Article pubs.acs.org/est
Carbonaceous Aerosols Emitted from Light-Duty Vehicles Operating on Gasoline and Ethanol Fuel Blends Michael D. Hays,* William Preston, Barbara J. George, Judy Schmid, Richard Baldauf, Richard Snow, James R. Robinson, Thomas Long, and James Faircloth Office of Research and Development National Risk Management Research Laboratory, United States Environmental Protection Agency, Research Triangle Park, North Carolina 27711, United States S Supporting Information *
ABSTRACT: This study examines the chemical properties of carbonaceous aerosols emitted from three light-duty gasoline vehicles (LDVs) operating on gasoline (e0) and ethanolgasoline fuel blends (e10 and e85). Vehicle road load simulations were performed on a chassis dynamometer using the three-phase LA-92 unified driving cycle (UDC). Effects of LDV operating conditions and ambient temperature (−7 and 24 °C) on particle-phase semivolatile organic compounds (SVOCs) and organic and elemental carbon (OC and EC) emissions were investigated. SVOC concentrations and OC and EC fractions were determined with thermal extraction-gas chromatography−mass spectrometry (TE-GC-MS) and thermaloptical analysis (TOA), respectively. LDV aerosol emissions were predominantly carbonaceous, and EC/PM (w/w) decreased linearly with increasing fuel ethanol content. TE-GC-MS analysis accounted for up to 4% of the fine particle (PM2.5) mass, showing the UDC phase-integrated sum of identified SVOC emissions ranging from 0.703 μg km−1 to 18.8 μg km−1. Generally, higher SVOC emissions were associated with low temperature (−7 °C) and engine ignition; mixed regression models suggest these emissions rate differences are significant. Use of e85 significantly reduced the emissions of lower molecular weight PAH. However, a reduction in higher molecular weight PAH entities in PM was not observed. Individual SVOC emissions from the Tier 2 LDVs and fuel technologies tested are substantially lower and distributed differently than those values populating the United States emissions inventories currently. Hence, this study is likely to influence future apportionment, climate, and air quality model predictions that rely on source combustion measurements of SVOCs in PM.
■
INTRODUCTION In the United States (U.S.) and elsewhere ethanol is being used as a renewable fuel in motor vehicles in an effort to reduce emissions of greenhouse gases and select pollutants and dependence on imported oil. Bioethanol use in on-road, motor vehicles is projected to escalate in the U.S. due to federal policy. Yet, the net benefit, if any, of producing corn-based ethanol as a renewable motor fuel is uncertain currently due to poorly understood life-cycle, energy, and environmental metrics.1 Air pollution is among the many environmental and public health concerns associated with increased ethanol use in vehicles. Jacobson2 showed for the U.S. market that full conversion to e85 ((85% ethanol, 15% gasoline)the maximum standard blend used in modern dual fuel vehicles) from 100% gasoline may cause excess morbidity and mortality risk. Higher ozone, peroxy acetal nitrate (PAN), and volatile carbonyl concentrations in urban air are associated with ethanol fuel combustion in vehicles (ref 3 and refs therein). Of the studies examining combustion emissions from light-duty vehicles (LDVs) operating on ethanol-gasoline blends, the vast majority characterize air toxics (e.g., benzene, 1,3-butadiene, acetaldehyde, and formaldehyde), criteria pollutants (e.g., CO, NOx, particulate matter (PM), and O3), or greenhouse gases.4,5 Fewer examine the PM composition despite This article not subject to U.S. Copyright. Published XXXX by the American Chemical Society
its climate relevancy, toxicological importance, and use in regulatory-based air quality and predictive dispersion models.6 Regarding relevant past studies, Maricq et al.7 used a thermal analysis to fractionate carbon in PM emissions from a light-duty gasoline direct injection (GDI) vehicle burning e0 thru e48 fuel blends. While this study showed high elemental carbon (EC) associated with GDI engines and increasing organic aerosol production commensurate with fuel ethanol content, pyrolysis which can bias OC-EC ratioswas unaccounted for with their thermal analysis technique and no further compositional information was presented. A second LDV-ethanol fuel study conducted in Brazil measured individual polycyclic aromatic hydrocarbons (PAH) from LDVs burning e20 and e100.8 It was concluded from this study that PAH emissions from the LDV burning e100 were substantially lower than LDVs burning either e0 or e20. Although, exhaust emissions can vary among LDVs, and use of different LDV models to burn each fuel type complicates these fuel-based comparisons. Dutcher et al.9 Received: July 16, 2013 Revised: November 5, 2013 Accepted: November 18, 2013
A
dx.doi.org/10.1021/es403096v | Environ. Sci. Technol. XXXX, XXX, XXX−XXX
Environmental Science & Technology
Article
Table 1. Properties of the Tier 2 Vehicles Testeda vehicle (mfr./model)
fuel system
odometer (km)
inertia weight (kg)
fuel capacity (L)
cylinders
displacement (L)
Chevrolet Impala LS Chrysler Town & Country Honda Civic LX
MFI flex-fuel SFI flex-fuel MFI
23 785 78 283 26 459
1814 2155 1361
64.4 75.7 49.2
6 6 4
3.5 3.3 1.8
All vehicles tested were model year 2008 and had automatic transmissions. MFI − multiport fuel injection; SFI − sequential fuel injection. The Chevrolet and Chrysler vehicles were flex fuel vehicles.
a
Dynamometer Testing. Vehicle road load simulation was conducted on a 48 in. roll electric chassis dynamometer (BurkePorter model 4100). Vehicle and dynamometer operation conformed to procedures found in the Code of Federal Regulations, title 40, part 86. Target road load curve coefficients and test inertias used for each LDV were identical to those used in vehicle emissions certification. Actual dynamometer road load coefficients were derived from the target coefficients using coastdown procedures.11 A temperature-controlled climate chamber surrounding the dynamometer was used to precondition and test each vehicle at −7 and 24 °C. Test vehicles remained parked in the chamber overnight prior to each cold- start test to ensure engine oil and coolant temperatures were stable at test temperature. Test vehicles were operated over the three-phase LA-92 unified driving cycle (UDC) (also termed the California Unified Cycle LA92 Dynamometer Driving Schedule). Figure S1 in the SI shows the LA-92 UDC, which includes a 300 s cold-start transient phase (phase 1), a 1135 s hot-stabilized phase (phase 2), and, following a 10 min engine-off “soak” period, a 300 s warm-start transient phase combined with a 1135 s hot-stabilized phase (phase 3). Each LDV-fuel-temperature combination was tested in duplicate sequentially. A third test was added if the continuous emissions monitors for regulated emissions (NOx, CH4, total hydrocarbons, CO2, or CO) indicated a greater than 10% difference or 2% for CO2. A change in protocol was followed when switching between the fuels. It included draining and refilling the LDV fuel tank, purging the catalyst of sulfur via wideopen-throttles, and performing multiple preparatory coast-down and LA-92 UDC cycles with the fuel and LDV to be tested. PM2.5 Emissions Sampling and Chemical Analysis. LDV exhaust was directed to a dilution tunnel and constant volume sampling (CVS) system using an insulated transfer tube. A schematic of the CVS sampling system is provided in SI Figure S2. Briefly, dilution air (21 m3/min; 21 °C) passed through a charcoal bed to stabilize hydrocarbons and then through a HEPA-filter to remove PM before mixing with the LDV raw exhaust. A mixing orifice positioned immediately downstream of the exhaust entry point ensured a homogeneously diluted exhaust mixture (dilution ratio 1:∼15−30). For this study, the SVOC composition of PM2.5 was the focus. Particle emissions were collected on both Teflon (R2PJ047, Pall Corporation, Ann Arbor, MI) and prefired (550 °C, 12 h) quartz fiber filters (Pallflex 47 mm diameter, Pall Corporation, Ann Arbor, MI) positioned downstream of a PM2.5 cyclone (93 L/min; URG2000−30-EP, URG, Chapel Hill, NC). Multiple four-point isokinetic probe assemblies were used that allowed separate and proportional PM2.5 sampling of each UDC phase. Each probe was dedicated to a specific filter type or substrate array and was heated to 47 °C in an in-house fabricated steel box. The gas and particle sampling system and its operation were compliant with the Code of Federal Regulations, title 40, part 1065. Teflon filters were analyzed gravimetrically using a microbalance (MC5, Sartorius Corp., Bohemia, NY) and procedures
confirmed that black carbon (BC) and PAH in individual particles decrease with fuel ethanol content. However, while informative, this study is largely qualitative and used an obsolete engine type without a three-way catalyst. To our knowledge, composition of PM in the exhaust of multiple Tier 2 LDVs each burning e0, e10, and e85 is unavailable presently. Testing different fuels with the same vehicle set should produce an improved understanding of PM compositional changes due to fuel type differences. Moreover, the selection of e0, e10, and e85 fuels and their subsequent use in LDVs allows bounding of past, present, and potential U.S. emissions scenarios. The present study aims to examine the effects of fuel ethanol content (e0 versus e10 versus e85), operating conditions, and ambient temperature (−7 and 24 °C) on the semivolatile organic compound (SVOC) composition of the carbonaceous PM emitted from a set of three modern LDVs meeting U.S. Tier 2 emissions standards. PM mass sampled from exhaust during dynamometer testing is fractionated into organic and elemental carbon using thermal optical analysis. Particle SVOC composition (e.g., nonpolarsPAH, hopanes and steranes, n-alkanes) is determined using thermal extraction-gas chromatography mass spectrometry, a method that has exhibited sufficiently high sensitivity despite the low PM sample mass generated by modern LDVs.10 Statistical models are applied to assess the effects of phase, temperature, and fuel covariates on emissions.
■
EXPERIMENTAL SECTION All vehicle testing and emissions characterization measurements were conducted within the LDV emissions test facility located at the U.S. Environmental Protection Agency (EPA) Research Triangle Park Campus, North Carolina. Vehicles and Fuels. Three Tier 2 compliant vehicles were selected based on anticipated high market share in the U.S. fleet. Specifications for the Tier 2-certified LDVs examined as part of this study are given in Table 1. The 2008 model year LDVs have an odometer range of 23 78578 283 km and were equipped with automatic transmissions and either multiport or sequentialport fuel injection systems. Table S1 (see Supporting Information (SI); note that table navigation is achieved using the tabs at the bottom of the worksheet) offers an ASTM fuel properties summary for summer- and winter-grade e0 (100% gasoline), e10 (10% ethanol, 90% gasoline), and e85 fuel blends. The winter e85 fuel was formulated with slightly less ethanol (78% v/v; near the recommended maximum), and butane was added to all winter fuels to improve volatility for cold-start cranking, marking the only compositional differences between the winter and summer grade fuels. The gasoline contained predominantly parrafin and aromatic moieties that decreased with ethanol content. The benzene content in the fuels was fixed at roughly 1% v/v. Each fuel blend was examined before and after LDV testing, over which time minimal compositional change was observed. Only the flex-fuel vehicles (Impala LS and Town and Country) were tested with the e85. B
dx.doi.org/10.1021/es403096v | Environ. Sci. Technol. XXXX, XXX, XXX−XXX
Environmental Science & Technology
Article
identical to those described earlier.12 PM2.5 filter mass loads ranged from below detection limits (∼ 3 μg) to 328 μg. Dilution air background levels were checked daily before LDV testing. PM2.5 mass was rarely detected on these filters (≤23 μg); although, if it was, the PM2.5 mass values were background corrected. The organic and elemental (OC-EC) carbon fractions in PM collected on the quartz filters were determined by thermaloptical analysis (TOA) and NIOSH Method 5040. TOA results showed total carbon load per sample ranged from 0.4 μg to 320.1 μg. Background correction included predominantly OC and accounted for 25% of the total detected carbon on average. Positive sampling artifacts associated with the quartz filters were unaccounted for in this study. May et al.13 showed for gasoline vehicles that these artifacts generally decreased with increasing primary organic aerosol mass in the emissions. Due to its sensitivity, thermal extraction gas chromatography− mass spectroscopy (TE-GC-MS) was selected to further examine the PM SVOC composition. Individual PAH, saturated hydrocarbon (branched-, cyclic-, and normal-alkanes), and hopane/ sterane compounds (see SI Table S2) were targeted for the TEGC-MS analysis. An exhaustive description of the TE-GC-MS equipment and procedures used was provided in Herrington et al.10 and references therein. Briefly, quartz filters were introduced to a glass TE tube and spiked with internal standard solution (see SI Table S1 for deuterated internal standards). After tube insertion, the TE unit (TDS2, Gerstel Inc., Germany) temperature was ramped from 25 to 300 °C at 10 °C/min and held constant at 300 °C for 5 min. A splitless-mode extraction was performed with He flowing continuously (50 mL/min) over the sample. Extract was directed through a heated (300 °C) capillary transfer line to a cryo-cooled (−100 °C) PTV inlet also operating in splitless mode. Following the TE step, the inlet was flashheated to 300 °C at a rate of 702 °C/min. Sample was chromatographed on an ultralow bleed capillary column (DB5, Agilent Technologies, 30 m length, 0.25 μm film thickness and 0.25 mm i.d.). Helium was used as the carrier gas (1 mL/min). The GC oven temperature was programmed at 65 °C for 10 min and then ramped to 300 °C at 10 °C/min and held fixed for 41.5 min. The MS (5973, Agilent Technologies) was operated in scan mode (50−500 amu, 3 scans/s). Further data acquisition and instrument control details are given in Herrington et al.10 Method detection limits (0.01−0.7 ng) were determined in accordance with SW-846 guidelines14 and are also shown in SI Table S2. Average response factors from multilevel calibration curves (N = 4, ranging from ∼0.6 ng/μL to 6 ng/μL) were used for quantification in tandem with the internal standard method. Only response factors with relative standard deviations ≤30% were accepted. TE-GC-MS analysis of laboratory tube and sample media blanks revealed slight or negligible contamination. All SVOC concentrations were background corrected using a weekly CVS system blank. Any detected CVS system background was placed in units of μg/m3 and treated as constant for each phase over the three test days. Statistical Analysis. SAS 9.3 (SAS Institute Inc. 2011. SAS/ STAT 9.3 User’s Guide. Cary, NC) was used for all descriptive statistical analyses and linear models. The descriptive statistics included n, mean, median, minimum, and maximum for the concentration, log-concentration, proportion, and log-proportion measurements for each compound by all combinations of phase, temperature, and fuel (SI Table S3). Tests for normality favored the log-scale measurements for the linear models. Maineffect linear mixed models were fit separately for both logconcentration and log-proportion measurements for each
compound to assess their relationships with the covariates of phase, temperature, and fuel; and these models accounted for the repeated measures structure of testing each vehicle in all three phases and for the random effect of vehicle. Main-effect linear mixed models were fit separately for the log ratio of OC and EC to background-subtracted PM. These models accounted for the repeated measures structure of testing each car in all three phases and for the fixed effect of vehicle. All these main-effect ANOVA models would arguably be improved by incorporating interaction terms for the covariates, but the data set is too small to estimate larger-order models.
■
RESULTS Thermal-Optical Analysis Results. SI Table S4 provides descriptive statistics and the OC and EC emission factors and ratios (μg/km and mg/kg fuel burned) by vehicle, temperature, fuel type, and UDC phase. The vehicle PM emissions are predominantly carbonaceous with a mean total carbon to PM mass ratio of 1 ± 0.7. UDC phase-integrated (denoted in SI Table S4 as “123”) OC and EC values varied from 30 to 618 μg/ km and from 1 to 2784 μg/km, respectively. Integrated phase data also clearly exhibit the EC/PM ratio decreasing linearly (0.6−0.1) with increasing fuel ethanol content (r2 = 0.97). An OC/PM trend is less apparent. The relatively low OC/EC values from vehicles operating on e0 and e10 (study average = 0.6) are a noteworthy feature of the aerosol emissions. Figure 1 shows the OC-EC data (log-transformed); main effects model results for OC and EC are provided in Table 2. Significantly different OC and EC emissions rates are observed for all UDC phases (p < 0.05), showing a general trend of phase 1 > phase 2 > phase 3. The −7 °C dynamometer cell temperature produces significantly higher aerosol OC and EC emissions. Use of e0, e10, and e85 in these vehicles produce similar OC emissions. Compared with e0 and e10, use of e85 produces significantly less EC emissions and a decrease in EC particle proportions on average. The proportions of OC and EC in the aerosol particles emitted from e0 and e10 are relatively unchanged, and UDC phase shows no affect on the proportion of particle OC. Although, EC proportions decrease as the vehicle warms (phase 3 < phase 2 < phase 1) with phase 3 being significantly lower than phases 1 and 2. GC-MS Determined SVOC Emissions Rates. The UDC phase-integrated sum of GC-MS identified SVOCs emitted from the LDVs ranges from 0.703 μg km−1 to 18.8 μg km−1, accounting for up to 4% of the PM2.5 mass. The PAH and saturated hydrocarbon (sat-HC) class emissions ranges are 0.01−15.0 μg km−1 and 0.3−3.2 μg km−1, respectively. SVOC class means by UDC phase, temperature, and fuel are presented in SI Table S5 with more descriptive statistics. The highest PAH emissions are typically associated with the e0 and e10 fuels, the cooler −7 °C dynamometer cell temperature, and engine ignition (phase 1). Organic markers of vaporized oil, the steranes and hopanoids, are detected in the aerosol emissions from all LDVs at both ambient temperatures but only for the tests performed with e0, during the warm-running UDC phases 2 and 3. Logtransformed SVOC concentration data (ng km−1) grouped by compound class, fuel type, UDC phase, and dynamometer cell temperature are shown in Figure 2. While the lower cell temperature produces a marked increase in the PAH emissions, the effect observed for the sat-HC is less profound. High relative SVOC emissions associated with engine ignitionUDC phase 1and a limited decrease (for PAH at phases 2 and 3) in mean SVOC emissions with fuel oxygen content are two general data trends that emerge from Figure 2A. Overall, the PAH class is C
dx.doi.org/10.1021/es403096v | Environ. Sci. Technol. XXXX, XXX, XXX−XXX
Environmental Science & Technology
Article
grouped as in Figure 2A. A noteworthy feature of Figure 2B data is the high proportion of saturated hydrocarbons in the PM emitted upon LDV ignition (UDC phase 1) at 24 °C relative to −7 °C; an effect that is most evident for e0 and e10. If considering the effects due to temperature at only UDC phase 1, observed differences are significant for all sat-HC SVOCs except pristane (p = 0.068), see SI Table S4 results. Otherwise, the UDC phases did not strongly influence the SVOC composition of PM. With regard to the effect of fuel ethanol content on SVOC proportions, e0 and e10 differences are minimal. The SVOC proportions produced using e85 show differences compared with both the e0 and e10 fuels, but these were significant for less than one-third of the SVOCs under study, representing a mixed group of lower molecular weight PAH and saturated hydrocarbons.
■
DISCUSSION SPECIATE (http://www.epa.gov/ttnchie1/software/speciate/) is a U.S. VOC and PM emission factor database of air pollution sources. It is used to produce emissions inventories and for photochemical and source-receptor modeling. It contains LDV studies conducted in 2005 and earlier, and shows that OC and EC proportions in LDV PM emissions can vary substantially. In SPECIATE, OC (50.3% ± 40.3%) is significantly (p > 0.0001) higher than EC (30.7% ± 23.3%). The present LDV emissions study produced higher mean pooled EC (60% ± 30%), showing that modern vehicles have different weight fractions of OC and EC. Black carbon (BC) increases with LDV acceleration and speed (SI Figure S1), and the aggressive nature of the LA92 UDC likely contributes to our comparatively high EC fractions. Also, OC is reduced via catalysts which lower nonmethane organic gases, while EC is not targeted for reduction per se. The fuel oxidation potential increases with the addition of ethanol to petroleum, explaining why the EC/PM ratio decreases with increasing fuel ethanol content.9,15 Phase-averaged EC values (SI Table S4) are virtually identical to those BC values produced in the past dynamometer studies (∼10−40 mg kg−1 of fuel) as compiled in Liggio et al.,16 who point to a low bias associated with dynamometer BC measurements compared with those performed on a Canadian highway. Phase 1 of the LA92 UDC poorly simulates highway driving, but produces EC values as high as 252 mg kg−1 of fuel (SI Table S4) near the top end of the onroad LDV BC range given in Liggio et al.16 In another on-road study conducted in Los Angeles, California lower LDV EC values of 5.3 mg kg−1 of fuel are observed only in PM0.25.17 Vehicle technology and driving behavior and conditions play a larger role in EC/BC emissions than accounted for currently, stressing the importance of producing and understanding driving-phase and -activity based emissions factors that reveal underlying causes of the substantial variability concealed in emissions test averages residing in the inventories as Lam et al.18 aptly allude to. Note that the BC and EC terms are interchanged for this discussion; Lam et al.18 provide the conditions for which BC is equivalent to EC. Greater than 95% of U.S. gasoline contains ethanol (www.afdc. energy.gov). Yet, the vast majority of studies to date that investigate SVOC emissions from LDVs use e0 fuel. The de Abrantes et al.8 study is an exception. It examines particle- and gas-phase PAH emissions from a 1998 gasohol (e20-e25) LDV and a 2004 e100 LDV both of which had traveled less than 80 000 km. As noted, PAH are of toxicological interest and their physical state is critical to understanding secondary organic aerosol formation and for predicting the atmospheric chemistry used as input to climate and air quality models, justifying the need for
Figure 1. Results of thermal-optical analysis with each fuel blend grouped by UDC phase. Data for both dynamometer operating temperatures (−7 and 24 °C) are provided. Data points are background subtracted.
more sensitive to the fuel, ambient cell temperature, and LDV driving and operating condition variables. The significance of these data trends was evaluated on a per compound basis using the statistical analysis described, the results of which are given in Table 2. Fourteen of the 18 PAH targeted in this study show significantly higher emissions at −7 °C. In contrast, only 2 of the 21 sat-HCs emitted exhibit a significant rate difference due to the dynamometer cell temperature. Regarding the influence of fuel ethanol content, the measured PAH and sat-HC emissions produced from LDVs burning e0 and e10 show no significant difference. Use of e85 significantly reduces lower molecular weight PAH emissionsanthracene, chrysene, fluorene, methyl fluorene, fluoranthene, phenanthrene, and pyrenebut exhibits surprisingly limited influence on the SVOC emissions otherwise. Significantly higher SVOCs are associated with UDC phase 1 with just two exceptionsheptacosane and triacontane (phase 1 vs 2). A minor subset of sat-HC and phenanthrene show differences when comparing phases 2 and 3, but the overall differences between phases 2 and 3 are largely inconsequential. Proportion of SVOCs in Particle Matter. Again, this study defines the proportion term as the weight fraction (w/w) of an individual SVOC in PM (i.e., quartz filter-collected SVOC/ Teflon filter-collected PM mass). Figure 2B (also see SI Table S5) shows the proportions of individual SVOCs in the LDV PM D
dx.doi.org/10.1021/es403096v | Environ. Sci. Technol. XXXX, XXX, XXX−XXX
1v2
