Technical note: Emission factors, chemical composition, and morphology of particles emitted from Euro 5 diesel and gasoline light-duty vehicles during transient cycles

Abstract. Changes in engine technologies and after-treatment
devices can profoundly alter the chemical composition of the emitted
pollutants. To investigate these effects, we characterized the emitted
particles' chemical composition of three diesel and four gasoline Euro 5
light-duty vehicles tested at a chassis dynamometer facility. The dominant
emitted species was black carbon (BC) with emission factors (EFs) varying
from 0.2 to 7.1 mg km−1 for direct-injection gasoline (GDI) vehicles,
from 0.02 to 0.14 mg km−1 for port fuel injection (PFI) vehicles, and
0.003 to 0.9 mg km−1 for diesel vehicles. The organic matter (OM) EFs varied from 5 to 103 µg km−1 for GDI gasoline vehicles, from 1 to 8 µg km−1 for PFI vehicles, and between 0.15 and 65 µg km−1 for the diesel vehicles. The first minutes of cold-start cycles contributed the largest PM fraction including BC, OM, and
polycyclic aromatic hydrocarbons (PAHs). Using a high-resolution time-of-flight mass spectrometer (HR-ToF-AMS), we identified more than 40 PAHs in both diesel and
gasoline exhaust particles including methylated, nitro, oxygenated, and amino
PAHs. Particle-bound PAHs were 4 times higher for GDI than for PFI
vehicles. For two of the three diesel vehicles the PAH emissions were below
the detection limit, but for one, which presented an after-treatment device
failure, the average PAHs EF was 2.04 µg km−1, similar to the
GDI vehicle's values. During the passive regeneration of the catalysed diesel particulate filter
(CDPF) vehicle, we measured particles of diameter around 15 nm mainly
composed of ammonium bisulfate. Transmission electron microscopy (TEM) images revealed the presence of
ubiquitous metal inclusions in soot particles emitted by the diesel vehicle
equipped with a fuel-borne-catalyst diesel particulate filter (FBC-DPF).
X-ray photoelectron spectroscopy (XPS) analysis of the particles emitted by the PFI vehicle showed the presence
of metallic elements and a disordered soot surface with defects that could
have consequences on both chemical reactivity and particle toxicity. Our findings show that different after-treatment technologies have an
important effect on the emitted particles' levels and their chemical
composition. In addition, this work highlights the importance of particle
filter devices' condition and performance.



Introduction 35
On-road diesel and gasoline vehicles are an important source of urban air pollution, releasing fine particulate matter (PM1) and gaseous pollutants into the atmosphere (Dallmann and Harley, 2010;Borbon et al., 2013;Platt et al., 2014;Argyropoulos et al., 2016;Hoffman et al., 2016;Gentner et al., 2017). Light-duty vehicle pollutants have been associated with adverse effects on human health inducing cardiovascular, respiratory and also cognitive diseases (Hime et al., 2018 and references therein).
Improved information about the chemical composition of PM is essential to understand the source contributions, to implement mitigation measures and to assess health protection programs. PM vehicle emissions are mainly composed of BC and organic aerosol (OA) due to the incomplete combustion of fuel and lubricating oil. Less abundant components of vehicles exhausts include sulfate and metal in traces (Maricq, 2007;Cheung et al., 2010). Sulfur and trace elements such as Fe, Zn, P, Mg and 50 Ca are commonly used as additives to lubricant oils (Maricq, 2007;Rönkkö et al., 2014). The above elements have been correlated to the oxidative potential indicating the toxic nature of these emissions (Cheung et al., 2010).
PAHs have been measured in both modern diesel and gasoline engine exhaust (Zielinska et al., 2004;Cheung et al., 2010; Particles emitted by the PIF4 vehicle were further analysed by XPS recorded under ultra-high vacuum using a Resolve 120 hemispherical electron analyser (PSP Vacuum) and an un-monochromatized X-ray source (Mg Kα at 1253.6 eV, PSP Vacuum) operated at 100 W at an incidence angle of 30° with respect to the analyser axis. This X-ray excitation energy and detection geometry correspond to an analyzed depth of about 1 nm at the C1s and O1s lines. Survey spectra were collected at a pass 130 energy of 50 eV and an energy step of 0.2 eV, while the other lines were collected at 20 eV pass energy and a step of 0.1 eV.
The XPS lines were deconvoluted with the CasaXPS program, after Shirley-type or linear background subtraction. Quantitative estimations of the samples composition were done after correction by the relative sensitivity factors (RSF) provided in the program.

Organics and PAHs
The AMS data were analyzed with SQUIRREL v1.60A and PIKA v1.20A with Igor Pro 6.37 (Wave-Metrics). For the organic species, we used the fragmentation table of Aiken et al. (2009). For the vehicles D1, D3 and GDI3 the mass spectra are provided in unit mass resolution (UMR), while for the vehicles D4, PFI4 and GDI5 the mass spectra are given in high resolution (HR).
(taking account the minimum and the maximum fA,i reported for the AMS measured PAH spectra). An example of the HR mass spectra fitting for naphthalene, methyl-naphthalene, anthracene and nitro-anthracene is given in Figure S1. 160

Emission factors
The EF of each species during the cycle was calculated using the Eq. (2): were C(t) is the mass concentration of the pollutant, Qex(t) is the exhaust flow rate at the tail pipe measured by the CVS or the FPS, DR is the external dilution before the entrance of the instrumentation, and D is the distance of the cycle (4.51 km for 165 Artemis urban cycle, 23.8 km for Artemis motorway cycle and 23.25 km for WLTC). Figure 1 shows the particle mass concentration transient profile of the BC, organics, sulfate, nitrate and ammonium measured 170 by the HR-ToF-AMS for the GDI5 vehicle during Artemis cold urban, hot urban and motorway cycles. The mass concentrations have been corrected for the dilution in front of the instrumentation. The particle phase emissions were mainly composed of BC (96.8-98%), while the organic fraction accounted for 1.9-3.1%; ammonium, sulfate and nitrate were approximately 0.1%. The highest mass concentrations of BC and organics, 1600 µg m -3 and 120 mg m -3 respectively, were observed during the first 1-2 minutes of the cycle, due to the cold engine and thus low catalyst efficiency, which is in agreement 175 with previous studies (e.g., Weilenmann et al., 2009;Clairotte et al., 2013;Collier et al., 2015;Karjalainen et al., 2016;Louis et al., 2016;Pieber et al., 2018). GDI3 particulate mass concentrations in the exhaust flow were measured during WLTC ( Figure S2): BC contributed 83-98 % to the total PM mass, while the organic fraction ranged from 1.8 to 14% of the PM. The remaining fraction 0.2-3% was composed of ammonium, sulfate and nitrate. The emitted PM concentrations were comparable to the values measured for the GDI5 vehicle during Artemis/Cold Urban. 180 PM emissions from the PFI4 vehicle are shown in Figure S3. The organic and nitrate mass concentrations were a factor of 10 lower in comparison to those for GDI5 vehicle. High GDI PM emissions have been also reported in previous studies (e.g., Zhu et al., 2016;Saliba et al., 2017;Du et al., 2018), and were explained by the incomplete volatilization and mixing of the fuel in the combustion chamber (Fu et al., 2014;Chen et al., 2017;Saliba et al., 2017). Figure 2 shows the PM emissions of the D1 car (equipped with a CDPF) in terms of (a) chemical composition and (b and c) 185 particle size distribution during a cold urban and three consecutive motorway cycles. The cold cycle was characterized by relatively high BC and organic matter emissions, reaching concentrations of 300 and 50 µg m -3 , respectively. During the cold urban cycle BC accounted for 94% of the total mass concentration and organics only for 4%, while during the motorway cycle https://doi.org/10.5194/acp-2020-842 Preprint. Discussion started: 4 September 2020 c Author(s) 2020. CC BY 4.0 License.

Time series profiles
the contribution of BC decreased to 85% while ammonium bisulfate increased to (6%) and organics to (8%). The three motorway cycles showed good repeatability, characterized by a first release of BC followed by emissions of ammonium 190 bisulfate and organics nanoparticles (15 nm mean diameter). This behavior was interpreted as a passive regeneration of the DPF occurring at the high temperatures reached during the cycle. Similar observations have been reported during regeneration of diesel cars equipped with CDPF (R'Mili et al., 2018).
The emission profiles during WLTC cycles from a second CDPF vehicle (D3) are shown in Figure S4. This vehicle was characterized by very low emissions, demonstrating the efficiency of the after treatment devices. Emissions were observed 195 during few accelerations; the organic mass concentration remained always below 20 μg m -3 , while ammonium bisulfate concentrations reached maximum values of 50 μg m -3 .
The emission of sulfate containing particles from the two CDPF vehicles was explained by the presence of the catalyst on the DPF walls. It has been proposed that during acceleration or hot engine combustion periods sulfur can be released and converted into SO3 by the catalyst, forming successively sulfuric acid and/or bisulfate/sulfate ammonium ultrafine particles (e.g., Bikas 200 and Zervas, 2007;Bergmann et al., 2009;Arnold et al., 2012;R'Mili et al., 2018).
PM emissions from the D4 vehicle equipped with an FBC catalyst ( Figure S5) were relatively high: OA, nitrates and sulfate reached 300, 90 and 40 μg m -3 , respectively. The high PM concentrations were interpreted as a possible failure in the aftertreatment system and will be further discussed in section 3.3. This was supported by the relatively higher emissions of CO2, CO, NOx and THC in comparison to the rest diesel cars (Table S1). 205 Figure 3 shows the HR-AMS mass spectra for the GDI5 and the D4 vehicle during the first and last 2 minutes of each cycle.
During some periods of the cycle, the m/z ratios of 43/41, 57/55 and 71/69 were relatively high with values of 1.50, 1.72 and 1.19 during the GDI5 cold start (Figure 3a), and 1.30, 1.32 and 0.85 during the D4 hot engine regimes ( Figure 3b). Comparing 215 our mass spectra with pure gasoline, diesel and lubricant oil mass spectra analyzed with a similar instrument (R'Mili et al., 2018), and knowing that the fuels contain high concentrations of n-alkanes, while lubricating oils tend to contain mostly cycloalkanes (Tobias et al., 2001;Isaacman et al., 2012) we concluded that both GDI5 and D4 emitted randomly oil droplets (see also section 3.2 for TEM images).
( Figure S7 and Figure S8, UMR mass spectrum). The OA concentration emitted from the D3 car was very low and the high uncertainty was associated to the corresponding AMS mass spectrum.
For all of the gasoline cars, sulfur containing organic fragments at m/z 45 (CHS + , 44.979), 46 (CH2S + , 45.987) and 47 (CH3S + , 46.995) were detected. They accounted for approximately 2-4% of the organic mass fraction for the GDI5 (Figure 3a) and 6-225 7% to the organic mass for the PFI4 ( Figure S7). For the GDI3 car ( Figure S5), a high m/z 45 contribution was detected at the beginning of the hot start WLTC, but the spectrum was acquired with a c-ToF-AMS and therefore the signal can be assigned to both oxygenated (CHO2 + and C2H5O + ) and organosulfur (CHS + ) fragments. Sulfur containing ion fragments were mostly emitted from hot engines (end of urban cycle and motorway cycle) and are tentatively explained by the release of some lubricant oil. 230 Table 2 presents the correlations between the mass spectra of the tested vehicles with those of previous studies (AMS mass spectra database). A very good correlation was found between the mass spectra from diesel and gasoline vehicles (Canagaratna et al., 2004;Mohr et al. 2009;R'Mili et al., 2018) and PMF factors related to fresh traffic emissions (Mohr et al., 2012;Kostenidou et al., 2015;Kaltsonoudis et al., 2017). The R 2 ranged between 0.72 and 0.92 (Table 2) for all cases.

PAHs 235
In total, 45 PAHs were identified for the GDI5, PFI4 and D4 vehicles during Artemis cycles (Table S2). The mass concentrations of all the PAHs during the cold cycle were considerably higher for GDI5 and D4 than for PFI4, with values of 1.66, 2.21 and 0.47 µg m -3 , correspondingly. Slightly lower mass concentrations were observed during the hot cycles. For the D1 and D3 vehicles, the PAHs signal was close to the detection limit, demonstrating that after-treatments devices (DOC and DPF) efficiently reduce PAHs emission from light-duty diesel engines. The remarkable difference of the three diesel vehicles 240 confirms that D4 indeed presented a failure in the after-treatment device. For the GDI3 vehicle, the identification of individual PAHs was not possible since the data were collected with a c-ToF-AMS (UMR mass spectra). The UnSubPAHs represented the most abundant group, accounting for 52 to 66% of the total PAHs, followed by MPAHs (14-245 35%), then OPAHs (5-19%), NPAHs (1-11%) and finally APAHs (1-6%). Table S3 demonstrates the individual PAHs fractions during the cold-and hot-start cycles. For all three cars, naphthalene emissions dominated, contributing from 9.6 to 19.1% of the total PAHs, which is in agreement with previous studies (e.g., de Abrantes et al., 2004;Vouitsis et al., 2009;Huang et al., 2013;Alves et al., 2015;de Souza and Corrêa 2016;Muñoz et al., 2018). Among the 3-rings PAHs species, acenaphthylene (4.3-9.7%), anthracene and its isomer phenanthrene (4.1-15.9%) were the most abundant; concerning the 4-250 rings PAHs, the major contribution derived from pyrene and from the isomers fluoranthene and acephenanthrylene (1.3-13.9%), while among the 5-rings PAHs, benzo[a]pyrene and all its isomers (0.4-3.8%) and benzo[ghi]fluoranthene (1-3.3%) were the most significant compounds. Some heavier PAHs as indio[1,2,3-cd]pyrene its isomer benzo[ghi]perylene (0.4-6.6%) and coronene (0.06-5.3%) were mostly found in gasoline car emissions. Light PAHs have often been measured in exhaust https://doi.org/10.5194/acp-2020-842 Preprint. Discussion started: 4 September 2020 c Author(s) 2020. CC BY 4.0 License.
particles of light-duty vehicles (Ravindra et al. 2008;de Souza and Corrêa 2016;Muñoz et al., 2018), and their presence has 255 been tentatively explained by incomplete fuel combustion (Lea-Langton et al., 2008;Ravindra et al., 2008) since these compounds are present in the fuel composition (Marr et al. 1999;de Souza and Corrêa 2016). During the gasoline hot cycles an increase of the 3-and 4-ring PAHs (anthracene, pyrene, paracylene and all its isomers) contribution was observed.
MPAHs accounted for 14 to 35% of the total PAHs, and were more abundant for the D4 vehicle; major contributions arose from methyl-and dimethyl-naphthalene, methyl-phenanthrene, methyl-fluorene and ethyl-phenanthrene, which is in 260 agreement with Muñoz et al. (2018). All these compounds have been recently associated to carcinogenic potency (Samburova et al., 2017). BaP and its isomers (Benzo have been classified as carcinogenic or/and photomutagenic (compounds that cause mutagenicity after being exposed to visible or UV light) according to IARC 2010. Following BC and organics' emission trend, PAHs were also important in the first few minutes of the cold urban cycle and during acceleration periods of the motorway cycle or during fuel-rich combustion periods in agreement with previous studies (Muñoz et al., 2018). The mass fractions of these carcinogenic and/or photomutagenic 270 PAHs accounted for 27-49% for the GDI5, 29-30% for the PFI4 and 29-31% for the D4 vehicles.
A considerable fraction -up to 31% of the total PAHs -was functionalized and included OPAHs, NPAHs and APAHs. All technologies emitted an important fraction of OPAHs (up to 19%); anthraquinone was the most abundant in agreement with previous emission studies ( Karavalakis et al., 2011) followed by fluorenone, indanone, dibenzofuran and dibenzopyran.
APAHs accounted for 1 to 6 % of the total PAHs fraction and were mostly emitted by gasoline cars. Major NPAHs were 275 aminopyrene/carbazole and dibenzocarbazole/amino-benzopyrene, however very little is known about the car emissions of these compounds so far.
Nitro-anthracene and its isomer nitro-phenanthrene contributed up to 8% of the total PAHs in the GDI5 emissions, but only 1% in PFI4 and D4 vehicles. Nitro-fluorene, nitro-pyrene and nitro-chrysene were found in the car exhaust of all three vehicles, and accounted for less than 1% of the total PAHs mass fraction. Even if present in small amounts, some of these compounds, 280 as 6-nitrochrysene and 1-nitropyrene, are classified as possibly carcinogenic to humans (group 2B) (IARC, 2012;Bandowe and Meusel, 2017). Surprisingly, NPAHs, including nitro-pyrene, were considerably higher in GDI emissions than in those of diesel car, questioning the validity of using NPAHs such as 1-nitropyrene as markers of diesel emissions (Keyte et al., 2016). was collected for 300 seconds and the dilution was around 40. TEM images confirmed the quite higher emissions of soot particles (or BC) for the two GDI vehicles with respect to the diesel car, which is in agreement with BC emissions measured by the MAAP and the Aethalometer. As usually mentioned in the literature, soot particles are observed either as fractal 290 branched chains and or as bigger agglomerates made of primary soot spheres of different sizes (Lapuerta et al., 2020). Primary soot particles with diameter of around 25 nm were observed for the gasoline cars during cold cycles (Figure 6a and b), while the diameter was significantly smaller (below 20 nm) during hot cycles ( Figure S9c and S9f). The results are in a good agreement with previous literature, which reported primary soot particles with diameter in the range between 20-25 nm (Barone et al., 2012) and smaller sizes down to 16 nm (Mathis et al., 2004;Gaddam and Vander Wal et al., 2013) for gasoline exhaust 295 particles. A slight decrease in the primary particles size with increasing temperature was observed, in agreement with recent studies (Cadrazco et al., 2019). It has been shown that the engine load has no effect on soot morphology (Lapuerta et al., 2020) as many other parameters may favor opposite trends and compensate each other. Indeed, a higher fuel-air ratio would tend to extend primary particles growth while a higher engine temperature would favor their oxidation and thus lead to smaller particle sizes (Ye et al., 2014). Similarly, increasing the injection timing leads to a decrease of primary particles size due to an increase 300 of in-cylinder oxidation time (Xu et al., 2014). It is therefore difficult to unambiguously attribute the slight decrease in particle size observed only to the temperature effect. Figure S9 (i) depicts soot particles from the D4 car; tiny sparse dark spots were ubiquitous within the soot particles and were interpreted as metal inclusions. Unfortunately, EDX could not reveal their chemical nature due to the very small amount of material in these inclusions as they were very small (typically less than 0.5 nm) and their spatial density was low. Nevertheless, 305 we assume that these inclusions were metallic and resulted from the after-treatment device of the FBC-DPF vehicle (D4), which implies the use of additives made of metallic salts or organometallic compounds into the engine combustion chamber.

Off-line analysis: TEM and XPS
Upon combustion, the additive produces nanoparticles of metal oxides that are mixed with soot particles and accumulated on the DPF walls. The role of these metals will be to reduce the DPF regeneration temperature (Ntziachristos et al., 2005;Majewski and Khair, 2006;Song et al., 2006). 310 Nearly spherical particles were observed for some of the cars: GDI3 (Figure 6b, Figure S9), D1 ( Figure 6c) and D4 ( Figure   S9b). They were observed both during cold and warm cycles and they had variable sizes and shapes ranging from 100 nm to almost 1µm. EDX analysis revealed that on average the droplets presented C and O as major components, followed by S which was enriched in few droplets. Minor components accounted also for calcium, phosphorus, sodium, silica. Only minor traces of zinc, iron, copper, chromium, aluminum and nickel were observed. Analysis of the lubricant oils for D1, D3 and GDI1, GDI3 315 are presented in table S4. Sulfur accounted around 0.12 and 0.14 wt% of the lubricant oil. Other components of the lubricant oil were calcium, phosphorus and zinc, and only traces of iron, silica and copper were found. The iron found in the used lubricant oil suggests erosion of the engine wear and exhaust line for both D1 and GDI1 vehicles. These findings are in line with previous studies that reported emissions of lubricant oil particles during transient driving conditions Rönkkö et al., 2014).
Particles emitted by the PFI4 car were analyzed by XPS. Figure S10 (Table S4), while Ti might originate from the washcoat of the catalytic converter. The weak N1s signal showed typical energy of amino groups confirming the presence of APAHs as observed from AMS chemical analysis (Table S3). 330 Figure S10 (b and c) depicts the deconvolution of C1s (b) and O1s spectra (c). In the C1s spectrum the carbon speciation can be derived in terms of graphitic sp 2 carbon (at 284.5 eV), aliphatic sp 3 carbon (285.4 eV) and oxidized carbon in C-O-C bonds (ethers, alcohols; 286.4 eV), in C=O bonds (carbonyls, quinones, 287.5 eV), and acidic O=C*-OH bonds (288.9 eV) (Estrade-Szwarckopf, 2004). The analysis revealed a soot sample dominated by sp 2 hybridized carbon, the absence of the usual shakeup line associated with graphitic structures, and a significant "defect" contribution (at 283.5 eV, 12% of the C1s signal) associated 335 to carbon vacancies (Barinov et al., 2009), which indicates a significant concentration of carbon radical defects (Levi et al., 2015). All these elements hint to a structurally disordered soot surface, possibly having chemical toxicity or reactivity due to the presence of surface radicals. In addition, a rather high concentration of sp3 carbons (alkanes, 20 % of the total carbon) was detected at the surface of the particles, in agreement to what observed by AMS analysis ( Figure S6). From the O1s spectrum the relative contribution of the C=O carbonyl and carboxylic groups (532.1 eV), the C-O-C groups ethers and alcohols (533.2 340 eV), and the OH groups acids (534.3 eV) were derived. A strong contribution of Ti-O* in TiO2 was detected at 530.2 eV coming from ashes. Oxidized calcium and silicon also contributed to the O1s spectrum as Si-O* and Ca-O* lines in the 533-535 eV range (Ni and Ratner, 2008;Yang et al., 2011). Using the C=O contribution at 532.2 eV -the only line not overlapped by the Si, Ca and Ti oxides-and the integrated intensity of the C1s line, we evaluated a soot surface oxidation by the ratio O/(O+C), giving an oxidation rate of 10.8 %. This is in good agreement with Schuster et al. (2011) who found for Euro 4-5 345 soot particles oxidation rates between 5.5-11.5%. Figure 7 shows the emission factors (EFs) for BC, organics, PAHs, sulfate, ammonium and nitrate for all the cars tested in this study. Table S5 summarizes these EFs in µg km -1 . Most of the particles were emitted during the cold start cycles (both Artemis urban and WLTC) followed by motorway, hot WLTC, and hot urban cycle. Gasoline vehicles generally emitted higher BC 350 concentrations compared to the diesel vehicles. The average BC EFs for the GDI5, GDI1 (during Artemis cold urban cycle) and the GDI3 (during WLTC cold) vehicles were 3.18, 7.14, and 5.7 mg km -1 , confirming a relative small variability among the vehicles having the same injection technology. These results are in a reasonably good agreement with previous studies, https://doi.org/10.5194/acp-2020-842 Preprint. Discussion started: 4 September 2020 c Author(s) 2020. CC BY 4.0 License.

Emission factors
such as Saliba et al. (2017) who reported elemental carbon (EC) EFs between 0.08 and 5.8 mg km -1 for GDI light vehicles (models 2012-2014) during a unified cycle cold start. Taking into account the cycle distance, the fuel consumption and the 355 fuel density (Table S6), the EFs were converted into mg kg -1 fuel. The BC EFs during the cold cycles for GDI5, GDI1 and GDI3 were 51, 101 and 120 mg kg -1 fuel, respectively. These values are in good agreement with Pieber et al. (2018) and Platt et al. (2017), who reported BC EFs for GDI Euro5 light duty vehicles between 10 and 100 mg kg -1 fuel during cold start WLTC and 10-250 mg kg -1 fuel during cold start New European Driving Cycle (NEDC). The BC EF for the diesel D1 vehicle during Artemis cold start cycle was 0.07 mg km -1 , while for the D3 was 0.01 mg km -1 during a WLTC cold start in agreement with Platt et al. 360 (2017).
The OA EFs emitted by the GDI5, GDI3 and PFI4 gasoline cars were 66.3, 103.5 and 8.4 μg km -1 respectively during the cold start cycle. Converting these EFs into mass of fuel the corresponding EFs were 1.1, 2.2 and 0.2 mg kg -1 fuel for the GDI5, GDI3 and PFI4, respectively. This is in agreement with Pieber et al. (2018) who reported EFs between 1 and 10 mg kg -1 fuel for cold start WLTC. The corresponding OA EFs of the D1, D3 and D4 vehicles were 11, 0.7 and 61 μg km -1 (or 0.21, 0.02 and 0.94 365 mg kg -1 fuel), correspondingly. The OA EF of the D3 was quite low and this is generally in agreement with Platt et al. (2017), who found that the diesel vehicles equipped with a DPF emitted very low amounts of OA (less than 0.01 g km -1 fuel). The D4 OA EF is close to the values of gasoline vehicles underlining the impact of DPF condition on PM emissions (see also below).
The PAHs EFs were measured for only 3 vehicles: GDI5, PFI4 and D4. The EFs for GDI5 were 1.5 and 1.1 μg km -1 during cold urban and motorway cycles correspondingly. The PFI4 vehicle emitted considerably lower PAHs, only 0.4 μg km -1 for 370 cold urban and 0.04 μg km -1 for motorway cycles. These values are in relatively good agreement with Cheung et al. (2010) who reported PAHs EFs of 0.67 μg km -1 from PFI gasoline cars. Alves et al. (2015) measured low PAHs EFs of 0.002 μg km -1 during Artemis urban cold start and much higher values of PAHs up to 3.9 μg km -1 during Artemis road, but surprisingly none of the tested gasoline cars in the present work showed similar trends. Much higher values from 108 to 489 μg km -1 during WLTC cold start and from 27 to 102 μg km -1 during WLTC hot start were recently reported for both gaseous and particle-375 bound PAHs for three Euro 5 GDI cars (Muñoz et al., 2018). The PAHs EFs for the D4 car were 2.0 μg km -1 for the cold urban and 1.7 μg km -1 for the motorway cycle. These findings are similar to the values of 2.3 μg km -1 for a cold urban cycle and 0.6 μg km -1 for a motorway cycle for a diesel Euro 5 measured by Alves et al. (2015). Vouitsis et al. (2009) reported PAHs EFs of 7.7 and 2.0 μg km -1 for cold urban and motorway cycles, correspondingly, for a Euro 4 car equipped with a CDPF (which actually corresponds to a Euro 5 technology). When Cheung et al. (2010) added a DPF to a Euro 4 diesel car (converting it 380 into a "Euro 5") the PAHs were below detection limit, underlining the high variability of PAHs emissions from different Euro 5 cars. As already mentioned in the previous sections both OA and PAHs EFs for D4 were surprisingly high for a diesel vehicle equipped with a DPF suggesting a failure in the after-treatment device, as also supported by TEM images Figure S9 (b).
Sulfate, ammonium and nitrate EFs were generally low. The highest sulfate emissions were observed for the D3 during the cold start WLTC cycle with 4.2 μg km -1 and for the D1 during the Artemis motorway cycle with values of 1.3 μg km -1 . Both 385 D1 and D3 cars were equipped with a CDPF, while the D4 was equipped with an FBC-DPF and had lower sulfate EFs (0.18 -0.22 μg km -1 ), underlying the determining influence of the DPF technology on PM chemical composition. Nitrate was mostly https://doi.org/10.5194/acp-2020-842 Preprint. Discussion started: 4 September 2020 c Author(s) 2020. CC BY 4.0 License. emitted by gasoline cars, the highest EFs were measured for the GDI3 car with 2.2 and 4.9 μg km -1 for cold and hot start WLTC, respectively.

Conclusions 390
We characterized the chemical composition and we evaluated the emission factors of three diesel (two CDPF and one FBC-DPF) and four gasoline (three GDI and one PFI) Euro 5 light duty vehicles during transient cycles. Most of the particulate matter was emitted at the beginning of the cold start cycle due to the incomplete combustion and low catalyst efficiency. BC was always the dominant species accounting for 83-98% of the total particle mass concentration, while the corresponding OA fraction ranged between 1.8 and 14%. The OA emissions of the GDI gasoline cars were 5 to 12 times higher compared to the 395 gasoline PFI vehicle OA emissions. Organosulfur containing ion fragments were detected for the first time at the exhaust of gasoline cars, probably from the release of lubricant oil, and accounted for 2-7% of the total organic mass concentration. In total 45 PAHs were identified and quantified. Similar to the OA, the PAHs emitted by the GDI car were considerably higher in comparison to those emitted by the PFI car. Approximately 52-66% of the PAHs were unsubstituted PAHs, followed by methylated PAHs (14-21% of the PAHs), oxygenated PAHs (5-19%), nitro-PAHs (1-11%) and amino PAHs (1-6%). 400 Unexpectedly, the GDI car emitted the higher concentrations of nitro-PAHs, questioning the validity of using some NPAHs as marker of diesel emissions.
Oil droplets associated to metallic components as calcium, phosphorus, sulfur and zinc were also observed in PM from both gasoline and diesel vehicles. Analysis of particles emitted from the PFI vehicle revealed a disordered soot surface, which could affect the chemical reactivity and toxicity of the PM. 405 Nanoparticles in the size range of 15-20 nm, mainly composed of ammonium bisulfate, were measured during the motorway cycle, suggesting passive regeneration of the DPF for CDPF diesel vehicles. This behavior was not observed for the FBC-DPF vehicle, indicating that the different after-treatment strategy highly affects the PM size and composition.
Diesel cars equipped with well-functioning after treatment-devices generally emitted far less pollutants than the gasoline vehicles, but in the case of a DPF failure, very high levels of PM, similar to those reported for the GDI vehicles,were measured. 410 This indicates that the DPF condition is important and special attention should be given to its maintenance during the lifetime of the vehicle. All particle characteristics investigated in this work should be taken into account in emission control strategies and in the assessment of the impact of light duty particle emissions on the environment and on human health.
Data availability. All data from this study are available from the authors upon request. 415 Supplement. The supplement related to this article is available on-line at: (link will be included by Copernicus).