To capture information on the submicron aerosol composition, the ERICA instrument was deployed on the HALO aircraft during PHILEAS 2023 . Aerosol particles outside the aircraft were sampled using the HALO Aerosol Submicrometer Inlet (HASI), which is a forward-facing aerosol inlet. This inlet system was developed specially for sampling submicron aerosol particles aboard the HALO aircraft and was combined with the HASI Flow Control Unit (FCU). The FCU controls the flows in the inlet system to provide isokinetic sampling. An additional bypass pump was deployed to maintain an isokinetic flow to the ERICA instrument. Further details about the flow system can be found in the Supplement (Sect. S2).

The measurement principle of the ERICA instrument is briefly described in the following. Particles enter the system through a constant-pressure inlet, maintaining a constant pressure within the aerodynamic lens by a varying volume flow into the instrument . Subsequently, the particles are focused to a narrow beam with the help of the aerodynamic lens . After passing the lens, the particles are detected by two light scattering signals (λ=405 nm), allowing the measurement of the size-dependent particle velocity. We used manufactured polystyrene latex particles of different sizes to calibrate and determine the vacuum-aerodynamic diameter (dva) of the atmospheric particles. In the next step, the particles enter the high-vacuum system. Here, with the ERICA-LAMS (Laser Ablation Mass Spectrometer) technique, the particles are ablated and ionized by single-triggered laser shots (frequency-quadrupled Nd:YAG laser with a wavelength of λ=266 nm and an energy between 3.8 to 4.5 mJ). The resulting positive and negative ions are extracted by electrical fields to generate bipolar mass spectra for each individual aerosol particle. Overall, this technique provides information on the single particle chemical composition, the mixing state, and the size of individual particles. This technique further enables the differentiation between refractory and non-refractory components within each aerosol particle. Due to the limitations of the aerodynamic lens and the detection unit, the ERICA-LAMS covers a particle size range from approximately 180 to 3000 nm .

A total of 37 401 single-particle spectra obtained from ERICA-LAMS were analyzed using the CRISP software package Concise Retrieval of Information from Single Particles;. The analysis involved the calibration of the ion mass-to-charge ratio (m/z ratio) for each spectrum, followed by the classification of particle mass spectra into certain types. In this context, NO+-rich particles were identified by a predominant ion signal at m/z +30 corresponding to NO+, following the methodology outlined in , in which this particle type was named “secondary Type 1”. Additionally, we differentiated between primary and secondary particle components present within the NO+-rich particles, as described in . For this study, we limit the size range of the secondary type NO+-rich particles to between 200 and 500 nm (dva), which reflects ∼ 94 % and ∼ 86 % of the particles for the StratoClim and PHILEAS missions, respectively. The particle size selection is based on the same particle types defined in (StratoClim campaign in 2017), which also serves as the basis for particle selection in this study. We excluded larger and smaller particles sizes from the evaluation due to the effect of particle size on the ratio of ion peak intensities (Supplement Sect. S3.4). The average spectrum of the NO+-rich particle type for the Early and Late Phase is given in Fig. .

Furthermore, potassium-dominated particles were identified by a pronounced K+ ion signal at m/z +39/+41, which is characteristic of biomass burning (BB) sources (e.g., ). The denotation NO+-rich and potassium-dominated used in this study refers to the comparatively strong signal intensities of NO+ and K+. It does not imply that the particles consist exclusively of these components, since contributions from organic material, sulfate, and others can also be present (Figs.  and S4). The particle fraction (PF) for each particle type was calculated by binning the number of particles of each type relative to the total number of particles within 5 K potential temperature and 5° equivalent latitude intervals. To additionally account for variations in the absolute abundance of ERICA-LAMS detected particles, particle type resolved number concentrations were calculated following . Here, N0 denotes the averaged ERICA-LAMS particle number concentration at the first detection stage in each bin. The scaled number concentration (N0⋅PF) of a given particle type was obtained by multiplying N0 with the corresponding PF. This quantity, N0⋅PF, was used to evaluate whether changes in the relative contribution of a particle type are also reflected in changes in its number concentration within the ERICA-LAMS detected particle population. Comprehensive details regarding the ERICA-LAMS data post-processing and analysis are provided in the Supplement Sect. S3.

In this study, we use the ratios and relative proportions of ion peak intensities in the average mass spectra of particle types to characterize the internal mixing state and composition of single particles e.g. . Table  summarizes particle types investigated in this study: NO+-rich and potassium-dominated particles.

NO+-rich particles are internally mixed with ammonium, nitrate, sulfate, and organic compounds (Fig.  and Table ), consistent with earlier single particle measurements in the center of the AMA StratoClim 2017 mission;. To analyze the internal mixing state of NO+-rich particles during PHILEAS, three key ion signal ratios were considered: the nitrate-to-sulfate ratio (m/z -62/-97), the ammonium-to-nitrate ratio (m/z +18/+30), and the ratio of less-to-more oxidized organic signals (m/z +27/+43).

The second part of the instrument comprise the so-called ERICA-AMS (Aerosol Mass Spectrometer) technique. Here, non-refractory particle components, such as nitrate, sulfate, ammonium, and organic matter, are vaporized at around 600 °C on a tungsten vaporizer. The resulting vapor is then ionized by electron impact ionization. A compact time-of-flight mass spectrometer (C-TOF-MS) generates unipolar positive mass spectra of small particle ensembles with a time resolution of 10 s. The ERICA-AMS provides a quantitative analysis of the bulk chemical composition of these non-refractory aerosol components . The ERICA system operates in alternating mode: during 5 s of laser ablation operation with ERICA-LAMS, the ERICA-AMS measures the background signal within the ionization chamber, which is closed by a shutter. It then switches to ERICA-AMS mode (with ERICA-LAMS in standby) to measure the aerosol beam for the following 5 s. This cycle repeats every 10 s.

The ERICA-AMS data were processed using TofWare 2.5.7 (Tofwerk), as detailed in Supplement Sect. S4. For this study, 9472 ERICA-AMS data points for each species corresponding to the Early and Late Phases of stratospheric background air masses were analyzed. Background noise calculations and detection limits for all species were performed following the methodology described by , further information about the ERICA-AMS detection limits is provided in Supplement Sect. S4.3. The average detection limits for the Early and Late Phases flights during this campaign, are summarized in Table .

The University of Mainz Airborne QCL-based Spectrometer (UMAQS) is an advanced instrument designed for precise in-situ measurements of trace gases, such as CO and nitrous oxide (N2O), in the UTLS. Developed for atmospheric research, UMAQS operates with a temporal resolution of 1 s, enabling the detailed study of small-scale mixing and exchange processes across the tropopause. Utilizing quantum cascade laser (QCL) spectroscopy, UMAQS ensures high accuracy and reliability, as demonstrated during the PHILEAS campaign. We used the measurements of N2O and CO to identify stratospheric air masses, in particular stratospheric unperturbed air masses (see Sect. ; ). Based on in-situ calibrations against secondary laboratory standards, which were calibrated before and after the campaign against standards traceable to the NOAA calibration scales for atmospheric trace gases to ensure global comparability, the total uncertainty (2σ, 1 Hz) is estimated to be 0.3 ppbv for N2O and 1.4 ppbv for CO.

The Backscatter Cloud Probe with Polarisation Detection (BCPD) was operated on HALO to identify the presence of cloud hydrometeors in the size range from 2 to 42 µm. The instrument detects particles by measuring light scattered in the backward direction and uses a polarization filter to distinguish between spherical and non-spherical shapes . For PHILEAS, a cloud flag was derived from the BCPD based on the occurrence of particles within this size range that exceeded a number concentration of 0.02 cm⁻³. Such signatures indicate the presence of liquid or mixed-phase cloud elements and were used to exclude cloud-contaminated periods from the aerosol analysis. This approach provides a robust indicator for cloud influence and served as the primary cloud flag throughout the PHILEAS dataset.

CLaMS is a Lagrangian chemical transport model, with the advective transport scheme based on the calculation of three-dimensional forward trajectories and an additional parameterization of small-scale mixing based on deformations in the large-scale flow e.g.,. The simulations for this paper have been driven with meteorological data from ERA5 reanalysis .

CLaMS simulations of surface origin tracers allow quantifying the amount of air originating from specific surface regions. In detail, surface origin tracers have been defined for different source regions (e.g., South Asia, western Pacific, entire Model Boundary Layer (MBL)). The definition of the regions is based on and the origin tracer mixing ratios are set to unity in the lowest model layer (approximately the boundary layer) from May 1 on throughout the simulation. As these tracers are chemically inert, their mixing ratio at a given location and time equals the fraction of air originating in the respective surface region (for further details see ).

A second CLaMS simulation provides information on the age of air spectra, the transit time distribution for an air mass since leaving the tropical surface (lowest model layer between 30° N/S). The CLaMS-derived age spectra are calculated from pulse tracers in the model following . The resolution of age spectra along the transit time axis is 1 month and therefore allows analysis of fast transport time scales into the UTLS, down to time scales of a few months. Both model diagnostics, including surface origin tracers as well as age of air spectra, have been interpolated along the PHILEAS research aircraft flight tracks to enable interpretation of the measurements.

Additionally, the diabatic CLaMS backward trajectories were used to investigate also the origin and transport pathways of the air masses sampled during the PHILEAS campaign. These trajectories were driven by ERA5 reanalysis data with a horizontal resolution of 0.3°×0.3° and were initialized along the flight paths at one-second resolution using aircraft positions. All trajectories were calculated backward in time until 1 May 2023, which allowed for the identification of source regions and transport histories on synoptic to seasonal timescales. However, the initial position of the CLaMS backward trajectories along the aircraft flight path were selected separately for the Early and Late Phases, based on the trace gas criteria for ExLS background air masses defined in Sect. . The trajectories were then traced backward until they reached a vertical hybrid coordinate (ζ) =120 K. Here, ζ is used for the CLaMS simulations, which follows potential temperature in the stratosphere and gradually transforms into a pressure based coordinate in the troposphere, thereby accounting for surface pressure and orography . Following , ζ=120 K was used as threshold for the model boundary, which corresponds to an altitude of approximately 2 to 3 km above the Earth's surface. Finally, all backward trajectories reaching this threshold were defined as trajectory endpoints and counted within 1°×1° latitude and longitude grids. This approach was used to identify low tropospheric source regions that were connected to the studied ExLS background air masses via backward transport. The diabatical formulation provides a consistent representation of the transport process, which is particularly important for analyzing air mass pathways into the UTLS during the ASM. The back trajectory dataset follows the methodology described by and complements the CLaMS surface origin tracers and age of air diagnostics by providing a Lagrangian perspective on individual air parcel histories.

Trace gases such as N2O and CO are commonly used to distinguish stratospheric air masses from tropospheric air and tropospheric pollution events e.g.,. In the troposphere, N2O exhibits almost uniform concentrations because of its long atmospheric lifetime and the absence of significant removal processes. In the stratosphere, however, N2O is gradually depleted by photolysis and reaction with excited oxygen atoms O(¹D). As a result, mixing ratios decrease systematically with altitude and increasing potential temperature above the tropopause, allowing for an identification of stratospheric air masses e.g.,.

During the PHILEAS campaign, stratospheric air was identified using a criterion of N2O mixing ratios being below 336 ppbv . In addition, air masses in the background stratosphere were further classified by applying a CO threshold of 40 ppbv, with CO serving as a tracer for pollution and combustion processes. The CO threshold was determined by analyzing the relative frequency distribution of all CO values measured in the stratosphere (N2O<336 ppbv; Fig. ), which revealed two distinct modes representing undisturbed and polluted stratospheric air. Air masses with CO mixing ratios below this threshold are considered undisturbed by recent transport from polluted tropospheric air masses. The applied threshold is consistent with previous findings by .

For the following analysis of CLaMS data, it is necessary to distinguish between unperturbed stratospheric and tropospheric air masses. To evaluate whether the N2O and CO thresholds, originally developed for in-situ measurements, are also usable to the CLaMS dataset, a comparison was made with potential vorticity (PV) values. The PV values and equivalent latitudes were derived from the ERA5 data with 1°×1° horizontal resolution. The PV values consistently exceed 7 PVU with major contribution from PV above 8 PVU when N2O and CO concentrations are below 336 and 40 ppbv, respectively. As a result, values above 8 PVU are consistent with the N2O and CO thresholds that were applied to the CLaMS data. These PV values are used only as an additional consistency check for the N2O and CO based classification and for the dynamical interpretation of the sampled air masses. The transport and origin analysis itself is based on CLaMS data driven by ERA5 meteorological data with a horizontal resolution of 0.3°×0.3°. Further, details on the PV threshold determination are provided in Supplement Sect. S5.

The following findings provide evidence that the changes in stratospheric aerosol composition between summer and autumn are significantly influenced by the northward outflow of ASM-influenced air masses. Figure  provides a schematic overview of the large-scale AMA circulation and the associated filamentary transport pathway toward the ExLS.

First, the NO+-rich particle type observed in this study closely resembles the particle type identified in the UTLS region over the ASM (within the ATAL) reported by . In their study, it was demonstrated that particles containing ammonium nitrate are formed in the upper troposphere above the ASM convection. A key finding was that ammonium nitrate mass concentrations predominantly increased above 370 K potential temperature. Furthermore, they reported that aerosols in the ATAL primarily consist of nitrate, ammonium, and organic compounds, with nitrate concentrations exhibiting a pronounced peak between 370 to 390 K. These results are in good agreement with our PHILEAS measurements, which reveal a similar increase in nitrate concentration and in the fraction of NO+-rich particles at comparable potential temperature levels (see Figs. a, b and a). This consistency underscores the comparable chemical characteristics of aerosol populations observed during both campaigns.

Second, we analyzed CLaMS model results to investigate the origin and source regions of the enhanced particulate ammonium nitrate in the ExLS in autumn 2023. Figures  and provide results from CLaMS surface-origin tracer and the CLaMS averaged age spectra as a comparison between the Early and Late Phases. The analysis reveals that the ExLS background air during the Late Phase, was influenced by recent transport of boundary layer air from the South Asia and western Pacific region (Fig. ). The increase in the surface-origin tracer for the entire MBL between the Early and Late Phases indicates the growing influence of boundary layer air since 1 May 2023 on the ExLS background air between autumn and summer 2023. Specifically, air masses from the South Asia and the western Pacific region contributed to the ExLS background composition during the Late Phase (Fig. ).

Furthermore, the averaged CLaMS age spectra show differences in the ExLS background air between the Early and Late Phases (Fig. ). The general shift in the age spectra between the Early and Late Phases reflects the time difference between the campaign phases, with the peaks at transit times of more than five months lagging by about one month in the Late Phase spectra compared to the Early Phase spectra. However, during the Late Phase in autumn, the ExLS background air exhibits a higher fraction of young air masses (<3.5 months; Fig. ). In addition, Fig.  shows CLaMS-derived backward trajectory statistics for air masses observed during the Early Phase (Fig. a) and Late Phase (Fig. b). It is demonstrated that differences in backward trajectory endpoints (trajectories reached the boundary layer) exists between the Early and Late Phases, indicating changes in the pre-dominant source regions and transport characteristics of young air masses reaching the ExLS. The Late Phase, in comparison to the Early Phase, exhibits a pronounced, spatially coherent maximum in trajectory endpoints over northern Indian Subcontinent and along the southern margin of the Tibetan Plateau. It is thus obvious that the ExLS background air during the Late Phase was strongly influenced by relatively young (<3.5 months) air masses from ground-level sources in South-East Asia. Additional trajectory analyses, shown in the Supplement Sect. S6, indicate that within the potential temperature range of 350 to 380 K, the trajectories suggest the AMA acts as a reservoir for ASM-influenced air. This allows the masses to reside for extended periods before being exported poleward. The enhanced nitrate concentrations observed during the Late Phase should therefore not be interpreted solely as a local accumulation of particulate nitrate with time. Instead, they likely reflect the progressive seasonal accumulation of ASM-influenced air masses in the ExLS, driven by repeated eddy shedding and air mass transport from the AMA, which is most frequent during July and August but continues into September e.g.,. This result goes along with a recent study by , which provides direct evidence for the northward transport of pollution aerosol (including ammonium nitrate) from the ASM region and irreversible mixing of this aerosol into the ExLS. The authors conclude that the quasi-horizontal isentropic advection occurring above 370 K potential temperature in the lower stratosphere is the most important transport pathway of pollution aerosol associated with the ASM region into the ExLS. In combination with their findings, this study provides additional evidence that polluted aerosol from Asia containing ammonium and nitrate are enriched in the ExLS background air and can persist there for a few months.

Third, we observed a significant increase of methane (CH4) mixing ratio in the ExLS background air during the Late Phase compared to the Early Phase. Figure  shows the probability density functions of CH4 and CO. Both trace gases can be related to tropospheric sources. In detail, enhanced CO mixing ratios are typically associated with combustion processes e.g.,; CH4 is mainly related to agricultural activity in Asia e.g.,. Earlier studies demonstrated that both trace gases show higher mixing ratios within AMA confined air masses e.g.,. In particular, demonstrated that CH4 exhibits a pronounced and persistent enhancement within the ASM related air masses due to strong surface emissions and its long atmospheric lifetime, making it a robust tracer of ASM-influenced air masses in the UTLS. Previous studies have further shown that CO is an effective tracer for recently uplifted tropospheric air within the ASM dynamics and its subsequent export into the extratropical UTLS, particularly during late summer and early autumn . For CH4 (Fig. a), the Late Phase exhibits a clear shift toward higher mixing ratios compared to the Early Phase. The Late Phase distribution primarily spans from 1850 to 1890 ppbv, while the Early Phase peaks at around 1820–1850 ppbv. This result reflects the accumulation of methane-rich air in the ExLS background air during autumn 2023. Compared to CH4, the difference between the Early and Late Phase CO distributions is less pronounced (Fig. b). The Late Phase CO distribution is only marginally shifted toward higher mixing ratios, while both distributions strongly overlap and lower CO mixing ratios are also present during the Late Phase. Overall, the distribution appears broader in the Late Phase than in the Early Phase and a local maximum of the probability density in the Late Phase occurs near 30–32 ppbv, whereas the Early Phase peaks closer to 27–29 ppbv. This concurrent shift toward higher mixing ratios of CO and CH4 during the Late Phase provides further evidence for an enhanced influence of Asian boundary layer air on the ExLS background in autumn 2023.

Together, the observed changes of particulate ammonium nitrate in the ExLS background (above 370 K potential temperature) in autumn 2023 can be attributed to the presence of young air masses (<3.5 months) from Asian ground-level sources and its subsequent northward transport and outflow within the ASM circulation. This result is consistent with earlier studies, highlighting the ASM outflow as an important process to understand the increase of anthropogenic pollutants in the ExLS late summer and autumn e.g.,.

During the campaign, we observed a decrease in the concentration of particulate organics in the ExLS background air, particularly below a potential temperature of 370 K (Fig. c). One reason for the observed changes in aerosol composition could be the declining influence of BB in the Northern Hemisphere in the course of the summer and autumn 2023. The Canadian wildfire season in 2023 was record-breaking with its unprecedented number of pyrocumulonimbus events between May and August . However, mentioned that stratospheric aerosol composition was not significantly perturbed by the pyrocumulonimbus events in 2023. In contrast, described that pyrocumulonimbus clouds can perturb stratospheric aerosol composition. Our measurements show that the ExLS background composition below 370 K was influenced by BB emissions during this Early Phase, as indicated by the abundance of organic matter (Fig. c, f) and potassium-dominated particles (Fig. b, d). Earlier studies showed that organic matter is largely enhanced in BB smoke observed in the stratosphere e.g.,. Potassium-dominated particles are known to originate from BB emissions e.g.,. This particle type should therefore not be interpreted as a marker for secondary organic aerosol associated with the ASM-influenced air mass. Previous observations within the ATAL indicate that organic matter in ASM-influenced air masses is mainly associated with secondary aerosol formation from precursor gases, whereas direct transport of primary particles from near surface sources is strongly limited by efficient wet scavenging during convective transport e.g.,. Consequently, ASM-related organic aerosols are not necessarily enriched in potassium components. Potassium-dominated particles are specifically used here as an indicator of BB influence. The PF and N0⋅PF of potassium-dominated particles decreased progressively from August (Early Phase) to September (Late Phase) (Fig. b, d). This trend suggests a decreasing influence of BB emissions on the ExLS aerosol composition as well as a dilution and removal of BB-derived particles as the season advanced. Our observations are in line with measurements described by , who reported a reduction in aerosol elemental concentration of carbon from August to September in the lowermost stratosphere.

Overall, our results provide evidence that the ASM outflow can alter the aerosol composition in the ExLS background air in summer/autumn. We observed the increasing abundance of ammonium nitrate in the ExLS background during the summer-to-autumn transitions. However, the organic content was likely dominated by BB events in summer (Early Phase of PHILEAS), rather than by the outflow from the ASM.