Laboratory evaluation of the volatility and composition of ultrafine particles generated from jet engine lubrication oil

Abstract We present a laboratory evaluation of the volatility and organic compositions of ultrafine particles (UFPs) generated from unused jet engine lubrication oil. The particle generation method was based on the evaporation of jet engine lubrication oil droplets in a heated quartz tube followed by the nucleation of oil vapors by cooling. Our experiments were not aimed at strictly simulating the particle formation processes in jet engines but exploring the possibility of generating UFPs having chemical properties similar to those of UFPs observed in the real atmosphere. The particle volatility was estimated by measuring the particle number concentrations upstream and downstream of a heated stainless-steel evaporation tube. The organic compositions of the UFPs were measured by thermal desorption – gas chromatography mass spectrometry analysis of size-segregated aerosol samples collected on quartz fiber filters. The volatility and compositions of the UFPs varied significantly with the particle generation (quartz tube) temperature (TPG). We show that the evaporation of the jet engine lubrication oil droplets at TPG = 250 °C and subsequent nucleation by rapid cooling could form UFPs having volatility similar to that of the particles observed at Narita International Airport and composed of nearly intact forms of jet engine lubrication oil. Our results suggest that it is possible to generate UFPs that can consistently explain the chemical properties of the UFPs observed in the real atmosphere by the evaporation and nucleation of jet engine lubrication oil droplets.


Introduction
The characterization of ultrafine particles (UFPs; diameter < 100 nm) emitted from jet aircraft is an important issue in assessing the impacts of aviation on climate and human health (Masiol and Harrison 2014;Stacey 2019;ICAO 2022).UFP emissions from turbofan engines include the formation of nonvolatile particles during combustion processes and the nucleation of volatile particles during plume expansion (e.g., K€ archer et al. 2000;Onasch et al. 2009;Timko et al. 2010).
We conducted field measurements of aerosols near a runway at Narita International Airport (NRT) in February 2018 (Fushimi et al. 2019;Saitoh et al. 2019;Takegawa et al. 2021Takegawa et al. , 2023)).Our measurements provided new insights into the volatility and organic compositions of UFPs emitted from aircraft under real-world operating conditions.Takegawa et al. (2023) showed that 50% and տ 90% (number basis) of the volatile particles observed at NRT were removed at an evaporation tube temperature of 150 C and 250 C, respectively.They estimated that a large fraction (more than half by number) of the volatile UFPs emitted from aircraft was equally or less volatile than 30-nm pure triacontane (n-C 30 H 62 : C 30 ) particles and only a small fraction of the volatile UFPs was equally or less volatile than 30-nm pure tetracontane (n-C 40 H 82 : C 40 ) particles.Fushimi et al. (2019) revealed that organic compounds of UFPs (10-30 nm) sampled at NRT were dominated by nearly intact forms of jet engine lubrication oil.Considering that organic compounds originating from jet fuels (corresponding to the volatility of n-alkanes with carbon numbers in the C 11 -C 18 range) can be well separated from those from jet engine lubrication oil (corresponding to the volatility of n-alkanes with carbon numbers in the C 27 -C 36 range), the volatility and composition of UFPs obtained from the above two studies collectively suggest that jet engine lubrication oil is the major source of aircraft exhaust UFPs.The volatility and organic compositions of diesel exhaust UFPs and their relation to diesel engine lubrication oil have been extensively studied (e.g., Sakurai et al. 2003;Tobias et al. 2001).Such studies have not been performed for aircraft exhaust UFPs.
The major population of volatile particles emitted from aircraft was generally found at particle diameters smaller than 50 nm near runways (e.g., Moore et al. 2017;Takegawa et al. 2021).For some types of turbofan engines, submicron-to micron-sized droplets originating from jet engine lubrication oil (oil droplets) could be released to the atmosphere from breather vent systems (Timko et al. 2010;Yu et al. 2010).The generation of liquid or solid particles by mechanical processes such as spraying and nebulizing tends to yield droplets in the size range from submicron to a few hundred microns, whereas the generation of liquid or solid particles by evaporation and nucleation processes can yield high concentrations of nanometer-sized particles (Hinds 1999).Following the discussion by Fushimi et al. (2019) and Takegawa et al. (2023), we hypothesize that the evaporation of jet engine lubrication oil droplets by mixing with high-temperature combustion gas and subsequent cooling during plume expansion may lead to the formation of UFPs having organic compositions similar to those of unused jet engine lubrication oil.A simple approach to test this hypothesis is to generate UFPs from jet engine lubrication oil in the laboratory and measure the chemical properties of the UFPs.
The physical and chemical characteristics of UFPs generated by evaporation and nucleation of a source material depend on various factors including the thermochemical properties of the material, the heating and cooling temperatures, and the concentrations of vapors evolved from the source material.These well-known characteristics have been used for the generation of test particles in the laboratory.The generation of particles from pure materials is straightforward, whereas that from mixed materials is rather complex because some compounds may be affected by thermal decomposition and distillation at selected heating and cooling temperatures.
The purpose of this study was to investigate the volatility and organic composition of UFPs generated from unused jet engine lubrication oil via the evaporation and nucleation processes in the laboratory.Our experiments were not aimed at strictly simulating the particle formation processes in jet engines by laboratory-based approaches, which is highly challenging and unrealistic.The key question was whether it is possible to generate UFPs having volatility similar to that of the particles observed at NRT (Takegawa et al. 2023) and composed of nearly intact forms of jet engine lubrication oil.

Overview
Figure 1 shows the experimental apparatus for investigating the volatility and compositions of UFPs generated from jet engine lubrication oil.We used a lubrication oil sample from Mobil Jet Oil II, which is widely used in commercial aircraft (Winder and Balouet 2002).According to the material safety datasheet provided by the manufacturer, the flash and autoignition points of Mobil Jet Oil II are 270 C and 404 C, respectively.The thermochemical properties of Mobil Jet Oil II were measured using a thermogravimetricdifferential thermal analysis instrument (TG-DTA; Model TG-DTA8122, Rigaku, Japan) at the Fujitsu Quality Laboratory Environment Center Ltd.The analysis was performed in air atmosphere at a heating rate of 10 C min −1 .
The experimental targets included the optimization of particle generation by size distribution measurements, evaluation of the number-based particle volatility, and qualitative analysis of the organic compositions by filter sampling.The optimal conditions for each experiment were determined sequentially: i.e., the conditions for the particle volatility measurements were determined based on the particle size distribution results, and those for the particle composition measurements were determined based on the particle size distribution and volatility results.

Particle generation
Figure 1a shows the experimental apparatus for generating particles from the jet engine lubrication oil.The main components included a glass nebulizer for generating particles from jet engine lubrication oil (hereafter referred to as the oil droplets), a quartz tube consisting of heating and cooling sections for the evaporation and nucleation processes, and mass flow controllers.The quartz tube was heated by an electric furnace with a length of 300 mm.The outer and inner diameters of the quartz tube in the heated section were 30 and 26 mm, respectively.A thermocouple sensor was attached to the outer wall of the quartz tube at the axial center position for controlling the temperature of the quartz tube.The outer and inner diameters of the quartz tube were reduced to 10 and 7 mm, respectively, for connecting to a T-junction composed of a 3/8-inch perfluoroalkoxy alkane (PFA) T-union.The air stream containing the oil droplets and their vapors (oil droplet flow) was mixed with a particle-free air stream (dilution flow) supplied from an air compressor and a mass flow controller at the T-junction for generating supersaturation conditions for the oil vapors.A copper tube and a buffer tank were placed downstream of the T-junction to bring the temperature of the aerosol flow to room temperature.We also tested another experimental setup during preliminary experiments: jet engine lubrication oil droplets generated by the nebulizer were mixed with particlefree air before being introduced into the heated quartz tube.The details of the preliminary experiments are described in the Supplemental Information (SI).
Figure 1b shows the experimental apparatus for measuring the size distributions of particles generated from the jet engine lubrication oil.The number size distributions of the generated particles were measured using a scanning mobility particle sizer (SMPS) consisting of a differential mobility analyzer (DMA; Model 3081, TSI, Inc., Shoreview, MN, USA) and a condensation particle counter (CPC; Model 3772, TSI; d 50 ¼ 10 nm).Furthermore, an aerodynamic particle sizer (APS; Model 3321, TSI) was temporarily used to measure the population balance of nucleation-and coarse-mode particles.An additional dilution flow was needed upstream of the APS for a flow balance.Synchronized SMPS and APS measurements were performed independently of the series of the size distribution measurements shown in Figure 1b.
The major population of the oil droplets was in the size range larger than 100 nm.The size distributions of the oil droplets could be shifted to smaller diameters by diluting a liquid sample of the lubrication oil with highly volatile solvents (e.g., methanol, n-hexane).Assuming that the initial droplet diameter generated from the nebulizer is constant, the dilution factor of 1000 would result in a decrease in the particle diameter by an order of magnitude.However, small contaminants in the solvents of the order of 0.1% could yield non-negligible artifacts in the production of nanoparticles.To avoid this ambiguity, we used undiluted lubrication oil samples for the nebulizer.The number size distributions of UFPs generated via the evaporation of the oil droplets followed by the nucleation of oil vapors exhibited complex dependency on the oil droplet and dilution flow rates.We set the oil droplet and dilution flow rates to 2.3 and 2.0 L min −1 , respectively.These values were experimentally determined to increase the fraction of particle number and volume concentrations in the UFP size range.
The Reynolds number of the oil droplet flow and dilution flow was estimated to be of the order of 10 2 , indicating that the flow regimes were laminar.The residence time of air in the heated section at the radial center of the quartz tube was estimated to be Շ 2 s, depending on the quartz tube temperature.A previous study on fluid dynamics experiments showed that the mixing of two air streams at a T-junction may cause vortices after the mixing when the flowrates of the two streams are comparable (Hibara et al. 2004).We assume that similar features could occur upon the mixing of the oil droplet and dilution flows at the T-junction, resulting in rapid cooling of the heated oil droplet flow.Although the current configuration may be still insufficient to simulate the heating and cooling processes in real jet engines, we consider that it is tentatively an optimized setup for testing the volatility and composition of UFPs generated from jet engine lubrication oil.
As described in Section 1, we hypothesize that the evaporation of jet engine lubrication oil droplets by mixing with high-temperature combustion gas and subsequent cooling during plume expansion may lead to the formation of UFPs having organic compositions similar to those of unused jet engine lubrication oil.The air temperature at the engine exit plane can reach 900 C (Kittelson et al. 2022).The actual temperature conditions to which jet engine lubrication oil droplets are exposed in real jet engines are unknown.The set point of the particle generation temperature (quartz tube temperature) (T PG ; C) was altered from room temperature (17-25 C; uncontrolled) to 600 C, which was the upper limit in the current experimental setup.We analyzed the data obtained at the T PG values of room temperature, 200 C, 250 C, 300 C, 400 C, and 600 C.
The air temperature inside the quartz tube was measured for T PG ¼ 200-300 C independently of the series of the main experiments.The air flow was introduced from the entrance port of the quartz tube, and a thin thermocouple sensor was inserted from the exit port (vented to the atmosphere) to point approximately toward the center of the quartz tube along the axial direction.There was a large temperature gradient from the end of the heated section of the quartz tube to the T-junction, at which the major part of the nucleation likely took place.We measured the temperature of the outer surfaces of the tube and T-junction to ensure that it did not exceed the melting point of PFA.We observed false particle counts downstream of the T-junction without a supply of oil droplets when the quartz tube temperature was increased from a certain set point to higher ones (e.g., from 250 C to 300 C).This was probably due to adsorption and desorption of evaporated compounds at either the quartz tube or the T-junction.Potential artifacts due to these effects were estimated by passing the oil droplet flow through a particle filter upstream of the quartz tube and measuring the particle number concentrations downstream of the DMA with a mobility diameter of 30 nm.The particle volatility and organic composition measurements were conducted after reducing the artifacts by baking the quartz tube at higher temperatures than the target T PG on the day of each experiment.This prebaking procedure decreased the false counts to 0.1 cm −3 .

Particle volatility measurements
Figure 1c shows the experimental apparatus for measuring the particle volatility.Monodisperse particles classified by the DMA were introduced into the CPC 3772.An ultrafine condensation particle counter (UCPC; Model 3776, TSI; d 50 ¼ 2.5 nm) and another CPC (Model 3771, TSI; d 50 ¼ 10 nm) were used to measure the particle number concentrations downstream of a heated stainless-steel evaporation tube.The UCPC 3776, CPC 3771, and evaporation tube were the same as those used at NRT (Takegawa and Sakurai 2011;Takegawa, Iida, and Sakurai 2017, Takegawa et al. 2021, Takegawa et al. 2023).The evaporation tube was used for investigating the volatility of the jet engine lubrication oil particles.The temperature of the evaporation tube (T ET ) was varied from room temperature to 350 C for this purpose.This approach is similar to those used for particle volatility measurements by previous studies (Sakurai et al. 2003;Huffman et al. 2008;Cappa 2010), in which particle volume or mass fraction remaining downstream of an evaporation tube or thermodenuder was used to quantify the particle volatility.The UCPC 3776 and CPC 3771 were corrected based on calibrations at the National Institute of Advanced Industrial Science and Technology (AIST) (Takegawa and Sakurai 2011;Takegawa, Iida, andSakurai 2017, Takegawa et al. 2021).The detection efficiencies of the two CPCs were nearly plateau for particle diameters of 30 nm (Takegawa and Sakurai 2011;Takegawa, Iida, and Sakurai 2017).The performance of the CPC 3772 was not tested at AIST, and only the corrections for the sample flow rate were applied.The overall consistency between these three instruments was evaluated by comparing the particle number concentrations obtained at room temperature.
The major population of volatile particles emitted from aircraft was often found at particle diameters smaller than 50 nm and the vast majority of the particle number concentrations were found at diameters smaller than 30 nm (e.g., Moore et al. 2017;Takegawa et al. 2021).Following the discussion by Takegawa et al. (2021Takegawa et al. ( , 2023)), we set the mobility diameter of 30 nm as a representative diameter (number basis) of the particles formed by evaporation and nucleation.The CPC 3772 and 3771 measured the particle number concentrations larger than 10 nm upstream and downstream of the evaporation tube, respectively.They are referred to as the upstream and downstream CPC, respectively.The downstream CPC is simply referred to as the CPC unless otherwise noted.The monodisperse aerosol flow was mixed with particle-free air before being introduced into the upstream CPC and the evaporation tube for the flow rate matching.
The particle number fraction remaining (PNFR) downstream of the evaporation tube for a selected particle diameter (D p ; 30 nm in the current study) was calculated as a function of T PG and T ET .The PNFR was derived from the UCPC and the downstream CPC data normalized by the upstream CPC data, which are referred to as F UCPC (T PG , D p , T ET ) and F CPC (T PG , D p , T ET ), respectively.We used a modified sigmoid function to fit the relationship between F and T ET determined from the experiments: where f 0 , f 1 , a, and T 50 are fitting parameters (obtained through iteration processes) and X is either the UCPC or the (downstream) CPC.f 0 and f 1 are the asymptotic particle number fractions at lower (room temperature) and higher (տ 350 C) T ET values, respectively, a is an indicator of the steepness of the curve (smaller a values indicate steeper curves), and T 50 is the evaporation tube temperature at which the F X (T PG , D p , T ET ) value becomes (f 0 þ f 1 )/2.The f 0 value may not be equal to unity due to differences in the instrument responses of the upstream CPC, the downstream CPC, and the UCPC.Non-zero f 1 values indicate the formation of particles that were not fully vaporized at T ET ¼ 350 C (operationally "nonvolatile" particles).In order to stabilize the iteration processes, the f 1 value was forced to zero when the experimental F X (T PG , D p , T ET ) values at T ET ¼ 350 C were < 0.01.
Possible contamination of multiply charged particles at the selected DMA voltage may affect the particle volatility measurements because the removal efficiencies for multiply charged particles tend to be lower than those for singly charged particles.The details of the effects of multiply charged particles were described by Takegawa et al. (2023).The diameter of doubly charged particles for the set point of 30 nm was 43 nm.The contributions from doubly charged particles in the monodisperse aerosol flow were estimated by doubling the DMA voltage corresponding to 30 nm.The contributions from triply charged particles were not measured but likely negligible considering the Boltzmann equilibrium charge distribution.
The operation mode (e.g., heated and unheated sampling or doubling the DMA voltage) was altered every 2-4 min.The data were averaged over 3 min (1 min for the data with doubling the voltage) after the particle number concentrations had stabilized.The experimental errors were estimated from the standard deviation of the number concentrations.
The PNFR values for volatile particles depend not only on the thermochemical properties of the particles but also on the loss of the particles due to Brownian diffusion and thermophoresis (Romay et al. 1998;Durand, Crayford, and Johnson 2020).These are instrument-specific, and therefore the absolute values of PNFR may not be directly compared with those obtained by other volatility measurement systems.We used the same evaporation tube system for the ambient measurements and laboratory experiments; therefore quantitative intercomparison of the PNFR values between these two datasets can be performed.Takegawa et al. (2021Takegawa et al. ( , 2023) ) presented the size-resolved penetration and detection efficiencies for nonvolatile particles through the evaporation tube at T ET ¼ room temperature, 150 C, 250 C, and 350 C using the UCPC to test the effects of Brownian diffusion and thermophoresis.Potential uncertainties in the PNFR values due to these effects are given in the SI.

Particle composition measurements
Figure 1d shows the experimental apparatus for collecting the particles generated from the jet engine lubrication oil.We used two-stage, multi-nozzle cascade impactors (MAIS, Tokyo Dylec Corp., Japan) and a 47-mm filter sampler (NL-I-01, Tokyo Dylec Corp.) for size-resolved aerosol sampling.The flow rate through the impactors and the sampler was 9.0 L min −1 regulated by a diaphragm pump (DAP-12S, ULVAC, Inc, Japan).The structure of the cascade impactors is the same as that presented by Takegawa et al. (2013).The nozzle diameter and the number of nozzles for the first stage were 1.1 mm and 20, respectively, and those for the second stage were 0.25 mm and 30, respectively.Quartz-fiber filters were used for the collection substrates.The effective deposit area for the filter sampler was 40 mm.The quartz filters were baked at 400 C for 2 h in a nitrogen atmosphere before particle collection.Blank filter samples were also prepared using the same procedure (except for the collection of particles).The cutoff diameters of the first and second stages under the sampling conditions were theoretically estimated to be 1.9 and 0.19 lm, respectively.Therefore, we obtained aerosol samples in the size ranges of տ 2 lm, 0.2-2 lm, and Շ 0.2 lm for each T PG setting.The first size range (տ 2 lm) is representative of the original oil droplets (or thermally processed ones by heating) generated by the nebulizer, whereas the third size range (Շ 0.2 lm) is assumed to be representative of the UFPs generated by the evaporation and nucleation processes (except for T PG ¼ room temperature).The second size range is an intermediate of the oil droplets and nucleated UFPs.The collected samples were stored in a freezer before being analyzed using thermal desorptiongas chromatography mass spectrometry (TD-GC-MS).The analytical procedures followed those presented by Fushimi et al. (2019), except for the use of the GC-MS instruments and internal standards.The samples were thermally desorbed from 30 C (held for 0.5 min) to 350 C (held for 3 min) at 50 C min −1 .A 7890A GC (Agilent Technologies, Palo Alto, CA, USA) and a 7200 quadrupole high-resolution time-of-flight MS (Agilent Technologies) were used.The m/z range of 33-1000 was monitored with a mass resolution of 10,000.The current study focused on the qualitative analysis of the aerosol samples; therefore quantification of the collected mass was not performed and internal standards were not added to the samples in the TD-GC-MS analysis.We roughly estimated the mass of the particles collected on the filters based on the particle volume size distributions measured by the SMPS.We also visually inspected the color of the particle-loaded filters to determine the portion of the filters to be analyzed.

Particle generation
Figure 2 shows the TG-DTA results for Mobil Jet Oil II.The data indicate a thermally stable feature up to 200 C and significant thermal decomposition or combustion at around the flash point (270 C).The heating rate for the TG-DTA analysis (10 C min −1 ) was much slower than that for vaporizing the oil droplets in the quartz tube (100 C s −1 ).Furthermore, the evaporation processes of bulk materials and aerosol particles may differ significantly due to the effects of the surface curvature and the latent heat of vaporization.Nevertheless, these results suggest that the effective temperature of the oil droplets needs to be comparable to or lower than the flash point to generate nearly intact forms of jet engine lubrication oil particles.
The air temperature near the axial center of the heated section of the quartz tube was 50 C lower than the corresponding T PG values in the range of 200-300 C. The air temperature inside the quartz tube decreased toward the downstream of the tube.These results indicate that the effective temperature of the oil droplets was lower than the bulk flash point of Mobil Jet Oil II at T PG Շ 300 C.
We performed the current experiments within 16 wk.We refreshed the jet engine lubrication oil liquid in the nebulizer at the beginning of the experiment period.We assume that the chemical properties of the oil samples did not significantly change during the experiment period (16 wk).The basis for this assumption is described in the SI.
Figure 3 shows the particle number and volume size distributions obtained at T PG ¼ room temperature (19 C), 200 C, 250 C, 300 C, 400 C, and 600 C measured using the SMPS.The shape of the particle number and volume size distributions exhibited reasonable reproducibility for a given T PG .The data obtained at T PG ¼ 19 C, which correspond to the size distributions of the oil droplets generated by the nebulizer, showed the number peak diameter at 400-500 nm (near the upper limit of the SMPS size range) but did not show the volume peak diameter within the SMPS size range.The data obtained at T PG 200 C showed that the number and volume concentrations of particles in the submicron size range (10-400 nm) increased significantly as compared to those at T PG ¼ 19 C. The peak diameters of the particle number size distributions tended to decrease with increasing T PG .The number and volume size distributions in the submicron size range likely consisted of multiple modes at T PG 200 C. Bimodal lognormal fitting analysis for the particle number and volume size distributions is given in Section S3 in the SI.
Although the SMPS detected only the tail of the population of coarse mode particles in Figure 3, the data obtained at T PG 200 C suggest that the number and volume concentrations of coarse mode particles decreased significantly with increasing T PG .This point can be more clearly illustrated by the SMPS and APS measurements.Figure 4 shows the particle number and volume size distributions obtained at T PG ¼ room temperature (20 C) and 250 C measured by the SMPS and APS.It should be noted that the size distributions of the coarse mode might have been affected by the detectable size range of the APS (< 20 lm).A small but non-negligible number of nucleation-mode particles were observed at particle diameters of < 30 nm even for T PG ¼ 20 C. We tested SMPS measurements at room-temperature conditions (20-23 C) several times but the presence of the nucleation mode was not reproducible.Although the mechanisms of the formation of the nucleation mode at room temperature are currently unidentified, it would not affect the major results in the following sections.
From Figures 3 and 4, it is obvious that shrinkage of the oil droplets generated by the nebulizer does not explain the high number concentrations of UFPs.Therefore, except for the data at room temperature, the majority of the particles in the UFP size range should have been formed by the nucleation (and condensation) of organic compounds evaporated from the oil droplets.These results support the assumption that the third size range of the size-resolved aerosol sampler (Շ 0.2 lm) is representative of particles generated by the evaporation and nucleation processes for T PG 200 C (Section 2.4).Although only limited APS data was available, the comparison between T PG ¼ 20 C and 250 C in Figure 4 suggests that the major population of the volume concentrations of the oil droplets generated by the nebulizer was present in the coarse mode and that a significant volume fraction of the oil droplets was vaporized at T PG 250 C.

Particle volatility
Considering the particle number size distributions presented in Section 3.1, we selected T PG ¼ 250 C, 300 C, 400 C, and 600 C for the particle volatility measurements.Figure 5 shows the F UCPC (T PG , 30, T ET ) and F CPC (T PG , 30, T ET ) values as a function of T ET for the selected T PG conditions.The difference between the UCPC and CPC can be interpreted as the fraction of particles that were not fully vaporized but shrunk to sizes between 2.5 and 10 nm.Table 1 summarizes the fitting results for Figure 5.
Potential uncertainties due to the nucleation of gaseous compounds vaporized from particles in the evaporation tube (i.e., nucleation artifacts) might be a concern.The particle number concentrations classified by the DMA (30 nm), which were used for estimating the PNFR values in Figure 5, were of the order of 10 3 -10 4 cm −3 , corresponding to particle mass concentrations of 0.01-0.1 lg m −3 .Therefore, the nucleation artifacts were likely negligible under these conditions (Giechaskiel and Drossinos 2010;Takegawa et al. 2021).In fact, we did not find evidence for the artifacts in the UCPC and CPC data downstream of the evaporation tube.
Other uncertainties may originate from the effects of multiply charged particles, consistency among the three CPCs, and the loss of particles due to Brownian diffusion and thermophoresis.The contributions from doubly charged particles in the monodisperse aerosol flow were < 0.10, < 0.15, < 0.07, and < 0.04 for the datasets obtained at T PG ¼ 250 C, 300 C, 400 C, and 600 C, respectively.The errors in the experimental PNFR values due to contamination by doubly charged particles were the largest at T ET T 50 , where the experimental PNFR values exhibited the largest changes within the temperature range tested.They were estimated to be 0.01-0.04 at T ET T 50 and were mostly negligible at lower and higher T ET values.The f 0 values (the asymptotic PNFR at room temperature) for the UCPC ranged from 0.96 to 0.99.These values indicate the degree of agreement between the UCPC and the upstream CPC for the particle diameter of 30 nm.Similarly, the f 0 values for the downstream CPC ranged from 0.89 to 0.94.The lower f 0 values for the downstream CPC may be due to the longer sampling tube from the flow splitter to the downstream CPC than that from the flow splitter to the UCPC (Takegawa et al. 2021).The loss of particles We set the diagnostics for the "similarity" in the particle volatility between the laboratory-generated particles and that of Takegawa et al. ( 2023) as (1) the Table 1.Fitting parameters for the particle number fraction remaining.(250,30,150) and F CPC (250, 30, 150) values were 0.60 and 0.52, respectively, and the calculated F UCPC (250,30,250) and F CPC (250,30,250) values were both < 0.001 (Table 1).Although the F UCPC (250, 30, 150) value of 0.60 exceeded the range of the median-based volatile PNFR values (0.49-0.54) observed at NRT, it falls within the expanded range incorporating the uncertainty of 18-44% (0.29-0.76).The calculated F UCPC (250,30,150) and F CPC (250,30,150) values become 0.64 and 0.61, respectively, if we consider the potential uncertainties in the calculated PNFR values described above.They still fall within the expanded range incorporating the uncertainty (0.29-0.76).These results suggest that the volatility of the 30-nm particles generated at T PG ¼ 250 C was similar to that of the volatile UFPs observed at NRT (Takegawa et al. 2023).
The calculated F UCPC (300, 30, 150) value of 0.82 did not fall within the expanded range of the median-based volatile PNFR values observed at NRT incorporating the uncertainty (0.29-0.76).The disagreement becomes even worse if we consider the potential uncertainties in the calculated PNFR values.For T PG ¼ 400 C and 600 C, there was a long tail of F UCPC (T PG , 30, T ET ) at higher T ET values, and the f 1 values of the UCPC data were found to be 0.18 and 0.25, respectively, significantly exceeding those of the CPC data.These results suggest the formation of sub-10 nm "nonvolatile" cores at T PG տ 400 C. It is currently not clear whether these "nonvolatile" cores were composed of soot or some other high-molecular-weight organic compounds that were not fully vaporized at T ET ¼ 350 C. The T 50 values at T PG ¼ 600 C were lower than those at T PG ¼ 400 C for both the UCPC and CPC, suggesting that the 30-nm particles formed at T PG ¼ 600 C were more volatile than those at T PG ¼ 400 C. Consequently, the volatility of the 30-nm particles generated at T PG ¼ 300 C, 400 C, and 600 C did not meet the diagnostic criteria for similarity with the ambient results.

Particle compositions
Considering the particle volatility results presented in Section 3.2, we selected T PG ¼ room temperature (19 C), 250 C, 400 C, and 600 C for the particle composition measurements.Figure 6 shows mass chromatograms at m/z 113 (a marker of fatty acid esters) of Mobil Jet Oil II and those of the particles generated from the jet engine lubrication oil in the third size range (Շ 0.2 lm) at each temperature setting.The signals for the blank filter samples were much smaller than those for the particle samples (not shown).Fushimi et al. (2019) used m/z 85 instead of m/z 113 as a marker of fatty acid esters.These two peaks are essentially the same as a marker of jet engine lubrication oil (the signals at m/z 85 are more intense and include more contributions from alkanes than those at m/z 113).The mass chromatograms for the other size ranges are presented in the SI.The mass chromatograms for unused Mobil Jet Oil II can be characterized by a series of distinct peaks at the retention time (RT) of 21-29 min.It turned out that the peak shapes depended on the mass loadings of the collected particles injected into the TD-GC-MS.The peak widths appeared broadened when the mass loadings were too high, which was probably due to saturation of the signals.Because quantification of the collected mass was not performed and visual inspection of the filter color was qualitative, it was rather difficult to predict the amount of particle mass in the sample before each analysis.Small differences in the shapes of the individual peaks might contain some information on the organic compositions but are not discussed here.
We set the diagnostics for similarity in the mass chromatograms between the laboratory-generated particles and unused Mobil Jet Oil II as (1) presence of distinct peaks at RT 21-29 min; (2) no significant biases in the relative intensities of the distinct peaks at smaller and larger RT values within the RT range of 21-29 min; (3) absence of significant peaks at either RT < 21 min or RT > 29 min; and (4) absence of "humps" (baseline elevations) around the distinct peaks.The lack of point ( 2) is indicative of distillation by incomplete vaporization of the oil droplets.The lack of point (3) or ( 4) is indicative of some changes in the organic compositions by thermal decomposition or combustion.
The mass chromatograms for the particles in the third size range (Շ 0.2 lm) at T PG ¼ 19 C and 250 C were similar to that for unused Mobil Jet Oil II, characterized by a series of distinct peaks at RT 21-29 min.The particle volume size distributions (Figure 3) and the mass chromatograms (Figure 6) suggest that the particles in the third size range generated at T PG ¼ 250 C originated mainly from the evaporation and nucleation processes and were dominated by nearly intact forms of Mobil Jet Oil II.The size dependency of organic compositions in the nucleated particles (10-400 nm) needs to be carefully considered in comparing the particle volatility at the diameter of 30 nm and the organic compositions in the third size range.The particle volume size distributions measured by the SMPS and APS shown in Figures 3 and  4 suggest that the particle mass loadings in the second (0.2-2.0 lm) and third (Շ 0.2 lm) size ranges were both dominated by the nucleated particles rather than the original oil droplets for T PG 250 C. The size dependency of organic compositions in the nucleated particles has not been precisely measured, which may limit the interpretation of the data.However, it is unlikely that there was a strong size dependency at least for T PG ¼ 250 C because the mass chromatograms of the second size range were similar to those of the third size range for T PG ¼ 250 C (see Section S2 in the SI).
By contrast, the mass chromatograms for T PG ¼ 400 C and 600 C were significantly different from that for unused oil.We found the presence of humps superimposed on the marker peaks for 400 C and 600 C and significant peaks at RT < 21 min, suggesting the formation of various organic compounds by thermal decomposition.The thermal decomposition could have formed relatively volatile compounds from the oil droplets, and a portion of the peaks at RT < 21 min might have originated from these compounds.
Overall, the chemical characteristics of the particles in the third size range for T PG ¼ 250 C, 400 C, and 600 C were qualitatively consistent with the particle volatility results presented in Section 3.2, indicating significant alteration of the organic compositions and possible formation of nonvolatile particles at T PG ¼ 400 C and 600 C.

Discussion
The volatility and organic compositions of the UFPs generated from Mobil Jet Oil II commonly exhibited a significant transition at T PG of 300-400 C (Figures 5 and 6).Considering the temperature difference between the heated air and the quartz tube (50 C), this transition temperature seems to correspond to the range of the flash and autoignition points of the oil (270 C and 404 C, respectively), where the TG-DTA of the bulk oil sample showed a large weight loss associated with exothermal reactions (Figure 2).It is likely that thermal decomposition or combustion of organic compounds in the oil droplets occurred at T PG ¼ 400 C and 600 C. Furthermore, it seems that some portions of organic compounds in the UFPs were converted to nonvolatile particles in the quartz tube at T PG ¼ 400 C and 600 C.These results suggest that the temperature conditions to which jet engine lubrication oil droplets are exposed in real aircraft exhaust gases should not be too high to generate UFPs having organic compositions similar to those of bulk oil.The lower T 50 values at T PG ¼ 600 C than those at T PG ¼ 400 C might be due to the possible formation of lighter-molecular-weight compounds by further thermal decomposition at higher T PG .
The results of the preliminary experiments, in which the supersaturation of oil vapors was made by diffusion cooling, might also provide some insights into the particle generation processes (see the SI).The volatility and organic compositions of the UFPs generated at T PG ¼ 250 C for the preliminary experiments were more or less similar to those at T PG ¼ 250 C for the current experiments and also those observed at NRT.By contrast, the volatility and organic compositions of the UFPs generated at T PG տ 400 C (450 C and 800 C) for the preliminary experiments were largely different from those at T PG տ 400 C (400 C and 600 C) for the current experiments and those observed at NRT.We speculate that the chemical properties of the UFPs could significantly depend on the cooling processes when the generation temperature exceeded the range of the flash and autoignition points of the jet engine lubrication oil.
As mentioned earlier, it is highly challenging to simulate the evaporation and cooling processes of jet engine lubrication oil droplets in real jet engines.There remains a large gap between our experimental setup and real-world conditions, and the detailed mechanisms of the formation of volatile UFPs in aircraft emissions are still uncertain.Nevertheless, we have shown that the evaporation of jet engine lubrication oil droplets at 250 C and subsequent nucleation and condensation by rapid cooling formed UFPs having volatility similar to that of UFPs sampled at NRT (Takegawa et al. 2023) and composed of nearly intact forms of jet engine lubrication oil.These results suggest that it is possible to generate UFPs that can consistently explain the volatility and organic compositions of the UFPs observed in the real atmosphere by evaporation and cooling of jet engine lubrication oil droplets.

Conclusions
We investigated the volatility and organic composition of UFPs generated from unused jet engine lubrication oil (Mobil Jet Oil II) in the laboratory.Our experimental results showed that the evaporation of jet engine lubrication oil droplets at 250 C and subsequent nucleation and condensation by rapid cooling formed UFPs having volatility similar to that of UFPs sampled at NRT (Takegawa et al. 2023) and composed of nearly intact forms of jet engine lubrication oil.The volatility and organic compositions of the UFPs generated at the particle generation temperatures of 400 C and 600 C exhibited a remarkable difference from those observed at NRT.A significant transition of the volatility and organic compounds occurred at the particle generation temperature of 300-400 C. The transition temperature likely corresponded to the flash and autoignition points of the bulk Mobil Jet Oil II (270 C and 404 C, respectively).
To summarize, our experimental results suggest that it is possible to generate UFPs that can consistently explain the volatility and organic compositions of UFPs observed in the real atmosphere (Takegawa et al. 2023;Fushimi et al. 2019) by evaporation and cooling of jet engine lubrication oil droplets.Further investigations are needed to reveal the actual temperature conditions and residence time to which jet engine lubrication oil droplets are exposed in real jet engines.

Figure 1 .
Figure 1.Schematic diagram of the experimental apparatus for testing the volatility and compositions of particles generated from jet engine lubrication oil: (a) particle generation section; (b) particle number size distribution measurements; (c) particle volatility measurements; (d) particle sampling for chemical composition measurements.The numbers beside the arrows represent the flow rates (L min −1 ).

Figure 3 .
Figure 3. (a) Particle number (solid) and volume (dashed) size distributions measured using the SMPS for the particle generation temperature (T PG ) of 19 C (room temperature).Each size distribution is the average of 3 sets of a 2-min scan obtained every 3 min.(b) Same as (a) but for T PG ¼ 200 C. (c) Same as (a) but for T PG ¼ 250 C. (d) Same as (a) but for T PG ¼ 300 C. (e) Same as (a) but for T PG ¼ 400 C. (f) Same as (a) but for T PG ¼ 600 C.

Figure 4 .
Figure 4. (a) Particle number size distributions measured using the SMPS and the APS for T PG ¼ room temperature (20 C) and 250 C. (b) Same as (a) but for particle volume.
due to Brownian diffusion and thermophoresis may yield additional uncertainties in the f 0 values of about þ0.02 and those in the T 50 values of about þ2 C (see the SI).Considering the factors affecting the uncertainties described above, we estimated the potential uncertainties in the PNFR values calculated from the fitting results by forcing the f 0 values to unity and shifting the T 50 values by þ2 C. Takegawa et al. (2023) estimated the PNFR values for the volatile particles from median-based calculations (median-based volatile PNFR) classified by the geometric mean diameter (GMD) of the total particles and the detectable size ranges of the instruments.The temperature dependency of the median-based volatile PNFR did not vary significantly with altering the diameter range of the data classification (smaller or larger GMD values; inclusive or exclusive of sub-10 nm particles).The range of the median-based volatile PNFR was 0.49-0.54for T ET ¼ 150 C and 0.01-0.11for T ET ¼ 250 C with the estimated uncertainties ranging from 18 to 44%.

Figure 5 .
Figure 5. (a) Particle number fraction remaining (PNFR) and the fitting curve as a function of T ET at T PG ¼ 250 C for the UCPC (F UCPC (250, 30, T ET ); solid circles and line, respectively) and the CPC (F CPC (250, 30, T ET ); open circles and dashed line, respectively).The fitting parameters (f 0 , f 1 , T 50 , a) are given in Table 1.(b) Same as (a) but for T PG ¼ 300 C. (c) Same as (a) but for T PG ¼ 400 C. (d) Same as (a) but for T PG ¼ 600 C.

Figure 6 .
Figure 6.Mass chromatogram at m/z 113 (a marker of fatty acid esters) obtained from TD-GC-MS analysis for (a) the unused Mobil Jet Oil II samples and for the particle samples in the third size range (Շ 0.2 lm) generated at T PG ¼ (b) room temperature (19 C), (c) 250 C, (d) 400 C, and (e) 600 C.
The F X (T PG , D p , T ET ) values at T ET ¼ 150 C and 250 C calculated from the fitting results.b Forced to zero (the experimental F X (T PG , D p , T ET ) values at T ET ¼ 350 C were < 0.01).F UCPC (T PG , 30, 150) and F CPC (T PG , 30, 150) values calculated from the fitting results fall within the range of the median-based volatile PNFR values for T ET ¼ 150 C observed at NRT (0.49-0.54); and (2) the F UCPC (T PG , 30, 250) and F CPC (T PG , 30, 250) values calculated from the fitting results are smaller than 0.1.The calculated F UCPC a