Identification of secondary aerosol precursors emitted by an 1 aircraft turbofan 2

. Oxidative processing of aircraft turbine-engine exhaust was studied using a potential aerosol mass 13 (PAM) chamber at different engine loads corresponding to typical flight operations. Measurements were 14 conducted at an engine test cell. Organic gases (OGs) and particle emissions pre/post PAM were measured. A 15 suite of instruments, including a proton-transfer-reaction mass spectrometer (PTR-MS) for OGs, a multi-gas 16 analyzer for CO, CO 2 , NO X , and an aerosol mass spectrometer (AMS) for non-refractory particulate matter (NR- 17 PM 1 ) were used. Total aerosol mass was dominated by secondary aerosol formation, which was approximately 18 two orders of magnitude higher than the primary aerosol. The chemical composition of both gaseous and 19 particle emissions were also monitored at different engine loads and were thrust dependent. At idling load 20 (thrust 2.5-7%), more than 90% of the secondary particle mass was organic and could be explained by the 21 oxidation of gaseous aromatic species/ OGs; e.g. aircraft NMOG emissions in Zurich for 2010 in the range of 90−190 tons/year (Kilic al., on the average SOA bulk yields (SOA/total NMOG) obtained herein (~5-8%), we estimate a total SOA production potential from airport emissions for the area of Zurich to range from 5.4 to 13.2 tons/year. SOA production potential values can be directly compared to emissions from on-road vehicles derived from the emission inventory, which provides worldwide temporally and spatially resolved NMOG from road vehicles with a grid size of ~200 km 2 . For the grid cell containing Zurich (47.25° North, East) the NMOG from on-road is to be SOA are than car volatility suggest comparably this yield value for emissions from both types of vehicles (Platt the total SOA production potential from on road vehicles for the area of Zurich to be tons/year, higher than SOA from aircraft emissions. However, the airport is a point source within this region and thus the relative contribution of the airport emissions to a specific location downwind of this source is higher than implied by this calculation. Although this estimate applies to a specific airport, it does indicate that aircraft NMOG emissions may constitute significant SOA precursors downwind of airports, while other fossil combustion dominate urban areas in general.

Aging of fossil fuel combustion exhaust leads to SA/PA ratios higher than 1. Single-ring aromatics are 43 traditionally thought to be the most important secondary organic aerosol (SOA) precursors from combustion 44 emissions. While this has been shown to be the case for some emissions, e.g. from 2-stroke engines (Platt et al.,45 2014), in other cases non-traditional precursors were assessed to be responsible for the bulk of the SOA mass 46 formed, e.g. for biomass smoke ( authors showed the dominance of secondary organic aerosol (SOA) at low loads, while at high loads sulfate was 50 the main SA produced. While single-ring aromatic compounds determined using gas-chromatography/mass 51 Atmos. Chem. Phys. Discuss., https://doi.org/10.5194/acp-2017-907 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 8 November 2017 c Author(s) 2017. CC BY 4.0 License. spectrometry seemed to be important precursors of the SOA formed, a greater part of SOA was believed to 52 originate from non-traditional precursors, whose nature remains to be identified (Miracolo et al., 2011;. 53 In this study, we measured the SA production potential of aircraft jet engine exhaust as a function of engine load 54 and examined the bulk gas-phase organic emissions and their SOA formation potential. SOA was produced by 55 OH-initiated oxidation of aircraft NMOG emissions in a potential aerosol mass (PAM) flow reactor (Kang et al., 56 2007). Primary and secondary PM mass was characterized for different engine loads, using an aerosol mass 57 spectrometer (AMS). SOA precursors were analyzed in real-time by a proton-transfer-reaction mass 58 spectrometer (PTR-MS) and SOA closure was examined under different conditions. The impact of these 59 emissions and their SOA potential in typical urban atmospheres, at the proximity of airports is assessed and 60 compared to other mobile sources. 61 2 Methods 62

Experimental setup 63
Exhaust measurements were conducted to characterize NMOG and non-refractory submicron particulate mass 64 (NR-PM 1 ) emissions from an in-production CFM56 variant turbofan in the test cell of SR Technics at Zurich 65 Airport. The test engine was fueled with standard JET A-1 fuel, and was operated at several engine loads, 66 selected to represent aircraft activities during a typical landing/take-off (LTO) cycle. Engine loads were set by 67 specifying the combustion chamber inlet temperature values which correlate with a specific thrust (lbf) at 68 standard atmospheric conditions. The selected loads included idle-taxi (3-7% of the maximum thrust), approach 69 (30% of the maximum thrust), and an approximated cruise load (50-65% of the maximum static thrust). After 70 starting the engine, a warm-up sequence of 25 minutes ran before each test, consisting of five minute-long steps 71 at thrusts of 5%, 15%, 7%, 65% and 85% in sequence.

72
A simplified scheme of the experimental setup is shown in Figure 1 and is discussed in detail elsewhere (Kilic et  73 al The condensed phase was continuously monitored before and after the PAM using a high resolution time-of-134 flight aerosol mass spectrometer (AMS) and a scanning mobility particle sizer (SMPS). The reader is referred to 135 DeCarlo et al. (2006) for a more detailed description of the AMS operating principles, calibrations protocols, 136 and analysis procedures. Briefly, a particle beam sampled through an aerodynamic lens is alternately blocked 137 and unblocked, yielding the bulk particle mass spectra (MS mode) of the non-refractory (NR) species, including 138 organic aerosols (OA), NO 3 -, SO 4 2-, NH 4 + , and Cl -. The NR particles are flash vaporized by impaction on a 139 heated tungsten surface (heated to ~ 600°C) at ∼ 10 −7 Torr. The resulting gases are ionized by electron 140 ionization (EI, 70 eV) and the mass-to-charge ratios (m/z) of the fragments are determined by the ToF mass 141 spectrometer. The AMS was operated in the V-mode, with a time resolution of 30 sec. The AMS data were 142 analyzed using the SQUIRREL (version 1.52L) and PIKA (1.11L) analysis software in Igor Pro 6.3 143 (WaveMetrics). Standard relative ionization efficiencies (RIE) were assumed for the organic aerosol and 144 chloride (RIE = 1.4, and 1.3, respectively) and experimentally determined for sulfate and ammonium (RIE = 145 ~1.1 and ~4, respectively). The collection efficiency due to the particle bounce was determined to be ~1 under 146 our conditions for organic rich aerosols by comparing the AMS mass to the SMPS volume (assuming an OA 147 density of 1.4). 148

Data analysis 149
Emissions from the aircraft turbofan were measured at different thrust levels referred to as "test points" 150 hereinafter. Results and discussion 178

SOA formation as a function of OH exposure 179
The evolution of the chemical composition of the primary organic gases and NR-PM 1 components with 180 increasing OH exposure is shown in Figure 2 for engine idling operation (thrust 3%). Measurements were 181 conducted for primary emissions, as well as for OH exposures of 59x10 6 , 88x10 6 , and 113x10 6 molecules cm -3 182 h, which correspond to approximately 39, 58, and 75 hours of atmospheric aging under an average tropospheric 183 OH concentration of 1.5 x 10 6 molecules cm -3 (Mao et al., 2009). The OH exposure, calculated using d9-184 butanol as a tracer, was varied by varying the light intensity. 185 Figure 2a shows the OG composition under these conditions with compounds classified as a function of their 186 molecular composition, as described in Kilic et al. (2017). A stepwise increase of the OH exposure reduced the 187 NMOG mass detected in the chamber by 35%, 40% and 50%. Except for carboxylic acids, the concentrations of 188 all NMOGs decreased during aging, indicating that their loss rate exceed their production from other NMOGs. 189 For example, aromatic compounds and carbonyls were oxidized in the PAM by up to 90% and 50%, 190 respectively, while the acids doubled after 75 hours of daytime-equivalent aging. 191 Figure 2b shows a time series of secondary NR-PM 1 composition, as well as the concentrations of two of the 192 most abundant aromatic gases, C 10 H 14 and C 11 H 16, for the same experiment. Here stable oxidation conditions 193 were alternated with sampling of primary emissions, with OH exposures indicated in the figure. Secondary 194 aerosol, especially SOA, dominated the total NRPM 1 . By increasing the OH exposure from 59x10 6 to 88x10 6 , 195 the generated SOA increased by approximately 14%. However, increasing the OH exposure further to 113x10 6 196 molecules cm -3 h yielded only an additional 3% increase in SOA mass. This suggests that at these OH 197 exposures, the bulk of SOA precursors have reacted and the additional SOA production did not significantly 198 exceed its loss. Under these conditions, the formed SOA may be considered as a reasonable estimate for the 199 total SOA potential. The observed production rate of SOA against OH exposure is consistent with precursor 200 reaction rates of 8 × 10 -12 molecule -1 cm 3 s -1 . This estimate is based on the assumption of a constant SOA mass 201 yield with aging and instantaneous equilibrium partitioning of the condensable gases, and is therefore lower than 202 the reaction rates of the main identified precursors (see below). SOA production rates are thus expected to be 203 faster in the ambient atmosphere. increased from 3-5% to 90%. At thrust 3-5%, the emissions of gaseous aromatic-hydrocarbons were highest 208 (with an EI of ~5g/kg fuel) and decreased with increasing thrust (with an EI of ~0.15 g/kg fuel at thrust 90%). 209 Similar to aromatic gases, SOA were formed mostly at 3-5% thrust and had a declining trend with thrust. In hydrocarbons emitted were single-ring aromatics. 75 -95% of these aromatics were oxidized with an OH 223 exposure of ~90 x 10 6 molecules cm -3 h in the PAM. 224 By using previously reported SOA yields (Table 2) for NMOGs, SOA production was predicted from individual 225 precursors according to Eq. 1. Figure 5 shows Results in Figure 5 indicate that the most important SOA precursors emitted by turbine engines at idle are 231 aromatic hydrocarbons such as benzene derivatives but also oxygenated aromatics such as phenol. The predicted 232 SOA formed by aromatics alone, both by aromatic hydrocarbons (60-70%) and oxygenated-aromatics (15-25%), 233 explained all AMS-determined SOA at low loads (thrust 3-5%) and most of the SOA formed (by 80%) at idle 6-234 7% ( Figure 5). Predicted aromatic SOA from benzene (C 6 H 6 ), C2-benzenes (C 8 H 10 ), C3-benzenes (C 9 H 12 ), C4-235 benzenes (C 10 H 14 ), dimethylstyrenes (C 10 H 12 ), toluene (C 7 H 8 ), methylbenzaldehydes (C 8 H 8 O) and phenol 236 (C 6 H 6 O) accounted for 60% of the AMS-determined SOA at 3-5% thrust ( Figure 5). These results are consistent 237 with those previously obtained using a smog chamber, confirming that aromatic compounds are indeed 238 important SOA precursors in jet-engine emissions (Miracolo et al., 2011). Only a small fraction of these 239 compounds was determined in previous experiments using GC/MS measurements and therefore traditionally 240 considered as SOA precursors in models. Here, compared to previous experiments we show that non-traditional 241 aromatic and oxy-aromatic compounds, including naphthalene and its alkyl derivatives, C>3 alkyl derivatives of 242 single ring aromatics, and phenols, can explain the gap between measured SOA and SOA predicted based on 243 traditional precursors. 244 Exhaust-aging experiments were repeated 6 times at thrust 3-5% and the oxidation of NMOGs varied during 245 each of these aging experiments. Error bars shown in Figure 5 denote this variability in NMOG oxidation (in the 246 PAM) during aging experiments of the same thrust level. Indeed, errors related to yield values used may 247 significantly influence the results. These errors may be systematic and are complex to assess. They can be 248 affected by potential differences between the oxidation conditions in chambers and in the PAM (e.g. NOx, RH, 249 particle mass). Yields obtained with the PAM are consistent with those obtained from chambers (Bruns et al.,  250 2015), therefore we do not expect large systematic errors in the SOA predicted. However, based on the 251 variability of yields in previous chamber experiments we estimate the accuracy of our prediction to be within a 252 factor of 2, indicating that within our uncertainties a significant fraction of the precursors was identified. 253 NMOGs, including aromatic gases, were reduced with increasing thrust (from thrust 3-5% to thrust 6-7%) due 254 to more efficient operation of the turbine engine. This decrease amounted to 40% for the sum of aromatic HCs 255 and corresponded to a 30% decrease in SOA EI. Therefore, a more efficient engine operation implies less 256 NMOG emissions and reduced SOA formation potential at idle. 257

SOA formation at an approximated cruise load 258
A comparison of the predicted SOA with the SOA determined by the AMS is presented in Figure 6 at cruise 259 loads (top panel). Figure 6 also shows the SOA contribution predicted by the oxidation of NMOGs in the PAM 260 (bottom panel) under the same engine conditions. The SOA EI was 0.07 g/kg fuel for cruise load. The predicted 261 SOA fraction accounted for only 30% of the AMS-determined SOA (green bar, Figure 6) during cruise load 262 experiments. Aromatic SOA (predicted) accounted for only 4% of the AMS-determined SOA during these 263 experiments. The major fraction of the remaining SOA mass that was assigned to the identified precursors was 264 predicted to be from oxygenated NMOG molecules ( Figure 6). Another 6% of the determined SOA may 265 originate from non-aromatic HCs (aliphatics and HC fragments > C6). 266 Predicted SOA was significantly lower compared to the measured SOA. While SOA precursors remain 267 unidentified under these conditions, several hypotheses might explain the observation. First, we could not 268 determine the contribution of alkanes smaller than 9 carbon atoms to the formed SOA, because these 269 However, our data do not suggest that a great part of the observed SOA is from non-measured alkanes, as we do 273 not observe any increase in the contribution of hydrocarbon fragments in the PTR-MS compared to idling 274 emissions. Second, the oxidation of primary semi-volatile compounds may yield significant SOA, because of 275 their elevated yields of near unity (Robinson et al., 2007). However, we note that these semi-volatile precursors 276 would play an important role at low aerosol concentrations, when most of these precursors reside in the gas-277 phase where they can be oxidized. Under our conditions, concentrations range between 10 and 50 µg m -3 and a 278 substantial fraction of these products resides already in the particle phase. Therefore the oxidation of these 279 products in the gas-phase by OH is unlikely to explain the observed entire 10-fold increase in the OA mass upon 280 oxidation, but only part of the mass. Finally, the PTR-MS data suggest that a great part of the precursors 281 measured are highly oxygenated gases, with O:C ratios ranging from 0.2 to 0.7, including, among others, 282 anhydrides (e.g. phthalic, succinic and maleic) and quinone derivatives. Unlike aromatic compounds and 283 alkanes present in the fuel, these compounds are likely formed at high temperature during combustion. The SOA 284 yields of these compounds remain unknown and it is likely that the yield value of 0.15 used here is a lower 285 estimate, which would result in an underestimation of the contribution of these compounds to the observed 286 SOA. We also note that unlike precursors detected under idle conditions, the ionization efficiency and the 287 fragmentation pattern of these compounds in the PTR-MS are highly uncertain, resulting in large uncertainties 288 in our predicted SOA. Therefore, results in Figure 6 should be considered with care. Notwithstanding these 289 uncertainties, we note that at cruise conditions the SOA contribution to the total secondary PM is minor 290 compared to sulfate and therefore these uncertainties have little impact on the implications of our results. 291

Conclusions and implications for ambient air quality 292
Gas-phase primary emissions and SA formation from an in-production turbofan were investigated in a test cell. 293 The engine loads (thrusts) during experiments were selected to simulate different aircraft operations. These 294 operations are summarized as landing take-off (LTO) cycle under four modes taxi/idle, approach, climb and 295 take-off with corresponding engine loads of 3-7%, 30%, 85% and 100%, respectively. In addition an 296 approximated cruising load (60%) was selected. 297 At idle conditions, SOA formation was mostly attributed to the oxidative processing of aromatic gases. Benzene 298 derivatives together with phenol were predicted as the major SOA precursors for an idling aircraft. Meanwhile, 299 during cruise load the emission of aromatic compounds was much lower and only explained a minor fraction of 300 SOA (4%). During these conditions, however, sulfate was found to dominate SA, contributing ~85% of the total 301 mass of aged aerosols and therefore its fraction is more relevant aloft. 302 The oxidation of NMOGs in the PAM yielded a SOA EI 100 times greater than POA under idling conditions 303 and 10 times greater at cruise load. According to our calculated production rates SOA from airport emissions 304 (idling jet engines) exceeds POA by a factor of 10 after only 3 hours of atmospheric aging and therefore 305 considerably impacts urban areas downwind of airport emissions. Compared to idling aircraft emissions, aging 306 of vehicle exhaust emissions results in much lower enhancements, ranging between factors of 5-10 and 1.5-3, 307 for gasoline and diesel vehicles, respectively (Gordon et al., 2014a; 2014b). 308 The NMOG emission factors and SOA potential can be used in conjunction with emission inventories and fuel 309 use data to assess the impact of aircraft emissions on air quality in comparison with other mobile sources. Here, 310 we have considered the Zurich international airport as an example (Switzerland, 23 million passengers in 2010). 311 Combining Based on the average SOA bulk yields (SOA/total NMOG) obtained herein (~5-8%), we estimate a total SOA 314 production potential from airport emissions for the area of Zurich to range from 5.4 to 13.2 tons/year. These 315 SOA production potential values can be directly compared to emissions from on-road vehicles derived from the 316 EDGARv4.2 emission inventory, which provides worldwide temporally and spatially resolved NMOG 317 emissions from road vehicles with a grid size of ~200 km 2 . For the grid cell containing Zurich (47.25° North, 318 8.75° East) the NMOG emissions from on-road vehicles is estimated to be 631 tons/year. While SOA yields 319 from diesel vehicle emissions are expected to be more elevated than those from gasoline car emissions, due to 320 the presence of intermediate volatility species, recent reports suggest these yields to be comparably high, ~15% 321 (Gentner et al., 2017). Using this yield value for emissions from both types of vehicles , we 322 estimate the total SOA production potential from on road vehicles for the area of Zurich to be ~94 tons/year, 10 323 fold higher than SOA from aircraft emissions. However, the airport is a point source within this region and thus 324 the relative contribution of the airport emissions to a specific location downwind of this source is significantly 325 higher than implied by this calculation. Although this estimate applies to a specific airport, it does indicate that 326 aircraft NMOG emissions may constitute significant SOA precursors downwind of airports, while other fossil 327 fuel combustion sources dominate urban areas in general.