Size distributions (d > 15 nm) and volatile properties of combustion particles were measured during test-rig experiments on a jet engine, consisting of a combustor and three simulated turbine stages (HES). The combustor was operated to simulate legacy (inlet temperature 300°C) and contemporary (500°C) cruise conditions, using kerosene with three different fuel sulfur contents (FSC; 50, 400 and 1300 μg g−1). Measurements found that contemporary cruise conditions resulted in lower number emission indices (EIN15) and higher geometric mean particle diameter (dG) than for legacy conditions. Increasing FSC resulted in an overall increase in EIN15 and decrease in dG. The HES stages or fuel additive (APA101) had little influence on EIN15 or dG, however, this is uncertain due to the measurement variability. EIN15 for non-volatile particles was largely independent of all examined conditions.
 Ground and in-flight studies have observed that aircraft exhaust emissions contain both primary emitted soot particles as well as volatile particles, formed by nucleation [Anderson et al., 1998; Petzold et al., 1999; Schröder et al., 2000]. High concentrations of nucleation mode (d < 10 nm) volatile particles are believed to be composed of H2SO4 and unburned/partially oxidized non-methane hydrocarbons (NMHCs). Despite progress on microphysical models [Vancassel et al., 2004], the effect of jet engine operating conditions on aerosol properties still needs to be further characterized.
 The PartEmis project (Measurement and Prediction of Emissions of Aerosols and Gaseous Precursors from Gas Turbine Engines) was initiated to extend current knowledge of jet engine emissions. An overview of all measured parameters and test-rig technical aspects is given in Wilson et al. [2004, and references therein]. The present study deals with measurements of particle size spectra and their volatile properties.
 Two test-rig campaigns were conducted at the QinetiQ facilities in Farnborough, UK. The first campaign (Jan.–Feb. 2001) focused on characterizing emissions from a jet engine combustor (previously used during AEROTRACE [see Wilson et al., 2004]) and the second (March 2002) focused on the same combustor, integrated to a so-called Hot End Simulator (HES) unit. The HES passively simulated thermodynamic conditions at three pressure stages within a turbine (high, intermediate, low; hp, ip, lp; ∼3.7, 2.5, 1.1 bar). Combustor operating conditions were chosen to simulate legacy (referred to as Old) and contemporary (Modern) aircraft cruise conditions at 33,000 ft (∼11000 m) altitude for which inlet temperatures were 300 and 500°C, respectively. These conditions conformed to the ICAO (International Civil Aviation Organization) engine emissions smoke standard. Jet fuel kerosene with fuel sulfur contents 50, 400 and 1300 μg g−1 (low, mid and high FSC) was used where 400 μg g−1 FSC approximates the contemporary fleet average. A Shell additive (APA101) was used in the kerosene in a second round of tests at concentrations of 256 and 1280 μg g−1. APA101 has recently been shown to result in lower particle concentrations amongst other beneficial properties [e.g., Corporan et al., 2002]. Emissions were sampled using a 5-hole equal area water-cooled gas sampling probe to map the lateral distribution at 11 positions across the combustor exit plane. However, average combustor properties are only presented due to a lack of data. After exit from the combustor, exhaust emissions were cooled to ∼150°C using a water jacket, diluted by a factor of ∼65, and then allowed to cool naturally to 20–25°C in the sample line (length ∼38 m, tube diameter ∼21 mm. Dilution factors were determined from CO2 mixing ratios, and temperature and pressure sensors which were distributed throughout the system. Aerosol losses (C. Hurley, internal report) were estimated to total 20% but were not modeled as a function of particle diameter (d). In the absence of size-resolved correction factors, data have not been corrected here. Particle size spectra were measured using two standard TSI 3071 DMAs (differential mobility analyzer) and a TSI CPC 3022 (condensation particle counter). The instrument was operated in two modes: i) SMPS (scanning mobility particle sizer) mode for polydisperse particles, or ii) TDMA (tandem DMA) mode to select and measure monodisperse particles. The addition of a custom-built thermodesorber [Burtscher et al., 2001] allowed particle volatile properties to be measured at temperatures T = 25, 120 and 300°C. The ranges, i) T = 25–120 (T25–120), ii) 120–300 (T120–300), and iii) >300°C (T>300) are commonly attributed to the following aerosol components, respectively: i) H2SO4, ii) (bi)sulfate, and iii) elemental carbon (EC) or soot, mineral dust and sea salt. Category iii) is commonly referred to as refractory or (NV) non-volatile aerosol, and is composed almost entirely of EC in our case. It should be noted that NMHCs may result in a positive artifact at all temperatures, and is discussed later. Size-dependent corrections for thermodesorber losses were applied. The effect of particle-bound water on particle size determination was considered negligible due to the low sample line RH (<10%). Polydisperse size spectra were measured with a 3–5 minute resolution per thermodesorber channel to give the number concentration N15 over the range d = 15–550 nm. Monodisperse particles were measured at 11 nominal diameters do = 15–100 nm. The ratio d3/do3 therefore gives the volume shrinkage factor where 1 − d3/do3 equals the volume fraction of volatile material assuming particle sphericity. Emitted particle concentrations (number integral of SMPS spectra) are reported as number emission indices per kg burned fuel (EIN15). As the measurement variability was larger than the uncertainty, standard deviations for the former are given.
3. Combustor Measurements
 An overview of EIN15 and geometric mean diameter dG at the combustor exit as a function of FSC and cruise conditions is given in Figures 1 and 2, respectively. Combustor-HES results are also shown and are discussed in Section 4. The largest difference lies in the grand average of EIN15 for Old versus Modern conditions which gave 7.0 ± 1.4 and 4.0 ± 1.4 × 1014 kg−1, respectively. Corresponding values of dG were 41.0 ± 1.6 and 43.4 ± 1.8 nm. FSC resulted in higher EIN15 for the high FSC case, while a decrease in dG was observed on average with increasing FSC. The differences between high and mid or low FSC were significant at the 95% level (Students T-test) in most cases. The effect of APA101 additive on EIN15 and dG on the other hand was uncertain when considering the measurement variability.
 An interesting observation during both campaigns was that only monomodal size spectra were observed. The geometric standard deviation during the combustor campaign was σG = 1.67 ± 0.05. However, a nucleation mode at dG = 4–7 nm for high FSC cases was observed for d < 15 nm in other simultaneous measurements [Petzold et al., 2003]. A distinct minimum in particle concentration at d ∼ 10 nm occurred between the nucleation mode and the lower tail of the Aitken mode (d = 10–100 nm). The above results are similar to previous in-flight studies where EI in the range 0.1–1 × 1015 kg−1 and dG = 30–60 nm were found [Anderson et al., 1998; Hagen et al., 1998; Petzold et al., 1999; Schröder et al., 2000]. The decrease in EIN15 and increase in dG with increase in combustor inlet temperature (i.e., Old to Modern conditions here) is consistent with other studies [e.g., Petzold et al., 2003]. This reflects the fleet-average trend of civilian aircraft towards a reduction in particulate emissions.
 Measurements of particle volatility on monodisperse particles gave similar results for both cruise conditions. Figures 3 and 4 illustrate d3/do3 (Modern conditions) at both thermodesorber temperatures without and with APA101 additive, respectively. Particles tend to exhibit greater shrinkage (i.e., increasing volatility) with decreasing do except for the low FSC case. In addition, mid and high FSC curves also appear to illustrate greater shrinkage. Although each data point consists of 3–6 size scans, the measurement variability does not allow any clear conclusions to be drawn. Hence, polynomial fits are illustrated as solid curves in both figures (see captions for details). In Figure 3, small particles (do < 20 nm) exhibited volatile volume fractions ∼10 and ∼5% for the T25–120 and T130–300 fractions, respectively, which decreased to <3% for do ≥ 50 nm particles. Particle volatility for do < 30 nm is therefore mainly due to the T25–120 fraction, and is most probably due to H2SO4/NMHCs. Above this size, volatility appears to be independent of FSC and cruise conditions. The above observations are supported in part by NMHC desorption measurements on particulate samples. Only 0.4 wt% of NMHCs were found to desorb at T < 130°C, whereas 99.6 wt% desorbed at T = 130 to 350°C (H. Puxbaum, personal communication, 2004). These results are not size-resolved but do suggest that the T25–120 fraction is mainly H2SO4 and T130–300 is composed of NMHCs and not just (bi)sulfate. In Figure 4, the absence of particle shrinkage for do < 30 nm is obvious when compared to the mid and high FSC results in Figure 3, again suggesting that H2SO4 is responsible for the T25–120 fraction. The influence of APA101 appears to be negligible, if any. Only low FSC was chosen for tests with APA101 as it represents future FSC trends in fuel kerosene.
4. Combustor-HES Measurements
 Particles were sampled at all three pressure stages of the HES. An inter-comparison of HES stages suggests little effect on EIN15 for constant FSC or APA 101 additive. A similar conclusion is reached for dG in Figure 2. In order to increase the statistical significance, results averaged over all HES stages are also shown in Figures 1 and 2. A clear difference is evident between HES Old and Modern conditions which gave EIN15 = 4.3 ± 0.7 and 2.5 ± 0.2 × 1014 kg−1, and dG = 40.4 ± 0.8 and 45.9 ± 0.9 nm.
 When comparing Combustor with Combustor-HES measurements, only one conclusive trend is apparent. Lower EIN15 concentrations are observed in the HES for both Old and Modern conditions, while dG is slightly higher only under Modern conditions. Cloud condensation nuclei concentrations during the Combustor-HES campaign were also significantly lower than during Combustor measurements (unpublished data). Particle coagulation between the combustor exit and HES entrance was examined as a possible reason for lower EIN15. The geometric standard deviation for Combustor-HES measurements was 1.78 ± 0.03, and compares with 1.67 ± 0.05 for the Combustor campaign. However, if σG values from the 11 probe positions at the combustor exit are weighted with the appropriate gas density then a value 1.74 is found. Hence, the observed broadening between combustor exit and HES exit can be explained by mixing effects rather than coagulation. In conclusion, the most likely explanation for the difference in EIN15 is differences in nominal combustor settings resulted from Combustor and Combustor-HES campaigns not being back-to-back.
 The volatile properties of monodisperse particles during Combustor-HES measurements (not shown) were largely similar to those for the Combustor campaign. However, due to the increased measurement variability (due to fewer size scans) no clear conclusions can be drawn concerning the effect of FSC or APA101.
5. Discussion and Conclusions
 An important aspect of PartEmis was not only to characterize the physical but also chemical properties of particles. Our measurements are only able to suggest but not conclusively determine the composition of the T25–120 and T130–300 fractions. Further evidence comes from PartEmis measurements of nucleation mode EIN [Petzold et al., 2003] which not only suggest the presence of H2SO4, but also demonstrate a clear effect with high FSC. More direct measurements with size-selected impactor samples and ion chromatography analysis [Puxbaum et al., 2003], and a chemical ionization mass spectrometer [Katragkou et al., 2004] conclusively confirm sulfur species to be present. In-flight studies also indicate H2SO4 particles as being a component of exhaust emissions. Nucleation mode EIN have been found to increase with FSC, and is attributed to H2SO4 droplets [e.g., Petzold et al., 1999; Schröder et al., 2000]. An inter-mode minimum at d ∼ 10 nm separates a mostly volatile nucleation mode from a non-volatile Aitken mode in young exhaust plumes. Such volatile particles nucleate in cooling exhaust gases up to 1–2 s after emission. In our case, nucleation primarily occurred after dilution in the main sampling line [Wilson et al., 2004]. However, volatile particle EIN are apparently not linearly related to FSC. For instance, Schröder et al.  reported little effect when FSC was decreased below 100 μg g−1. Modeling studies suggest that NMHCs may substantially contribute to the T25–120 fraction [e.g., Kärcher et al., 2000]. The increase in monodisperse aerosol volatility for d < 30 nm (Figure 3) may therefore be due to: 1) a thicker coating of volatile material, and/or 2) external H2SO4/NMHC particles belonging to the upper tail of the nucleation mode. The former is in agreement with hygroscopicity TDMA results which also suggest an increasingly thick coating with decreasing dG and increasing FSC [Gysel et al., 2003].
 Previous studies on APA101 indicate a decrease in d, and EIN by 20–35%, although the mechanism of action, whether chemical, physical or other, is at present poorly understood [see Corporan et al., 2002, and references therein]. It is unclear why similar effects were not observed here. APA101 appears to have a greater influence on fuels with high aromatic and sulfur content (P. Bogers, personal communication, 2004). APA101 was only tested here with low FSC having a low aromatic content (18.1% v/v), and may partly explain the lack of a reduction in EIN15 and dG. In addition, our back-to-back measurements with and without APA101 additive may not be comparable with other studies where tests were first conducted 50–100 hours after the addition of APA101. The maintenance of clean engine components is believed to result in optimum fuel combustion conditions (i.e., fuel spray distribution etc) and hence in reduced emissions [Corporan et al., 2002]. The lack of an observable effect therefore suggests the above mechanism to be more probable than either chemical or physical mechanisms, which may be expected to cause an immediate reduction in EIN.
 Several in-flight studies have observed that non-volatile EIN are relatively unaffected by FSC level [Hagen et al., 1998; Petzold et al., 1999]. PartEmis measurements of non-volatile emission indices (EIN15,NV) reflected the trends for EIN15 in Figures 1 and 2. For instance, EIN15,NV for Old and Modern conditions during the Combustor campaign were ∼4.8 ± 1.5 and 2.4 ± 1.3 × 1014 kg−1 (average over all FSC), respectively, and compare with EIN15 of 7.0 ± 1.4 and 4.0 ± 1.4 × 1014 kg−1. High FSC resulted in a higher EIN15 only for Old conditions (6.9 ± 0.1 × 1014 kg−1. EIN15,NV may thus be insensitive to FSC level and APA101 additive, although the measurement variability should be taken into account.
 In summary, this study investigated the size spectra and volatile properties of particle emissions from a jet engine combustor and turbine simulation stage. As the dataset is multi-dimensional (polydisperse and monodisperse aerosol properties with respect to aerosol volatility, combustor conditions, FSC, and APA101 additive) definite trends were somewhat difficult to ascertain. Nevertheless, certain trends were discernible: 1) Modern versus Old conditions resulted in lower EIN15 and higher dG; 2) high FSC versus mid or low FSC led to increased EIN15 and a decrease in dG; 3) EIN15 and dG were unaffected by APA101 additive, however, this cannot be stated with any certainty due to the measurement variability; 4) monodisperse particles (do = 15–30 nm) exhibited increasing volatility with decreasing do and also with increasing FSC.
 PartEmis was kindly funded by the European Commission (G4RD-CT-2000-00207) and the Swiss Bundesamt für Bildung und Wissenschaft (99.0632). We also thank the QinetiQ team, and two anonymous Reviewers for their useful comments.