Article pubs.acs.org/est
Volatility of Primary Organic Aerosol Emitted from Light Duty Gasoline Vehicles Toshihiro Kuwayama,† Sonya Collier,‡ Sara Forestieri,† James M. Brady,§ Timothy H. Bertram,§ Christopher D. Cappa,† Qi Zhang,‡ and Michael J. Kleeman*,† †
Department of Civil and Environmental Engineering, University of California at Davis, One Shields Avenue, Davis, California 95616, United States ‡ Department of Environmental Toxicology, University of California at Davis, One Shields Avenue, Davis, California 95616, United States § Department of Chemistry and Biochemistry, University of California, San Diego, California 92093, United States S Supporting Information *
ABSTRACT: Primary organic aerosol (POA) emitted from light duty gasoline vehicles (LDGVs) exhibits a semivolatile behavior in which heating the aerosol and/or diluting the aerosol leads to partial evaporation of the POA. A single volatility distribution can explain the median evaporation behavior of POA emitted from LDGVs but this approach is unable to capture the full range of measured POA volatility during thermodenuder (TD) experiments conducted at atmospherically relevant concentrations (2−5 μg m−3). Reanalysis of published TD data combined with analysis of new measurements suggest that POA emitted from gasoline vehicles is composed of two types of POA that have distinctly different volatility distributions: one lowvolatility distribution and one medium-volatility distribution. These correspond to fuel combustion-derived POA and motor oil POA, respectively. Models that simultaneously incorporate both of these distributions are able to reproduce experimental results much better (R2 = 0.94) than models that use a single average or median distribution (R2 = 0.52). These results indicate that some fraction of POA emitted from LDGVs is essentially nonvolatile under typical atmospheric dilution levels. Roughly 50% of the vehicles tested in the current study had POA emissions dominated by fuel combustion products (essentially nonvolatile). Further testing is required to determine appropriate fleet-average emissions rates of the two POA types from LDGVs.
■
INTRODUCTION Emissions tests have determined that POA generated from combustion sources behave like a series of semivolatile compounds when the particulate phase concentration ranges between 100−10 000 μg m−3.1 Recent tests suggested a single volatility distribution approach to characterize POA emissions from a fleet of LDGVs.2 These findings are based on quadrapole aerosol mass spectrometer (Q-AMS, Aerodyne Research Inc.) measurements of vehicular POA emissions diluted to near-atmospheric concentration ranges in a Teflon reaction chamber. Although this experimental design is far more realistic than original tests used to establish the potential volatility of POA,3 the methodology could still be improved in several areas: (1) POA volatility was derived by measurements of compound volatility at temperature below 320 °C and then verified through a comparison of traditional thermal-optical organic carbon (OC) measurements and Q-AMS organic © 2014 American Chemical Society
aerosol (OA) measurements; (2) measurements were made from moderately high emitting vehicles and then extrapolated to low emitting vehicles; (3) POA emissions were assumed to be internally mixed and with identical volatility properties for all particles in the distribution. These features produce a model with a large amount of scatter in the comparison between the POA volatility predictions and measurements. The present study analyzes an additional set of POA volatility measurements from a fleet of LDGVs with particulate matter (PM) emission rates in the lower third of the measured distribution of on-road vehicles in the United States (U.S.),4 which is typical for California. Emissions were diluted to Received: Revised: Accepted: Published: 1569
February 28, 2014 December 5, 2014 December 10, 2014 December 10, 2014 DOI: 10.1021/es504009w Environ. Sci. Technol. 2015, 49, 1569−1577
Article
Environmental Science & Technology atmospherically relevant concentrations (2−5 μg m−3) through two stages of dilution. Average elemental carbon (EC) and OC concentrations for each test were measured at the first and the second stages of dilution using thermal-optical carbon analysis. Real-time OA measurements were made at the second stage of the dilution using high-resolution time-of-flight aerosol mass spectrometer (HR-AMS, Aerodyne Research Inc.). The sampling configuration after the second stage of dilution allowed for collection of diluted vehicle exhaust exposed to variable temperatures, up to 100 °C, to assess the POA volatility. Results were interpreted with a model that allowed for two distinct types of POA emissions with different volatility distributions. This framework is consistent with the concept that POA from motor vehicles has contributions from both the unburned lubricating oil and the products of fuel combustion, which may have dissimilar properties.5,6 In addition, the motor oil contribution to tailpipe emissions can vary significantly based on the age and maintenances of the vehicles. Volatility of POA therefore may not be a constant throughout the fleet. The results of these measurements provide an improved estimate of POA volatility for the LDGVs with low-emission vehicle (LEV) technology operating in California. The multidimensional volatility distribution model may be incorporated into regional air quality calculations to better predict POA behavior compared to the one-dimensional volatility model when performing calculations outside the exact range of experimental conditions.
collected on a set of aluminum foil substrates installed in Micro-Orifice Uniform Deposit Impactors (MOUDIs) immediately after the second stage of dilution. The remaining emissions were then aged for approximately 1.2 min in a dark residence time chamber (RTC) to allow for rapid transformations that modify POA emissions on time scales that could not be reasonably simulated in reactive chemical transport models. The aging time achieves this goal, and the exhaust enters a regime where slower transformations may still be occurring. The reactive chemical transport models can simulate these slower time scales, and therefore the measurements can provide an initial condition for these models. After aging, the emissions were split into two sets of four heated lines with variable temperature (25 °C, 50 °C, 75 °C, and 100 °C) to perturb the gas-particle partitioning. The first set of heated lines supplied aerosol to real-time instruments that included a HRAMS, a cavity ring-down/photoacoustic spectrometer (CRD/ PAS), scanning mobility particle sizer (SMPS), and a time-offlight chemical ionization mass spectrometer (ToF-CIMS) that drew a total of approximately 5 L min−1 from the RTC. The current study mainly utilizes measurements from the HR-AMS and the results from other real-time instruments are reported in separate publications. Refer to Forestieri et al. 2013, Collier and Zhang 2013, Collier et al. 2014, and Brady et al. 2014 for additional characterization of tailpipe emissions.9−12 The second set of heated lines supplied aerosol to four dedicated offline sampling trains that collected composite samples from all eight vehicles during each test day using a denuder-filterpolyurethane foam (DFP) configuration for semivolatile aerosol sampling. Each DFP train included one eight-channel annular denuder (URG, Chapel Hill, NC) followed by a threechannel annular denuder to prevent breakthrough of gases. The annular denuders were coated with ground XAD-4 polystyrene resin to increase collection capacities.13,14 The total combined flow rate from the RTC through the four DFP sampling trains was 68 L min−1. The residence time in both sets of heated lines was identical. Detailed schematics of the secondary dilution system and the DFP sampling trains are provided in the SI, Figure S1 and S2. Analytical Procedures. Quartz fiber filters used in the DFP trains and aluminum foil impaction substrates used in the MOUDIs were prebaked at 550 °C for 48 hours prior to use and stored in Petri dishes lined with aluminum foil that had also been baked at 550 °C. All of the Petri dishes were sealed with Teflon tape and stored in a −30 °C freezer. The EC/OC content of each quartz filter collected after the first stage of dilution was measured using a Thermal-Optical Analyzer (Desert Research Institute) following the IMPROVE_A temperature protocol.15,16 The EC/OC content on each quartz filter and aluminum impaction substrate collected after the second dilution was measured using thermal-optical analysis with a Sunset Lab OCEC Aerosol Analyzer following the NIOSH temperature protocol.17 Measurements from the current study show that a negligible amount of carbon emitted from gasoline-powered motor vehicles pyrolyzes during the analysis, making the thermal-optical measurements essentially equivalent despite the differences in temperature protocols. HR-AMS measurement methods, background corrections, and mass spectrometer interpretations are described in detail by Collier and Zhang 2013 and Collier et al. 2014.10,11 PAS measurement methods are described by Forestieri et al.9 Both real-time instruments provide measurements of carbonaceous
■
EXPERIMENTAL METHODS Vehicle Test Fleet and Driving Cycle. Eight LDGVs with LEV technology were recruited from the California public and/ or a pool of vehicles maintained by the California Air Resources Board (CARB) (see Supporting Information (SI), Table S1). The average age of the test fleet (9.9 years) approximately matched the average age of vehicles in the U.S. (10.9 years) in 2011, the measurement year.7 All vehicles were pretested prior to use to ensure proper operation and PM emission rates representative of the California fleet ( 600 °C. The amount of OC that evolved at T > 600 °C during the thermal-optical analysis accounts for 30 ± 11% of the OC under-prediction by the HRAMS. Additional factors seem to prevent measurement of some fraction of OA emitted from LDGVs using HR-AMS, including lower than unity transmission of particles smaller than ∼80 nm.19 SMPS measurements suggest that approximately 9 ± 6% of the PM mass was below 80 nm throughout the driving schedule, assuming spherical particles. Note that comparisons between measurements made with Q-AMS and quartz filters at different stages of dilution have been employed in previous measurements of LDGV exhaust volatility.2 The current results suggest that some of the volatility attributed to differences in
particles that can help interpret emissions behavior within different portions of the UC driving schedule.
■
RESULTS
Sampling and Measurement Techniques. EC and POA measurements made using quartz filters from CVS and DFP were compared to measurements made using MOUDI, HRAMS, and PAS to determine the reproducibility of the data set across different instruments. EC is nonvolatile and can be compared at any stage of dilution after accounting for intermediate dilution factors. Comparisons for POA were only made between two techniques at the same stage of dilution. POA measurements made after the first stage of dilution were not compared to measurements after the second stage of dilution due to the likelihood of semivolatile materials in the condensed phase after dilution by a factor of ∼12.7 that would further evaporate after dilution by a factor of ∼61. The comparison in Figure 1a shows good agreement between the thermal-optical measurements of EC/OC made at different locations in the dilution system with different collection media, different analysis temperature protocols, and different carbon analyzers. All EC/OC measurements are in agreement within experimental uncertainty. Figure 1b shows the comparison between the thermal-optical measurements of EC and PAS measurements of black carbon (BC), and the thermal-optical measurements of OC and the HR-AMS measurements of OA (OC calculated using the mass weighted average organic matter (OM) to OC ratio, OM/OC, of 1.23 determined in this study11). BC concentrations were calculated as babs,532 nm/MAC532 nm, where babs,532 nm is the absorption coefficient measured at 532 nm, and MAC532 nm is the mass absorption coefficient for BC at 532 nm.18 In both cases, the real-time measurements detected less carbon than the offline instruments. A comparison of CO2 measurements made after the first stage of dilution by CARB agrees well with the CO2 measurements made alongside the real-time instrument sampling lines.9,10 The differences between the HR-AMS and the DFP train filter measurements (33−89%, median = 51%) are larger than the differences between the PAS and the DFP train filter measurements (12−69%, median = 28%), suggesting that systematic bias caused by differential flow rates to the real1571
DOI: 10.1021/es504009w Environ. Sci. Technol. 2015, 49, 1569−1577
Article
Environmental Science & Technology
Figure 2. PM emissions factors measured after the first stage of dilution (panels a and b) and POA measured using HR-AMS after the second stage of dilution (panels c and d). Panel a and c represents emissions of seven vehicles used by CMU in their thermodenuder experiments. Dashed red line indicates California emissions standard in upper panels. The final dilution ratios were (c) ∼300−1500 and (d) ∼61. The white columns represent the emissions rates measured from +200 vehicles by U.S. EPA. ACQ = adsorption corrected quartz. The gray columns represent the emissions rates from the vehicles tested in the two studies discussed in this manuscript. ntot is the number of vehicles in each fleet.
these previous measurements may actually be caused by differences between AMS measurements vs traditional EC/ OC measurements. Test Fleet Emissions Characteristics. The PM and POA emission rates from the 8 California vehicles utilized in the current chassis dynamometer tests were compared to the emission rates from seven California vehicles used in a recent study that also employed increased temperature to assess POA volatility. 2 May et al. utilized thermal desorption gas chromatography mass spectrometer (TD-GC-MS) analysis on emissions from 30+ vehicles to develop a fleet average volatility distribution. Of these, 19 vehicles were utilized in calculating the final fleet average volatility distribution due to possible biases (signal-to-noise ratios 10 mg/mile, which exceeds the current California standard. For these reasons, the volatility distribution based on this fleet of 19 vehicles may not have fully characterized the lowest volatility POA. The seven vehicles mentioned previously were tested separately using thermodenuder experiments to evaluate and validate their TD-GC-MS volatility distribution analysis. As shown in Figure 2a, five out of the seven LDGVs used in the thermodenuder experiments conducted in the previous study also had PM emission rates >10 mg/mile. These thermodenuder measurements from seven vehicles were also used as a comparison point in this manuscript to determine the validity of the multidimensional volatility distribution. It is important to note that these comparisons will be imperfect due to inherent differences in the fleet composition between the two studies. The emissions distribution for the set of vehicles utilized in the previous study for the development of the volatility distribution (using the TD-GC-MS method) is presented in SI Figure S5. Figure 2b shows that the highest emitting vehicle used in the current study had a PM emission
rate of 3.2 mg/mile. The distribution of emission rates from +200 vehicles tested by the U.S. Environmental Protection Agency (EPA) is included in the background of Figure 2a and b as an indicator of a typical fleet behavior.4 The U.S. EPA fleet was composed of vehicles from Kansas City residents with variable age, emissions control technologies, and maintenance. The fleet tested in Kansas City did not operate on fuel with ethanol as an additive while the current fleet of California vehicles did use fuel with ethanol. The use of ethanol is expected to lower PM emissions.20 Thus, being in the lower half of the U.S. EPA fleet emissions distribution is expected for current California fleet emissions. The eight vehicles tested in the current study (Figure 2b) fall under the lower third of the log-normal portion of the vehicle distribution. Figure 2c and d illustrates the POA emission rates from the vehicles used in the previous study and the vehicles used in the current study, respectively.2 POA emission rates in the previous study were substantially larger than POA emission rates from the fleet recruited for the current study based on the measurements made by the Q-AMS and the HR-AMS. Both sets of POA measurements are based on AMS and likely underestimate the absolute concentration of low-volatile POA in the emissions, but the comparison between the different fleets is not affected by the issue since they presumably have the same bias relative to thermal-optical OC measurements (see SI Figure S3). POA Volatility Measurements. The evaporation of OA at elevated temperatures is one of a number of measures that has been used to estimate how POA might evaporate when diluted downwind of emissions sources. The real-time HR-AMS measurements made in the current study were used to determine the POA mass fraction remaining (MFR) after heating to different temperatures for particles emitted from the test fleet. Figure 3 shows the POA MFR, averaged across all eight vehicles, as a function of temperature as measured during 1572
DOI: 10.1021/es504009w Environ. Sci. Technol. 2015, 49, 1569−1577
Article
Environmental Science & Technology
Figure 3. (top panels) Primary organic aerosol (POA) mass fraction remaining (MFR) from the HR-AMS measurements as a function of diluted exhaust heating temperature during different phases of the UC driving schedule on the base case test day (eight individual vehicles): (a) Phase 1, cold-start acceleration cycle, (b) Phase 2a, high velocity, (c) Phase 2b, hard acceleration, (d) Phase 2c, hot-running acceleration cycle. Box and whisker plots illustrated in the top panel represent vehicle-to-vehicle variability in the OM emissions where bottom darker boxes represent the 25th percentile, the top lighter gray boxes represent the 75th percentile, and the whiskers represent the min and the max emissions observed. (middle panel) An example of the OM measured by the HR-AMS throughout the UC driving schedule. (bottom panel) The vehicle velocity profile (left axis) and acceleration (right axis) for the UC driving schedule.
temperatures. The EC/OC ratio in the secondary dilution system was >3 but the MFR of POA was not influenced by the addition of generated EC into the dilution air during a separate test. The Reynold’s number was >4000, which suggests fully turbulent conditions where the generated EC and the LDV exhaust would be well mixed. This result suggests that the additional EC and the existing POA did not become chemically mixed, in other words, adsorption did not play a significant role in the partitioning mechanism, which may have made POA appear less volatile when heating occurs. The median POA MFRs at all temperatures are >0.7 with MFRs from individual experiments reaching a maximum of 1.5 (Phase 2b at T = 75 °C) and a minimum of 0.2 (Phase 1 at T = 100 °C). Modeling POA Volatility. Previous studies have interpreted POA MFR measurements using a model of particle evaporation as a function of temperature with input data describing the POA volatility distribution obtained from independent experiments.2 These calculations assumed mono-
different phases of the UC driving schedule: (1) cold-start acceleration, (2a) high velocity, (2b) hard acceleration, and (2c) hot-running acceleration. The velocity and acceleration profile shown in Figure 3 was maintained throughout the duration of each experiment. All of the data shown in Figure 3 are from a base case test in which filtered ambient air was used for dilution without additional manipulation of humidity or the introduction of background particles. POA emissions during the cold-start phase were higher and more volatile than during other portions of the driving schedule (as demonstrated by a lower MFR at T = 100 °C compared to other portions of the driving schedule). Emissions of volatile organic compounds (VOCs), carbon monoxide (CO), and nitrogen oxides (NOx) are also typically larger during the cold-start period due to the suboptimal operating temperature of the combustion chamber/ combustion process and the emissions control equipment.20 The POA was less volatile during Phases 2a−c after the emissions control equipment reached normal operating 1573
DOI: 10.1021/es504009w Environ. Sci. Technol. 2015, 49, 1569−1577
Article
Environmental Science & Technology dispersed OA with a particle diameter Dp = 200 nm, diffusion coefficient of 5 × 10−6 m2 s−1, accommodation coefficient of 1, and a surface tension of 50 dyn cm−1 to account for the Kelvin effect. Table S2 in the SI lists the enthalpy of vaporization at each saturation concentration (C*) bin. A similar model21 was used in this study to determine POA volatility distributions that are most consistent with the current MFR observations. MFR profiles were calculated for two cases: (1) using the single vehicle emissions volatility distribution recommended by May et al.,2 and (2) assuming that two distinct types of POA exist in the exhaust that have unique volatility distributions and that can vary in their relative abundances from vehicle to vehicle and throughout the UC driving schedule. For the two volatility distribution model, one of the POA types is taken to be motor oil (POAMO), with a volatility distribution from May et al.2 based on used motor oil mass fractions in each C* bin measured using a TD-GC-MS method. The second POA type is assumed to be products derived from fuel combustion that are less volatile than motor oil, hereafter referred to as fuelderived (POAFD). The POAFD is assumed to have a volatility distribution primarily comprised of low volatility species, discussed further below. Calculations were performed for two different starting OA concentrations, 0.2 μg m−3 and 30 μg m−3, to reflect differences in the OA concentrations observed between different vehicles and studies. Results from the model based on a single volatility distribution are shown in Figure 4 with the current observations
two OA concentrations (red and blue lines) bound the median MFRs (median = dividing line between light and dark regions of central box) measured in both studies but significant scatter is evident for individual vehicle MFRs (upper and lower bounds of box and whiskers). This suggests that a single volatility distribution model does not robustly capture the behavior for POA from individual vehicles. A scatter plot of predicted versus observed MFRs further illustrates the behavior of the single distribution model (Figure 5a). The coefficients of determination shows weak correlation (R2 = 0.52) with a regression line slope of only 0.24. The single volatility distribution model typically underpredicts observed MFR values close to one and overpredicts MFR values smaller than 0.6. The single volatility distribution and the two-component volatility distributions are shown in Figure 5. The performance of the TD model can be significantly improved by adopting a framework in which two distinct types of POA are emitted from LDGVs with separate volatility distributions for POAMO and POAFD. Previous studies that analyzed the detailed chemical composition of emissions from gasoline-powered motor vehicles have determined that lubricating oil and fuel combustion products make separate and distinct contributions to the POA emissions.5,6 In those studies, hopanes and steranes were found to be associated with a fraction of the POA (lubricating oil emissions), while heavy PAHs (MW > 300 amu) were found to be associated with another fraction of POA (fuel combustion products) emitted from gasoline powered motor vehicles. The absence of heavy PAHs from diesel powered vehicles suggests that they are associated with the gasoline fuel used in light duty vehicles. It is therefore reasonable to think that fuel combustion products contribute strongly to the POA that evolves at the higher temperatures of the thermal-optical analysis (SI Figure S3). A volatility distribution for the POAFD was developed with very low volatility (Figure 5), that is, MFR values close to unity are obtained at all temperatures for 100% POAFD. The results were found to be relatively insensitive to the exact distribution of materials into volatility bins with C* ≤ 0.01 μg m−3. The volatility distribution for POAMO was based on measurements made by May et al.2 The exact balance between POAMO and POAFD for a given vehicle is not known a priori. Therefore, the relative amounts of POAMO and POAFD were calculated on a vehicle-by-vehicle basis to achieve the best agreement between the predicted and observed MFR at 100 °C, discussed further in the following section. In most cases, a near perfect match could be obtained, as illustrated in the scatter plot of the model predictions based on two volatility distributions versus measured MFRs (Figure 5b). Coefficients of determination for the model with two volatility distributions were R2 = 0.94 when the regression line is not forced through zero with a regression line slope of 0.75. A comparison between the model and measured MFRs for individual vehicles at all temperatures is provided in SI Figure S6. Treating the POA emissions as a mix of two POA types with variable relative abundances and very different volatility distributions allows for overall better model/measurement agreement. The residual disagreement between the two volatility distribution model predictions and the observations likely reflects limitations of the TD model itself, not the input distributions. The model treats only the physical evaporation process under an assumption of absorption partitioning of semivolatile compounds and, as such, does not account for any chemical reactions that may occur at the elevated temperatures.
Figure 4. POA MFR from (a) UCD-AMS data set with denuder residence time of 1 s and (b) CMU-AMS data set with denuder residence time of 30 s using single volatility distribution model.2 Box and whiskers represent the measured MFRs from individual vehicles (n = number of represented vehicles). The red and blue dashed lines illustrate the sensitivity of the modeled MFR to the range of dynamic POA concentrations observed in the two studies.
(Figure 4a) and the CMU observations (Figure 4b). The MFR calculations for the current study use an experimentally constrained particle diameter of 100 nm, as opposed to the 200 nm used by May et al., and differences in the heating residence times between studies (1 s for the current study versus 30 s for May et al.) are accounted for. Shorter denuder residence time suggests a state further away from equilibrium and the POA may in fact be actively evaporating. Therefore, differences in the residence time must be explicitly accounted for when comparing the two data sets. The model MFRs at the 1574
DOI: 10.1021/es504009w Environ. Sci. Technol. 2015, 49, 1569−1577
Article
Environmental Science & Technology
Figure 5. Predicted vs observed UCD POA MFRs at 100 °C using (a) TD model with single volatility distribution and (b) TD model with two volatility distributions. Dashed lines represents 1:1 relationship.
For a subset of vehicles in both this study and the May et al. study an increase in the MFR to values above 1 (i.e., POA increased) was observed at the intermediate temperatures (T = 50 and 75 °C). This observation suggests the possibility of continued chemical reactions in the diluted exhaust that caused an increase in the POA concentration measured by AMS that outweighed evaporation. Consistent with the AMS results, thermal-optical measurements of POA and gas chromatography mass spectrometer (GC-MS) measurements of individual carbonyl compounds from quartz filters that collected emissions from multiple vehicles at different temperatures also reflect increasing concentrations at elevated temperatures (see SI, Figure S7). Fuel and Oil Contributions to POA Emissions. The fits summarized in Figure 5 based on the model with two volatility distributions were determined by iteratively adjusting the relative amounts of POAMO and POAFD in the TD model to progressively minimize the residual error in the cycle-average MFR at 100 °C. Figure 6 shows the relative amount of POAMO and POAFD determined for each of the eight vehicles tested in the current study (top panels). Four vehicles had emissions dominated by motor oil derived POA while the other four vehicles had emissions that were predominantly fuel-derived POA. These results are consistent with the inability of the single volatility distribution model to match the observations. The total POA emissions rate for the vehicles tested in the current study was not a predictor of the emissions composition (oil vs fuel). The fitting process for the model with two volatility distributions was repeated based on the MFR(100 °C) values observed during each stage of the driving schedule to examine how the relative fractions of POAMO and POAFD
Figure 6. Emissions of OA attributed to POAMO and POAFD: Top set represents UCD vehicle fleet and bottom set represents CMU vehicle fleet.
varied (SI Figure S8). No clear dependence on UC phase is evident. The POA TD model with two volatility distributions was also applied to the May et al. measurements of LDGV exhaust2 to determine POAMO and POAFD contributions. Figure 6 (bottom panel) shows that emissions from this previous vehicle fleet are 1575
DOI: 10.1021/es504009w Environ. Sci. Technol. 2015, 49, 1569−1577
Article
Environmental Science & Technology
Figure 7. Motor oil and fuel-derived EFPOA for individual vehicles used in the UCD study (a and b) and in the CMU study (c and d) measured using AMS. Each column represents the number of vehicles in the indicated emissions range. Dashed lines represents the vehicle fleet average EFPOA determined from the analysis of quartz filters collected using the DFP system.
thermograms in the present study, the result in Figure 6 conveys a similar message where average POAFD and POAMO contributions to total POA are similar.
dominated by motor oil POA (>60% contribution). The significantly larger PM emission rates for these “moderately high emitting” vehicles relative to the vehicles tested in the current study (Figure 2) appear to mainly consist of motor oil. Measurements of a motor oil molecular marker (17α(H)21β(H)-hopane)6 were made on extracted, nonderivatized samples collected in the current study as a method to independently determine the motor oil contribution to the overall fleet POA emissions. The ratios of OC/30-norhopane in the previous smoker tests matched an independent analysis of used motor oil conducted by Desert Research Institute (DRI). Therefore, OC/30-norhopane ratios from both (i) LEV and (ii) the used oil tests6 were used as the uncertainty limits in the hopane analysis. The results suggest that approximately 14− 33% of the fleet averaged POA was derived from motor oil, which compares favorably with the two volatility distribution model estimate of 24−86%. In addition, a customized temperature ramp was specified for the thermo-optical OCEC analyzer to study the effect of higher temperature and/or longer residence time (at T = 320 °C) on quartz filters collected in the current study to imitate the TD-GC-MS analysis performed by May et al. EC/OC analysis of the samples collected by May et al. found that only 20% of the POA was retained on the quartz filters at a temperature of 320 °C for 5 min. A similar test conducted on the samples collected during the current study found that 33% of the organic mass remained in the particle phase even after the exposure to T = 320 °C for 5 min (refer to SI Figure S9). This customized temperature ramp was calibrated for the UCD instrument22 and the extended hold time eliminated the possibility of transient temperature ramp effects. The difference in apparent volatility at elevated temperatures likely reflects the cleaner fleet and the lower sampling concentrations in the current study. This is also apparent from the differences between the fleet emissions presented in Figure 2 and SI Figure S5. It is likely that the TDGC-MS method employed in previous studies only characterized the more volatile fraction of the POA mass emitted from gasoline-powered motor vehicles, which would modify the inferred volatility distribution. Although there were no attempts to create a full volatility distribution directly from the
■
DISCUSSION Figure 7 summarizes the distribution of POAMO and POAFD emission rates measured in the current study (eight vehicles) and in the previous study (seven vehicles)2 calculated as the products of the relative fractions of POAMO and POAFD and the overall POA emission rate. The emission factor for POA (EFPOA) from fuel combustion products in the current study spans ∼1 order of magnitude while the EFPOA from motor oil spans ∼2 orders of magnitude. The median POAMO and POAFD emission rates measured in the current study are 1−2 orders of magnitude lower than the corresponding emission rates measured during the previous study, which suggests consistency with the comparison of PM emission rates reported prior to the multicomponent comparison (see Figure 2). Median total PM (=OA+EC) emission rates in the current study are 1 order of magnitude lower than corresponding emission rates measured during the previous study. The results shown in Figures 6 and 7 also suggests that motor oil contributed to a larger fraction of the total POA emission from the previous study, likely due to the composition of the test fleet that included seven vehicles (out of 19) that had PM emission rates >10 mg/mile. A multidimensional volatility distribution model that simultaneously incorporates volatility distributions of two POA types - motor oil and fuel derived - is able to reproduce observed range of POA volatility much better (R2 = 0.94) than models that uses a single average distribution (R2 = 0.52). The ability to vary contributions from each POA type allows a modular approach in developing future emissions inventories and will be useful for future regional modeling applications where on-road fleet composition differs from the test fleet composition. However, further measurements with larger sample size are needed to adequately characterize and separate emissions of motor oil and fuel-derived POA in support of 1576
DOI: 10.1021/es504009w Environ. Sci. Technol. 2015, 49, 1569−1577
Article
Environmental Science & Technology
from light-duty gasoline and heavy-duty diesel vehicles. Environ. Sci. Technol. 2008, 42 (1), 235−242. (7) DOT. National Transportation Statistics; United States Department of Transportation, 2013. (8) CARB. Characterizatoin of Driving Patterns and Emissions from Light-Duty Vehicles in California, Final Report; California Air Resources Board, 1993. (9) Forestieri, S.; Collier, S.; Kuwayama, T.; Zhang, Q.; Kleeman, M. J.; Cappa, C. D. Real-time black carbon emission factor measurements from light duty vehicles. Environ. Sci. Technol. 2013, 47, 13104−13112. (10) Collier, S.; Zhang, Q. Gas-phase CO2 subtraction for improved measurements of the organic aerosol mass concentration and oxidation degree by an aerosol mass spectrometer. Environ. Sci. Technol. 2013, 47 (24), 14324−14331. (11) Collier, S.; Kuwayama, T.; Forestieri, S.; Brady, J.; Zhang, M.; Bertram, T.; Cappa, C.; Kleeman, M.; Zhang, Q. Organic PM emissions from vehicles: Composition, O/C ratio, and dependence on PM Concentration. Aerosol Sci. Technol. 2014, Submitted. (12) Brady, J. M.; A, C. T.; Collier, S.; Kuwayama, T.; Forestieri, S.; Perraud, V.; Zhang, Q.; Kleeman, M.; Cappa, C.; Bertram, T. Realtime emission factor measurements of isocyanic acid from light duty gasoline vehicles. Environ. Sci. Technol. 2014, Submitted. (13) Gundel, L. A.; Lee, V. C.; Mahanama, K. R. R.; Stevens, R. K.; Daisey, J. M. Direct determination of the phase distributions of semivolatile polycyclic aromatic-hydrocarbons using annular denuders. Atmos. Environ. 1995, 29 (14), 1719−1733. (14) Jakober, C. A.; Robert, M. A.; Riddle, S. G.; Destaillats, H.; Charles, M. J.; Green, P. G.; Kleeman, M. J. Carbonyl emissions from gasoline and diesel motor vehicles. Environ. Sci. Technol. 2008, 42 (13), 4697−4703. (15) CARB. Procedure for Organic Carbon and Elemental Carbon (OC/EC) Analysis of Vehicular Exhaust Particulate Matter (PM) on Quartz Filters; California Air Resources Board, m2011. (16) Chow, J. C.; Watson, J. G.; Chen, L. W. A.; Chang, M. C. O.; Robinson, N. F.; Trimble, D.; Kohl, S. The IMPROVEA temperature protocol for thermal/optical carbon analysis: Maintaining consistency with a long-term database. J. Air Waste Manage 2007, 57 (9), 1014−1023. (17) NIOSH. NIOSH Manual of Analytical Methods, 4 ed.; National Institute for Occupational Safety and Health: Cincinnati, OH, 1996. (18) Bond, T. C.; Bergstrom, R. W. Light absorption by carbonaceous particles: An investigative review. Aerosol Sci. Technol. 2006, 40 (1), 27−67. (19) Liu, P. S. K.; Deng, R.; Smith, K. A.; Williams, L. R.; Jayne, J. T.; Canagaratna, M. R.; Moore, K.; Onasch, T. B.; Worsnop, D. R.; Deshler, T. Transmission efficiency of an aerodynamic focusing lens system: Comparison of model calculations and laboratory measurements for the Aerodyne Aerosol Mass Spectrometer. Aerosol Sci. Technol. 2007, 41 (8), 721−733. (20) Hays, M. D.; Preston, W.; George, B. J.; Schmid, J.; Baldauf, R.; Snow, R.; Robinson, J. R.; Long, T.; Faircloth, J. Carbonaceous aerosols emitted from light-duty vehicles operating on gasoline and ethanol fuel blends. Environ. Sci. Technol. 2013, 47 (24), 14502− 14509. (21) Cappa, C. D. A model of aerosol evaporation kinetics in a thermodenuder. Atmos Meas Tech 2010, 3 (3), 579−592. (22) Phuah, C. H.; Peterson, M. R.; Richards, M. H.; Turner, J. R.; Dillner, A. M. A temperature calibration procedure for the sunset laboratory carbon aerosol analysis lab instrument. Aerosol Sci. Technol. 2009, 43 (10), 1013−1021.
future emissions inventories used in regional modeling calculations.
■
ASSOCIATED CONTENT
S Supporting Information *
Table S1 summarizes the vehicle fleet used in current experiments, Table S2 presents the enthalpy of vaporization used in TD model at each C* bin, Figure S1 presents a schematic of the exhaust dilution and sample collection configuration, Figure S2 presents a schematic of the DFP sampling train, Figure S3 displays a thermogram of carbon evolution from a quartz filter sample, Figure S4 shows measured OC concentration at four thermodenuder temperatures, Figure S5 presents the emissions distribution of vehicles utilized by May et al. for their POA analysis using TD-GC-MS method, Figure S6 shows two-distribution model fits for MFR from individual vehicles, Figure S7 shows MFR measurements using HR-AMS, thermo-optical OC/EC, and GC-MS, Figure S8 shows motor oil contribution to vehicle OA emissions separated by UC driving schedule phases, and Figure S9 presents the OCEC thermogram of underivatized samples using a customized temperature ramp. This material is available free of charge via the Internet at http://pubs.acs.org.
■
AUTHOR INFORMATION
Corresponding Author
*Phone: +1 530 752 8386; fax: +1 530 752 7872; e-mail:
[email protected]. Notes
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS This research was supported by the California Air Resources Board under contract No. 10-313. We thank Nehzat Motallebi for project management. We thank Mang Zhang and the staff at Haagen-Smit Laboratory in El Monte, California (Shiyan Chen, Tuyen Dinh, Paul Moon, Sulekha Chattopadhyay and others involved) for their time and effort in operating and conducting the dynamometer experiments. We thank Michael Hays and Bill Preston from EPA for their technical assistance in molecular marker analysis. We also thank Dr. Shantanu Jathar for helpful technical discussions.
■
REFERENCES
(1) Shrivastava, M. K.; Lipsky, E. M.; Stanier, C. O.; Robinson, A. L. Modeling semivolatile organic aerosol mass emissions from combustion systems. Environ. Sci. Technol. 2006, 40 (8), 2671−2677. (2) May, A. A.; Presto, A. A.; Hennigan, C. J.; Nguyen, N. T.; Gordon, T. D.; Robinson, A. L. Gas-particle partitioning of primary organic aerosol emissions: (1) Gasoline vehicle exhaust. Atmos. Environ. 2013, 77, 128−139. (3) Grieshop, A. P.; Miracolo, M. A.; Donahue, N. M.; Robinson, A. L. Constraining the volatility distribution and gas-particle partitioning of combustion aerosols using isothermal dilution and thermodenuder measurements. Environ. Sci. Technol. 2009, 43 (13), 4750−4756. (4) EPA, U.S. Analysis of Particulate Matter Emissions from Light-Duty Gasoline Vehicles in Kansas City; United States Environmental Protection Agency, April, 2008. (5) Sonntag, D. B.; Bailey, C. R.; Fulper, C. R.; Baldauf, R. W. Contribution of lubricating oil to particulate matter emissions from light-duty gasoline vehicles in Kansas City. Environ. Sci. Technol. 2012, 46 (7), 4191−4199. (6) Kleeman, M. J.; Riddle, S. G.; Robert, M. A.; Jakober, C. A. Lubricating oil and fuel contributions to particulate matter emissions 1577
DOI: 10.1021/es504009w Environ. Sci. Technol. 2015, 49, 1569−1577
