Chamber simulation on the formation of secondary organic 1 aerosols ( SOA ) from diesel vehicle exhaust in China 2

25 In China primary particulate matter emission from on-road vehicles is predominantly 26 coming from diesels, yet secondary organic aerosols (SOA) formed from diesel 27 emission may be also of greater significance due to more intermediate volatile organic 28 compounds (IVOC) in the exhaust. Here we introduced exhaust from in-use diesel 29 vehicles under warm idling condition directly into an indoor smog chamber with a 30 30m 3 Teflon reactor, and investigated the SOA formation as well as chemical aging of 31 organic aerosols during photo-oxidation. The emission factors of primary organic 32 aerosol (POA) and black carbon (BC) for the three typical Chinese diesel vehicles 33 ranged 0.18-0.91 and 0.15-0.51 g kg-fuel -1 , respectively; and the SOA production 34 factors ranged 0.50-1.8 g kg-fuel -1 with an average SOA/POA ratio of 1.6. Aromatic 35 hydrocarbons could only explain less than 3% of SOA formed during aging, and 36 IVOC and oxygenated VOC might contribute substantially to SOA formation. High 37 resolution time-of-flight aerosol mass spectrometer (HR-ToF-AMS) resolved that 38 POA dominated by CH classes (alkanes, cycloalkanes and alkenes) with high 39 abundances of the CnH2n+1 and CnH2n-1 fragments, and after photo-oxidation the 40 fraction of CH classes and the H/C ratios decreased, while the fraction of CHO, as 41 well as the ratios of O/C and of organic matter to organic carbon (OM/OC), all 42 increased. The plot of f44 (ratio of m/z 44 to the total signal in a mass spectrum) versus 43 f43 indicated that diesel SOA were semi-volatile oxygenated organic aerosols 44 (SV-OOA). The slopes of O:C versus H:C element ratios in the Van Krevelen diagram 45 ranged from -0.47 to -0.68, suggesting a combination of carboxylic acid and 46 2 Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-50, 2016 Manuscript under review for journal Atmos. Chem. Phys. Published: 1 March 2016 c © Author(s) 2016. CC-BY 3.0 License.


Introduction
Air pollution by particulate matter not only adversely affects human health by causing respiratory and cardiopulmonary diseases (Pope et al., 2009;Brook et al., 2010;Liu et al., 2015b；Lelieveld et al., 2015), but also impacts regional and global climate (Ramanathan et al., 2001;Parrish and Zhu, 2009;Wang et al., 2014b).Health risks are of particular concern when heavy fine particle (particulate matter with dynamic diameter less than 2.5 m, PM 2.5 ) pollution occurs in densely populated megacities, such as China's capital city Beijing, which is hard-hit by frequent heavy haze episodes with a large body of people exposed to severe PM 2.5 pollution (Guo et al., 2014;Huang et al., 2014).In urban agglomerations, vehicle exhaust contributes substantially to PM 2.5 , with mass fractions ranging from ~22% in southeastern US (Chen et al., 2012), ~37% in Guangzhou in the Pearl River Delta during wet season (Cui et al., 2015), to as high as 49% in Mexico City (Stone et al., 2008).In particular, People usually expose to much higher air pollutants in urban roadside microenvironments due to traffic-related emission (Zhao et al., 2004;Xu et al., 2008).
Nevertheless, the contribution of vehicle exhaust to PM 2.5 is often a debatable issue.In Beijing, for example, previous studies revealed that contributions of vehicle exhaust to PM 2.5 might range from 4% to 16.3% (Zheng et al., 2005;Song et al., 2006aSong et al., , b, 2007b;;Zhang et al., 2013;Wu et al., 2014), whilst very recently Beijing Municipal Environmental Protection Bureau announced that vehicle exhaust alone accounted for 31% of PM 2.5 mass (http://www.bjepb.gov.cn/bjepb/413526/331443/331937/333896/396191/index.html).One crucial reason for the discrepancies is the lack of understanding about secondary aerosols formed from vehicle exhaust.
Direct motor vehicle emission of PM is predominantly from diesel vehicles (Reff et al., 2009;Zhang et al., 2009).In China diesel vehicles contributed more than 99% of primary vehicle emission of PM although they only account for 15.2% of China's on-road vehicles (MEPC, 2014).Recent studies in Beijing revealed that diesel vehicles contribute 80%-90% of PM emissions from on-road sources (Huo et al., 2011;Wu et al., 2010;Wang et al., 2010).Hence, restriction of diesel vehicles into the core urban areas has become a control measure widely adopted by municipal governments to improve air quality.Besides primary particle emission, vehicle exhaust also contributes substantially to gaseous pollutants, such as volatile organic compounds (VOCs) and nitrogen oxides (NO x ), which can form secondary organic and inorganic aerosols via photo-oxidation (Weitkamp et al., 2007;Robinson et al., 2007;Nordin et al., 2013;Liu et al., 2015a).Nordin et al. (2013) reported that secondary organic aerosols (SOA) formed from gasoline exhaust can reach as high as 500 times that of primary organic aerosols (POA).Although primary PM emission factors of diesel vehicles are typically orders of magnitude higher than gasoline vehicles, recent studies demonstrated that for diesel vehicles the SOA/POA ratios could reach about 3 based on chamber simulations (Chirico et al., 2010;Gordon et al., 2014b).Consequently, contribution of vehicle exhaust to ambient fine particles would become more complicated if considering secondary aerosol formation.
In China, a large portion of gasoline vehicles are produced in Sino-Foreign joint ventures and due to transfer of gasoline engine technology from abroad, chamber simulation study showed that the SOA/POA ratios for China's gasoline vehicle exhaust are quite similar with those reported in the Europe or in the US (Liu et al., 2015a).However, engines equipped on China's diesel vehicles are mainly designed and produced domestically with their technology lagging behind the developed nations.According to previous studies (Yanowitz, 2000;Cheung et al., 2009;Liu et al., 2009), the emission factors of both hydrocarbon and particulate matter for diesel vehicles in China were much higher than those in the developed nations.Therefore, the SOA formation from China's diesel exhaust may be different with those in Europe and the US as well.Furthermore, most diesel vehicles in China are not equipped with emission control aftertreatment devices, which can significantly reduce both POA emission and SOA formation (Chirico et al., 2010;Gordon et al., 2014b).As previous study indicates that even for diesel vehicles SOA might dominate over POA, formation of SOA from diesel vehicles in China would be an issue of wide concern.
In this study, we chose three typical types of diesel vehicles made in China, introduced the exhaust from the diesel vehicles under warm idling condition into an indoor smog chamber with a 30 m 3 reactor, and investigated the SOA formation under photo-oxidation.The main purpose of this study is to obtain a more comprehensive evaluation of diesel vehicle's contribution to carbonaceous aerosols by studying SOA formation from the primarily emitted exhaust.

Vehicles and fuel
Table 1 lists the three diesel vehicles used for our chamber experiments.They

Experimental setup
The experiments were carried out in the indoor smog chamber at Guangzhou Institute of Geochemistry, Chinese Academy of Sciences (GIG-CAS) with a ~30 m 3 Teflon reactor suspended in a temperature-controlled room.Details of setup and facilities about the chamber were described elsewhere (Wang et al., 2014a).Briefly, 135 black lamps (1.2 m long, 60 W Philips/ 10R BL, Royal Dutch Philips Electronics Ltd., the Netherlands) are used as light source, providing a NO 2 photolysis rate of 0-0.49min -1 .
Temperature can be set in a range from -10 to 40C with an accuracy of ±1C, and is measured by eight sensors inside the enclosure and the other one inside the Teflon reactor.In this study, temperature and relative humidity (RH) for all experiment were set to 25C and less than 5%, respectively.Prior to each experiment, the Teflon chamber was flushed with dry purified air for at least 48 hours, which represents at least 5 whole exchanges of the reactor volume.Before each experiment, the chamber Research Inc., USA) operated in alternating mode were used to measure nonrefractory submicron aerosol mass and chemical compositions (Jayne et al., 2000;DeCarlo et al., 2006).The average operating time was 1 min for the high-sensitivity V mode and 1 min for high-resolution W mode.The toolkit Squirrel 1.53G was used to analyze time series of various mass components, and Pika 1.12G was used to determine the average element ratios (http://cires1.colorado.edu/jimenez-group/ToFAMSResources/ToFSoftware/index.html).For elemental analysis, the data were analyzed based on the method described in Aiken et al. (2007Aiken et al. ( , 2008)).The fragmentation table from Allan et al. ( 2004) was used to interpret the AMS data.The contribution of gas phase CO 2 to the AMS m/z 44 signal was corrected by analyzing HEPA filtered air from the smog chamber after filling the exhaust.

Operation Steps
Each experiment consisted of five steps: 1) Introducing exhaust into the chamber from t=-2h.With the injection of exhausts, concentrations of NO x , BC and OA were climbing.Their concentrations when the injection stopped are shown in Table 2.The mixing ratio of NO x increased from 0 to ~1 ppmv; the particle number concentration increased fast from ~2 to ~350,000 particles cm -3 ; the total particle mass concentrations increased from ~0 to over 100 μg m -3 ; and the VOC concentrations also slightly increased at this step.2) Characterizing primary emissions from t =−1.5h.
After completion of injection, the increase of NO x , BC, OA and VOCs were measured against that of CO 2 and CO, and the emission factors were further calculated based on equation (1).3) Adding HONO and propene at approximately t =−0.5h, leading to a

Data analysis
The emission factors (EF) for various pollutants and the production factors (PF) for SOA were calculated on a fuel basis (g kg-fuel -1 ): Where [△P] is the background corrected pollutant concentration in μg m -3 , [△CO 2 ] and [△CO] is the background corrected concentration of CO 2 and CO in the chamber in μg m -3 .MW CO2 , MW CO and MW C are the molecular weights of CO 2 (44.1 g mol -1 ), CO (28 g mol -1 ) and carbon (12 g mol -1 ), respectively.C f is the carbon intensity of the fuel, which was adopted as 0.87 kg C kg-fuel -1 for diesel (Chirico et al., 2010).
Equation ( 1) assumes that all carbon in the fuel was converted to CO 2 and CO, and the contribution from VOC was negligible.This assumption was reasonable, because [△CO 2 ] and [△CO] after introducing exhaust were approximate 100 ppmv and 1 ppmv, respectively, while the increase of VOC was below 5ppbv.The concentrations of hydroxyl radical (OH) during the experiments were inferred from the decay of deuterated butanol measured with the PTR-MS (Atkinson and Arey, 2003).The average OH levels during our experiments were calculated to be approximate 2-510 6 molecules cm -3 , which approached to the levels in the ambient and that in the previous study by Gordon et al. (2014b).The loss of particles and condensable organic vapors onto the reactor walls need to be corrected to accurately quantify particle concentrations in the smog chamber.In this study, the AMS and SMPS data were corrected for wall loss using the method of Gordon et al. (2014b).Briefly, particulate losses were quantified by assuming that the aerosol was internally mixed and thus, organic aerosol (OA) had the same wall-loss rates with BC.Two limiting cases were considered: ω=0, no organic vapors condense to wall-bound particles; ω=1, organic vapors remain in equilibrium with both wall-bound and suspended particles (Weitkamp et al., 2007).
For ω=0, the loss rate of OA to the chamber wall is Where OA wall and OA sus are the wall-bounded and the suspended OA measured at time t, respectively, and k is the wall loss rate constant of BC.
For ω=1, the total concentration of OA at time t (OA total,t ) was estimated as: Where BC(t 0 ) was the initial BC concentration measured before lights were turned on and BC(t) was the BC concentration after lights were turned on for a time span t.
In the experiments with low BC concentrations, we corrected the wall loss effect using exponential fit to the BC data rather than the actual BC data themselves (Gordon et al., 2014b): Where k is the wall loss rate constant of black carbon.Ion enhancement ratios (IER) of the selected ions from the mass spectra of AMS were calculated to evaluate the chemical evolution of POA with aging.The method described by Chirico et al. ( 2010) is used in this study, and the IER is defined as: Where Ion(t) and Ion(t 0 ) are the ion signals at time t and at the time when the lights were just turned on (t 0 ), respectively.
3 Results and discussions

Emission factors of carbonaceous aerosols (BC and POA)
The emission factors of POA and BC are shown in Figure 1.The EF BC were 0.15-0.51g kg-fuel -1 in this study, comparable with those of 0.466-0.763g kg-fuel -1 reported by

SOA formation from diesel vehicle exhausts
Figure 2 shows the typical temporal evolution of gas and particle phase species during PF SOA (0.50-1.8 g kg-fuel -1 ) from this study were much higher than those from previous studies, as the highest PF of SOA reported by Chirico et al. ( 2010) for a medium-duty diesel vehicle (MDDV) at idling condition in Switzerland was merely 0.461 g kg-fuel -1 .The SOA/POA ratios for all the experiments ranged from 0.6 to 2.4 in this study, lower than that of ~3 for a MDDV at idling condition as reported by Chirico et al. ( 2010) and the value of approximately 10 for a heavy-duty diesel vehicle (HDDV) at creep condition in the US as reported by Gordon et al. (2014b).
Comparatively, the highest EF POA and PF SOA for gasoline vehicle exhaust in China were reported to be 0.0004 g kg-fuel -1 and 0.044 g kg-fuel -1 , respectively (Liu et al., 2015a), which are 1-3 orders of magnitude lower than those for diesel vehicle exhaust according to this study.For this reason, diesel vehicle exhaust would still account for

SOA yield from precursor VOCs
Aromatic hydrocarbons were considered as very important anthropogenic SOA precursors (Odum et al., 1997).For gasoline vehicle exhaust, aromatics account for 51-90% of formed SOA (Nordin et al., 2013;Liu et al., 2015a).SOA production from aromatics, including benzene, toluene, C 2 -benzene, C 2 -benzene, C 4 -benzene, was estimated by the following formula: Where SOA predicted (μg m -3 ) is the predicted SOA concentration from precursor i; X i (μg m -3 ) is the mass of the reacted precursor i which was inferred from PRT-ToF-MS data; and Y i (%) is the SOA yield of precursor i.In this study, SOA yields for benzene, toluene, and m-xylene were estimated using the two-product model curves taken from  al. (1997).
As presented in Table 3, the predicted SOA concentrations from traditional aromatic precursors accounted for less than 3% of the observed SOA production.Similarly, Weitkamp et al. (2007) reported that SOA formed from 58 known precursors, including aromatics, alkanes and alkenes, just explained less than 8% of the new particle mass in diesel exhaust simulation.It demonstrated that traditional VOC precursors could not explain the amount of diesel SOA formation.One possible reason is that the yields of aromatic hydrocarbons in complex mixture condition might be higher than those in single precursor condition (Song et al, 2007a), thus the SOA mass was probably underestimated.However, even if we took a higher aromatics yield, such as the effective SOA yield of ~30% reported by Gordon et al. (2014b), the discrepancies between predicted and measured SOA were still huge.The unexplained part is probably from the photo-oxidation of the intermediate volatile organic compounds (IVOCs) (Weitkamp et al., 2007;Robinson et al., 2007), such as C 13 -C 20 n-alkanes (Miracolo et al., 2010).
It worth noting that there would be other oxygenated species, like glyoxal and methyl glyoxal, were considered as potential SOA precursors (Volkamer et al., 2006;Carlton et al., 2007;Fu et al., 2008;Kamens et al., 2011).Glyoxal and methyl glyoxal can be primarily emitted from vehicles (Zhang et al., 2016), or secondarily formed by the photo-oxidation of VOCs (Carlton et al., 2007;Healy et al., 2008;Volkamer et al., 2009).The PTR-ToF-MS measured ion m/z 59 represents acetone and glyoxal, and Very similar to those in previous studies (Weitkamp et al., 2007;Chirico et al., 2010;Presto et al., 2014) indicates that PAHs, probably along with other POA species, were oxidized through heterogeneous reaction, or vaporized to gas phase, oxidized and then condensed back onto particles (Donahue et al., 2006;Robinson et al., 2007).
The f 43 versus f 44 triangle plot for OA in our experiments is presented in Figure 8

Conclusions
In this paper, chamber simulations were conducted to investigate SOA formation from diluted exhaust of three types of diesel vehicles widely used in China.EF POA and PF SOA in this work were 0.19 and 0.61 g kg-fuel -1 for JAC, 0.18-0.34and 0.56-0.76g kg-fuel -1 for Foton, 0.72-0.91 and 0.50-1.8g kg-fuel -1 for Changan, respectively, which were all higher than those reported in previous studies in Europe and in the US.
These EF POA and PF SOA values were also 2-3 and 1 orders of magnitude higher than those of gasoline vehicle exhaust.The color scale represents time evolution.

13
Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2016-50,2016   Manuscript under review for journal Atmos.Chem.Phys.Published: 1 March 2016 c Author(s) 2016.CC-BY 3.0 License.a smog chamber experiment (Experiment 8).As showed in Figure2(c), wall-loss corrected OA started to climb with the formation of SOA after turning on the lights.After 5 h photo-oxidation, the wall loss corrected OA in Experiment 8 increased from 64 μg m -3 to 112 μg m -3 for =0 case and to 166 μg m -3 for =1 case.However, while the median diameters were increasing with aging (Figure3(a)), there was no sign of increasing particle numbers, indicating few new particles were formed.Moreover, as shown in Figure3(b), after about half an hour photo-oxidation (t=0.4), the number of small particles dropped fast, whereas the number concentration of larger particles almost did not change, and the peak diameter slightly increased.It suggested particle growth by coating of the newly formed SOA on existing particles, consistent with the results ofWeitkamp et al. (2007).

14
Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2016-50,2016   Manuscript under review for journal Atmos.Chem.Phys.Published: 1 March 2016 c Author(s) 2016.CC-BY 3.0 License.a larger portion of traffic-related primary and secondary OA, despite higher faction of gasoline vehicles (83.5%) against that of diesel vehicles (15.2%) in China(MEPC,     2014).As reported in China Energy Statistical Yearbook (2013), the fuel consumption of diesel and gasoline for transportation is 107.27 and 37.53 million tons in 2012, so with the POA emission factors and SOA production factors available in this study and in a previous study for gasoline exhaust(Liu et al., 2015a), the diesel derived OA would dominate overwhelmingly over the gasoline derived OA in the traffic-related OA.

Figure 5
Figure5showed the average AMS mass spectra for the total OA measured in the (a).Compared with POA, SOA had higher f 44 and lower f 43 values.According to Ng et al. (2010), ambient low-volatility oxygenated OA (LV-OOA) and SV-OOA factors fall in the upper and lower portions of the triangle, respectively.Similar with that reported by Presto et al. (2014) for diesel SOA, the plots of SOA were within the range of Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2016-50,2016 Manuscript under review for journal Atmos.Chem.Phys.Published: 1 March 2016 c Author(s) 2016.CC-BY 3.0 License.SV-OOA.Figure 8(b) translates the AMS data into van Krevelen space.The van Krevelen slopes for this study ranged from -0.47 to -0.68, which were similar to that reported by Presto et al. (2014) for diesel exhaust experiments, and also in the range of slopes for ambient OOA factors observed by Ng et al. (2011).It indicates that the SOA chemistry observed in present smog chamber experiments is atmospherically relevant (Presto et al., 2014).According to Heald et al. (2010) and Ng et al. (2011), slopes of -1 and 0 in the Van Krevelen diagram represent chemical reaction for addition of carboxylic acid and alcohol/peroxide, respectively.Therefore, the van Krevelen slopes in this study suggest that SOA formed was a combination of carboxylic acids and alcohols/peroxides.

Figure 1 Figure 2
Figure 1 Emission factors of BC and POA and production factors of SOA for

Figure 3
Figure 3 (a) Particle size-number concentration distributions as a function of time.745

Figure 8
Figure 8 (a) The fractions of total organic signal at m/z 43 (f 43 ) versus m/z 44 (f 44 ) for769 771 , POA spectra in the beginning were dominated by the C n H 2n+1 C and OM/OC ratios and decrease of H/C ratios after photo-oxidation were found during all the experiments (Table2), further confirming the increased oxidation state of OA during aging.It should be pointed out that the increase of oxidation state is due to not only the formation of SOA, but also further oxidation or degradation of POA.For example, as shown in Figure7, the IERs of polycyclic associated with cycloalkanes and alkenes(Canagaratna et al., 2004).The POA from all the experiment show that the largest signal contribution is from m/z 43 (47% of which is C 3 H 7 + , and 52% of which is C 2 H 3 O + ), followed by m/z 41 (97 % of which is Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2016-50,2016 Manuscript under review for journal Atmos.Chem.Phys.Published: 1 March 2016 c Author(s) 2016.CC-BY 3.0 License.CH class fell from 67.6% to 48.3%, whereas that of the CHO class grew from 29.2% to 47.6%, indicating a more oxidized chemical property.The CHN and CHON classes contributed 2.3% and 1.7%, respectively; they were also minor components although Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2016-50,2016 Manuscript under review for journal Atmos.Chem.Phys.Published: 1 March 2016 c Author(s) 2016.CC-BY 3.0 License.
Therefore, although diesel vehicle population is much less than that of gasoline in China, it still plays a vital role in the contribution of primary and secondary OA.It should be noted that all the experiments for SOA formation from both gasoline and diesel vehicles in China were conducted under 20 Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2016-50,2016 Manuscript under review for journal Atmos.Chem.Phys.Published: 1 March 2016 c Author(s) 2016.CC-BY 3.0 License.idling condition.The emission of POA as well as formation SOA under different operating modes, especially on-road conditions, deserves further investigation.Atmos.Chem.Phys.Discuss., doi:10.5194/acp-2016-50,2016 Manuscript under review for journal Atmos.Chem.Phys.Published: 1 March 2016 c Author(s) 2016.CC-BY 3.0 License.
a .measured SOA were wall loss corrected at ω = 0.