AIDA (Aerosol Interaction and Dynamics in the Atmosphere) is a cloud simulation chamber which has extensively been described elsewhere e.g., , so only a brief description is provided here. Figure  presents a schematic of the experimental setup used for this study. The AIDA chamber mainly consists of an aluminum vessel (84.3 m3 volume) located inside a thermostatic housing. Air ventilation through heat exchangers inside the housing allows us to control the temperature of the cloud chamber from +60 to -90 ∘ (accuracy ±0.3 ∘C). Mechanical pumps allow us to control and reduce the chamber pressure down to 0.01 hPa, with different pumping speeds. The chamber is cleaned by flushing several times with clean and dry synthetic air at low pressures between 1 and 10 hPa. Before refilling to ambient pressure with synthetic air, ultra-pure water is evaporated into the evacuated chamber to add the amount of water required for the following experiment. With this procedure, the walls can be partially covered with an ice layer to maintain ice-saturated conditions and to provide sufficient water vapor during the expansion cooling experiments (described in Sect. ). The water vapor concentration inside the chamber is measured with an in situ multi-path tunable diode laser absorption spectrometer (APicT) with an accuracy of ±5 % . The water saturation ratios with respect to water and ice are computed by using the liquid water and ice saturation pressures from .

A second vessel is located in proximity to the AIDA chamber. It is called aerosol preparation and characterization (APC) chamber, is made of stainless steel, has a volume of 3.7 m3, is operated at ambient temperature, can also be evacuated down to 1 hPa and flushed with synthetic air for cleaning purposes, and is refilled with clean and dry synthetic air before starting the experiments. The APC chamber allows us to prepare the aerosol particles in a reproducible way (i.e., under well-controlled and repeated experimental conditions) and to transfer them into the AIDA chamber. Furthermore, the evacuated APC chamber can be used as a buffer volume for cloud simulation experiments at a high expansion rate in the AIDA chamber (more details are provided in Sect. ).

Standard aerosol particle instrumentation is present at the facility to measure the aerosol number concentration and number size distribution. Both the AIDA and the APC chambers are equipped with condensation particle counters (models 3010 and 3022 from TSI) and scanning mobility particle sizers (SMPS, TSI). Two optical particle counters (OPC, model welas, Palas GmbH) are installed at the bottom of the AIDA chamber to measure the size distribution of large aerosol particles, cloud droplets, and ice crystals in two different size ranges (0.7–46 and 5–240 µm, ). The two OPCs are located inside the thermal housing to prevent evaporation of cloud droplets and sublimation of ice crystals. The additional instrumentation used to characterize the aerosol population in terms of chemical composition and phase state is described in the next section.

The aerosol particles were generated, characterized, and aged in the AIDA chamber and in the aerosol preparation and characterization (APC) chamber. The coating procedure performed in the APC chamber aimed at generating a thick organic coating layer on the AS seed particles with high reproducibility. The ice nucleation ability of these particles was either directly probed in a reference experiment (experiments of type B) or after subjecting the particles to different processing steps (experiments of type D and E). The coating experiments performed in the AIDA chamber, instead, aimed at quantifying the effect of a thin coating on the inorganic seeds (experiments of type C). In this case, the coating procedure consisted of several steps, each with a small amount of organics condensing on the seed aerosol particles. The coating procedure in the APC chamber was performed at ambient temperature and low relative humidity and in the AIDA chamber at low-temperature and ice-saturated conditions. Table  provides a list of the performed experiments together with information on the thermodynamic conditions at which the particles were generated, the concentration of the gas precursors, the median diameter of the seed particle size distributions before the coating procedure, and the organic mass fraction after the coating procedure.

An aqueous solution was prepared by dissolving ammonium sulfate (Merck, 99.5 %) in ultra-pure water to obtain a 1 wt % solute concentration. The solution was then aerosolized by means of an ultrasonic nebulizer (GA 2400, SinapTec). A series of diffusion dryers (Topas GmbH) ensured a relative humidity with respect to water in the injection line lower than 1 %, well below the efflorescence relative humidity of ammonium sulfate . Secondary organic material (SOM) was generated from the ozonolysis of α-pinene (Aldrich, 99 %). Another important pathway for the formation of α-pinene SOM in the atmosphere is photooxidation reactions, but the ice-nucleating ability of the organic material generated via dark ozonolysis or photooxidation of α-pinene was found to be very similar . Ozone was produced with a silent discharge generator (Semozon 030.2, Sorbios) in pure oxygen.

The generation of AS particles with thick SOM coating was started by injecting crystalline AS particles (∼ 105 particles cm-3) into the APC chamber. Then ozone was added (∼2 ppm), followed by a continuous flow of α-pinene for approximately 30 min (experiments B1 and D2) or 1 h (experiments B2, D1 and E1), resulting in a total concentration between 100 and 300 ppb of injected precursor gas (see Table  for the concentrations used in each experiment). The high initial aerosol concentration in the APC chamber during the α-pinene injection provided a sufficiently large total surface area for the organic reaction products to condense on the existing AS particles and to suppress, in the majority of cases, the nucleation and growth of pure organic aerosol particles. At the end of the coating procedure, a fraction of the aerosol particles were transferred into the AIDA chamber.

For the thin coating experiments performed in the AIDA chamber (experiments of type C), the coating was performed in three to four steps, with the injection of smaller amounts of α-pinene. Each step consisted in the injection of 100 to 700 ppt of organic gas precursor, summing up to a total concentration of approximately 1 ppb (see Table  for the concentrations used in each experiment). A new injection step of gas precursor was performed approximately every hour to allow the reaction to be completed and to simultaneously perform the ice nucleation measurements (see Sect. ).

A high-resolution time-of-flight aerosol mass spectrometer (HR-ToF-AMS, Aerodyne) measured the chemical composition of the particles, with an uncertainty of 5.3 % regarding the measured mass concentrations (based on ionization efficiency calibration). The particles' organic content was measured after the transfer of the particles from the APC chamber to the AIDA chamber (experiments type B, D, and E) or continuously during the thin coating experiments in AIDA (experiments of type C). During the coating procedure of experiments B2 and C2, some pure organic particles nucleated and eventually grew to a size detectable in the SMPS and the mass spectrometer. For these experiments, to obtain the organic mass concentration of the organic coating on the AS particles, the total organic mass measured by the AMS was corrected for the mass of the smaller mode of nucleated pure organic particles. This was done by independently deriving the total mass concentration of the smaller particle mode from the size distribution measured with the SMPS (estimating the density of the organic material as ρorg=1.25 gcm-3 ) and then subtracting it from the total mass concentration measured with the AMS. The organic mass fraction values forg reported in Table  refer to the organic content at the end of the coating procedures; i.e., regarding experiments of type C, they refer to the organic content after the last coating step. The size-resolved measurements of the chemical composition of the particles by the AMS indicate that the organic mass fraction is constant and not a function of the particle size.

To estimate the thickness of the organic coating, d, we combined the size distribution of the pure crystalline AS particles from the SMPS measurements and the organic mass fraction forg from the AMS. Thereby, we assume the crystalline AS seeds to be spherical particles with diameter Dp and the organic material to be evenly distributed on their surface, leading to a spherical organic shell. The coating thickness d is thus calculated assuming a perfectly concentric core–shell morphology as follows: 1d=Dpcoated-Dp2. The diameter of the coated particles Dpcoated is calculated by considering the size dependence of the particles' organic mass Morg(Dp): 2Dpcoated(Dp)=[Dp3+6πMorg(Dp)ρorg]13, with 3Morg(Dp)=forg1-forg⋅ρAS⋅π6Dp3.

The resulting estimated organic coating thickness corresponding to the median diameter of the particle population is reported in Table . In Table , we also indicate between brackets the estimated coating thickness for seed particles with diameters of 300 and 500 nm, i.e., the size range which comprises the major particle mode in the number size distribution (see below and Fig.  for the number size distribution measurements). The thickness estimated for the thickly SOM-coated AS particles (experiments of type B, D and E) ranged from 24 to 54 nm. Thinner coatings were obtained at the end of the coating experiments performed in the AIDA chamber (experiments of type C) with 6, 5, and 8 nm coating thicknesses.

Exemplary particle size distributions for the pure crystalline AS and for the SOM-coated AS particles are reported in Fig. . The effect of the organic coating on the particle diameter is clearly visible for the thickly coated particles (Fig. a, experiment D1). Instead, no change is detectable in the SMPS measurements for the thin coating experiments, in which a lower amount of organic material was produced and condensed on the particles (Fig. b, experiment C3). However, due to the numerous processes taking place during the coating procedures (e.g., change of the particles' shape factor), we cannot directly infer the coating thickness from the SMPS measurements.

The phase state of the ammonium sulfate content in the mixed particles (i.e., crystalline or liquid) has been determined with depolarization measurements of back-scattered laser light and the analysis of infrared extinction spectra recorded by means of a Fourier transform infrared spectrometer (FTIR, type IFS66v, Bruker) . The presence or absence of specific liquid water absorption bands and changes in the peak position and band width of the ν3 sulfate mode at about 1100 cm-1 between crystalline and liquid AS allowed us to infer the phase state of the AS content in the mixture . The SIMONE instrument measures the intensity of the light scattered by the aerosol particles in the forward (2∘) and backward (178∘) directions from a linearly polarized laser (λ=488 nm). In the backward direction, the polarization-resolved scatter light intensity is detected so that the linear depolarization ratio of the aerosol particles can be determined. Non-spherical and inhomogeneous particle morphologies, like a crystalline core of AS or islands of crystalline AS in the mixed particles, can induce a change in the incident polarization state of the laser, thereby causing a non-zero value for the depolarization ratio. In contrast, homogeneously mixed, spherical aqueous solution droplets would show a depolarization ratio of zero. A detailed example of the use and interplay of the SIMONE and FTIR data in the interpretation of our experiments is given in Sect. .

The ice nucleation ability of the aerosol particles was probed by means of expansion cooling experiments inside the AIDA chamber and with a continuous flow diffusion chamber sampling the aerosol from the AIDA chamber. Details about these two methods are provided below.

The ice nucleation ability of pure and SOM-coated solid ammonium sulfate particles at cirrus cloud conditions was probed with experiments of type A, B, and C (Fig. ). The particles were coated in the APC chamber for experiments of type B and in the AIDA chamber for experiments of type C. In both cases, we assume that the coating procedure led to particles with a core–shell morphology, with the solid AS as the core and the organic material surrounding it. This hypothesis is supported by the ice nucleation measurements as discussed below. We conducted the experiment of type B two times with different coating time and amount of α-pinene added to the APC chamber. In experiment B1, we generated SOM-coated AS particles with an organic mass fraction of 26.9 % and a coating thickness of 28 nm; in experiment B2, the organic mass fraction was 39.1 % and the coating thickness 50 nm. As explained in Sect. , these coating thicknesses were computed for the median particle diameters assuming a uniform coating.

Time series of the AIDA expansion cooling runs for experiments A1 (pure crystalline AS) and B2 are presented in Fig. . The upper panels show the pressure and temperature drop during the expansion runs. The middle panels show the saturation ratio with respect to ice (Sice) and the homogeneous freezing threshold . The lower panels depict the fraction of aerosol particles activated as ice crystals (ice-nucleating fraction, INF). Black dotted vertical lines indicate when the INF reaches the threshold of 0.1 %, defined here as the ice nucleation onset. Crystalline AS particles induce heterogeneous ice nucleation at saturation values lower than those required for the homogeneous freezing of solution droplets. For crystalline AS the ice onset is measured at Sice=1.29 (Fig. a middle panel). For SOM-coated AS particles, instead, the ice onset is measured at Sice=1.44 (Fig. b middle panel). The organic coating clearly suppresses the ice nucleation activity of the crystalline AS core and shifts the ice nucleation onset close to the homogeneous freezing threshold. The strong decrease in the particles' ice nucleation ability suggests that most of the IN-active sites of the bare AS have been shielded by the condensed organic material distributed in a core–shell morphology.

Figure a summarizes the ice onset saturation ratios for pure and thickly SOM-coated crystalline AS particles from this study (experiments A1, B1, and B2). The water saturation (RHw=100 %), the homogeneous freezing threshold for Δaw=0.3 , and the AS deliquescence relative humidity are also shown in the figure. Results from AIDA and INKA are shown as squares and circles, respectively, and the symbol colors refer to the different aerosol types used in the different experiments (orange for pure AS particles and light and dark green for experiments B1 and B2, respectively). Both methods clearly show the suppression of the heterogeneous ice nucleation activity of the crystalline ammonium sulfate core by the organic coating. The AIDA results show lower ice onset saturation ratios for both pure and SOM-coated AS particles compared to the INKA results. already noted this discrepancy in previous experiments with inorganic particles at cirrus conditions and encouraged inter-comparison studies of different methods and instruments at low temperatures. It is worth mentioning that secondary organic material is more viscous at low temperature, and this also affects water diffusion. Therefore, different experimental techniques could be subjected to different kinetic limitations, inducing ice nucleation at different conditions. During an AIDA expansion run, for example, the thermodynamic conditions inside the chamber change relatively slowly and allow long equilibration times. However, inside the continuous flow diffusion chamber, new particles are continuously sampled, suddenly exposed to a supersaturated environment, and only have a few seconds (∼10 s) to activate as ice crystals.

In Fig. b, we compare our data to literature results of Sice at the ice nucleation onset for pure crystalline AS particles (red colored symbols) and pure secondary organic aerosol (SOA) particles (green colored symbols). One can see that the pure AS and pure SOA data form clearly separated blocks on the Sice vs. T diagram, with the heterogeneous nucleation onsets for AS ranging from 1.1 to 1.3 and for pure SOA from 1.4 to 1.7. With regard to our experiments with the thickly SOM-coated AS particles, experiment B1 shows a slightly lower ice onset saturation ratio (Sice=1.4, AIDA experiment) compared to the measurements of the pure SOA particles. However, this early onset was only representative of a small fraction of the aerosol particle population, and if we had chosen an INF of 1 % as the onset condition, the related Sice value would have already increased to 1.46. When analyzing in detail the results of the processing experiments on the particles' ice nucleation ability (experiments of type D and E), we will not only compare the onset conditions but show the full spectrum of the INF as a function of Sice.

Experiments B1 and B2 have shown that thick SOM-coating layers almost completely suppress the heterogeneous ice nucleation ability of the AS component and shift the particles' ice nucleation onset close to that observed for the pure SOA particles. We thus conducted experiments of type C to investigate and quantify the effect of a thin organic coating layer on the ice nucleation activity of crystalline AS particles. As described in Sect. , the coating procedure for these experiments was performed directly in the AIDA chamber at different temperatures (-45, -50, and -60 ∘C) and at ice-saturated conditions by stepwise oxidizing small amounts of α-pinene. After each coating step, the aerosol particles were sampled and their ice nucleation ability measured with the continuous flow diffusion chamber at the same, or slightly lower, temperature of the AIDA chamber (∼-50, ∼-54, and ∼-60 ∘C). Figure a shows the ice-nucleating fraction (INF) as a function of the saturation ratio with respect to ice for humidity ramps performed at -54 ∘C with particles characterized by different organic mass fractions with INKA. The activation curves from experiments C1 and C3 are shown with solid symbols (circles and down-pointing triangles, respectively), and data from experiments B1 and B2 are shown as empty circles. The gradual shift of the ice onset towards higher saturation values with the increase in the condensed organic material is clearly visible. Already after the second coating step of experiment C1, which corresponded to an organic mass fraction of only 2.4 wt %, most of the ice crystals that had formed at Sice<1.4 on the pure crystalline AS particles have disappeared from the records. After the fourth coating step of experiment C1, when the organic mass fraction has reached a value of 5.7 wt %, the ice nucleation onset was shifted close to the value observed for pure α-pinene SOA particles. The ice nucleation ability of particles with ∼8 wt % organic material (experiment C3) is almost identical to the one of the thickly coated particles from experiments B1 and B2 (with 27 and 39 wt % organic mass fractions, respectively), indicating that such a low organic content, if uniformly distributed, is sufficient to suppress the ice nucleation ability of the solid AS particles. The progressive increase in the ice onset saturation ratios measured by the continuous flow diffusion chamber highlights its sensitivity to detect the effect of the organic coating.

Figure b summarizes the ice onset saturation ratios, corresponding to an INF of 0.1 %, obtained from the INKA measurements as function of the organic mass fraction of the particles. Results for experiment A1 (left side) and for all experiments of type C (right side) are shown. Different symbols correspond to different experiments, and colors indicate at which temperature the ice nucleation ability was measured. The data confirm the results from the experiments presented in Fig. a, showing that the ice nucleation ability of the solid AS component is gradually suppressed with the increase in organic material condensed on its surface. In most cases organic mass fractions of only 4 wt %–8 wt % (yielding estimated coating thicknesses of 5–8 nm) shift the ice nucleation onsets of the coated AS particles to values above Sice=1.45 at temperatures between -45 and -55 ∘C, i.e., to the regime where also pure SOA particles would nucleate ice. These small organic mass fractions were sufficient to completely suppress the heterogeneous ice nucleation ability of the crystalline AS core. The suppression of the particles' ice nucleation ability during and at the end of the coating procedure suggests, as for experiments of type B, that the organic material is evenly distributed on the AS surface and progressively covers its IN-active sites.

The temperature trend of the ice nucleation onsets for organic mass fractions larger than about 4 wt % revealed by experiments C, i.e., higher Sice values with decreasing temperature, could point to a homogeneous ice nucleation pathway of the coated particles, meaning that at least the outer layer of the organic material has liquefied during the particles’ residence time in INKA. A similar temperature trend has been observed in ice nucleation studies with pure α-pinene SOA particles e.g., . Furthermore, we need to consider that the condensed organic material could have a different chemical composition and viscosity at the different temperatures of the in situ coating experiments , thereby affecting the water diffusion into the particles and the ice nucleation pathways .

In view of the results from Sect.  that the ice nucleation onset saturation ratios of SOM-coated crystalline AS particles are already at or above Sice=1.45 for small organic mass fractions of 10 wt %, one might ask the question whether this type of internally mixed particles should still be considered as candidates for inducing heterogeneous ice nucleation in cloud models. Furthermore, mass spectrometer measurements by show that the average organic mass fraction of internally mixed particles is even higher (ranging from 30 wt % to 70 wt %). Experiments of type C, however, just tested one specific pathway for the formation of the mixed particles, where the crystalline AS was already present and then coated with the organic substances. As already discussed in the introduction, different particle morphologies, like partially engulfed structures, might form when we start from homogeneously mixed, aqueous mixtures of sulfate and organics and then induce the crystallization of the AS component. This pathway was investigated with experiments of type D, where liquid cloud processing is simulated. Here, we started from thickly SOM-coated AS crystals as investigated in experiments of type B, temporarily activated these particles to homogeneously mixed, aqueous droplets in a short expansion run conducted in the AIDA chamber at warm and humid conditions, and then reduced the relative humidity to induce the recrystallization of the ammonium sulfate component.

Figure a shows a schematic of the liquid cloud processing together with the possible phase states and morphologies of the particles during the experiment. The thickly coated ammonium sulfate particles were generated as described in Sect.  in the APC chamber (stage I in Fig. a) and then transferred into the AIDA chamber (stage II), which was held at -5 ∘C and RHw=85 %. These are conditions where the ammonium sulfate component is already in a liquid state (the AS deliquescence relative humidity at -5 ∘C is 83 % RHw, ) but where the entire particle is probably still in a liquid–liquid phase-separated state (separation relative humidity can be as high as ∼90 %, ). Figure b presents the time series of the AIDA pressure (black line), AIDA temperature (red line), AIDA relative humidity with respect to water (blue line), and linear depolarization ratio measured with the SIMONE instrument (green line) during the liquid-cloud-processing experiment. The liquid state of the ammonium sulfate fraction upon transfer from the APC chamber into the AIDA chamber is verified by the FTIR measurements, reported in Fig. c. Spectrum A (in orange) is a reference spectrum of crystalline ammonium sulfate particles from experiment A1 in the regime of the ν3 (SO42-) mode at about 1114 cm-1. Spectrum B (in blue) was recorded after addition of the organic-coated AS crystals into the AIDA chamber. Here, the ν3 mode is clearly broadened and shifted to lower wavenumbers. These spectral changes are indicative of the transition of AS from the solid to the liquid state . Also the low value of the depolarization (δ∼1 %) is indicative of a particle morphology that is close to that of a homogeneous sphere. However, the non-zero value of δ indicates that some particle inhomogeneity still remains. We suggest that the particles are characterized by two phase-separated, slightly eccentrically arranged liquid components, because δ would also be zero for concentric spheres .

To then transform the phase-separated particles in the AIDA chamber into homogeneously mixed liquid particles, we performed a short expansion run in order to reach RHw=100 % and thereby exceed the relative humidity threshold for the liquid–liquid phase separation (stage III in Fig. a). During the expansion run, the depolarization δ indeed dropped to the background value of about 0 % when the relative humidity increased above 90 %, indicating that we had successfully transformed the initially SOM-coated AS crystals to homogeneously mixed aqueous droplets. After the expansion, we refilled the AIDA chamber to ambient pressure and induced the recrystallization of the AS component by reducing the relative humidity to 32 % (stage IV). This was done by extracting the humid air from the chamber and refilling it at the same rate with dry synthetic air. Note that in this specific experiment the chamber walls were not ice-coated, because the ice layer would have acted as a source of water vapor and prevented the reduction of the relative humidity. To support the reduction in relative humidity, we also slightly warmed the AIDA chamber to -3.5 ∘C. The recrystallization of the AS fraction becomes evident from the increase in the depolarization δ over time and the associated change in the infrared signature of the particles. Spectrum C in Fig. c (in purple), which was recorded at RHw=32 % at the end of the recrystallization step, clearly shows that the ν3 mode of particles has shifted back to the peak position typical for crystalline AS. This procedure has been applied in two experiments with particles of different organic mass fraction (experiments D1 and D2; see Table ).

The ice nucleation ability of the in situ crystallized particles has been probed with INKA at -54 ∘C immediately after the AS crystallization (for experiments D1 and D2) and with an expansion cooling experiment after overnight cooling of the AIDA chamber to -50 ∘C (for experiment D2). As the INKA measurements were performed by sampling from the AIDA chamber at -5 ∘C and RHw∼32 %, a diffusion dryer was used to prevent frost formation in the instrument inlet. To infer the possible effect of the dryer on the ice nucleation ability of the recrystallized particles, an additional INKA measurement was performed after the cooling of the AIDA chamber, thus without the need of a diffusion dryer. The ice nucleation ability measured with and without the dryer (i.e., before and after the cooling) are comparable (not shown), indicating that the dryer did not influence the ice nucleation ability of the particles. Figure d and e present the ice-nucleating fraction (INF) as a function of the saturation ratio with respect to ice Sice for pure AS crystals (orange data, experiment A1), for unprocessed thickly SOM-coated AS crystals (green, experiment B1), and for the internally mixed particles that were subjected to the droplet activation and in situ crystallization process (yellow data, experiment D2). Very similar results were obtained for experiment D1. The AIDA and INKA results indicate that the ice nucleation ability of the processed particles lies between those of the pure crystalline AS and of the thickly SOM-coated AS crystals. The onset of the heterogeneous ice nucleation mode of the processed particles is observed at Sice∼1.35. The in situ crystallization of initially homogeneously mixed aqueous AS / SOM particles can therefore lead to particle morphologies that are heterogeneously ice nucleation active, even if the organic mass fraction is as high as 39.1 wt % or 24.3 wt % (experiment D1 and experiment D2, respectively). Most probably, the particles adopt a partially engulfed structure, where some ice-nucleating sites of the AS component remain uncovered by the organic material . This result again supports the hypothesis of a core–shell morphology for the coated particles of experiments B, for which the heterogeneous ice nucleation ability of the AS component was completely masked by the condensed organic material.

The measured ice onsets agree with previous measurements of and . As noted in the introduction, investigated mixed particles composed of AS and the water-soluble fraction of SOM derived from α-pinene. The particles were generated from liquid solutions with inorganic-to-organic mass ratios of 1:1 and 4:1, dried, and then probed in a CFDC on their ice nucleation ability at -55 ∘C. They also found an intermediate ice nucleation onset of the mixed particles (at Sice=1.4) compared to reference experiments with pure crystalline AS (ice onset at Sice=1.25) and pure secondary organic particles (ice onset at Sice=1.52). The results from are shown for comparison with the INKA data in Fig. e (triangles, AS in orange, mixed particles in yellow, and pure SOA in green). The results for the pure SOA particles from and for the thickly SOM-coated particles from this study agree, confirming that the coated particles have an ice nucleation behavior similar to the purely organic particles, consistent with our assumption of a core–shell morphology. Also the ice nucleation behavior of the mixed particles obtained from homogeneously mixed aqueous AS / SOM solutions upon crystallization of the AS component nicely agrees. have shown that complex particle morphologies are formed when AS crystallizes from solution droplets composed of methylglyoxal, methylamine, and AS. In their experiment, the ice nucleation onset of the internally mixed particles was lower compared to the pure organics, in agreement with our results.

We note, however, that the underlying heterogeneous ice nucleation mode might be different in the various studies. In addition to deposition nucleation, crystalline AS particles might also be ice nucleation active in the immersion freezing mode . This activity may be controlled by the viscosity and water solubility of the organic material . The finding that α-pinene SOM mass fractions greater than ∼ 5 wt %–10 wt % completely suppress the heterogeneous ice nucleation ability of crystalline AS (Sect. ) rules out that immersion freezing is a prevalent nucleation mode in our experiments. The higher heterogeneous ice nucleation activity of internally mixed AS / SOM particles formed by cloud droplet activation and recrystallization of the AS component must therefore be related to a change in the particle morphology to a partially engulfed structure. For organic materials with a lower viscosity and higher water solubility, however, the mixed particles might also be heterogeneously ice nucleation active in a core–shell morphology due to immersion freezing by the crystalline AS core. assumed that immersion freezing was indeed responsible for the heterogenous ice nucleation mode observed in their experiments. As only used the water-soluble fraction of the generated SOM, its hygroscopic behavior might be different compared to our study so that their observed heterogeneous ice nucleation mode of the mixed particles after the AS crystallization might also be due to immersion freezing. The viscosity and amount of the organic material will also determine whether the crystallization of AS in internally mixed aqueous AS / SOM particles can occur at all. In particular, at lower temperatures and/or higher amounts of organics, the AS efflorescence could also be inhibited . If recrystallization of the AS component and the formation of a partially engulfed particle morphology did not occur at lower temperatures (in contrast to the -5 ∘C condition simulated in this study), liquid cloud processing would not increase the ice nucleation ability of the AS / SOM particles. This behavior could be investigated in future AIDA experiments where the recrystallization process is performed at lower temperatures.

The enhanced ice nucleation ability of the liquid-cloud-processed particles compared to the pure organic and organic-coated particles clearly indicates that the ice nucleation ability of atmospheric aerosol particles can strongly change during their lifetime in the atmosphere. Cloud processing of internally mixed aerosol particles is a common phenomenon, whose impacts on the particles' microphysical properties need to be investigated in future studies.

After an atmospheric freeze-drying process in a convective cloud system (described below), organic particles can adopt a highly porous morphology . The porous morphology results from the fact that the initially liquid organic material concentrates and vitrifies upon ice formation, leaving behind a porous structure when the ice sublimates. This might provide a pathway to increase the ice nucleation ability of internally mixed AS / SOM particles with high organic mass fraction, because a porous organic coating might less efficiently cover the ice-active sites of the crystalline AS component. This pathway is investigated by our experiment of type E.

The freeze-drying process of the thickly SOM-coated AS particles was started at -30 ∘C in the AIDA chamber. At this temperature, we can run through the droplet activation (liquefaction of the organic material) and freezing (re-vitrification of the organic material) within a single expansion cooling experiment. Figure a summarizes the cloud-processing steps and the potential phase and morphology changes of the particles. The time series of the AIDA records during the freeze-drying expansion run, started at time 0 s, are shown in Fig. b. When the relative humidity exceeded the water saturation (first vertical line in Fig. b), aerosol particles were activated as cloud droplets, detected by their larger diameter in the optical particle counter data. When the temperature approached the homogeneous freezing threshold (∼-36 ∘C), a further fast drop in pressure was achieved by opening a valve in the pipe connection between the AIDA chamber and the evacuated APC chamber. This additional fast drop in pressure almost instantly reduced the gas temperature by another 2.5 K and caused the entire droplet population to freeze homogeneously. The ice crystal formation is clearly visible in the optical particle counter data (second vertical line in Fig. b). The chamber was then refilled to ambient pressure with dry synthetic air, and the ice crystals thereby quickly sublimated. To probe the ice nucleation ability of the freeze-dried aerosol particles at cirrus cloud conditions, we further cooled the AIDA chamber to -50 ∘C. After the chamber cooling, the infrared spectrum of the freeze-dried AS / SOM particles in the regime of the ν3 (SO42-) mode was similar to spectra A and C in Fig. c, indicating that the sulfate was effloresced. The AIDA expansion run started at -50 ∘C, and the INKA supersaturation scan was performed at -54 ∘C with the particles sampled from the AIDA chamber at -50 ∘C. The results are shown in Figs. c and d with yellow symbols. As a reference, we also show the ice-nucleating fraction (INF) curves for pure ammonium sulfate crystals (experiment A1, orange symbols) and unprocessed, thickly SOM-coated AS crystals (experiment B2, green symbols). The ice nucleation ability of the freeze-dried particles lies between those of the reference systems, similar to the in situ crystallized particles (experiment of type D). The heterogeneous ice nucleation mode of the freeze-dried particles has the ice onset for an ice-nucleating fraction of 1 % at a saturation ratio Sice of 1.41, while the ice onset saturation ratio is 1.33 for pure AS crystals and 1.48 for thickly SOM-coated AS crystals.

The activation curve of the freeze-dried particles measured with INKA (Fig. d, yellow data) has a different profile compared, for example, to the thickly coated particles (green data). The slower increase in the ice-nucleating fraction as a function of the ice saturation ratio (similar to the activation curve for AS in orange) suggests that heterogeneous freezing is the dominant ice nucleation mechanism occurring in INKA up to Sice∼1.6. This effect could be related to the limited residence time of the particles in the INKA instrument, which together with a slower water uptake from the particles could have shifted the detected homogeneous freezing onset to higher Sice values. This kinetic limitation is not evident in the AIDA data, probably due to longer equilibration time during the expansion run. Particles in experiment E1 will likely be more viscous than particles from experiment B2 due to their larger organic mass fraction and/or due to a change in the chemical–physical properties of the particles after the freeze-drying process.

The enhanced ice nucleation ability of the freeze-dried particles may be explained by morphology changes. First, as already suggested above, a porous organic coating could less efficiently cover the ice-nucleation-active sites of the crystalline AS core. Second, the organic material might have been redistributed during the freeze-drying process, so that a similar morphology change as discussed in experiment D could have occurred (like the formation of partially engulfed structures). The similar heterogeneous ice onsets measured in experiments D and E support this hypothesis. Third, the porous organic material formed during the freeze-drying process could be a better ice nucleus on its own via the pore condensation and freezing mechanism or by retaining an imprint of the sublimated ice on the highly viscous organic material. Previous studies observed an efficient preactivation of glassy organic aerosol particles after ice cloud processing . investigated in the AIDA chamber four different organic solutes, raffinose, 4-hydroxy-3-methoxy-DL-mandelic acid (HMMA), levoglucosan, and a multicomponent mixture of raffinose with five dicarboxylic acids and ammonium sulfate, and observed that the ice-cloud-processed glassy aerosol particles catalyzed ice formation at ice saturation ratios between 1.05 and 1.30. confirmed the improved ice nucleation ability of cirrus-cloud-processed raffinose particles in dedicated CFDC ice nucleation measurements. This raises the question of why the ice-cloud-processed AS / SOM particles (observed ice onset at Sice=1.41) do not reveal an even better ice nucleation ability. However, repeated ice nucleation experiments with secondary organic aerosol particles produced from the ozonolysis of α-pinene performed by showed a much lower susceptibility towards preactivation compared to the abovementioned organic solutes. For example, the ice nucleation onset of the ice-cloud-processed α-pinene SOA particles at 220 K was at an ice saturation ratio of about 1.45. A similar ice nucleation threshold was found in an experiment where highly porous α-pinene SOA particles were formed by a freeze-drying process at 243 K (as in the current study) and then cooled to cirrus temperatures to probe their ice nucleation ability. Furthermore, showed that in mixed AS / OM particles with a 1:1 molar ratio, the formation of a porous structure was reduced compared to the pure organic particles, adding to the effect that the freeze-dried AS / SOM particles are not extremely efficient INPs.

In spite of the strong effect of thin SOM coating layers revealed by experiments of type C, we have shown that the ice nucleation ability of AS / SOM internally mixed particles is also strongly linked to their thermodynamic history. Uplifting of particles to the upper troposphere via convective clouds, for example, could lead to particles with an enhanced ice nucleation ability.