In this study, all measurements were performed on a vehicle chassis dynamometer (Peng et al., 2017) using an in-use light-duty gasoline direct inject (GDI) engine vehicle meeting the China V standard (similar to Euro 5). Tests were conducted with two fuels: commercial China standard V gasoline and E10 fuel (10 % ethanol by volume). The test cycles included the Worldwide Harmonized Light Vehicles Test Cycle (WLTC) and the China Light-Duty Vehicle Test Cycle (CLTC). Furthermore, different typical acceleration rates were also tested. A detailed description and speed profiles of WLTC and CLTC are in Fig. S1 in the Supplement. CLTC was specifically designed to simulate the driving patterns in Chinese cities, while WLTC referred to the Euro 6 standard and adopted it as the China VI testing protocol. Prior to tests, the tested vehicle was preconditioned with an overnight soak, without an evaporative canister purge. Different acceleration rates were selected based on their frequency in both CLTC and WLTC, i.e., 1.2, 3.6 and 6.0 kmh-1s-1 (written as ACR1.2, ACR3.6 and ACR6.0), to investigate the effects of acceleration rates on IVOC emissions. All three acceleration “cycles” last for 600 s with a maximum velocity of 70 kmh-1. The acceleration driving cycles were set according to the criteria of an identical cycle period and maximum velocity, and hence the mean velocity for each acceleration cycle is the same (Fig. S2). We also measured IVOC emission factors (EFs) when the test vehicle was idling.

Tailpipe emissions were introduced to a constant-volume sampler (CVS) that diluted the exhaust by a factor of 20 to 40. For WLTC and CLTC tests, IVOCs emissions were collected by sampling the diluted exhaust through a quartz filter followed by two tandem Tenax TA-filled glass tubes (Gerstel 6 mm o.d. and 4.5 mm i.d. glass tube filled with ∼180 mg Tenax TA). Sampling tubes and transfer lines from the CVS were kept at a constant temperature (27±2 ∘C). The flow rate for the quartz filter was 10.0 Lmin-1, and the flow rate for the Tenax tube was set as 0.5 Lmin-1. Dynamic blanks were also collected when the CVS was operated with only dilution air (no exhaust) to estimate the contribution of background organic vapors. Prior to sampling, the quartz filters were preheated to 550 ∘C in air for 6 h in clean aluminum foil using a muffle furnace to remove contaminations. Tenax tubes were preconditioned by using a tube conditioner (BCT700, BCT Technology LTD) at 300 ∘C for 3 h in pure nitrogen with a constant flow rate of 100 mLmin-1. All samples were sealed after sampling and stored in a freezer at -20 ∘C.

Quartz filters and Tenax tubes were analyzed using a gas chromatography–mass spectrometer (Agilent 6890GC/5975MS) equipped with a capillary column (Agilent HP-5MS, 30m×0.25mm) coupled to a thermal desorption system (Gerstel, Baltimore, Maryland, USA). The detailed method was described by Zhao et al. (2014). Prior to analysis, 5 µL of the internal standards (d10-acenaphthene, d12-chrysene, d4-1,4-dichlorobenzene, d8-naphthalene, d12-perylene, d10-phenanthrene and seven deuterated n-alkanes) were injected into each adsorbent tube to track the IVOC recovery.

For each test, particulate matter samples were also collected using independent Teflon and quartz filters. The Teflon filters were weighted using a microbalance (Toledo AX105DR, USA) after equilibration for 24 h in an environmentally controlled room (temperature of 20±1 ∘C and relative humidity of 40±3 %) (Guo et al., 2010). A punch (1.45 cm2) from each quartz filter was analyzed for organic carbon (OC) and elemental carbon (EC) via a thermal–optical method using a Sunset Laboratory-based instrument (National Institute for Occupational Safety and Health, NIOSH, protocol thermal optical transmittance, TOT) (Guo et al., 2013). VOCs were sampled in SUMMA^® polished stainless-steel canisters and analyzed using GC-MS (gas chromatography–mass spectrometry) with a flame ionization detector. Total hydrocarbon (THC), nitrogen oxide, CO and CO2 emissions under operation scenarios were measured using a HORIBA OBS 2200 portable emission system.

A total of 20 IVOC compounds were quantified using authentic standards (Table S1). However, the majority of the IVOC mass appears as a broad hump of co-eluting hydrocarbons and oxygenated organics. These compounds could not be resolved at the molecular level and were therefore classified as an unresolved complex mixture (UCM) and grouped based on their volatilities.

The total mass of IVOCs was determined following the method of Zhao et al. (2014) (Supplement). In short, the TIC (total ion current) of each sample was divided in to 11 retention time bins corresponding to C12-C22 n-alkanes. The total mass in each bin was estimated using the instrument response to the n-alkane in that bin. UCM was determined as the difference between total IVOCs and speciated IVOCs in each bin. UCM was then further classified into unspeciated branched alkanes (b-alkanes) and unspeciated cyclic compounds following the approach of Zhao et al. (2016) (Supplement). The uncertainty of the IVOCs could be ascribed to both sampling and analysis. The sampling uncertainty was assumed as 10 % (Huang et al., 2019). The uncertainty of using n-alkanes as surrogate standards for the total IVOC mass was estimated to be less than 6.0 % for alkanes and 30.6 % for polycyclic aromatic hydrocarbons (PAHs) based on the analysis of a suite of standard compounds (Supplement). Therefore, combined with the above uncertainty, we consider a maximum IVOCs mass uncertainty of 32.2 % (Supplement).

Fuel-based IVOC emission factors (EFs, mgperkilogramfuel) were calculated using the carbon mass balance method as follows: 1EFIVOCs=[ΔIVOC]ΔCO2fc, where [ΔIVOC] represents the background-corrected mass concentration of IVOCs, [ΔCO2] is the background-corrected CO2 concentration in the CVS expressed in units of carbon mass and fc is the measured mass fraction of carbon in the gasoline (0.82).

Figure 1 depicts IVOC EFs of the tested China V gasoline vehicle and compares them with previous studies. The IVOC EFs ranged from 12.1 to 226.3 mgperkilogramfuel, with a median value of 83.7 mgperkilogramfuel. The median IVOC value was ∼3 times higher than that of the US LEV-2 (low-emission vehicle) gasoline vehicles (21.9 mgperkilogramfuel) and 1 order of magnitude lower than diesel-fueled nonroad construction machinery and a diesel-fueled large cargo vessel (971.1 and 800 mgperkilogramfuel, respectively) (Qi et al., 2019; Huang et al., 2018).

Figure 1 summarizes the influences of fuel type, starting mode, operating cycles and acceleration rates on the total IVOC EFs. Various operating conditions may cause different IVOC emissions and fuel consumption. In order to get a relatively reliable comparison, what we show here is all described in IVOC EFs which consider both IVOC mass and the fuel consumption. Among all of the factors, acceleration rate has the largest influence on the IVOC EFs. The fuel consumption at a high acceleration rate (6.0 kmh-1s-1) would be higher than that at a low acceleration rate (idling). Although not emitted in IVOCs, the high consumption of the fuel would exist as other types of carbon, e.g., VOCs and CO2 which may also have great effects on the atmosphere. Therefore, the usage of IVOC EFs can moderately balance the effects of the IVOC emissions and fuel consumption and get a comprehensive comparison among different acceleration rates. As the acceleration rate increases, the IVOC EF decreases, with the median IVOC EF of ACR6.0 being 1 order of magnitude lower than that at idling. Qi et al. (2019) and Zhao et al. (2016) report similar results for nonroad construction machinery and on-road diesel vehicles, where idling conditions emitted significantly higher IVOCs than those under higher-speed cycles. They proposed that the higher IVOC EFs at idling were the result of less efficient fuel combustion. An additional factor in these tests may be the efficiency of the catalytic converter varying with operating conditions (i.e., lower efficiency at idle operations).

When using commercial China standard V gasoline, the median IVOC EF was 1.4 times greater than that using Ethanol gasoline, i.e., E10 (10 % ethanol, v/v), with median values of 91.5 and 67.6 mgperkilogramfuel, respectively. The median THC EFs for gasoline and E10 were 485 and 589 mgperkilogramfuel, respectively, showing no significant difference.

As expected, the IVOC EFs for cold-start tests were higher (83.7 mgperkilogramfuel) than those for hot-start tests (58.7 mgperkilogramfuel). This reflects the reduced efficiency of the catalytic converter during cold-start operation. The cold-start-to-hot-start IVOC emission ratio is about 1.4, which is similar to the previous study (Zhao et al., 2016). The median THC EFs for cold-start and hot-start tests are 556.2 and 507.8 mgperkilogramfuel, respectively. Previous studies also show that cold-start operation has higher THC EFs than hot-start operation, but cold-to-hot ratios can span a wide range due to differences in operating conditions and model years (Jaworski et al., 2018; Drozd et al., 2016). The ratio is generally larger for more modern, heavily controlled vehicles (Saliba et al., 2017; May et al., 2014).

The median IVOC EF for CLTC was about 1.7 times of that for WLTC (103.5 versus 60.9 mgperkilogramfuel). Similar results were also found for THC emission, with median THC EFs for CLTC and WLTC of 617.3 and 420.3 mgperkilogramfuel, respectively. Previous studies also show test cycles influence THC EFs. For example, Suarez-Bertoa et al. (2015) and Marotta et al. (2015) found that the New European Driving Cycle (NEDC) has higher THC EFs than WLTP (Worldwide Harmonized Light Duty Test Procedure) or WLTC. One possible explanation for the differences between the CLTC and WLTC IVOC EFs is the differences in acceleration rates. A histogram of acceleration rates of the two cycles (Fig. S3) shows that CLTC has a frequent low-acceleration process compared to WLTC; 76.9 % of CLTC has acceleration rates ranging from -1.5 to 1.5 kmh-1s-1 versus 69.6 % for WLTC. CLTC has no acceleration rate higher than 4 kmh-1s-1, suggesting that the gasoline vehicles frequently drive in congested conditions in China.

The results from the acceleration rate cycles suggest that the frequent low acceleration rate in CLTC is responsible for the differences of the IVOC EF between CLTC and WLTC. The effect of acceleration on IVOC EFs is probably especially important in urban areas in China, which frequently have substantial traffic congestion. These results underscore the importance of developing cycles that simulate real-world Chinese driving conditions, e.g., CLTC, instead of using WLTC or other cycles to get relevant emission data.

Figures 2 and S4 compare the chemical composition of IVOC emissions from the tested China V vehicle under different operating conditions. In general, IVOC chemical composition was similar across all the tests. Unspeciated IVOCs (UCM) dominate the total IVOCs mass (85.6±4.9 %), including 65.2±5.2 % for unspeciated cyclic compounds and 20.4±0.7 % for unspeciated b-alkanes. n-Alkanes and speciated aromatics contribute 10.9±4.7 % and 3.5±1.7 % of the total IVOC mass, respectively. These results are similar to previous studies. For example, Zhao et al. (2016) found the consistent composition of IVOC emissions across a wide set of vehicles.

Since the majority of the IVOC mass appears as UCM, the average mass spectra provide additional insight into its composition. A similar distribution of mass fragments was observed across all tests. Figure 2b shows the average IVOC mass spectrum collected during an E10 CLTC test. Mass fragments associated with aliphatic hydrocarbons (m/z 43, 57, 71 and 85) are the most abundant followed by those associated with aromatics (m/z 91, 105 and 119 for alkylbenzenes (Pretsch et al., 2013) and m/z 115, 165 and 189 for polyaromatic species) (Dall'Osto et al., 2009; Spencer et al., 2006).

Figure 2c and d exhibit the contribution of selected mass fragments in low- and high-volatility ranges, i.e., B12–B16 and B17–B22. Aliphatic fragments are higher than aromatic fragments in both B12–B16 and B17–B22 bins. Compared to the higher-volatility (B12–B16) bins, the ratio of selected aromatic to aliphatic fragments is lower in the lower-volatility (B17–B22) bins (0.8 versus 1.7), which suggests a different weighting of compounds in a different volatility range. Therefore, unspeciated IVOC UCM in B12–B16 is predominantly aromatic, while B17–B22 are more abundant in cyclic alkanes.

Figures 3 and S5 show the volatility distribution of IVOC emissions over the 11 retention time bins (B12–B22). IVOC emissions are more heavily weighted towards the more volatile end of the distribution, with more than 50 % of the emissions in B12–B14 bins. After B14, the IVOC emission decreases significantly.

Although the IVOC EFs varied by an order of magnitude across the set of tests (Fig. 1), the volatility distributions of the emissions were largely the same. When the vehicle is fueled by gasoline, the median IVOC fractions in the B12–B14 bins are slightly higher than when fueled by E10 (Fig. S5a). Cold-start operation has a higher median percentage of IVOC in B12–B14 bins compared to hot-start operation (Fig. S5b). There are no distinct differences in volatility differences between CLTC and WLTC (Fig. S5c). Compared with idling conditions, acceleration cycles have a higher median percentage of IVOC in lower-volatility bins (B17–B22) (Fig. S5d), similar to previous studies (Qi et al., 2019; Cross et al., 2015). The modest variations of volatility distributions of the IVOC emissions may be due to differences in combustion efficiency and/or catalytic converter efficiency as a function of volatility.

Considering the similarity of volatility distribution for different conditions and the importance of the volatility distribution in model input for SOA simulation, Fig. S6 and Table S3 present the volatility distribution of SVOC and IVOC emissions from the tested China V gasoline vehicle, using effective saturation concentration (C*) as a classification: IVOCs (C*=300–3×106 µgm-3) and SVOCs (C*=0.3–300 µgm-3). IVOCs are the dominant part of the low-volatility organics (IVOCs + SVOCs), with a median contribution of ∼95 %.

Previous studies have used different scaling approaches to estimate IVOC emissions using other primary emission data, e.g., POA (primary organic aerosol) and NMHC (non-methane hydrocarbon) (Murphy et al., 2017; Woody et al., 2016; Koo et al., 2014). However, these ratios depend on fuel, engine technology and operating conditions (Lu et al., 2018). Therefore, it is important to quantify the relationships between IVOCs and other pollutants using data collected from Chinese vehicles. Our results show that the IVOC-to-THC ratio does depend on fuel composition. The average IVOC-to-THC ratios for the gasoline-fueled and E10-fueled gasoline vehicle are 0.07±0.01 (R2=0.87) and 0.11±0.02 (R2=0.78), respectively (Fig. S7). The IVOC-to-THC ratios in this study are higher than US vehicles (IVOC-to-NMHC ratio of 0.04) (Zhao et al., 2016) but much lower than diesel-fueled vehicles (IVOC-to-THC ratio of 0.67) (Huang et al., 2018). The IVOC-to-POA ratio was 5.12±1.30 across all tests but with only a modest correlation (R2 of 0.66 for gasoline-fueled vehicle and 0.43 for E10-fueled vehicle). This ratio is similar to US data for gasoline vehicles. The correlation of IVOC to THC or POA in our dataset is lower than that of the on-road gasoline and diesel vehicles measured in the US. This may be because the US data are from a large fleet of vehicles, while our data are from a single vehicle operated over a range of conditions.

Figure 4 presents PM, NOx, THC and IVOC EFs of the tested gasoline vehicle (China V) and compares them to US vehicles tested by Zhao et al. (2016) and May et al. (2014). For this comparison, we combined all of the CLTC and WLTC data together. The US vehicles are grouped by model year, where pre-LEV refers to vehicles manufactured prior to 1994, LEV-1 represents vehicles manufactured between 1994 and 2003, and LEV-2 indicates vehicles manufactured between 2004 and 2012.

The emissions of NOx and THC from the tested vehicle are comparable with those from the newer (LEV-2) US vehicles tested by Zhao et al. (2016) and May et al. (2014). However, the PM EF (44.8 mgperkilogramfuel) of the tested vehicle is higher than the LEV-2 vehicles tested (17.0 mgperkilogramfuel). It is comparable to a pre-LEV vehicle (61.0 mgperkilogramfuel). In addition, we compared our results with those from European vehicles and found that the NOx and THC EFs for the tested vehicle were lower than a Euro 5 gasoline vehicle, while the PM EF was higher (Fontaras et al., 2014). This suggests that compared with US and European vehicles, the stringent emission standards implemented by the Chinese government have been effective at controlling NOx and THC but might be inefficient for PM emissions. For the past 30 years, the Chinese government has adopted a series of emission control policies and measures for light-duty vehicles, including implementation of emission standards for new vehicles for the promotion of sustainable transportation and alternative fuel vehicles and traffic management programs (Wu et al., 2017; Zhang et al., 2014). Wu et al. (2017) summarizes the implementation of the vehicle control policies in China, which shows that control for the vehicular pollutants is becoming stricter step by step. For example, the NOx emission standard changed from 0.15 to 0.035 gkm-1 while the standard changed from China III to China VI. Different from NOx and THC, which have been controlled since China III, only in 2017, when the China V standard was first introduced did the control of PM come into the emission control scope. Yang et al. (2020) investigated the effects of the gasoline upgrade policy on migrating the PM pollution in China and found that there is not much space for significantly reducing the PM concentration by simply improving the gasoline quality. Therefore, for PM control, more policies, i.e., developing cleaner alternatives to fossil fuels, replacing traditional vehicles with new energy and building developed public transport system, should be implemented.

The IVOC EFs for the tested China V vehicle are between the US Pre-LEV and LEV-1 vehicle. Therefore, Chinese regulations may also appear to be ineffective at controlling IVOC emissions. The IVOC-to-THC ratio measured here (0.07 for gasoline and 0.11 for E10) is higher than US vehicles (0.04), which means that IVOCs contribute a larger fraction of the THC emissions from China V vehicles than from the US vehicles. A detailed comparison of the individual VOC emissions between China V and US LEV-2 vehicles is in the Supplement (Fig. S9).

UCM accounts for large fraction of IVOCs for both China V and US gasoline vehicles. However, the speciated compounds exhibit different characteristics. The China V exhaust has less speciated IVOC aromatic compounds (3.5 %) and more alkanes (10.9 %) compared to US exhaust (12.9 % and 2.5 %, respectively). This is also reflected by the IVOC mass spectrum, where Chinese vehicle exhaust has higher signals at m/z 43, 57, 71 and 85. In addition, the specific aromatic mass fragments were not the same for China V and US IVOC emissions. For example, the dominant aromatic fragments in US gasoline exhaust are m/z 128, 119, 105 and 133 versus m/z 135, 91, 181 and 189 for China V (Fig. 2c and d).

Figure 3 compares the volatility distribution of the IVOC emissions from the China V and US vehicles. There are significant differences of volatility distribution between China V and US vehicles. Both distributions decrease with the increase of the retention time, but the IVOC volatility distribution of US vehicle exhaust exhibits heavier weight of the lower-volatility bin, i.e., B12 bin compared to the China V vehicle. In US exhaust the B12 fraction is more than double that of the B13 fraction. However, the contributions of B12–B14 bin volatility bins are comparable for Chinese vehicle exhaust. US vehicle exhaust has a similar IVOC volatility distribution as the unburned gasoline, indicating that the evaporate of IVOCs from fuel is non-neglectable.

The differences between the IVOC volatility distribution between the Chinese vehicle exhaust and unburned gasoline were further investigated. The higher emission factor and broader distribution of IVOCs in exhaust from China V compared with US vehicles may be due to differences in fuel composition, operating conditions, and engines and aftertreatment technology, as the tests of US vehicles were all performed using California commercial fuel, which is, in fact, E10 fuel. Therefore, in this study, the US (unburned) fuel or US gasoline means E10. Lu et al. (2018) demonstrated that IVOC emissions depend strongly on fuel composition. In our study, IVOCs contributed ∼2.0 wt % (2.1 wt % for gasoline and 1.9 wt % for E10) of the total fuel mass, which is ∼60 % higher than the California fuel (1.2 wt % for E10) (Gentner et al., 2012). Therefore, the higher IVOC fractions in China V exhaust (e.g., IVOC-to-THC ratio of 0.07 and 0.11 versus 0.04 in US exhaust) may lead to higher amounts of IVOCs in China V gasoline. When considering volatility distribution, Zhao et al. (2016) and Lu et al. (2018) reported similar distributions of IVOC between gasoline vehicle exhaust and unburned fuel, which demonstrates the significant influence of unburned fuel on exhaust volatility distribution. As a result, in Fig. 3, we use US gasoline vehicle exhaust to both represent the exhaust and the unburned (E10) fuel and compare the Chinese E10 fuel with US fuel to get a comparative study. However, the volatility distribution of the China V gasoline vehicle exhaust are different from that of the unburned fuel (Fig. 3). The difference might be related to the operating conditions and engine aftertreatment system.

Although operating conditions strongly influence the total IVOC EFs (Fig. 1), Fig. 3 indicates the volatility distribution of the IVOC emissions were largely consistent across the set of test conditions. Therefore, operating conditions cannot explain the difference in the IVOC volatility distribution between the China V vehicle, unburned gasoline and the US vehicles.

The engine aftertreatment system also influences IVOC emissions (Drozd et al., 2019; Alam et al., 2019; Zhao et al., 2018; Saliba et al., 2017). In order to investigate the efficiency of the aftertreatment system, we normalized the IVOC distributions of the fuel and exhaust to the sum of C8-C10 n-alkanes. It is believed that the C8-C10 n-alkanes can serve as the indicators for VOCs in fuel (Lu et al., 2018). For both US and the China V vehicles, IVOCs are enriched in the exhaust relative to the fuel. However, the enrichment factor is much smaller in Chinese exhaust with a median value of 4.0 than that for US vehicles (median value of 8.5) (Lu et al., 2018). The enrichment factor also varies with different compounds, with the enrichment factors following the order of n-alkanes (9.3) > b-alkanes (6.6) > unspeciated cyclic compounds (3.1) > aromatics (0.4). These results are consistent with previous studies stating that the aftertreatment devices have different removal efficiency towards different compounds (Ma et al., 2019; Hasan et al., 2018; Hasan et al., 2016; Alam and Harrison, 2016). Our results suggest that the Chinese three-way catalytic (TWC) converter has compound-dependent efficiency (better removal of aromatics compared to alkanes), which might explain the difference in compound composition between Chinese and US vehicle exhaust. Furthermore, Fig. S10 shows that the catalytic converter has different removal capacity towards different volatility bins, in which B14–B16 works much worse compared to other volatility bins, i.e., B12. Consequently, the SOA formation would be relatively high. In sum, the compound-dependent capacity and lower B14–B16 removal efficiency of the Chinese TWC converter is responsible for the volatility distribution differences between China V and US vehicles shown in Fig. 3.

After considering all the factors above, we can draw the conclusion that fuel type, starting mode and operating conditions can all affect the IVOC EFs. The only factor that impacts the volatility distribution is the engine aftertreatment system.

With the measured IVOC and VOC emissions, we estimated the SOA formation potential by using the yield method as follows (Yuan et al., 2013): 2ΔSOA/ΔCO=∑ER[HC]i×(1-e-(kOH,i-kCO)×OH×Δt)×Yi, in which ER[HC]iis the emission ratio of SOA precursor i (mgperkilogramfuel); kOH,i is the OH reaction rate constant of precursor i at 298 K (cm3permolecules-1); kCO is the OH reaction constant of CO at 298 K (2.4×10-13 cm3permolecules-1); [OH] is the OH mixing ratio, which is assumed to be 1.5×106 moleculescm-3 (Lu et al., 2019); Δt is photochemical age (h); and Yi is the SOA yield determined from chamber studies. Previous studies have shown that the SOA yield of individual hydrocarbons can be influenced by NOx level, due to the competition reactions among RO2 radicals, NO and HO2 radicals. Usually SOA yields under low-NOx conditions are independent on the OA loading. However, under high-NOx conditions, SOA yields highly depend on OA mass concentration, which can be described using two-product or multi-product models (Presto et al., 2010; Chan et al., 2009; Ng et al., 2007). In this study, we estimated SOA formation under low- and high-NOx conditions with OA concentrations of 10, 20 and 80 µgm-3 to represent the influence of NOx level and OA loading on SOA formation.

In this estimation, we include speciated C6-C9 single-ring aromatics (SRAs) as typical VOCs for SOA precursors, and the corresponding kOH and SOA yields are extrapolated according to two-product relationships from chamber studies (see Supplement) (Ng et al., 2007). The SOA yields under low- and high-NOx conditions and the OH reaction rates of speciated IVOCs and SRAs are from the previous studies (see Supplement) (Presto et al., 2010; Lim and Ziemann, 2009; Chan et al., 2009). In brief, surrogate species were used to represent the unspeciated b-alkanes and cyclic compounds in each of the volatility bins.

Figure 5 shows the POA emission and estimated SOA production under different operating conditions and NOx level after 48 h of photo-oxidation. The estimated SOA/POA ratio is between 4.0 to 5.0 under low-NOx conditions, and the SOA-to-POA ratios ranged from 1.8–2.2 to 3.8–4.4 when the OA loading increased from 10 to 80 µgm-3 under high-NOx conditions. The OA enhancement under low-NOx conditions is similar to that under high-NOx conditions with the OA loading of 80 µgm-3. Considering the high POA concentration and SOA formation capacity of Chinese gasoline vehicles, the SOA/POA ratios at 80 µgm-3 are considered as a lower estimation. Compared with OA enhancement from US studies (∼3.6) (Zhao et al., 2016), our results showed higher SOA formation potential both under low- and high-NOx conditions for Chinese gasoline vehicles.

Scenario-based analysis shows a similar tendency of SOA formation potential at different OA loading under low- and high-NOx conditions. Though the POA emission for the gasoline-fueled vehicle was higher than that fueled by E10, comparable SOA formation is estimated using gasoline and E10 as fuel. That means that the OA enhancement factor for E10 is higher than that for gasoline. This suggests that although the ongoing policy of ethanol gasoline will not exacerbate the POA emission in China, the SOA formation of E10 could not be neglected due to its high SOA enhancement capacity. Therefore, more research should be done to evaluate the effectiveness of using E10 as a surrogate to reduce the air pollution in China.

Cold-start operation has higher SOA potential with a higher OA enhancement factor than hot-start operation, due to the higher precursors EFs caused by the reduced catalytic converter effectiveness below its light-off temperature (Drozd et al., 2019). The IVOC EFs, the estimated SOA production and SOA/POA ratio of CLTC are all higher than those of WLTC, which further demonstrates the higher SOA formation potential of Chinese gasoline vehicles under typical driving conditions in China.

Figure S11 presents the contribution of different classes of precursors on the SOA production after 48 h of photo-oxidation under different OA loading and NOx conditions. The relative contributions of different chemical classes were similar across the different conditions, with the largest contribution from unspeciated cyclic IVOCs. This is different from the US gasoline vehicle SOA (Zhao et al., 2016), in which single-ring aromatics contributes the most.

In this section, we tried to establish parameterization methods to provide simple estimations of gasoline vehicle SOA based on our measurements of VOCs and IVOCs.

Figure S12 shows the average predicted SOA-to-POA ratio as the function of photo-oxidation time under different OA loading and NOx conditions. In general, SOA exceeds POA after first a few hours of oxidation and then levels off after 30 h. The SOA/POA ratio is influenced by OA concentration, NOx level and the photochemical age (OH exposure). At a certain OA loading and OH exposure, the SOA/POA ratio can be estimated and then used to quantify the contributions of gasoline vehicle SOA to the ambient OA. Therefore, we parameterized the SOA/POA variation under different OA and NOx conditions using a three-parameter-based logarithm equation: y=a-b×ln⁡(t+c), in which t represents the equivalent photochemical age (assume that the OH concentration is 1.5×106 moleculescm-3) and a, b and c can be described using a three-parameter logarithm equation, y=m-n×ln⁡([OA concentration]+p). Table 1 shows the parameterization results of compound-based SOA/POA variation under the different OA and NOx conditions. The fit quality could be found in Fig. S13.

The above photochemical-based parameterization method provides a conservative way to quantify the evolution of SOA from Chinese gasoline vehicle VOC and IVOC oxidation. However, there are still some uncertainties which may lead to discrepancies between predicted and measured SOA. In general, the positive or negative artifacts of quartz filters, n-alkane equivalent method in estimating the IVOC concentration, uncertainty in SOA yield, surrogate method to substitute SOA yield and kOH for UCM, and lack of semi-volatile organic compounds will exert influence on the SOA prediction.