Hydrocarbon Condensation in Heavy-Duty Diesel Exhaust - American

Aug 10, 2007 - imply that condensation, or diffusion-limited mass transfer, plays a major role in driving the volatile matter to the diesel exhaust pa...
0 downloads 0 Views 173KB Size
Download PDF
Environ. Sci. Technol. 2007, 41, 6397-6402

Hydrocarbon Condensation in Heavy-Duty Diesel Exhaust J Y R K I R I S T I M A¨ K I , † , § K A T I V A A R A S L A H T I , †,⊥ M A I J A L A P P I , ‡ A N D J O R M A K E S K I N E N * ,† Aerosol Physics Laboratory, Institute of Physics, Tampere University of Technology, P.O. Box 692, FI-33101 Tampere, Finland, and Emission Control, VTT Technical Research Centre of Finland, P.O.Box 1000, FI-02044 VTT, Finland

The semivolatile mass fraction of diesel exhaust particles was studied using size-resolved on-line techniques (DMAELPI; TDMA-ELPI). The average density of the semivolatile liquid on the particles was measured to be approximately 0.8 g/cm3. The measured size resolved values of mass transfer imply that condensation, or diffusion-limited mass transfer, plays a major role in driving the volatile matter to the diesel exhaust particles. The measured mass change values correspond to highly size dependent mass fractions for the semivolatile component, ranging from approximately 20-80%. Integrated over particle size distribution, the volatile mass fractions were 25 and 45% for the two load points studied. Calculation, based on the measured particle properties, indicates that only 10% volatile mass fraction could be explained by monolayer adsorption. The size resolved changes in particle effective density, fractal dimension, volatile mass fractions and mass are all in agreement with theoretical considerations of condensation.

1. Introduction Correlation between the ambient particulate matter (PM) mass and adverse health effects has been reported in several studies, as reviewed, e.g., by Brunekreef and Holgate (1). As no threshold value for the ambient PM has been found (1), there has been continuous decrease in the regulative PM emission limit values. The stringent emission limits make mass measurement more challenging as the filter artifacts may become significant compared to the soot mass (2). One interesting approach to increase the sensitivity is being developed in the UNECE particulate measurement program (PMP, Particle Measurement Programme; Transport division of United Nations Economic Commission for Europe (UNECE); http://www.unece.org) where, in addition to filtered mass, the number of solid particles is measured with a condensation particle counter (CPC). However, as the diesel PM mass consist of both solid and volatile mass fractions, the number of solid particles does not necessarily correlate well with the weighted total mass. Considering the adverse health effects, the PMP may provide the right approach as the adverse health effects have been found to increase with * Corresponding author phone: +358-3-31152676; fax +358-331152600; E-mail: [email protected]. † Tampere University of Technology. ‡ VTT Energy and Pulp&Paper. § Current address: Wa ¨ rtsila¨ Finland Oy, Ja¨rvikatu 2-4, P.O.Box 244, FI-65101 Vaasa, Finland. ⊥ Current address: Ecocat Oy, Vihtavuorentie 162, P.O.Box 20, FI-41331 Vihtavuori, Finland. 10.1021/es0624319 CCC: $37.00 Published on Web 08/10/2007

 2007 American Chemical Society

decreasing particle size even for inert solid particles (3, 4). However, it should be kept in mind that the diesel exhaust PM is comprised of externally mixed particles when the nucleation mode is present: some particles are completely volatile while some have solid core (5, 6). Additionally, the nucleation mode has frequently been observed in the roadside measurements (7, 8). In the presence of a nucleation mode, the total number count is controlled by the volatile nanoparticles not measured in the PMP protocol. It has been found that the nucleation-mode particles comprise organic or volatile compounds of lubricating oil (9).The volatile fraction of diesel PM may be the key component causing oxidative stress in living cells (10). As large particles can accommodate significant amounts of volatile mass, the soot mode particles could still dominate the adverse health effects caused by the volatile components. Therefore, the composition of the soot mode particles should also be known for correct evaluation of the adverse health effect potential of diesel PM. It is commonly stated that the volatile mass has transferred to diesel exhaust particles by adsorption and condensation. More detailed characterization of the processes is lacking. Size dependent volatility of diesel exhaust particles has been studied using volatility tandem-differential mobility analyzer (TDMA) (6, 11), with estimations for volatile volume fractions. However, the TDMA measurement assuming spherical particles underestimates the volatile volume fraction in the irregularly shaped accumulation mode particles, as stated by Sakurai et al. (5) They report the first size resolved mass fractions of volatile components measured by the TDMAAPM on-line method (APM stands for aerosol particle mass analyzer). In their measurements, a nucleation mode was present, resulting in an externally mixed aerosol consisting of both more and less volatile particles. In this paper, we concentrate on the accumulation particles. By measuring the aerodynamic diameter of TDMA classified particles with an electrical low-pressure impactor (ELPI), we characterize the volatile mass fraction of particles with this TDMA-ELPI method (12). While faster to operate, this method does not possess as good a mass resolution as the TDMA-APM. Consequently, we obtain average values for each mobility diameter. We therefore study cases where the nucleation mode is absent to avoid external mixtures of highly different particles. Based on the effective density, the method enables the determination of mass changes of diesel exhaust particles. The size resolved mass change of the particles is compared with simulation values to study the role of the adsorption and condensation processes. Additionally, the density of condensed matter is measured.

2. Experimental Section Tests were carried out with a 6-cylinder, 10 liter diesel bus engine (model year 1996, EURO II level) with and without oxidation catalyst (Finnkat FK23D). The fuel sulfur content (FSC) was analyzed to be 8 ppm. The lube oil sulfur level was analyzed to be 9330 ppm. More information of the engine and the lube oil can be found in ref 13. Two test points (Table 1) were used: 10% load (87 Nm 2000 rpm) and 25% load (245 Nm at 1790 rpm, mode 11 of the European Steady Cycle, ESC). The exhaust gas temperature was measured upstream of the sampling and dilution point, and the values were approximately 210 and 250 °C for the 10 and 25% load points, respectively. The 10% load point was measured because of other tests made with the engine at the same time. The 25% load point was selected based on knowledge from previous studies (13) that the nucleation VOL. 41, NO. 18, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

6397

FIGURE 1. Measurement system setups for a 10 and 25% load. Dashed arrow-lines present alternative piping for the second scanning mobility particle sizer (SMPS) or differential mobility analyzer (DMA) (DMA2 /SMPS).

TABLE 1. Test Matrix engine parameters

10%, load 25% load

primary target

8 Nm, effective 2000 rpm density + oxidation catalyst 245 Nm, 1790 rpm (ESC11)

secondary measurement target method mass transfer

density of mass volatile PM transfer

DMA-ELPI

TDMA-ELPI

mode is about to appear at this load point for this engine. In the present study, the absence of the nucleation mode was checked for each measurement. Diesel exhaust was first diluted with a porous tube primary dilutor with nominal dilution ratio of 12 as in the EU funded PARTICULATES2 program (Characterization of exhaust particulate emissions from road vehicles; Growth program, http://lat.eng.auth.gr/particulates/) (14, 15). Primary dilution was followed by secondary dilution with one or two ejector pump diluters. The actual primary dilution ratio was in the range of 8-20. The variation is due to small leak in the gas analyzer pump. This variation is not considered to have affected the results as the dilution ratio was stable in each measurement. The nominal dilution ratios of the ejector pumps were 8. The dilution air temperature was ∼26 °C and relative humidity was less than 5%. The overall system setup is presented in Figure 1. We used two measurement set-ups for the online instruments. In the first, differential mobility analyzer (DMA) classified particles are passed into an ELPI. This setup (DMAELPI) is a variant of the DMA-impactor method and measures the effective density of the particles (16-18). This method was used in studying the effect of the oxidation catalyst on the effective density of diesel exhaust particles at the 10% load point. In the second setup, the TDMA-ELPI method (12) was used to study the mass transfer of volatile matter to diesel exhaust particles at the 25% load point. In these measurements a thermodenuder (Dekati Ltd) was installed as an aerosol conditioner. The temperature of the thermodenuder was ∼270 °C. This setup measures the change in particle density and mass caused by the evaporation of semivolatile compounds in the thermodenuder. As a result, the density of the evaporated liquid can also be calculated. The error percentages in the results section for the measured effective density and mass change are approximated to be ∼10 and ∼20% respectively (95% confidence limits) (12). These values are used to estimate the detection limits in the figures.

3. Theoretical Considerations The measurements provide the changes in mobility diameter, effective density, and particle mass caused by transferred volatile material. It is, however, not easy to identify the 6398

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 18, 2007

FIGURE 2. Measured effective density values for 10% load, with and without oxidation catalyst (markers). The solid lines show the fractal law dependence of density on diameter (PP refers to primary particle size, df to fractal dimension, and G to bulk density). The dashed line shows density after calculated C28 monolayer adsorption on the soot particles having fractal dimension of 2.3. underlying processes as these quantities are linked in a complicated way for agglomerates growing by adding liquid material of different density. We next try to cover the theoretical aspects related to the processes, making numerous simplifications on the way. We hope this is justified as we are only looking for qualitative explanation of the measured results. 3.1. Structure of Diesel Exhaust Particles. The fractal dimension concept is commonly used to characterize the structure of aerosol particles (19-21), including diesel exhaust particles (22). The fractal dimension (df) scales the particle mass to the particle mobility diameter (22, 23). As the effective density (Feff) can be regarded as the ratio of mass to volume of mobility equivalent particle, the fractal dimension can also be used to scale effective density as a function of mobility diameter (eq 1).

log(Feff) ) -(3 - df)log

( )

dmob + log(Fref) dref

(1)

The familiar power law (Feff ∝ dmobdf-3) is presented above in logarithmic form to remain readable after introducing dref as the mobility diameter at a reference point. The reference point is selected at primary particle diameter (dref ) dpp) and bulk density (Fref ) Fbulk). This includes the (over)simplification that the fractal law can be extended down to the primary spheres. Based on electron microscope pictures, we have used 18 nm for the primary particle diameter and 1.8 g/cm3 is used for the bulk density of the solid matter24. Additionally, we have used 2.3 as a value for the fractal dimension. While this is the same number as reported by Park et al. (25), it is also the value measured for the 10% load case (see Figure 2). Interstitial volume exists within the structure of the agglomerates. This volume is highly dependent on the way the total volume of the particle is calculated. To obtain a first order approximation of interstitial volume, we estimate the total free volume (Vfree) formed as a function of mobility diameter by subtracting the volume of the mass equivalent particle from the mobility equivalent volume (eq 2).

Vfree ) Vmob - Vmass ) Vmob -

(

Feff m π ) d 3 1Fbulk 6 mob Fbulk

)

(2)

Here, m is the particle mass. We estimate that the particle becomes spherical-like when the transferred material volume exceeds the void volume. After this point, the mobility diameter should simply grow as the cube root of the transferred volume.

3.2. Transferring Species. Recent mass spectrometer analysis (9) indicates that most of the volatile compounds of diesel particles are heavy hydrocarbons. Although the particles are bound to contain some sulfate, we make the approximation that the volatile mass consists only of hydrocarbons. We use low sulfur fuel (less than 10 ppm) and low engine loads, which restrain sulfate formation. The total hydrocarbon content (THC) consists of a pool of hydrocarbons. Taking into account all transferring species would be impossible. On the other hand, a wide range of parameters can be used to obtain similar mass transfer curves, and we are only looking for visual agreement between the theoretical curves and measured values. Therefore, we use approximate values to describe the hydrocarbon properties. Based on the results of Sakurai et al., we approximate the properties of C28 hydrocarbon. We have used the parameters presented in the Supporting Information. Note that the values are presented only with one significant digit. The value for surface tension (0.02 Ns/m2) is typical for hydrocarbons, as is the value for density (0.8 g/cm3). The diffusion coefficient (0.002 cm2/s) is based on extrapolation of the results presented by Elliott and Watts (1971, ref 26). The mean free path (40 nm) is calculated using diffusion coefficient and molecule mass. The mean free path is the only quantity directly affecting the size dependence of the mass transfer. As the mean free paths for C24-C32 hydrocarbons differ only by some 10%, the choice of the hydrocarbon is not very critical for our purposes. 3.3. Mass Transfer. Adsorption, condensation, and evaporation are the main phenomena resulting in mass transfer to and from particles in the dilutor. We use basic approaches for adsorption and condensation as they are enough for the comparison between measurements and basic theories. To simplify the treatment, evaporation is not taken into account. In practice, evaporation is, at least to some extent, suppressed by the limited dilution ratio used in the experiments. 3.3.1. Adsorption. Sorption mechanisms in a multicomponent system may be complicated (see Pankow, 1994, ref 27). However, we consider a single adsorbing component on inert solid particles and take into account physical adsorption only. Adsorption may take place already within the hot exhaust, as supersaturation is not necessary. It is assumed that the first molecule layer(s) are always transferred through adsorption. If adsorption is the main mass transfer phenomenon, the transferred mass should be related to the surface area (A) of the particle and few monolayers would be formed (28). Unfortunately, the nonspherical structure of diesel soot complicates the relation between particle mobility diameter and surface area. However, one estimate of the upper limit for the surface area available (Asa) can be obtained by multiplying the surface area of the primary particle with the number of primary particles (Npp). In turn, the number of primary particles can be estimated from the particle mass with the help of primary particle diameter (dpp) and bulk density of the primary particle (Fbulk):

Asa ) Nppπdpp2 Npp )

mtotal - mvolatile mtotal - mvolatile ) mpp dpp3 πFbulk 6

(3) (4)

Here F refers to density, subscript “volatile” to volatile mass, and “total” to total mass of the agglomerate, whereas mpp refers to the mass of the primary particle. The number of adsorbed molecules is estimated by dividing the surface area of the particle by the 2D-projection area of a single molecule. The diameter of the molecule is estimated from the molecule mass and the bulk density. This

approach relates the volatile mass to the number of primary particles and yields a constant volatile mass fraction regardless of particle size. It is not clear how the adsorbed layer affects the mobility diameter, but as a first approximation we assume that the mobility radius grows by the thickness of the adsorbed layer. 3.3.2 Condensation. Although condensation could occur already in the exhaust line, it is fair to assume that it is most effective during the dilution when the exhaust gas cools down. Condensation mass transfer is diffusion limited once the quasi-steady-state concentration is reached. We restrict ourselves to this case, as the time scale to reach quasi-steady state is less than picoseconds for the size range of soot particles (29). In this case, calculation of condensational growth is rather straightforward and can be found in many textbooks (29, 30). We have used the Fuchs-Sutugin approach to calculate the flux of molecules to or from particle (J) as a function of the concentration (c) of the condensing species or the concentration difference ∆c:

J ) 2πdDgβ(c∞ - cs) ) 2πdDgβ∆c

(5)

Here β is the Fuchs-Sutugin correction factor depending on the mean free path of the hydrocarbon molecules. Subscript “g” refers to transferring gas, “∞” values far away from particle and “s” to surface. It is possible that Kelvin effect limits condensational growth of the particles. It is even possible that small particles first grow, and as the saturation ratio decreases, then evaporate again. We again simplify, and only take the Kelvin effect into account as limiting the growth by decreasing the effective concentration difference.

(

∆c ) c∞ - csat exp

( )) 4σMg RTFld

(6)

Here σ is the surface tension, M is the mole mass of the condensing species, R is the gas constant, and T is temperature. Subscript “l” refers to liquid phase, “sat” to saturation concentration, and “g” to condensing matter. Condensation growth is only calculated for positive values of ∆c. In simulation, we assume that adsorption precedes condensation, always forming a molecule layer. The mobility diameter and particle mass are increased accordingly. As particles grow by condensation, the mobility diameter will also grow. For simulation of condensation, we assume that there is growth in the diameter only after the condensed volume exceeds the estimated free volume of the agglomerate. This is obviously an underestimation. However, the molecule flux is not very sensitive to small changes in diameter. The initial concentration and the saturation concentration of the condensing species are treated as variables to obtain the best fit with the experimental data. The initial concentration is always lower than the measured THC value. Note that eqs 5-6 are for spherical particles, which is not the case with diesel exhaust particles. As diesel soot agglomerates are comprised of primary particles, it could be argued that all particle sizes equally experience the Kelvin effect. However, the concave structures in the particle connections are likely to make things more complicated. We will simply use mobility diameter to placed in the equations and emphasize that we are looking for qualitative agreement with measured results. In contrast to adsorption, condensation can result in a more complicated volatile mass fraction (VMF) dependence on particle size. Neglecting the Kelvin effect, condensation causes a 1/d dependence of volatile mass fraction: VOL. 41, NO. 18, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

6399

VMF )

mcond d 1 ∝ ) mcore + mcond ddf + d ddf-1 + 1

(7)

where mcore refers to the mass of the core particle and mcond to the condensed mass. One should note that the Kelvin effect can change this dependence for small particles. Mass transfer to particles also changes the dependence of density on particle diameter. For small agglomerates or heavy condensation, the particles will lose their fractal structure and the fractal dimension should tend to three. For large agglomerates it is possible that a major portion of the volatile matter fill the voids of the particle while causing only limited growth of the mobility diameter. Neglecting diameter growth and the Kelvin effect, the effective density after condensation would be as follows:

mcore + mcond FeffVmob + JMg∆t Ζ Feff_new ) ) ∝ Feff + Vmob Vmob d

2

mob

Ζ ) 12Dg∆c

(8)

If the size dependence of density is used as a basis of fractal dimension calculation (eq 1), this causes a counter-intuitive effect: for large agglomerates, condensation causes the fractal dimension to decrease (eq 8).

4. Results and Discussion 4.1. Effect of After-Treatment on Soot Particles. The effect of exhaust gas after-treatment on the particles was studied with the DMA-ELPI method. The size distribution and effective density of the particles both without after-treatment and with the oxidation catalyst have been presented earlier (31). Here, we will take a closer look at the change of effective density and calculate the changes in particle mass. 4.1.1. Change in Effective Density. Figure 2 shows the measured effective density values as a function of mobility diameter for the soot particles both without after-treatment and with the oxidation catalyst. The effective densities above 40 nm are considerably higher without the catalyst. The effect of monolayer adsorption was tested by simulation calculation using the parameters given above. The simulation result is presented by the lines in Figure 2. Particles having an effective density higher than that of the adsorbed hydrocarbon, experience a decrease in effective density (F1), whereas particles having effective density below the hydrocarbon density experience increase in the effective density (F2). This kind of monolayer adsorption could explain why it is common that effective densities, reported in many measurements (17, 18, 25, 32-34) seem to be below 1.8 g/cm3 even if extrapolated to primary particle size (18 nm). However, it can also be seen from Figure 2 that the measured change in density is much larger than that caused by the monolayer adsorption. As significantly more mass has transferred to particles, condensation is suggested to take place. Additionally, we also observe a decrease in the fractal dimension (from 2.3 to 2.0) as suggested by eq 8. However, we also see that for the smallest particles the fractal dimension is close to 3. This can be caused by particle growth to resemble spheres. 4.1.2. Change in Mass. The changes in the number distributions with and without catalyst are negligible (See Figure 1 in ref 31) and within the experimental error. We estimate that the maximum mobility diameter growth is 3 nm. Earlier, Virtanen also observed practically no change in measured SMPS number distribution caused by thermal treatment, although there was a change in ELPI distribution (35). Scheer et al. have briefly discussed their TDMA results where only a 2-3 nm change in the diameter of soot particles was observed due to thermal treatment (11). 6400

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 18, 2007

FIGURE 3. Measured mass change, calculated from particle mass values with and without oxidation catalyst (markers). Small markers refer to points that are close to the detection limit. The solid lines show estimated mass change values caused by monolayer adsorption of C28 (green) and by Kelvin effect limited condensation (orange line). The dashed lines indicate the estimated free volume and detection limit. 10% load point. If zero mobility size growth is assumed, it is possible to calculate the change in mass caused by the condensing species from the two sets of effective density values shown above. The mass change is presented as a function of diameter in Figure 3. For comparison, mass change caused by monolayer adsorption of C28 on primary particles is also presented as a solid green line. The measured mass change values are almost an order of magnitude higher than the adsorption curve for particles larger than 80 nm. It seems evident that most of the mass change is caused by condensation. The curvature of the measured curve below 80 nm suggests that condensation has been limited by the Kelvin effect. The solid red line shows a simulated mass transfer curve for the Kelvin effect limited condensation. With fitted saturation ratio, the condensation curve follows the measured values nicely, but not below 60 nm. These few points, on the other hand, agree with monolayer adsorption of C28. The measured mass change could, therefore, be explained by monolayer adsorption followed by condensation that is limited for smallest particles by the Kelvin effect. One should note that this simulation relies on zero growth of mobility diameter by condensation. However, within experimental error of the SMPS measurements it is possible that part of the finest particles have grown by more than the 3 nm. This would result in an externally mixed aerosol, as reported by ref 5. In this case, the present measurement method would give incorrect (low) average effective densities. This error would shift the calculated mass change values downward. It is, therefore, possible that the Kelvin effect is not as strong as it appears. 4.2. Size Dependence of Volatile Mass. The TDMA-ELPI method is much slower than the DMA-ELPI method used above. However, there is no need to assume anything about the change in mobility diameter as it is measured by the TDMA. Figure 4 shows the difference in particle mass caused by the thermodenuder treatment. The mass change is plotted as a function of “dry” diameter, which is the mobility diameter after thermal treatment. The mass change is assumed to be the same as the mass increase of soot particles caused by adsorption and condensation of volatile material, i.e., the volatile mass of a single particle. Figure 4 shows the results of two measurement series at 25% load point. For series 2, the dilution air flow was doubled from Series 1 (∼25 Lpm to 50 Lpm) while keeping the dilution ratio at 12. At the same time, the residence time after the dilutor was increased from ∼1 to ∼2.5 s. The change in dilution dynamics doubles the volatile mass for most of the

FIGURE 5. Volatile mass fraction values as a function of particle size for the measured mass changes (Figure 4). The lines show volatile mass fraction calculated using the simulated mass change. FIGURE 4. Measured mass change caused by thermal treatment (markers) for two series with different dilution dynamics. Small markers refer to points that are close to the detection limit. Solid lines show simulated mass change values caused by condensation. The dashed lines indicate the estimated free volume and detection limit. 25% load. size range. However, for smallest particles the difference is tenfold. The main difference caused by the different dilution dynamics is assumed to be in instantaneous saturation ratio. Admitting that this might cause a difference in the total amount of adsorption, we are reluctant to believe that it would change the size dependence of the adsorbed mass. We, therefore, interpret that the difference is caused by change in mass transfer by condensation and further that condensation is responsible for most of the transferred mass. This is in contradiction with some older studies (36, 37) suggesting adsorption contributes most to volatile PM without after-treatment. The solid lines again show simulated mass change values for condensation. At least qualitative agreement with the measured points is obtained. Different gas concentrations of the hydrocarbon were used for the two series. The high concentration (high saturation ratio) for Series 2 causes substantial growth of the smallest particles, as the estimated free volume is exceeded. For Series 1 the saturation ratio was lower. This causes the Kelvin effect to stop mass transfer to the smallest particles in the simulation. Note that this is a result of the chosen hydrocarbon properties. By changing these properties, a simulation curve that would fit equally well with the measurements but would show no Kelvin effect could probably be obtained. The measured mass transfer values for the particles larger than 90 nm within Series 2 are much higher than those predicted by simulation. However, this could be explained by the fact that in reality there are several condensing species. Our setup is optimized for automated calculation of particle density and mass and not for the measurement of mobility diameter change. Therefore, the diameter growth data is rather noisy. However, for Series 1, the mobility diameter change was approximately 4 nm, with slightly smaller values for larger particles. This is in line with the findings of Scheer et al. (11) For Series 2, we measured 10 nm changes for dry diameter above 50 nm, whereas below 50 nm the diameter change increased with decreasing dry diameter, reaching 16 nm at 30 nm. Compared to the standard filter measurements, DMA based methods (TDMA-ELPI, TDMA-APM, ref 5) enable the separation of volatile mass fraction as a function of particle size. This is shown in Figure 5 for the measured points of Figure 4. It can be seen that the volatile mass fraction decreases as a function of particle size as suggested by eq 7. Note the steep decrease in the volatile mass fraction with

increasing particle size for Series 2. The estimations in Figure 5 are based on the measured effective densities and the condensation curves presented in Figure 4. Note that for Series 1, the difference between the measured points and the estimated values originates from the measurement error of solid mass. In the case of bias error in the measurement, we would still obtain the correct mass change (Figure 4) but get error in the volatile mass fraction. Such a bias error could result, e.g., from errors in impactor calibration. As the experimental details are different, the value of direct comparison to the TDMA-APM results is limited. Nevertheless, the volatile mass fraction values are of the same order as those reported by Sakurai et al. (5) for the size range of 70-200 nm. The 3 nm mobility diameter growth reported by Scheer at al. (11) can also be used for comparison. Assuming that their soot agglomerates consisted of spherical primary particles having diameters of 18 nm, the diameter growth would result in a volatile mass fraction of 23%. Finally, the volatile mass found in the particulate phase corresponds to 0.23 and 0.17 ppmv of C28 in the gas phase at the 10 and 25% load points (Series 2), respectively. In turn, these values correspond to 0.5-0.6% of the measured THC (50/30 ppm). The volatile mass fraction can also be compared to the calculated volatile mass fraction resulting from adsorption. Monolayer adsorption of C28 on 18 nm primary particle surface area (Figure 2) would yield only ∼10% volatile mass fraction. One should note that the volatile fraction would in this case be very weakly dependent on particle size. We conclude that the fraction exceeding approximately 10% has condensed on particles. 4.3. Density of Volatile Matter. As the mobility diameter of the smallest particles for Series 2 (Figure 4) increases by 20%, they are assumed to become practically spherical. In which case the density of the condensing matter can be calculated from the TDMA-ELPI measurement (12) if some bulk density for the solid particle can be justified. Based on previous studies, we have used 1.8 g/cm3 for the bulk density of the solid material (24). For the two lowest dry diameter points, we obtain an average value of 0.8 g/cm3. For comparison, the density of the lubricating oil (0.89 g/cm3), diesel fuel, and C28 are in the same range. This agrees with the mass spectrometer analysis (9) indicating that diesel nanoparticles comprise mostly of unburned lubricating oil. Note that assuming an 18 nm primary particle size, the larger, 30 nm particles, used in the calculation are formed of only few of these primary particles. Therefore, these particles could still be tubular in shape, and it cannot be taken for granted that these particles have become spherical. This may result in a too low-density value for the evaporated matter.

Acknowledgments This research was funded by Tekes, the Finnish Funding Agency for Technology and Innovation and Finn Katalyt Ltd. VOL. 41, NO. 18, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

6401

We appreciate the help of Mr. Hannu Vesala and Mr. Timo Murtonen from VTT Energy and Pulp&Paper.

Supporting Information Available Parameters describing the hydrocarbon species used in the calculation of the simulation curves. This material is available free of charge via the Internet at http://pubs.acs.org.

Literature Cited (1) Brunekreef, B.; Holgate, S. Air pollution and health. Lancet 2002, 360, 1233-1243. (2) Chase, R.; Duszkiewicz, G.; Richert, J.; Lewis, D.; Maricq, M.; Xu, N. (2004) PM Measurement Artifact: Organic Vapor Deposition on Different Filter Media; SAE paper 2004-01-0967; SAE International: Washington, DC, 2004 (3) Brown, D.; Wilson, M.; MacNee, W.; Stone, V. Donaldson, K. Size-dependent proinflammatory effects of ultrafine polystyrene particles: A role for surface area and oxidative stress in the enhanced activity of ultrafines, Toxicol. Appl. Pharmacol. 2001, 175, 191-199. (4) Tin-Tin-Win-Shwe; Yamamoto, S.; Kakeyame, M.; Kobayashi, T.; Fujimako, H. Effect of intratracheal instillation of ultrafine carbon black on proinfammatory cytokine and chemokine release and mRNA expression in lung and lymph nodes of mice. Toxicol. Appl. Pharmacol. 2005, 209, 51-61. (5) Sakurai, H., Park K.; McMurry, P. H.; Zarling, D. D.; Kittelson, D. B.; Ziemann, P. J. Size-dependent mixing characteristics of volatile and nonvolatile components in diesel exhaust aerosols, Environ. Sci. Technol. 2003, 37, 5487-5495. (6) Kwon, S.-B., Lee, K. W.; Saito, K.; Shinozaki, O.; Seto, T. Sizedependent volatily of diesel nanoparticles: Chassis dynamometer experiments, Environ. Sci. Technol. 2003, 37, 1794-1802. (7) Morawska, L.; Thomas, S.; Bofinger, N.; Wainwright, D.; Neale, D. Comprehensive characterization of aerosols in a subtropical urban atmosphere: Particle size distribution and correlation with gaseous pollutants, Atmos. Environ. 1998, 32, 2467-2478. (8) Virtanen, A.; Ro¨nkko¨, T.; Kannosto, J.; Ristima¨ki, J.; Ma¨kela¨, J.; Keskinen, J. Pakkanen, T.; Hillamo, R.; Pirjola, L.; Ha¨meri, K. Winter and summer time size distributions and densities of traffic-related aerosol particles at a busy highway in Helsinki. Atmos. Chem. Phys. 2006, 2411-2421. (9) Sakurai, H.; Tobias, H. J.; Park, K.; Zarling, D.; Docherty, K. S.; Kittelson, D. B.; McMurry P. H.; Ziemann P. J. On-line measurements of diesel nanoparticle composition and volatility, Atmos. Environ. 2003 37, 1199-1210. (10) Koike, E., Kobayashi, T. Organic extract of diesel exhaust particles stimulates expression of Ia and costimulatory molecules associated with antigen presentation in rat peripheral blood monocytes but not in alveolar macrophages, Toxicol. Appl. Pharmacol. 2005, 209, 277-285. (11) Scheer, V.; Kirchner, U.; Casati, R.; Vogt, R.; Wehner, B.; Philippin, S.; Wiedensohler, A.; Hock, N.; Schneider, J.; Weimer, S.; Borrmann, S. Composition of Semi-Volatile Particles from Diesel Exhaust; SAE paper 2005-01-0197.; SAE International: Washington, DC, 2005 (12) Ristima¨ki, J.; Keskinen, J. Mass measurement of non-spherical particles: TDMA-ELPI setup and performance tests. Aerosol Sci. Technol. 2006, 40, 997-1001. (13) Vaaraslahti, K.; Keskinen, J.; Giechaskiel, B.; Solla, A.; Murtonen, T.; Vesala, H. Effect of lubricant on the formation of heavy-duty diesel exhaust nanoparticles. Environ. Sci. Technol. 2005, 39, 8497-8504. (14) Mathis, U.; Ristima¨ki, J.; Mohr, M.; Keskinen, J.; Ntziachristos, L.; Samaras, Z.; Mikkanen, P. Sampling conditions for the measurement of nucleation mode particles in the exhaust of a diesel vehicle. Aerosol Sci. Technol. 2004, 38, 1149-1160. (15) Ntziachristos, L.; Giechaskiel, B.; Pistikopoulos, P.; Samaras, Z.; Mathis, U.; Mohr, M.; Ristima¨ki, J.; Keskinen, J.; Mikkanen, P.; Casati, R.; Scheer, V.; Vogt, R. Performance Evaluation of a Novel Sampling and Measurement System for Exhaust Particle Characterization; SAE paper 2004-01-1439; SAE International: Washington, DC, 2004.

6402

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 18, 2007

(16) Kelly, W. P.; McMurry, P. H. Measurement of particle density by inertial classification of differential mobility analysergenerated monodisperse aerosol, Aerosol Sci. Technol. 1992, 17, 199-212. (17) Ahlvik, P.; Ntziachristos, L.; Keskinen, J.; Virtanen, A. Real Time Measurements of Diesel Particle Size Distribution with an Electrical Low Pressure Impactor; SAE technical paper series 980410; SAE International: Washington, DC, 1998. (18) Maricq, M.; Xu, N. The effective density and fractal dimension of soot particles from premixed flames and motor vehicle exhaust. J. Aerosol Sci. 2004, 35, 1251-1274. (19) Schmidt-Ott New approaches to in situ characterisation of ultrafine agglomerates. J. Aerosol Sci. 1988, 19 (5), 553-563. (20) Ku ¨ tz, Schmidt-Ott. Use of a lowspressure impactor for fractal analysis of submicron particles. J. Aerosol Sci. 1990, 21 (S1), S47-S50. (21) Schmidt-Ott, A.; Baltensperger, U.; Ga¨ggeler, H.; Jost, D. Scaling behaviour of physical parameters describing agglomerates. J. Aerosol Sci. 1990, 21 (6), 711-829. (22) Skillas, G.; Ku ¨ ntzel, S.; Burtscher, H.; Baltensperger, U.; Siegmann, K. High fractal-like dimension of diesel soot agglomerates. J. Aerosol Sci. 1998, 29 (4), 411-419. (23) Rogak, S.; Flagan, R.; Nguyen, H. The mobility and structure of aerosol agglomerates. Aerosol Sci. Technol. 1993, 18, 25-47. (24) Park, K.; Kittelson, D.; Zachariah, M., McMurry, P. Measurement of inherent material density of nanoparticle agglomerates, J. Nanopart. Res. 2004, 6, 267-272. (25) Park, K.; Cao, F.; Kittelson, D., McMurry, P. Relationship between particle mass and mobility for diesel exhaust particles, Environ. Sci. Technol. 2003, 37, 577-583. (26) Elliott, R.; Watts, H. Diffusion of some hydrocarbons in air: A regularity in the diffusion coefficients of a homologous series. Can. J. Chem. 1972, 50, 31-34. (27) Pankow, J. An adsorption model of gas/particle partitioning of organic compounds in the atmosphere, Atmos. Environ. 1994, 28 (2), 177-378. (28) Kittelson, D.; Dolan, D. Diesel Exhaust Aerosols. In Generation of Aerosols and Facilities for Exposure Experiments; Willeke, K., Ed.; Science Publishers Inc: Ann Arbor, MI, 1980; ISBN 0-25040293-9. (29) Seinfeld, J. H.; Pandis, S. N.Atmospheirc Chemistry and Physics, John Wiley & Sons Inc: New York, 1998; ISBN 0-471-17816-0. (30) Hinds, C. Aerosol Technology: Properties, Behavior, and Measurement of Airborne Particles, 2nd ed; Wiley-Interscience: New York, 1999. (31) Vaaraslahti, K., Ristima¨ki, J., Virtanen, A., Keskinen, J., Giechaskiel, B. On the effect of oxidation catalyst on diesel soot particles. Environ. Sci. Technol. 2006, 40, 4776-4781. (32) Maricq, M.; Podsiadlik, D.; Chase, R. Size distributions of motor vehicle exhaust PM: A comparison between ELPI and SMPS measurements. Aerosol Sci. Technol. 2000, 33, 239-260. (33) Virtanen, A.; Ristima¨ki, J.; Marjama¨ki, M.; Vaaraslahti, K.; Keskinen, J. Effective Density of Diesel Exhaust Particles as a Function of Particle Size; SAE Technical Paper Series 200201-0056; SAE International: Washington, DC, 2002. (34) Virtanen, A.; Ristima¨ki, J.; Vaaraslahti, K.; Keskinen, J. Effect of engine load on diesel soot particles. Environ. Sci. Technol. 2004, 38, 2551-2556. (35) Virtanen, A. Physical Characterization of diesel soot particles. Ph. D. Thesis. Tampere University of Technology, 2004. (36) Reichel, S.; Pischinger, F.; Lepperhoff, G. Influence on Particles in Diluted Diesel Engine Exhaust Gas; SAE paper 831333: SAE International: Washington, DC, 1983. (37) Plee, S.; MacDonald, S. Some Mechanisms Affecting the Mass of Diesel Exhaust Particulate Collected Following a Dilution Process; SAE Paper 800186; SAE International: Washington, DC, 1980.

Received for review October 10, 2006. Revised manuscript received May 18, 2007. Accepted July 5, 2007. ES0624319