimize particle bounce, greased substrates were used. Results and Discussion Mass-Mobility Relationships. The fractal dimension of diesel particles can be determined from the measured relationship between mobility size and mass (see eq 3). We measured this relationship at 10% and 75% engine loads over the size range from 50 to 300 nm. The results of these measurements are shown in Figure 4. The fractal dimension is obtained from the slopes of the observed log-log relationships (see eq 3). Note that the fractal dimension decreased with increasing engine load. The diesel particles at the lower engine load (10%) have a somewhat higher fractal dimension (2.41 ( 0.02) than at the higher engine load (75%, 2.33 ( 0.03). This suggests that particles are more compact at the lower engine load. We compare the fractal dimensions of diesel particles measured with the DMA-APM technique with values reported in other studies in Table 2. Skillas et al. (20) used a LPI and a DMA to determine the fractal dimension of diesel exhaust particles from mass-mobility relationship over the size range from 55 to 260 nm. The mass of the impacted particles was estimated from the mobility size and the critical Stokes number of the impactor that was calibrated with dioctyl sebacate (DOS). They found that the fractal dimension decreased with increasing engine load, but their results showed greater variability than ours. Virtanen et al. (13) ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 3, 2003

5 TABLE 2. Fractal Dimensions of Diesel Exhaust Particles from Various Studies engine type method engine load fractal dimenstion this study John Deere 4045 (75 kw) DMA-APM technique ( nm) 10% % % 2.33 ref 20 Yamaha EDA3000 (3.3 kw) DMA-impactor technique ( nm) 0% % % % % % 2.60 ref 13 Audi A4 TDI (66 kw) SMPS-ELPI technique normal road loads obtained the mass-mobility relationship by iteratively varying density to fit the ELPI and SMPS distributions. They showed that the fractal dimension varied from 2.58 to 2.84 at normal road load. It is also instructive to compare our measured values of fractal dimension with fractal dimensions obtained in modeling studies. Agglomerate formation in diesel engines occurs when gas densities are 5-15 times ambient values, so the mean free path during agglomeration (4-13 nm) is significantly smaller than the primary particle size. It follows that agglomeration in diesel engines is diffusion-controlled (6, 26). Our measured values of the fractal dimension ( ) are slightly below the value of 2.50 that is predicted for diffusion-limited particle-cluster agglomeration and significantly above the levels of that are predicted for diffusion-limited cluster-cluster agglomeration (26, 27). The models, however, do not account for vapor condensation, which may cause particles to be more compact than would occur if they only contained solid primary aggregates. The measured mass-mobility relationship also enables us to determine the effective density using eq 6. Effective densities of diesel particles measured with the DMA-APM technique in this study are shown in Table 1 and are compared with values reported for other engines in Figure 5. These results show that the effective densities decreased as the mobility equivalent size increased. As shown in Figure 5, results from other studies using different techniques and engines were qualitatively similar to ours in that the effective density decreased with increasing size. However, effective densities reported in the previous studies tended to be higher than ours, and several previous studies reported a distinct maximum in effective density at about 80 nm, which we did not observe. Maricq et al. (12) reported that the effective density varied by (20% depending on whether they attribute the particles collected on the upper impactor stages to truly aerodynamically large particles or to small particles that were deposited on these stages by electrostatic forces and diffusion. This would shift the aerodynamic distribution to larger sizes, leading to higher effective densities. It is also possible that the engines used in their studies produce particles with different physical properties. McMurry et al. (14) measured the effective density of 107 and 309 nm mobility diameter atmospheric aerosols (Atlanta, GA, August 1999) by using the DMA-APM technique. For particles of a given mobility size, they observed two distinct particle masses, one with a low effective density ( g/cm 3 ) and the other with a density of about 1.6 g/cm 3. They hypothesized that the low-density particles included carbonrich chain agglomerates produced from vehicular emissions. The average results for low-density particles from their study, shown in Figure 5, are in good agreement with the diesel particle densities from this study. We also examined the dependence of the effective density on engine load. As the engine load decreased from 75% to 10%, the effective density of the diesel particles increased somewhat as shown in Figure 5. For 50-nm particles, the FIGURE 5. Comparison of effective densities measured with the DMA-APM technique for the John Deere engine running at 10 and 75% loads and 1400 rpm with values measured using other techniques in previous studies. Method A of Ahlvik et al. (10) involved varying effective density to fit ELPI to DMPS distributions, and method B involved measuring the mobility diameter corresponding to the 50% collection efficiency of a given ELPI stage. Virtanen et al. (13) obtained effective densities by iteratively varying density so as to reconcile ELPI and SMPS distributions. effective density at 10% load was 1.20 g/cm 3 while it was 1.01 g/cm 3 at 75% load. There was not much difference between effective densities measured at 50% and 75% engine loads (see Table 1). The relative standard deviation of the effective densities as determined by repeated measurements was about 2.7% for 80-nm particles, which is less than the discrepancy between the measured effective densities at different engine loads. Therefore, we conclude that the reported sensitivity of effective density to engine load is valid. Ahlvik et al. (10) postulated that the variations in the effective density of diesel exhaust particles are due to the dependence of morphology on mobility size. In this study, we tested this hypothesis by examining TEM images of mobility-selected particles at several engine loads. Evidence for the relationship between particle shape and effective density is shown in Figures 6 and 7. Figure 6 shows TEM images of five typical diesel particles ranging in mobility diameter from 50 to 220 nm at 50% load. Over this size range, the diesel particles are chain agglomerates consisting of primary particles whose size lies between 20 and 40 nm. These TEM images show that the smallest particles (50 nm), which had the highest effective density, are the most compact in shape. Note that the 50-nm particles appear to be coated VOL. 37, NO. 3, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 581

6 FIGURE 8. Aerodynamic diameter of diesel particles as a function of the mobility equivalent diameter for John Deere engine (50% load, 1400 rpm, 360 ppm EPA fuel). FIGURE 6. TEM images of mobility classified diesel particles produced by John Deere engine (50% load, 1400 rpm, 360 ppm EPA fuel): the mobility sizes are shown on the images. FIGURE 7. TEM images of particles having a mobility size of 220 nm produced by the John Deere engine (10% load, 1400 rpm, 360 ppm EPA fuel). with condensed material, which leads to their compactness. As size increases, particles become more irregular and agglomerated, which is consistent with the lower effective density. Figure 7 shows TEM images of particles with mobility equivalent diameter of 220 nm generated at a low engine load (10% load). Note that these particles appear somewhat more compact than the 220-nm particles produced at a high engine load (see Figure 6). The necking that is observed between primary spheres suggests that these particles may be coated with condensed materials. We believe this may have occurred because much more volatile species (e.g., unburned fuel or lubricating oil) are available to adsorb or condense on soot at low engine loads (22). Data from the engine manufacturer suggest that, for a typical engine of this type operating at intermediate speed, the volatile organic fraction decreases with increasing load from about 60% at 10% load to about 15% at 75% load. A limitation of electron microscopy is that volatile species tend to evaporate from the particles when they are illuminated by the electron beam. We have not systematically investigated the effect of evaporation on particle morphology. The images suggest, however, that the gaps between primary particles are filled in, especially for the particles collected at the low engine load (Figure 7) where higher levels of semivolatile compounds are expected. We hypothesize that this is because vapors condensed on those particles. The measurements discussed above, which pertain to the effect of engine load and particle size on effective density, were done using the John Deere engine. We investigated the effect of fuel sulfur content with the Volkswagen engine. When high-sulfur fuel (EPA fuel, 360 ppm S) was used, we observed higher effective densities than for low-sulfur fuel (Fisher- Tropsh fuel, 0 ppm S), as shown in Table 1. For the 83-nm particles, the effective density was 1.10 g/cm 3 with highsulfur fuel, while it was 0.90 g/cm 3 with the Fisher-Tropsh fuel that contained no sulfur. This could occur because sulfuric acid produced from the fuel sulfur during combustion condenses on the diesel soot. This would increase the effective density because sulfuric acid is inherently denser than the condensed organic compounds (e.g., lubricating oil) and because liquid sulfuric acid would make the particles more compact. Also, the Fischer-Tropsch fuel has a lower density than the EPA fuel, and this may also affect the density of the exhaust particles. Comparisons of Mass Distributions Measured with the SMPS and MOUDI. The relationship between aerodynamic and mobility equivalent diameters of diesel exhaust particles for the John Deere engine is shown in Figure 8. We determined this relationship using the measured mass and mobility equivalent diameter with the DMA-APM (see eq 1). Note that the difference between the mobility diameter and the ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 3, 2003

7 Protection Agency through Grant R to the Georgia Institute of Technology and GIT Subcontract G-35-W62-G1 to the University of Minnesota, it has not been subjected to the Agency s required peer review and therefore does not necessarily reflect the views of the Agency, and no official endorsement should be inferred. FIGURE 9. Mass size distributions measured with the SMPS, the MOUDI, and the nano-moudi for the John Deere engine (50% load, 1000 rpm, 360 ppm EPA fuel). MOUDI data are the average of three distributions obtained under identical engine conditions. The error bars represent the standard deviation of these measurements. aerodynamic diameter increases with increasing mobility diameter because the effective density decreases as the mobility diameter increases. For 50-nm particles, the aerodynamic diameter is approximately equal to the mobility diameter because the effective density of 50-nm particles is around 1 g/cm 3. We used a MOUDI and a nano-moudi to measure mass distributions and a SMPS to measure number distributions of diesel particles. The MOUDI classifies particles according to aerodynamic size, and the SMPS classifies particles according to mobility size. These sizes can differ significantly (see Figure 8), so size distributions measured by different instruments can appear to be quite different. Using the relationship between mobility size and aerodynamic size shown in Figure 8, the MOUDI mass distributions as a function of aerodynamic size were converted to mass distributions as a function of mobility size as shown Figure 9. Figure 9 also shows mass distributions obtained by the SMPS-APM technique. The MOUDI data provide evidence of a small peak around 20 nm, which is not observed by the SMPS. We hypothesize that this peak is due to particle bounce from upper stages of the impactor or some other artifact formation in the low-pressure stages. The large error bar at 20 nm, however, reflects the large variability of our three mass measurements on this stage, suggesting that bounce may have occurred for some but not all measurements. Note also that the SMPS provides much higher size resolution than the MOUDI, but the upper size limit of the SMPS distribution is 500 nm, while the MOUDI measurements extend up to 10 µm. Acknowledgments Although the research described in this paper has been funded wholly or in part by the United States Environmental Literature Cited (1) Ehara, K.; Hagwood, C.; Coakley, K. J. J. Aerosol Sci. 1996, 27, (2) Liu, B. Y. H.; Pui, D. Y. H. J. Colloid Interface Sci. 1974, 47, (3) Knutson, E. O.; Whitby, E. R. J. Aerosol Sci. 1975, 6, (4) Hering, S. V.; Stolzenberg, M. R. Aerosol Sci. Technol. 1995, 23, (5) Kütz, S.; Schmidt-Ott, A. J. Aerosol Sci. 1990, 21, S47-S50. (6) Schleicher, B.; Kunzel, S.; Burtscher, H. J. Appl. Phys. 1995, 78, (7) Stein, S. W.; Turpin, B. J.; Cai, X.; Huang, P.; McMurry, P. H. Atmos. Environ. 1994, 28, (8) Kasper, G. Aerosol Sci. Technol. 1982, 1, (9) Allen, M. D.; Raabe, O. G. J. Aerosol Sci. 1982, 13, (10) Ahlvik, P.; Ntziachristos, L.; Keskinen, J.; Virtanen, A. SAE Tech. Pap. 1998, No (11) Kelly, W. P.; McMurry, P. H. Aerosol Sci. Technol. 1992, 17, (12) Maricq, M. M.; Podsiadlik, D. H.; Chase, R. E. Aerosol Sci. Technol. 2000, 33, (13) Virtanen, A.; Ristimaki, J.; Marjamaki, M.; Vaaraslahti, K.; Keskinen, J. SAE Tech. Pap. 2002, No (14) McMurry, P. H.; Wang, X.; Park, K.; Ehara, K. Aerosol Sci. Technol. 2002, 36, (15) Mandelbrot, B. The Fractal Geometry of Nature; Freeman: San Francisco, (16) Schmidt-Ott, A.; Baltensperger, U.; Gaggeler, H. W.; Jost, D. T. J. Aerosol Sci. 1990, 21, (17) Samson, R. J.; Mulholland, G. W.; Gentry, J. W. Langmuir 1987, 3, (18) Weber, A. P.; Baltensperger, U.; Gaggeler, H. W.; Schmidt-Ott, A. J. Aerosol Sci. 1996, 27, (19) Rogak, S. N.; Flagen, R. C. Aerosol Sci. Technol. 1993, 18, (20) Skillas, G.; Kunzel, S.; Burtscher, H.; Baltensperger, U.; Siegmann, K. J. Aerosol Sci. 1998, 29, (21) Abdul-Khalek, I.; Kittelson, D. B.; Brear, F. SAE Tech. Pap. 1999, No (22) Kittelson, D. B. J. Aerosol Sci. 1998, 29, (23) Gupta, A.; McMurry, P. H. Aerosol Sci. Technol. 1989, 10, (24) Hering, S. V.; Friedlander, S. K.; Collins, J. J.; Richard, L. W. Environ. Sci. Technol. 1979, 13, (25) Marple, V. A.; Rubow, K. L.; Behm, S. M. Aerosol Sci. Technol. 1991, 14, (26) Schaefer, D. W. Mater. Res. Soc. Bull. 1988, XIII, 22. (27) Meakin, P. On Growth and Form; Martinus Nijhoff: Dordrecht, The Netherlands, Received for review July 11, Revised manuscript received November 7, Accepted November 18, ES025960V VOL. 37, NO. 3, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 583