The measurements were performed at the engine test cell of SR Technics at Zurich Airport (Switzerland) using an in-service commercial turbofan CFM56-7B burning four different blends of Jet A-1 and HEFA fuel (HEFA vol. percentage of 0 %, 5 %, 10 % and 32 %). A schematic of the experimental set-up is shown in Fig. 1. The exhaust was sampled at the engine exit plane using a single-point sampling probe and then split into three sampling lines: the PM line for measurements of the particulate fraction, the GenTox line for the sampling of genotoxic compounds and the Annex 16 line for the measurements of the gaseous emissions and smoke number. The PM line was diluted with dry synthetic air (dilution factor ∼10) to prevent water condensation and coagulation of the particles in the sampling line. The non-volatile PM (nvPM) measurement system is compliant with the new ICAO standard (ICAO, 2017). The instruments relevant for this work are shown in blue in Fig. 1. The chemical composition was determined from the analysis of filter samples with a Sunset OC-EC Aerosol Analyzer (Sunset Laboratory Thermal/Optical Carbon Analyzer, Model 4L). The optical properties were monitored online with a Photo-Acoustic Extinctiometer (PAX; Droplet Measurement Technologies, λ=870 nm) and a Cavity Attenuated Phase Shift PM single scattering albedo monitor (CAPS PMSSA; Aerodyne, λ=532 nm). The particle size distribution was measured using a Scanning Mobility Particle Sizer (SMPS; TSI, Model 3938). The BC mass concentration was measured using a Micro-Soot Sensor (MSS; AVL, Model 483). The NO2 measurement (Eco Physics, CLD844 S hr) was used to perform an online correction of the interference in the optical measurements at 532 nm. The CO2 analysers (Thermo Fisher Scientific Model 410i in the PM line and Horiba PG-250 in the Annex 16 line) were used to calculate the dilution factors. Additional details of the measurement set-up are provided in the Supplement (Sect. S1.1).

Filter samples for EC/OC analysis were collected in the PM sampling line with a dual step stainless steel filter holder (URG, Series 2000-30FVT). Quartz fiber filters (Pall Tissuequartz, 2500QAT-UP) were used in both stages to collect the PM mass (main filter) and to determine the positive sampling artifact (back-up filter) from gaseous OC adsorbing onto the filter surface (Kirchstetter et al., 2001; Subramanian et al., 2004). Prior to sampling, the filters were baked for at least 6 h at 650 ∘C to remove possible contamination of adsorbed carbon. The sampling flow was 5 L min-1, and the duration was adjusted to provide optimal mass surface loadings for the EC/OC analyses (around 7 µg cm-2). Stainless steel masks were deployed to reduce the sampling area of the filters and, as a result, increase the mass loading per sample area when needed. Overall, 16 sets of filters were collected to cover the full range of thrust levels for the Jet A-1 fuel and the 32 % vol HEFA blend. The larger filter masks (24 mm inner diameter) were used to sample at high thrust levels (100 %–65 %), while the smaller masks (16 mm inner diameter) were required for the low thrust levels (50 %–7 %).

The thermo-optical analysis for the quantification of EC and OC was performed using a Sunset OC-EC Aerosol Analyzer (Sunset Laboratory Thermal/Optical Carbon Analyzer, Model 4L). A detailed description of the method is reported in the Supplement Sect. S1.2. For the analysis, a modified NIOSH 5040 thermal protocol, summarized in Table S1 (Birch and Cary, 1996), with a transmittance optical correction for pyrolysis was used.

The DMT PAX monitor simultaneously measures the aerosol optical absorption and scattering using a modulated diode laser (λ=870 nm). The light absorption is determined using the photo-acoustic technique. The modulated laser beam heats up the absorbing particles, which quickly transfer the heat to the surrounding air, generating a pressure wave that is measured with a sensitive microphone. The light scattering of the bulk aerosol is measured with a wide-angle integrating reciprocal nephelometer. Since there is relatively little absorption from gases and non-BC aerosol species at the 870 nm wavelength, the absorption measurement corresponds to the BC mass. Therefore, using an appropriate mass absorption cross section (MAC), the PAX absorption measurement can be used to determine BC mass (BCPAX). Vice versa, comparing the absorption measurement with the EC mass from filter measurements, we can infer the MAC870 for aircraft engine exhaust.

The CAPS PMSSA monitor provides simultaneous measurement of aerosol light extinction and scattering (Onasch et al., 2015). The extinction measurement is based on the cavity attenuated phase shift technique, which evaluates the phase shift of a LED light (532 nm) in a very long optical path (up to 2 km), created with very high reflectivity mirrors in the sampling cell (30 cm). In addition, the CAPS PMSSA includes an integrating sphere (integrated nephelometer) within the optical path for the measurement of particle scattering. A particle size-dependent truncation correction is required to take into account the light lost at extreme forward and backward scattering angles due to the apertures of the optical beam.

Laboratory calibrations of both instruments were performed prior to the measurement campaign using size-selected ammonium sulfate and nigrosine PM (see Sect. S1.3 in the Supplement). In addition to the standard calibrations, corrections for the CAPS scattering signal outside the instrument linear range and for the interference from gaseous NO2 at the measurement wavelength were also developed (Figs. S2 and S3). While the measured NO2 interference in the CAPS extinction is fairly consistent with previously reported values, we also found an unexpected non-linear interference in the CAPS scattering signal. This was initially attributed to a possible light leak in the instrument, but the scattering interference persisted after the light sealing of the instrument was renewed. To further investigate this issue, we compared laboratory calibration data of both optical instruments with results from Mie theory (Figs. S4–S6). Although there are several assumptions within the Mie model that may not be totally satisfied by the laboratory-generated calibration particles (e.g. spherical and homogeneous particles), the agreement with the PAX absorption and scattering measurements is fairly good for both ammonium sulfate and nigrosine. In contrast, the CAPS measurements only agree well with Mie theory for purely scattering particles, i.e. ammonium sulfate. In the case of nigrosine, the CAPS measurements agree with Mie theory in terms of total extinction, but the measured scattering is around 43 % higher than estimated from the model. Despite several experimental efforts, we could not find the origin of the discrepancies in the CAPS scattering measurement or a way to properly correct it. Instead, we derived the CAPS scattering coefficient from the PAX absorption measurement using a thrust-dependent absorption Ångström exponent (AAE) obtained from aircraft engine measurements with a seven-wavelength aethalometer. All the details of this calculation can be found in Sect. S1.5 in the Supplement.

The main differences between the conventional Jet A-1 fuel and the different HEFA blends used in this work are reported in Table 1. Most fuel properties were measured following the standard ASTM (American Society for Testing and Materials) methods, including the concentration of total aromatics, naphthalene and sulfur, the smoke point and the fuel density. In addition, the hydrogen mass concentration was determined by nuclear magnetic resonance, using a method equivalent to ASTM D7171. As expected, increasing the concentration of HEFA fuel in the blend corresponded to a reduction in the concentrations of the aromatic compounds (including naphthalene) and sulfur. In addition, with the addition of the HEFA fuel the hydrogen mass concentration and the smoke point increased, while the fuel density slightly decreased.

The SSA, MAC and MSC were calculated for the two measurement wavelengths (λ=532 and 870 nm) using the following relationships: 1SSAλ=bscat,λbext,λ2MACλ=babs,λEC3MSCλ=SSAλ⋅MACλ1-SSAλ. In addition, to estimate the instantaneous direct radiative effects of the PM from fresh aircraft engine exhaust, we evaluated the radiative transfer equation introduced by Chylek and Wong (1995), modified as in Bond et al. (2007), to express the wavelength-dependent SFE in terms of the MSC and the MAC: SFEλ=-S0λ4Tatm2λ,z1-Fc421-asλ2βMSCλ-4asλMACλ, where S0 is the top of the atmosphere solar irradiance, Tatm is the atmospheric transmission, Fc is the cloud fraction, as is the surface albedo, β is the backscatter fraction, λ is the wavelength and z is the height over sea level. The MAC and MSC determined for the two measurement wavelengths were fitted over the entire range of the solar radiation spectrum (280–4000 nm) using the power law relationships: 5MACλ=aλ-AAE6MSCλ=bλ-SAE, where a and b are fitting parameters, and AAE and SAE represent the absorption and scattering Ångström exponents, respectively. The selection of the parameters in Eqs. (4) to (6) is discussed in Sect. S2.4.

An overview of the main findings from the EC/OC analysis is presented in Figs. 2 and 3. The OC concentrations were corrected to take into account the positive sampling artifact as described in the Supplement (Sect. S2.1). The EC, OC and total carbon (TC) mass concentrations are reported in Table S3. Additionally, in Table S4 we also report the mass emission index of EC (EIm,EC, in mg kgfuel-1), together with the additional parameters required for the calculation of EIs (i.e. carbon dioxide, carbon monoxide and hydrocarbon concentrations) and the particles' size parameters (geometric mean diameter, GMD, and geometric standard deviation, GSD). Representative thermograms illustrating the EC/OC split for samples at low, medium and high thrust levels are reported in Fig. S9.

Figure 2 displays the mass concentrations of EC, OC and TC as a function of engine thrust and the concentration changes associated with the use of the 32 % HEFA blend in comparison to the base Jet A-1 fuel. For the Jet A-1 fuel, the concentrations of all three carbonaceous components increased with engine thrust, from a minimum of 0.1 mgTC m-3 at taxi (∼6 % thrust) to a maximum of 5.6 mgTC m-3 at take-off (∼95 % thrust). As the mass concentration increased with thrust, the GMD increased from 8 nm at taxi to 40 nm at take-off (Fig. S10). The slight increase in the mass concentrations between taxi and ground idle (∼3 % thrust) has been observed in previous works (Durdina et al., 2017) and is associated with a decrease in the combustion efficiency and the air / fuel ratio. As illustrated by the bar plots, the use of the 32 % HEFA blend induced a clear reduction in the EC concentrations at all thrust levels, in line with previous findings (Moore et al., 2015, 2017; Schripp et al., 2018). The HEFA effect was strongest at low thrust levels, inducing a decrease in EC mass of 50 %–60 % for thrust levels up to 30 %. An explanation for this thrust dependence can be found in Brem et al. (2015). However, very large uncertainties were associated with the EC measurements at ground idle due to low filter loading (down to 0.2 µg cm-2 despite the long sampling times and the use of the small filter mask). For thrust levels of 50 % and above, the decrease in EC mass with the HEFA blend became less significant, reaching a minimum EC decrease of 14 % at take-off. A similar trend was observed for the OC and TC concentrations at high thrust levels. In contrast, the OC, and consequently also the TC, seemed to be enhanced by the HEFA blend at ground idle.

Figure 3 shows the correlations of EC and OC with the BC mass concentrations measured with the MSS (BCMSS, reported as nvPM mass in regulatory measurements of aircraft engine emissions). EC and BCMSS were in excellent agreement for both fuel types, with a slope very close to unity (0.96±0.02) and Pearson's coefficient (R2) of 0.99. Although OC also increased with BCMSS, the correlation was weaker in this case, and small differences could be observed between the two fuels.

The main results from the measurement of the optical properties are reported in Figs. 4 and 5. For clarity, only the measurements with Jet A-1 fuel and the 32 % HEFA blend are shown, while the results from the intermediate HEFA blends (5 % and 10 %) are included in Fig. S11.

Figure 4 presents the thrust dependencies of the absorption, scattering and extinction coefficients, measured at 532 nm (CAPS, green squares) and 870 nm (PAX, blue circles). At both wavelengths, babs, bscat and bext showed similar thrust dependencies to those observed for EC, characterized by a large and continuous increase for thrust levels above 40 %. In addition, the use of the 32 % HEFA blend induced a decrease in the three optical coefficients at most thrust levels.

Figure 5 shows the correlation between babs at the two measurement wavelengths and the EC mass concentration determined from the thermo-optical measurements. From the linear regressions we derived the MAC values for aircraft exhaust at 532 and 870 nm, which appear to be independent of the particle size distribution, thrust or fuel type and will be further discussed in Sect. 3.4. The MAC870 was used to calculate BC mass from the PAX absorption measurement (BCPAX), which was strongly correlated with BCMSS (Fig. S12, R2=0.99, slope 0.98) and showed thrust-dependent reductions with the HEFA blends (Fig. S13), consistent with the reductions in EC mass.

To investigate the link between the chemical composition and the optical properties of the aircraft emissions, in Fig. 6 we compare the thrust dependencies of the OC/TC fraction and the SSA, calculated at the two measurement wavelengths using Eq. (1). The OC/TC fraction showed a large variability with the thrust level, with OC/TC between 0.75 and 0.90 at low thrust levels (3 %–30 %), decreasing to 0.25 at 50 % thrust and down to 0.17 at take-off. Similar trends in the OC/TC ratio with engine thrust have been observed in previous works (Delhaye et al., 2017). Although the use of the HEFA fuel had a strong effect on the EC and OC mass concentrations, it did not have any visible effect on the OC/TC fraction. The high OC content of the particles at low thrust levels resulted in very high SSA, which showed a maximum at ground idle (SSA532,base=0.88 and SSA870,base=0.55). Such high SSA values are common for particle emissions from biomass burning at low combustion efficiency (e.g. SSA532∼0.95 in wildfire emissions; Liu et al., 2014). The SSA decreased sharply between 30 % and 60 % thrust, likely due to the decreasing OC fraction, reaching a minimum at the combustor inlet temperatures and air / fuel ratios representative of cruise thrust (∼60 % thrust), where SSA532,base=0.29 and SSA870,base=0.07. These low SSA values are characteristic of primary on-road vehicle particle emissions (e.g. 0.22< SSA675<0.36 from tunnel measurements; Strawa et al., 2010). Above 60 % thrust we observed a slight increase in the SSA, which might be related to the increasing mean particle size (as the OC content remained constant in this thrust range). While there was no visible effect of the HEFA blend at the high thrust levels, it seems that slightly higher SSA were associated with the HEFA blend at low thrust levels.

The similarities observed in the thrust dependencies of the OC/TC and the SSA, as well as the high correlation between BCPAX, BCMSS and EC, indicate that the OC content in these particles strongly enhanced light scattering at both measurement wavelengths but did not have a substantial effect on the light absorption. The increased OC content of the PM with decreasing engine power is in agreement with the observations of Vander Wal et al. (2016) and could be explained by inefficient and incomplete combustion at the lower thrust levels. The thermograms from the EC/OC analysis show a large OC volatility range for all thrust levels, with a major fraction of OC evaporating between 200 and 310 ∘C, especially at low thrust (Fig. S9).

The MAC values at the two measurement wavelengths were determined from the linear fits between babs and EC mass, which yielded MAC532=7.5±0.3 m2 g-1 (R2=0.97) and MAC870=5.2±0.9 m2 g-1 (R2=0.94). As shown in Fig. 5, the MAC values appear to be independent of the thrust level and fuel type. The MAC870 is in good agreement with the filter-based determined MAC value by Petzold and Schröder (1998) for jet engine aerosol (MAC800=6 m2 g-1), which, using the inverse wavelength dependency of the cross section, leads to MAC870,calc=5.5 m2 g-1. Our results are also in line with the MAC value of freshly generated light absorbing carbon proposed by Bond and Bergstrom (2007) (MAC550=7.5±1.2 m2 g-1), which, converted to the wavelengths of interest, results in MAC532,calc=7.8±1.2 m2 g-1 and MAC870,calc=4.7±0.8 m2 g-1.

The MSC values were calculated using its relation with the SSA and MAC reported in Eq. (3). In contrast to the MAC, the SSA, and consequently the MSC, were found to be highly thrust-dependent. Using the SSA measured at 60 % thrust (SSA532=0.37±0.03; SSA870=0.09±0.01) and the MAC values reported above, this calculation yielded MSC532=4.5±0.4 m2 g-1 and MSC870=0.54±0.04 m2 g-1. The MSC532 falls within the higher end of MSCs measured for fresh biomass smoke by Levin et al. (2010). However, a more detailed comparison with literature values is hindered by the strong dependency of MSC on the particles size, morphology and chemical composition.

The wavelength-dependent MAC and MSC were determined by fitting the measurements with Eqs. (5) and (6), as shown in Fig. S14. The AAE in Eq. (5) was set to 1.0±0.2, as inferred for the BC emissions at cruise thrust level (60 %) from the aethalometer measurements described in the Supplement Sect. S1.5. This is a widely accepted value of the AAE, often used in literature for fresh BC particles. Lastly, SAE =4.5±0.7 was calculated for cruise conditions using the scattering coefficients at the two measurement wavelengths, i.e. SAE =ln⁡(bscat,532/bscat,870)/ln⁡(532/870). To put these results in context, the obtained MAC and MSC spectra were used to estimate the direct radiative effect of fresh aircraft exhaust PM emissions during cruise, using the SFE defined in Eq. (4). A detailed description of the SFE model and its results can be found in the Supplement Sect. S2.4; in the following we briefly discuss the main findings. For high surface albedo surfaces like snow, aircraft fresh PM emissions induced a strong warming effect (integrated SFE450-2000,snow=4700 W g-1). Other land surfaces, such as soil and grass, resulted in a more moderate warming (SFE450-2000,soil=1500 W g-1 SFE450-2000,grass=800 W g-1), while the effect of emissions over dark surfaces such as seawater was very low and had an overall cooling effect (SFE450-2000,sea=-90 W g-1). However, these results need to be taken with caution, as this simple radiative model does not consider the effect of underlying clouds. Moreover, we only consider the effect of fresh PM emissions, corresponding to an approximate time after emission of less than 0.6 s, when the jet is still conserved, and high temperatures prevent the condensation of volatile species. Previous studies have shown that sulfuric acid plays an important role in the formation of secondary PM in near-field aircraft plumes (Kärcher et al., 1996). Thus, plume evolution measurements of the particles' optical properties (if possible in-flight) and more complex models are needed to assess the overall radiative effects of aircraft PM emissions.