Secondary organic aerosol formation from gasoline and diesel vehicle exhaust under light and dark conditions †

Secondary organic aerosol (SOA) formed from vehicle exhaust contributes substantially to the atmospheric particulate matter in urban air but there still remain uncertainties in the simulation of the SOA by air quality models. This omission is due partly to uncertainties about the di ﬀ erences in SOA formation between vehicle types, exhaust aftertreatment devices, and oxidation conditions in the air. In this work, smog chamber experiments were conducted to investigate SOA formation from diluted exhaust gases emitted by two light-duty gasoline vehicles and one passenger and one medium-duty diesel vehicles. Emissions were aged in the smog chamber under light (ultraviolet) and dark (with high O 3 ) conditions. In these experiments, there were considerable emissions of elemental carbon and primary organic aerosol by a gasoline direct-injection vehicle and a diesel truck, but the corresponding emissions were relatively small from a gasoline vehicle with port fuel injection and a passenger diesel vehicle. However, signi ﬁ cant amounts of SOA were produced from all four vehicles. Box model simulations indicated that OH exposure was higher and the ratio of NO to peroxy radicals (HO 2 + RO 2 ) was lower under dark conditions in the presence of high added O 3 than under light conditions. The higher concentrations and yields of SOA under dark conditions than under light conditions could partly be explained by the lower NO/(HO 2 + RO 2 ) ratios under dark conditions. Simulations with a volatility basis-set model indicated that aromatic compounds were the dominant precursors of SOA from gasoline vehicles (64 – 84%), whereas unspeciated intermediate-volatility organic compounds (IVOC) were the dominant precursors for the production of SOA from diesel vehicles (75 – 87%). The observed O : C ratios were higher for gasoline SOA (0.7 – 0.8) than diesel SOA (0.3 – 0.4), and this di ﬀ erence was explained by the di ﬀ erences in their precursors. SOA formed under the light and dark conditions was similar in the dominant precursors and oxidants, as well as the O : C ratios, but di ﬀ erent in the NO/(HO 2 + RO 2 ) ratios during the oxidation processes. Thus, we cannot conclude that SOA formed under the light and dark conditions has similar chemical composition, and their characteristics should be further examined in future studies.


Introduction
][3][4][5] Carbonaceous aerosol, including elemental carbon (EC) and organic aerosol (OA), are major components of particulate matter (PM) from motor vehicles.1][12][13][14] Those studies have revealed that SOA production is strongly affected, in addition to the vehicle characteristics, by the experimental setups, such as reactor types (smog chamber or ow reactor) and oxidation conditions (e.g., light versus dark conditions). 15The estimated contributions of secondary PM from motor vehicles are therefore associated with large uncertainties.Because the SOA generated from the vehicle exhaust may be more toxic than SOA formed from the individual volatile organic compounds (VOC) or POA from vehicle exhaust, 16,17 there is a need to better understand the characteristics of SOA generated from vehicle exhaust.
Both diesel and gasoline vehicles have been shown to be important sources of atmospheric SOA, and their relative contributions differ among studies and regions.In the United States (U.S.), most previous studies have indicated that gasoline vehicles make greater contributions to SOA. 15,[18][19][20][21] For example, contributions of gasoline-exhaust are $56-79% to vehicle SOA over the U.S. 18 or $90% to vehicle SOA over the U.S. and Southern California. 19,21In contrast, Gentner, et al. 22 have indicated that diesel vehicles make higher contributions ($90%) to vehicle-SOA in the U.S. In Europe, Dunmore, et al. 23  have shown from their comprehensive measurements of organic compounds in London that diesel vehicles make higher contributions to atmospheric SOA.5][26] In a GDI engine, gasoline fuel is sprayed directly into the cylinder, and the compression ratio is relatively high compared to a GPI engine, in which gasoline is injected into the intake port.As a result, GDI vehicles have the advantages of better fuel efficiency and lower CO 2 emissions than GPI vehicles.However, primary PM emissions are higher from GDI vehicles than from GPI vehicles. 9,24,258][29][30][31] Understanding the SOA characteristics of different vehicle types is thus necessary for accurate estimation of the contributions of different sources to atmospheric SOA.
During the last decade, it has been shown that speciated VOC (e.g., single-ring aromatics and aliphatic compounds) cannot fully explain the production of SOA from vehicle exhaust and that the contributions of unspeciated precursors are critical. 12,13,19,28,29,32,33To estimate those contributions, emissions and formation rates of SOA from intermediate-volatility organic compounds (IVOC, with effective saturation concentrations (C*) of 10 4 -10 6 mg m À3 ) have been evaluated.A large part of IVOC is chemically unresolved and thus is not explicitly included in emission inventories. 34,35Lu, et al. 31 have estimated that unspeciated cyclic compounds make the predominant contributions to gasoline-exhaust IVOC, whereas unspeciated linear and branched alkanes make the greatest contributions to diesel-exhaust IVOC.In addition, the magnitude and chemical composition of the IVOC emitted from motor vehicles depends on the vehicle type and operating conditions, 13,19,28,29,36 and contributions of IVOC to the formation of SOA from vehicle exhaust are still uncertain.Because the yields of SOA produced from IVOC are high, 37,38 IVOC are important precursors of SOA from motor vehicles, 28,29,35 Lu et al. 39 have recently developed a parameterization of SOA formation from six lumped IVOC emitted from mobile sources; their model simulation has indicated that IVOC emitted from mobile sources make important contributions to the SOA simulated in Los Angeles.IVOC may make large contributions to ambient SOA, and thus, quantication of their contributions is important.
As analyses using numerical simulations are necessary to quantitatively evaluate the contributions of these emission sources to atmospheric PM, there have been attempts to simulate the production of SOA from motor vehicles. 40Robinson et al. 10 have shown that semi-volatile organic compounds (SVOC) make important contributions to SOA from diesel vehicle exhaust by using a volatility basis-set model (VBS) that explicitly calculates the partitioning and multigenerational reactions of POA, SVOC, and IVOC. 10,41Recent modeling analyses have indicated that IVOC is a key precursor to atmospheric SOA from diesel exhaust. 15,42,43revious experimental and modeling studies of SOA formation from vehicle exhaust have focused primarily on light conditions.Only a few studies have addressed the aging of vehicle SOA under dark conditions.Samy and Zielinska 44 have conducted chamber experiments of diesel exhaust aging under dark conditions with high concentrations of N 2 O 5 (for nitrate radical [NO 3 ] production).They have found that there is less formation of SOA under dark conditions than under conditions where photooxidation is possible.However, to our knowledge, the yield of SOA formed under dark conditions with high O 3 concentrations has not been well evaluated.Studies of the toxicity of SOA from vehicle exhaust have made use of SOA produced under dark conditions with high O 3 concentrations, partly because the large amounts of SOA necessary for the toxicological studies could be prepared relatively simply without the light sources required for photooxidation. 45,46ecause of the addition of high concentrations of O 3 , the main oxidants and the ratio of NO to peroxy radicals could be different between the light and dark experiments.Comparison of the formation mechanism and properties of SOA formed under light and dark conditions should provide insights into the toxicological properties of SOA from vehicle exhausts.For the analysis of SOA oxidation under dark conditions, simulations of gas-phase chemical mechanisms can provide useful insights into the processes involved in SOA formation.However, to our knowledge, no such analyses have been conducted.
In this study, SOA formation associated with dilute emissions from two light-duty gasoline vehicles as well as one lightduty and one medium-duty diesel vehicles was investigated under both light and dark conditions in a smog chamber.The aims of this study were (1) to compare the production of SOA and the oxygen : carbon (O : C) ratios of OA from different vehicle types and oxidation conditions (i.e., light and dark conditions), (2) to evaluate the ability of the VBS model to simulate the formation of SOA and O : C ratios, (3) to evaluate the contributions of different precursors to SOA formation.This experimental and modelling study allowed us to investigate the characteristics of SOA from both gasoline and diesel exhaust under light and dark conditions.

Experimental
Table 1 lists the four vehicles tested during this study.There were two light-duty gasoline vehicles, including one GDI vehicle and a GPI vehicle.The light-duty passenger diesel (Diesel-P) vehicle was equipped with a diesel particulate lter (DPF) and diesel oxidation catalyst (DOC).The medium-duty diesel truck (Diesel-T) was equipped with a DOC only.The loads of the test vehicles were consistent with Japanese emission regulation test protocols and equaled the vehicle weights +110 kg.The emission standards of the test vehicles were compliant with the 2009 emission regulations. 47The average vehicle age in Japan (8.7 year for passenger vehicles and 11.4 year for freight vehicles in 2020, Automobile Inspection and Registration Information Association) indicates that model year 2012 vehicles are still being used.The fuels and lubricating oils used in this study were those commonly used in Japan.The percentages of aromatics contained in the used regular gasoline fuel and lowsulfur-content (6 ppm) diesel fuel were measured to be 24.8% and 22.5%, respectively.The densities of the used gasoline and diesel fuels were 0.73 g cm À3 and 0.84 g cm À3 , respectively.The driving conditions corresponded to the Japanese official transient mode (JC08 mode) under cold-start conditions. 16,24,48ontrolled vehicle testing showed that during cold starts the DOC was not fully active and therefore unburned and incompletely combusted material passed through the exhaust aertreatment systems. 27,49he experiments were carried out in a 6 m 3 uorinated ethylene propylene (FEP) lm bag chamber (FBC).The dimensions of the FEP lm bag were 1.4 Â 1.4 Â 2.4 m 3 , and the lm thickness was 50 mm.Fig. 1 shows the experimental setup.All exhaust during the JC08 cycle was introduced directly into a 30 m 3 exhaust-gas diffusion chamber (EDC) that was lled with clean air.The clean air was supplied by a clean-air generator (DAR-2200, Horiba Ltd.) and contained less than 0.1 ppm of total hydrocarbon (THC) and less than 0.1 ppm of NO.The duration of the exhaust injection period was about 20 min.Aer exhaust injection, there was a period of 60 min to allow mixing in the chamber and to characterize the composition of the fresh exhaust.The partially diluted exhaust gases in the EDC were then transferred by three pumps (OSP-90W, As One Corp.), and nally the reaction air in the 6 m 3 FEP FBC was prepared with target concentrations.The total dilution factors from the raw exhaust to the 6 m 3 FBC were $25 for gasoline vehicle exhausts and $10 for diesel vehicle exhausts.
An array of instruments was used to characterize gas-and particulate-phase pollutants inside the 6 m 3 FBC.Gas-phase organic species were measured with two proton transfer reaction mass spectrometers (PTR-MS, Ionicon), as detailed in our previous studies 50,51 and Section S1 of the ESI.† The detection sensitivities are listed in Tables S1-S3 of the ESI.† Gas monitors were used to measure NO, NO 2 , CO, O 3 , non-methane hydrocarbons (NMHC), THC, and CH 4 .The monitors were zeroed daily and calibrated at least weekly.Concentrations of black carbon (BC) were measured with a black carbon monitor (Model AE51, Aethlabs).Nonrefractory PM 1.0 concentrations and the chemical composition of PM 1.0 were measured with a highresolution time-of-ight aerosol mass spectrometer (HR-ToFAMS, Aerodyne, Inc.). 52For the estimate of O : C ratios, the contributions of gas-CO 2 to the mass spectrum were subtracted, as in previous studies. 53The uncertainties of the OA concentrations and O : C ratios that were used in this study were 20-40% (ref.54 and 55) and $20%, 53 respectively.Particle number distributions were measured with scanning mobility particle sizers (SMPS Models 3034, 3936, TSI, Inc.).
The data shown in this study were obtained in the 6 m 3 FBC.The ratio of the CO concentration in the 30 m 3 EDC and 6 m 3 FBC (i.e., chambers before and aer passage through the pumps, respectively) was 0.183; the corresponding ratio of the THC concentrations was 0.194.As CO is unlikely to be lost in the pumps, the similar ratios for the CO and THC across the two chambers suggest that THC was not lost in the pumps either.In contrast, particle volume concentrations normalized to CO concentrations were reduced by a factor of $7 aer passage through the pumps (i.e., $86% of particles were lost in the pumps).The fact that the geometric mean diameter of the volume concentration changed marginally (284.0 nm and 273.0 nm before and aer passage through the pumps, respectively) suggested that evaporation reduced particle volume concentration only by 11% and particle loss was predominantly associated with wall loss.We should note that we analyzed POA in the 30 m 3 EDC and SOA in the 6 m 3 FBC.If we assume that the precursors of SOA were not lost in the pumps, then we can assume that the POA/SOA ratio was accurately estimated in this study.It is possible that some of the SVOC were deposited on the pump wall (materials of hard anodized aluminum), even though THC did not decrease aer passage through the pumps.Thus, the observed concentrations of SOA might be underestimated because of the loss of SVOC.
For photoreaction experiments, the 6 m 3 chamber was irradiated with 100 ultraviolet (UV) black-light bulbs (350BL, Hitachi Ltd.) from four sides (25 bulbs in each side).The wavelength of the peak intensity of the UV black lights was 350 nm, and the radiant ux of each bulb was 40 watts.The rate of NO 2  Quantication of SOA production in the 6 m 3 FBC required that losses of both particles and vapors to the chamber walls be corrected.In this study, the particle wall loss (PWL) rate was calculated from experimental data, whereas the volatilitydependent vapor wall loss (VWL) rate was estimated using a model calculation, as detailed in Section 2.2.The PWL was treated as a rst-order process with a rate constant determined from the measured rate of decay of aerosol volume concentrations aer the formation of SOA, as summarized in Table S4 of the ESI.† For the experiments of diesel vehicles, the geometric mean volume diameter of aerosol did not change aer 2.5 h from time t ¼ 0 (light on or O 3 injection).The decay rates of aerosol volume concentrations measured by SMPS were thus used to determine the PWL rates (0.21-0.45 h À1 ).The maximum diameter of the aerosol that the SMPS could measure in the diesel-vehicle experiments was $790 nm, but for the gasoline vehicle experiments, the corresponding maximum diameter was only $470 nm.Because the diameters of some aerosol exceeded 470 nm in the gasoline vehicle experiments, the volume concentration of the SOAs could not be fully captured by the SMPS.The PWL rate was therefore estimated not from the SMPS data, but from the decay rate of the EC concentrations (0.19 h À1 ).For the GDI experiment under dark conditions, the number concentration decreased at a high rate of 1.4 h À1 in the rst 1.5 h of the experiment.Because particle diameters increased during this experiment, the volume concentration could not be used to calculate the PWL rate; therefore, we made a second-best estimate by using number concentrations as alternatives to correct the PWL 57 in the rst 1.5 h of that experiment.Because the uncertainties associated with this estimate were undoubtedly large, the results of this experiment should be interpreted with caution (e.g., Fig. 2).Aer the rst 1.5 h of that experiment, the PWL correction rate was the same rate that was used in the other experiments (0.19 h À1 ).
Table 2 shows the initial experimental conditions.The NO x emitted from diesel vehicles was much higher than the NO x emitted from gasoline vehicles.Initial NO x concentrations were 0.27-0.36ppm for gasoline vehicles, 4.4-4.5 ppm for the Diesel-P vehicle, and 16.7-17.9ppm for the Diesel-T vehicle.In general, the NMHC/NO x ratio during the experiments for gasoline exhaust was >10 ppmC per ppm-NO x .Under those low-NO x conditions, peroxy radicals react primarily with HO 2 radicals to form organoperoxy species. 58,59In contrast, the NMHC/ NO x ratio during the experiments with diesel exhaust was <1 ppmC per ppm-NO x .Under those conditions, peroxy radicals react primarily with NO to form NO 2 and an alkoxy radicals.Peroxy radicals can also react with NO to form organic nitrates, and nitrate-containing SOA species have been observed under high-NO x conditions. 58ackground-corrected emission factors (EF in mg per kgfuel) for EC and POA and the production factors for SOA were calculated on a fuel-consumption basis using a carbon mass balance as follows: where [P] is the background corrected pollutant concentration (mg m À3 ); CO 2 , CO, and THC are the background-corrected CO 2 , CO, and THC concentrations (mg m À3 ); MW CO 2 , MW CO , MW THC , and MW C are the molecular weights of CO 2 , CO, THC, and carbon (g mol À1 ), respectively, and C f is the carbon content of fuel, which we assumed to be 0.85 (kg-carbon per kg-fuel). 60ecause the carbon contribution from THC was negligible compared to the contributions from CO 2 and CO in every experiment, it was unnecessary to include the THC species in calculating the emission factors.

Model
In this study, we conducted box model simulations with the VBS framework 41 coupled with the gas-phase chemistry mechanism SAPRC99. 61hree types of SOA precursors were considered: VOC (C* $ 10 7 mg m À3 ), IVOC (C* ¼ 10 4 -10 6 mg m À3 ), and SVOC (C* ¼ 10 À1 to 10 3 mg m À3 ).Table S5 of the ESI † lists the lumped VOC classes used in SAPRC99.The organic compounds measured by PTR-MS in this study were then mapped to each lumped VOC class for the initial conditions of the model simulation (Table S6 of the ESI †), as is usually conducted for air quality simulations.Any alkane or alkene with a carbon number (N C ) $ 12 and any single-ring aromatic with N C $ 11 measured by PTR-MS was not apportioned to the lumped VOC classes because these compounds are classied into IVOC.IVOC emissions were introduced into the model in proportion to NMHC emissions, as in previous studies. 7,19,28,29Initial concentrations of ve lumped IVOC were calculated by using the ratio of IVOC to NMHC estimated by Lu, et al.S6 of the ESI.† The IVOC collectively accounted for 4.3% and 50.1% of NMHC emissions of gasoline and diesel exhaust, respectively.The mass fraction distribution of SVOC in C* bins (10 À1 to 10 2 mg m À3 ) were taken from Zhao, et al. 28 Zhao, et al., 29 and the initial SVOC concentration was normalized so that the initial OA concentration could be reproduced using the equation ½OA ¼ , where [OA] is the OA concentration and C i is the total (gas + aerosol) concentration of compound i with a saturation concentration of C * i .Products of VOC oxidation were represented as four surrogate compounds with saturation concentrations ðC * i Þ of 1, 10, 100, and 1000 mg m À3 .NO x -dependent SOA mass yields of the oxidation products from C7-C11 alkanes (ALK5), monoaromatic compounds (ARO1 and ARO2), and isoprene were taken from Murphy and Pandis 62 and are listed in Table S7 of Fig. 2 Emission factors of elemental carbon (EC) and primary organic aerosol (POA) and production factors of secondary organic aerosol (SOA) from two gasoline-powered and two diesel-fuel motor vehicles (GDI: gasoline-direct injection, GPI, gasoline port-fuel injection: Diesel-P, diesel-fuel passenger car: Diesel-T, diesel-fuel truck).# The SOA data in the GDI experiments under dark conditions include large uncertainties associated with the correction for particle wall loss (PWL), as noted in Section 2.1.the ESI.4][65] In this study, we considered SOA formation from OLE2 (internal alkenes), but not from OLE1 (terminal alkenes), as discussed in Section S2 of the ESI.† Products of IVOC oxidation were represented as four surrogate compounds with C * i of 0.1, 1, 10, and 100 mg m À3 , and the SOA mass yields were taken from chamber experiments under high-NO x conditions (Table S9 of the ESI †). 39SOA mass yields from alkane IVOC under high-and low-NO x conditions are similar, 66 whereas those from aromatic IVOC are higher at low-NO x condition than at high-NO x condition. 38In our simulations, aromatic IVOC has a small contribution to SOA production (Section 3.3.2),and thus, NO x dependence of SOA yields from IVOC were not considered in this study.Formation of SOA from primary-SVOC followed Robinson, et al.10 Multi-generational aging reactions for products from VOC and IVOC were not considered as well.Gas-particle partitioning was calculated using absorptive partitioning theory 67 assuming that all of the organics formed a quasi-ideal solution and that the bulk gas and particle phases were always in equilibrium.The temperature dependence of the saturation concentration was estimated with the previously reported vaporization enthalpy. 39,68We should note that SAPRC99 is a lumped mechanism; hence it cannot represent the detailed behavior of individual VOCs.However, SAPRC99 and the later version of the SAPRC mechanism have been among the most widely used chemical mechanisms in air quality models, and we therefore used the SAPRC99 mechanism for the evaluation of the experimental data.
The O : C ratio of the POA was based on the individual experimental results.Murphy

Emission and production factors of carbonaceous aerosol
Fig. 2 and Table S11 † summarize EF of primary particles (EC and POA) and production factors (PF) of SOA.For the evaluation of variability, two experiments were conducted to measure concentrations of for POA and EC emitted by the GDI and Diesel-P vehicles.The standard errors ([standard deviation]/ [average]) of the POA and EC concentrations were 23% and 16%, respectively for the GDI vehicle.The relative error for the Diesel-P vehicle was 46% for POA, and the EC concentration was below the limit of detection.
For gasoline vehicles, primary emissions of EC and POA from GPI were small (<4 mg per kg-fuel), whereas the EF of EC and POA from GDI were 55 mg per kg-fuel and 19 mg per kg-fuel (2.4 mg km À1 and 0.83 mg km À1 ), respectively.The higher emissions of primary particles from GDI vehicles were consistent with results of previous studies, 9,14,25,74 though the reported contributions varied among vehicles.We should note that sample size of this study is small (one experiment for each condition and vehicle), and thus, this comparison with previous experiments is solely intended for the check of the consistency, but not for the demonstration of the representativeness of the tested vehicles.The ratio of EC to [EC + POA] of 0.75 from GDI in this study was within previous estimates of 0.20 in Du, et al., 25 0.80-0.86 in Platt, et al., 14 and 0.88 in Saliba, et al. 9 As summarized in Table S12 of the ESI, † the variability of the EF of EC and POA has been large in previous studies (>2 orders of magnitude).The EF differ, for example, among model years, engine types, emission control technologies, and operating modes.Discussion of this point is beyond the scope of this study because of the small sample size, although the EF of the EC and POA in this study were within the range of those in previous studies, both for GDI and GPI vehicles.In contrast to the EF of the POA, the PF of the SOA were higher for the GPI vehicle (40 mg per kg-fuel) than for the GDI vehicle (21 mg per kg-fuel) in the light conditions.These PFs were within the range of previous research for gasoline vehicles under light conditions.Gordon, et al.12 have reported SOA PFs of exhaust from 6 gasoline vehicles with LEV2 emission standards (manufactured in 2004 or later) that ranged from 2 to 55 mg per kg-fuel with an average of 18.8 mg per kg-fuel at an OH exposure of 5 Â 10 6 molecules per cm 3 per h.Gordon, et al.12 have also found that SOA is much more efficiently generated during cold-start tests than during hot-start tests.Nordin, et al. 75 have reported a higher SOA-PF (480 mg per kg-fuel) from an older regulation (2005) gasoline vehicle (Euro 4) with an OH exposure of around 5.0 Â 10 6 molecules per cm 3 per h under the cold-start conditions.Du, et al.25 have reported that the SOA-PF from GPI vehicle (with an emission standard equivalent to Euro 4) and GDI vehicle (with an emission standard equivalent to Euro 5) under light and cold-start conditions were 21 and 55 mg per kgfuel, respectively.As already noted, SOA-PF should depend on the type of vehicle, emission control technology (e.g., aertreatment devices), and driving conditions.In addition to these differences, oxidation conditions are also important determinants of SOA-PF.For both GDI and GPI vehicles, the SOA-PF were higher under dark conditions than under light conditions, even at similar levels of the OH exposure.This point is discussed in Section 3.3.1.
For diesel vehicles, the Diesel-P with DPF and DOC emitted negligible primary particles, whereas the EF of EC and POA from the Diesel-T with DOC were 153 mg per kg-fuel and 18.4 mg per kg-fuel, respectively.These ndings are consistent with previous work, which has indicated a signicant reduction of primary particles with the use of DPFs (Table S12 of the ESI †).For example, Gordon, et al. 13 have shown that emission factors of primary particles from heavy-duty diesel vehicles (HDDV) with DPF were <12 mg per kg-fuel, whereas the EF of mediumduty diesel vehicles (MDDV) without DPF exceeded 100 mg per kg-fuel.In contrast to the primary particles, the SOA-PF from the Diesel-P was even higher under both light and dark conditions (35 mg per kg-fuel and 288 mg per kg-fuel, respectively) than that from the Diesel-T (7 mg per kg-fuel and 235 mg per kgfuel).The SOA-PF from DPF-equipped vehicles varied among studies (0-85 mg per kg-fuel under light conditions, as in Table S12 †).As will be shown in Section 3.3.1,production yields of SOA from diesel vehicle exhaust were higher under dark conditions than under light conditions.We should note that the SOA-PF of this study under light conditions was within the range of previous studies with DOC or DPF (Tables S11 and S12 of the ESI †).

Gas-phase reactions
Box model calculations with the chemical mechanism, SAPRC99, revealed the different characteristics of major chemical reaction pathways between vehicle types (gasoline or diesel vehicles) and between oxidation conditions (light or dark).Fig. 3 indicates the production pathways of RO x radicals (sum of OH, HO 2 , and organo-peroxy [RO 2 ] radicals).Fig. S3 of the ESI † is the same gure on a linear timescale.
The box model calculations indicated that under the light conditions, the dominant radical production process in the case of the gasoline exhaust was photolysis of carbonyl compounds followed by ozonolysis of alkenes and HONO photolysis.Photolysis of carbonyl compounds has been reported to make large contributions to the production of radicals in the ambient air of urban areas 76,77 and of an oil and gas basin. 78In the experiments with diesel exhaust under the light conditions, the NMHC/NO x ratio was low (Table 2), and because we expected that irradiation of the exhaust would not produce sufficient HO x radical to facilitate SOA formation, we added H 2 O 2 .The photolysis of H 2 O 2 was therefore the dominant radical production pathway, followed by photolysis of HONO and carbonyl compounds.
With the addition of high concentrations of O 3 under the dark conditions (5-20 ppmv, Table 2), reactions of alkenes and O 3 /NO 3 were the major radical production pathways during the rst several minutes of the experiments.Over the rst half hour, reactions between alkenes and O 3 made the highest contributions to radical production from gasoline exhaust, whereas reactions involving carbonyls and alkenes + NO 3 made the highest contributions in the case of diesel exhaust.This contrast in the contributions of oxidants reected the different initial NMHC/NO x ratios (Table 2).One hour aer t ¼ 0, reactions involving carbonyl and NO 3 made the highest contributions to radical production, production and loss of peroxyacyl nitrates (PANs) balanced, and OH radical concentrations decreased (Fig. S4 of the ESI †).We should note again that these radical production pathways are not representative of the reactions under ambient conditions; the chemical reactions that occur under dark conditions are intended to simulate reactions that might occur during toxicological experiments, not chemical reactions that occur in the atmosphere at night.
The reaction pathways of RO 2 radicals are critical to the determination of the SOA yields.RO 2 radicals react with NO when NO x concentrations are high, whereas RO 2 radicals react with other RO 2 and HO 2 radicals when NO concentrations are low.Previous studies have revealed that NO x concentrations signicantly alter yields of SOA from biogenic VOC 79,80 and aromatic VOC 81 but perhaps not for alkane VOC. 66The difference between the SOA yields at high-and low-NO x conditions could be explained by the different product distributions.RO 2 radicals react with NO to form RO at high-NO x conditions.Then, carbonyl or hydroxycarbonyl products are formed aer RO reacts with O 2 , decomposes through C-C bond scission, or isomerizes through H-atom shi reactions. 82RO 2 radicals also react with NO or NO 2 to form organic nitrate or peroxynitrates.In contrast, RO 2 radicals react with HO 2 or other RO 2 radicals at low NO x conditions, and generate hydroperoxide, carbonyls, hydroxycarbonyls, or alcohol products.The NO x dependence of the SOA yield, a i , is parameterized as follows: where a i,high NO x and a i,low NO x are the mass-based stoichiometric yields of product i under high and low NO x conditions, respectively; b is the branching ratio of the fraction of RO 2 radicals that react with NO; and R(RO 2 + X) is the reaction rate of RO 2 and X.Under the light conditions, RO 2 radicals reacted predominantly with NO, and SOA was produced under the high-NO pathway (Fig. 3).In contrast, under dark conditions, b was below 0.5 at t # 0.5 h, when SOA production occurred.These low values of b were associated with high O 3 concentrations and low NO concentrations, and SOA was assumed to be produced under the low-NO pathway.This point is discussed further in Section 3.3.1.During all the experiments, SOA concentration increased with OH exposure.For gasoline exhaust, the concentrations of SOA were clearly higher under the dark conditions than that under the light conditions at the same OH exposure level.The concentrations of SOA from gasoline exhaust under light conditions started to increase about 1-2 h aer t ¼ 0 (Fig. S5 of the ESI †).The delay of SOA formation in the case of the gasoline vehicle exhaust was consistent with results of previous experiments, 75 which have shown that the formation of SOA from idling gasoline passenger vehicles is delayed by 30-45 min aer lights are turned on.In contrast, under the dark conditions, SOA formation occurred immediately aer t ¼ 0 and continued for about 30 min.This contrast results not only from the different behavior of OH radical, but also from the different NO x regime, as already discussed in Section 3.2.These differences of SOA concentrations (in turn, OA concentrations) reected the fact that SOA yields were also higher under dark conditions (20-35%) than under light conditions (2-15%) (Fig. 5).Our estimated SOA yields were within the range of those estimated for six gasoline vehicles with the LEV2 emission standard, 12 although the SOA yields of the six vehicles were highly variable (3-46%).
The concentration of SOA was higher from diesel exhaust than from gasoline exhaust at similar levels of OH exposure (Fig. 4).The SOA mass yield from diesel exhaust was higher under the dark conditions (60-80%) than under the light conditions (10-20%) (Fig. 5).The high mass yields of SOA when OA concentrations were high reected the semi-volatile nature of SOA (more oxidation products partition into the particle phase when OA concentrations are high).We should note that the NMHC monitor could not measure some portion of oxygenated VOC, 83 and the estimated SOA yield from VOC could therefore have been overestimated.Fig. 5 shows the SOA mass yields from unburned diesel and gasoline fuels along with similar yields reported by Jathar, et al. 84 The SOA mass yields for unburned fuel agreed with the SOA mass yields of the diesel engine exhaust both with and without aertreatment devices reported by Jathar, et al. 84 Our estimated SOA yields under the light conditions agreed within a factor of 2-3 with the yield curve of Jathar, et al., 84 except for the overestimation in the experiment of GPI.We should note that in the GDI experiment under light conditions, the SOA yield at t ¼ 1 h was low compared to that of previous studies and to the yield at t ¼ 2 h.This low yield at t ¼ 1 h could be explained by the high NO concentrations, as detailed in the next section.The SOA yields estimated by Gordon, et al. 13 for HDDV without DOC or DPF (hot-start experiments) were 4-19% at around 1-30 mg m À3 of OA concentrations.These SOA yields are also consistent with the yield curve of diesel fuel reported by Jathar, et al. 84 It is possible that SOA produced during dark conditions is lost by the photolysis of organoperoxide compounds; thus, SOA yields in the dark experiments may be reduced aer irradiation. 85.3.2.Simulated precursors' contributions to OA. Fig. 6 compares the simulated OA concentrations and O : C ratios with the observations.The observed OA concentrations were generally reproduced by the simulation, particularly during the rst hour of the experiments.As already noted, the SOA concentrations slowly increased during the experiments with gasoline exhaust under the light conditions, whereas SOA concentrations increased rapidly with both gasoline and diesel exhaust under the dark conditions.The simulation reproduced these trends.The simulation results indicated that this difference was caused by the different temporal variations of oxidant concentrations: the concentration of OH radical increased in the rst two hours during the light experiments, whereas the OH concentration reached a maximum immediately aer the start of the experiments and gradually decreased aerward during the dark experiments (Fig. S4 of the ESI †).We should note that the evolution of OA concentrations with diesel exhaust under the light conditions was not well captured by the simulation.The rate of increase of the observed OA concentrations slowed during the rst hour of the experiments, whereas the simulated OA concentrations increased for $2 h aer the start of the experiments.In the case of the experiments with a diesel truck, the decay rate of aromatic VOC was underestimated (Fig. S6 of the ESI †).This result suggests that the underestimation of SOA could be partly explained by the underestimation of the concentrations of the oxidants.
For gasoline vehicles, the largest contributor to SOA production was aromatic VOCs, followed by aromatic IVOC (light conditions) or alkene VOC (dark conditions).The simulated results indicated that aromatic VOCs accounted for 64-84% of the total SOA, and the sum of the alkane and alkene VOCs accounted for 11-15% of the SOA (Table 3).Unspeciated SVOC and IVOC accounted for 6-23% of the SOA.In contrast, the SOA from diesel exhaust was produced predominantly from the oxidation of IVOC (mostly alkane IVOC) (75-87%); the contribution of speciated VOCs (aromatics and alkanes) was less than 25%.
In previous studies, the contributions of aromatic VOCs to the SOA from gasoline exhaust were variable: several studies have estimated that single-ring aromatic VOCs are the dominant precursors of the SOA from gasoline exhaust; their contribution has been estimated to be 40-70%. 25,29,75,86,87Our results were similar to or slightly higher than these previous estimates.][88] Our estimate under light conditions (comparable with previous estimates) was 12-16%, which is similar to or less than previous estimates.For diesel vehicles, Eluri, et al. 43 have estimated that IVOC accounts for 81-90% of SOA.Jathar, et al. 19 and Zhao, et al. 28 have also indicated that unspeciated compounds make predominant contributions to the SOA from diesel exhaust.Under the light conditions, concentrations of SOA from diesel exhaust rapidly increased, whereas SOA formation from gasoline exhaust was slow, and there was a delay of 0.5-1.0h.This delay of gasoline-SOA formation was consistent with previous ndings of delays of SOA formation from aromatic VOC: 73,81,89 the formation of SOA from aromatic VOC has been shown to be suppressed in the presence of NO, but the concentrations of SOA start to increase aer NO has been consumed.As shown in Fig. S7, † concentrations of gasoline-SOA started to increase when NO concentration fell below 50 ppbv, while formation of diesel-SOA occurred even with high NO concentration (over 1 ppmv).
3.3.3.Oxygen-carbon ratio.We estimated the O : C ratio of the SOA (Fig. 4) by assuming that the O : C ratio of the POA did not change during the course of oxidation.The O : C ratio of the SOA was clearly higher for gasoline exhaust than for diesel exhaust.The O : C ratio of SOA from the diesel exhaust was around 0.3-0.4.O : C ratio of SOA from gasoline exhaust 2.5 h aer t ¼ 0 was 0.7-0.8 both in the light and dark experiments.
The O : C ratio of the SOA did not change during the course of oxidation, and it did not change signicantly as a function of vehicle type (GDI vs. GPI, or Diesel-P vs. Diesel-T) or oxidation conditions (i.e., light vs. dark).For both gasoline and diesel exhaust, the O : C ratio of the OA increased as the SOA/OA ratio increased, and it was higher under the dark conditions than under the light conditions (Fig. S5 †).
These O : C ratios are consistent with previous experimental results.For example, Pieber, et al. 90    photooxidation of diesel exhaust without a DPF are 0.1-0.3.The agreement of our experimental data with all of these results suggested that the differences of the O : C ratios between gasoline and diesel OA resulted from some common cause in addition to the differences in specic vehicles and oxidation conditions.The observed O : C ratios, including the contrast between the gasoline-and diesel-SOA, was generally reproduced by the simulation (Fig. 6).The observed contrast of the O : C ratios between the gasoline-and diesel-OA was consistent with the O : C ratios of the simulated precursors.The O : C ratio was 0.5-0.8 in the gasoline-OA, wherein SOA from aromatic VOC was dominant, and it was 0.2-0.3 in the diesel-OA, wherein SOA from alkane IVOC was dominant (Table 3).This difference of the O : C ratios between gasoline-and diesel-SOA clearly suggested that the major precursors of SOA from gasoline and diesel exhaust differed.The O : C ratios of the SOA from a single precursor could occasionally change during the course of oxidation of the precursor, 91,92 and the detailed evolution of O : C ratios during that oxidation has been simulated in some detailed SOA models. 93,94However, even our simple simulation could reproduce the contrast between the O : C ratios of the gasoline-and diesel-SOA.
Overall, the observed OA concentrations and O : C ratios were generally reproduced by the simulation.The normalized mean error during the rst two hours of the individual experiments was <15% for both diesel and gasoline exhaust in the dark conditions, 35-43% for gasoline exhaust in the light conditions, and 18-97% for diesel exhaust in the light conditions (Table S13 of the ESI †).Because we used the aggregated chemical mechanism, SAPRC99, the simulation did not rigorously represent the reaction rates and product yields of individual VOC (Table S5 of the ESI †).
3.3.4.Oxidation pathways of SOA formation.Most of the anthropogenic VOCs were oxidized only by the OH radical.In addition, we assumed that IVOC was also oxidized only by the OH radical as in the previous modeling studies. 19,42The only exceptions in our model setup were alkenes/dienes, which contain double bonds and are also oxidized by NO 3 radical and O 3 .In the simulation, OH radical was the dominant oxidant for the oxidation of OLE2 (internal alkenes and dienes) in the light experiments, though the contributions of O 3 were not small (16-45%) (Fig. S8 of the ESI †).In the dark experiments, the dominant oxidants of OLE2 were O 3 followed by NO 3 radical; the contributions of OH radical were negligible.Nonetheless, because the contributions of alkenes/dienes to the total SOA production were not large (6-10% for gasoline-SOA and 4-9% for diesel-SOA), the OH radical was the predominant contributor to SOA formation in both the light and dark experiments.
One unexplored consideration that may affect the chemical composition of SOA is organic nitrate.In our studies, the NO 3 radical chemistry did not contribute much to SOA formation.However, the experiments were conducted with NO x in the vehicle exhaust, and the SOA formed during our experiments could thus have included organic nitrate.As noted in Section 3.2, SOA was formed in the high-NO conditions during the light experiments, and in the low-NO conditions during the dark experiments.In addition, the NO x /NMHC ratio was higher for diesel exhaust than for gasoline exhaust (see Section 2.1).A comparison of the fractions of organic nitrate in SOA between vehicle types and light and dark conditions might provide information about the factors that control the amount of organic nitrate in SOA.
6][97] Organic nitrate in SOA from biogenic VOC (e.g., isoprene and monoterpenes) is produced via oxidation by the NO 3 radical or via oxidation by OH radical or O 3 followed by RO 2 + NO reactions.Formation of organic nitrate could result in high rates of SOA production because the vapor pressure is lowered with the addition of nitrate functionality.The behavior of organic nitrate is thus one of the important considerations in understanding SOA production yields.
The fraction of organic nitrate has been estimated by using the NO 2 + /NO + signal ratio measured with a high-resolution aerosol mass spectrometer (HR-AMS), 98 because the NO 2 + /NO + ratios differ between inorganic and organic nitrates. 99,100The NO 2 + /NO + signal ratios of biogenic SOA [101][102][103] and aromatic SOA 99,[104][105][106] are lower than the corresponding ratio of inorganic ammonium nitrate.These experimental data have been used for quantication of the organic nitrate fraction of ambient air in Europe and the U.S. 96,107 However, to our knowledge, there has been no measurement of the NO 2 + /NO + signal ratio in vehicleexhaust SOA.
The NO 2 + /NO + signal ratio of ammonium nitrate measured by the HR-AMS used in this experiment was 0.55.In contrast, the same ratio in aromatic SOA measured by the same HR-AMS was 0.17-0.26(ref.105) and was similar to the value reported for aromatic SOA in another study (0.13 for SOA from the photooxidation of 1,3,5-trymethylbenzene 99 ).Fig. 7 summarizes the relationship between NO 2 + and NO + signals during our experiments.Fig. 7 indicates that the NO 2 + /NO + ratio of the vehicleexhaust SOA was signicantly lower than that of ammonium nitrate, except for the gasoline-exhaust SOA during the rst 1-2 h aer the start of the dark experiments.The NO 2 + /NO + ratio was 0.36-0.50for the gasoline-exhaust SOA at more than 2 h aer t ¼ 0, and it was lower (0.17-0.36) for the diesel-exhaust SOA.To our knowledge, the NO 2 + /NO + ratio of the SOA from IVOC has not been previously reported.If we assume that the NO 2 + /NO + ratio of IVOC-SOA is similar to that of aromatic SOA, the fraction of organic nitrate was higher in diesel-SOA than in gasoline-SOA.This conclusion is consistent with the higher ratio of organic nitrate to total nitrate (¼inorganic + organic nitrate) in diesel exhaust (Fig. 7(d)) versus gasoline exhaust (Fig. 7(c)): the organic-nitrate/total-nitrate ratio estimated from an ion balance of the AMS unit-mass signals was higher for diesel exhaust (0.6-0.8) than for gasoline exhaust (0.3-0.4).In the gasoline experiments, NH 3 could also have been emitted, and the contribution of ammonium nitrate could thus be larger in gasoline exhaust than in diesel exhaust.In addition, as noted in Section 2.1, the NMHC/NO x ratio was higher in the gasoline exhaust than in the diesel exhaust, and it is thus possible that diesel-SOA contains more organic nitrate than does gasoline-SOA.The absolute concentration of NO 3 À determined from high-resolution mass spectrometric analysis of organic nitrate is inaccurate, 98,106 and only relative comparisons were therefore made here.Nonetheless, signicant amounts of organic nitrate appear to be present in both diesel-SOA and gasoline-SOA.
From the standpoint of oxidation conditions, the analysis of NO 2 + /NO + ratios indicated that the fraction of organic nitrate was higher in the light than the dark in the OA from both gasoline exhaust and diesel exhaust (Fig. 7).The fraction of organic nitrate has also been estimated by assuming an ionic balance of inorganic compounds.
In contrast, this analysis of ionic balance shows that the fraction of organic nitrate was lower in the light experiments.
We could not conclude whether the chemical composition differed between the SOA formed in the light and dark conditions.For this evaluation, chemical analyses of the SOA formed under light and dark conditions will be required in future studies.

Conclusions and implications
Emission and production factors of OA from the exhaust of gasoline vehicles and diesel vehicles were analyzed in smog chamber experiments under light and dark conditions.POA and EC were emitted from a GDI vehicle and an old diesel truck, whereas relatively smaller amounts of POA and EC were emitted from a GPI vehicle and a later-model diesel passenger car.Signicant amounts of SOA were produced from all four vehicles, and the SOA production factor was larger for the experiments in the dark conditions than in the light conditions.This difference was explained by the different SOA yields and may have been a result of the lower NO/(HO 2 + RO 2 ) ratio in the dark conditions.
The observed OA concentrations and O : C ratios were reproduced by using a 1-D VBS module with a model setup that included an up-to-date IVOC module and the O : C ratios of individual SOA.In both the light and dark conditions, aromatic VOC was the predominant precursor for the SOA from gasoline vehicles (64-84%), and IVOC (unspeciated alkanes) was the major contributor to the SOA from diesel vehicles (75-87%).The observed O : C ratios of the SOA were clearly higher in the SOA from gasoline vehicles (0.7-0.8) versus diesel vehicles (0.3-0.4).This difference could be explained by the differences between the precursors, i.e., aromatic VOC in case of the gasoline exhaust and alkane IVOC in case of diesel exhaust.
Under both light and dark conditions, the dominant oxidant was the OH radical, and the dominant precursors for gasoline and diesel vehicles were similar between the light and dark conditions.In addition, the O : C ratio of the SOA was similar in the light and dark conditions.However, because the NO/(HO 2 + RO 2 ) ratios were different, SOA was formed under the high-NO condition in the light experiments, whereas under the low-NO condition in the dark experiments.The differences of the chemical compositions of the SOA formed in the light and dark conditions may have been caused by these differences.We could not resolve the differences between the chemical compositions of the SOA produced in the light and dark conditions based on an analysis of organic nitrate.Because the SOA produced in the dark condition with high concentrations of O 3 has previously been used for toxicological studies, the differences of the chemical compositions of SOA should be further examined in future studies.9][110] Considering that the dominant SOA precursors differ between gasoline and diesel SOA, it is possible that the oxidative potentials of gasoline and diesel SOA are different.IVOC from diesel exhaust was dominated by alkanes: 31,39 thus it is recommended that toxicological impact of alkane IVOC is assessed for an informed understanding of the toxicological impact of vehicle exhaust.
photolysis was 0.198 AE 0.003 min À1 .H 2 O 2 (50% w/w) was injected as a radical precursor to boost the reaction for the diesel experiments; H 2 O 2 was not injected for the gasoline experiments to maintain hydroxyl radical (OH) concentrations similar to those for the diesel experiments.In the diesel experiments without H 2 O 2 injection, the estimated OH exposure was $5 times lower, and the produced SOA concentration was smaller by 94% (Diesel-P) or 50% (Diesel-T) than in the experiments with H 2 O 2 injection.Hereaer, we analyzed the experimental results with H 2 O 2 injection.The experiments without H 2 O 2 injection will not be further discussed.The average OH levels in the 6 m 3 FBC experiments were inferred from the decay of toluene and benzene measured with the PTR-MS.56For the ozonolysis experiments, ozone was generated by a corona discharge ozone generator (Iwasaki Electric, model OP-20W).The UV lights were turned on or O 3 was added at what is referred to as time t ¼ 0 throughout this paper.The temperature in the room containing the 6 m 3 FEP FBC was maintained at 25-28 C. The relative humidity (RH) inside the 6 m 3 FBC was estimated from the RH in the 30 m 3 EDC ($90%) and puried air ($40%) to be around 50%.Quantication of SOA production in the 6 m 3 FBC required that losses of both particles and vapors to the chamber walls be corrected.In this study, the particle wall loss (PWL) rate was calculated from experimental data, whereas the volatilitydependent vapor wall loss (VWL) rate was estimated using a model calculation, as detailed in Section 2.2.The PWL was treated as a rst-order process with a rate constant determined from the measured rate of decay of aerosol volume concentrations aer the formation of SOA, as summarized in TableS4of the ESI.† For the experiments of diesel vehicles, the geometric mean volume diameter of aerosol did not change aer 2.5 h

Fig. 3
Fig. 3 Budget of production and loss rates of the OH, HO 2 , and RO 2 radicals (left axis) and branching ratio, b, estimated with eqn (3) (right axis) in the simulation on a logarithmic timescale.

1 ..
Observed temporal variations and mass yields of SOA.Fig. 4 indicates the observed evolution of SOA concentrations and O : C ratios of SOA as functions of OH exposure.Here, the OH exposure was calculated based on the average decay rates of C8 and C9 aromatic VOCs measured by the PTR-Ratios of individual VOC compounds were not used because the temporal uctuations of individual VOC compounds were large.Because the PTR-MS does not quantify isomers, the estimated k OH values of the C8 and C9 aromatic VOCs (1.45 AE 0.66 and 4.06 AE 1.39 molecules per cm 3 per s, respectively 56 ) calculated from the average k OH values of the VOC were sources of uncertainty in the estimate of OH exposure.In addition, Fig. S5 of the ESI † summarizes the observed temporal variations of OA concentrations and O : C ratios of OA.

Fig. 4
Fig.4(a) Observed concentrations of secondary organic aerosol (OA) and (b) O : C ratio of SOA as a function of OH exposure under the light and dark conditions.Estimation of concentrations and O : C ratios of SOA and OH exposure is described in the main text.Observed SOA concentrations were corrected for particle wall loss but not for vapor wall loss, although vapor wall loss was considered in the simulation.

Fig. 5
Fig. 5 Scatterplots between (a) DOA and DNMHC and (b) SOA yields and OA concentrations at 1 h and 2 h from the start of the eight experiments.The SOA mass yields from unburned diesel and gasoline fuels in Jathar, et al.84 are also shown.
have shown from their experiments with a smog chamber and oxidation ow reactor that the O : C ratio of OA is 0.6-0.8aer photooxidation of gasoline exhaust with OH exposure of 8 Â 10 6 to 3 Â 10 7 molecules per cm 3 per h.They excluded data with high NH 4 NO 3 concentrations because these conditions are outside the AMS interference calibration and the O : C ratios in those cases may have been upward biased.Eluri, et al.43 have indicated that the O : C ratio of OA aer photooxidation of diesel exhaust with aertreatment devices (DPF and DOC) is around 0.2-0.4.Chirico, et al.11 have shown that the O : C ratios of OA aer the

Fig. 6
Fig. 6 Time-series of OA concentrations (left axis) and O : C ratios (right axis) in the simulation.Contributions of precursor groups to the simulated OA concentrations are indicated by hatching with different colors.

Fig. 7
Fig. 7 Relationship between NO 2 + and NO + signals (a and c) and time series of NO 2 + /NO + and organic-nitrate/total-nitrate ratio (b and d) during the experiments with gasoline exhaust (a and b) and diesel exhaust (c and d).The NO 2 + /NO + ratios for inorganic ammonium nitrate (NH 4 NO 3 ) and aromatic SOA 105 are also indicated.

Table 1
Characteristics of test vehicles used in this study

Table 2
Initial conditions of experiments.Concentrations were measured in 6 m 3 fluorinated-ethylene-propylene film bag , et al.69 have assigned O : C ratios for all the "anthropogenic" SOA.In this study, we assigned different O : C ratios to different precursors by using experimental O : C ratios from Zhao, et al.28 for SOA from aromatic VOC, Chhabra, et al.70 for SOA from aromatic-IVOC, and Cappa and Wilson 71 for SOA from alkane-IVOC (Table S10 of the ESI †).As noted in Section 2.1, the PWL rate in the SOA reaction chamber was calculated from experimental data.The simulation model with the formulation of Krechmer, et al.72 were used to calculate the volatility-dependent VWL rate.Morino, et al.73 have also provided a formulation of the VWL calculation.