Counterflow Virtual Impactor Based Collection of Small Ice Particles in Mixed-Phase Clouds for the Physico-Chemical Characterization of Tropospheric Ice Nuclei: Sampler Description and First Case Study

4.1 Location of the Sampling Site

The Ice-CVI was deployed at the high alpine research station Jungfraujoch (JFJ, 3580 m asl, 46°33′N, 7°59′E) within the framework of the international Cloud and Aerosol Characterization Experiment, CLACE-3, conducted from the end of February to the beginning of April 2004. Long-term aerosol measurements have shown that the JFJ can be regarded as representative for the continental lower free troposphere with only a minor boundary layer air influence in winter (CitationBaltensperger et al. 1997). In winter, the site is frequently wrapped in clouds at temperatures down to −30°C. Thus, the JFJ is well suited for studies of mixed-phase clouds in the lower troposphere (CitationHenning et al. 2004). Details about the cloud events presented in this study are given in Table 1, including mean temperature, wind velocity and direction measured by MeteoSwiss, the Swiss national weather service.

4.2. Ice-CVI Setup

The vertical Ice-CVI inlet system was mounted on a platform above the Sphinx laboratory with a distance of 2m between the omni-directional inlet and the surface. On the platform no wind measurements were available. Later tests showed that the wind velocity at the platform was significantly lower than the simultaneous MeteoSwiss values (especially for wind directions between WNW and NNW) measured at another location of the Sphinx Observatory. Thus, the values given in Table 1 should be considered as upper limits of the wind velocities at the Ice-CVI inlet.

All sensors for the residual particle analysis were located in the Sphinx laboratory directly beneath the platform. In parallel to the standard devices (CPC, PSAP, Lyman-α hygrometer) further instrumentation was coupled to the Ice-CVI (Figure 3).The particle number size distribution was measured with a Scanning Mobility Particle Sizer (SMPS, 20 nm < d_p < 800 nm) and an Optical Particle Counter (OPC, 0.3 μ m < d_p < 20 μ m), where d_p denotes the dry particle diameter. An Aerodyne Aerosol Mass Spectrometer (AMS, 20 nm < d_p < 1 μ m, Jayne et al. 2000; Schneider et al. 2006) was used to determine total mass concentration and mass size distributions of sulfate, nitrate, ammonium, and organic matter (OM). The aerosol particles are vaporized inside the AMS at a surface heated to 600°C. Hence all substances with a higher vaporization temperature cannot be detected. Moreover, aerosol samples were collected by a two stage impactor for off-line analysis with the Environmental Scanning Electron Microscope (ESEM, Ebert et al. 2002) in order to derive the morphology and elemental composition of single residual particles. Because all the sampling was carried out at temperatures below −5°C, the impaction plates of the PI were not actively cooled, i.e., they adopted the ambient temperature.

4.3. Further Aerosol Inlets

A total and an interstitial aerosol inlet were operated in parallel to the Ice-CVI (Figure 3) within a distance of 1.6 m and a similar inlet height above the Sphinx roof. The total aerosol inlet is heated to 25°C in order to evaporate/sublimate hydrometeors during sampling as early as possible to avoid transmission losses. This inlet is designed to collect hydrometeors with diameters smaller than 40 μ m up to a wind speed of 20 m s⁻¹ (CitationWeingartner et al. 1999). Thus it is possible to sample the entire dried aerosol size distribution inside clouds. The interstitial inlet segregates hydrometeors by a cyclone with a 2 μ m cut-off. It collects only the small non-activated particles inside clouds. Because the interstitial inlet was not heated clogging by snow and riming was possible. Therefore, it was continuously cleaned, which substantially reduces the occurrence of inlet blocking. The few remaining periods where clogging occurred have been identified by a pressure drop measured at the interstitial inlet, and are excluded from analysis. Downstream of both inlets, a similar set of instruments as on the Ice-CVI was connected (CPC, PSAP, SMPS, and OPC results will be shown in this study) to simultaneously measure the same parameters of the total and interstitial particle population (Figure 3). Under cloudless conditions the Ice-CVI inlet was operated as an aerosol inlet by switching off the counterflow, whereby all three inlets sample the same ambient aerosol. These periods were used to intercompare all sensors measuring the same parameter on the different inlets.

4.4. Cloud Microphysical Probes

A Particle Volume Monitor (PVM; Gerber 1991) measured the liquid water content (LWC). Its signal is only qualitatively used in this study, because it is not clear to what extent the PVM response is affected by the presence of ice particles. Images of cloud particles larger than 5 μ m were taken by the Cloud Particle Imager (CPI; Lawson et al. 2001) located on a separate platform 8 m apart and 2 m beneath the Ice-CVI installation. From this information, number size distributions of differently shaped hydrometeors are obtained up to ice particle sizes of 2500 μ m. The CPI analysis is carried out according to CitationConnolly et al. (2007) where the determination of ice particle concentration is significantly improved (by experimentally determining the size-dependent depth of field of the CPI) and the oversizing of ice particles is corrected. Both issues substantially increase the CPI sensitivity for small hydrometeors. All non-spherical particle shapes can be assigned to ice particles whereas the habit type “spherical” could contain liquid and recently frozen drops. The CPI analysis also allows one to infer the IWC for different habit types or for different size classes by means of shape and size dependent parameterizations of the ice particle mass (CitationHeymsfield et al. 2004).

Due to the low wind speeds on the ground-based platform compared to aircraft measurements, droplet and/or crystal shattering at the CPI instrument was minimal. Significant shattering, which could be identified by the CPI-measured ice particle shapes, occurred only during heavy precipitation periods. This was not the case for CPI data presented here.

5.1. Ice-CVI Setup Without VI and PI (Ice-CVI₍₀₎)

The modular design of the Ice-CVI inlet system allows the replacement of the VI or PI or both by cylindrical tubes having a diameter of 4 cm. An Ice-CVI configuration without any pre-impactor (Ice-CVI₍₀₎) was temporarily deployed, because only this setup allows the comparison of the Ice-CVI sampling to the difference of total and interstitial inlet aerosol collection. This implies the simultaneous sampling of small ice particles and supercooled drops under mixed-phase conditions but without a sharp upper cut size and with the possibility of contamination by precipitating and windblown particles. Four measurement examples of the total, interstitial (black and grey dashed lines, upper panels, left scale) and residual particle (black solid line, upper panels, right scale) mean number size distributions are shown in Figure 4. Additionally, the difference of total and interstitial number size distribution (TOT-INT, grey solid line, right scale) is plotted. In the lower panels the size resolved scavenging fraction F_N(d_p) is shown which is calculated in two different ways by Equation (Equation3) (open triangles) and Equation (Equation4) (open circles).

Here, N_tot and N_int are the number concentration of total and interstitial aerosol particles, respectively. F_N(d_p) denotes the fraction of activated particles as a function of particle size. When supercooled drops are present they usually dominate the ice particles in terms of number, so that the residual and TOT-INT number size distribution and F_N(d_p) should be almost completely ascribed to drop residual particles that served as cloud condensation nuclei (CCN). Indeed, the residual particle number size distribution measured with the CVI without VI and PI agrees qualitatively very well with the difference of total and interstitial number size distribution for all events. However, the Ice-CVI₍₀₎ measurements had to be scaled to achieve a good quantitative agreement for the CCN mode above 100 nm, which simultaneously causes the consistency of the differently derived F_N(d_p) curves. From this scaling factors a sampling efficiency for supercooled drops was inferred for each cloud event which is in the range of 5–30% (cf. Figure 4). It is supposed that the high wind speeds at south easterly winds (cf. Table 1), leading to a significantly reduced aspiration efficiency of the omni-directional inlet (Figure 1), in combination with drop freezing on the walls of the long tubes that replace VI and PI in front of the CVI tip, are responsible for the low drop collection efficiencies during the presented cloud events.

In warm clouds F_N(d_p) usually shows a rather steep increase from 0 to 1 in the region of the 50% activation diameter (CitationMertes et al. 2005). From Figure 4 it is obvious that F_N(d_p) is diverse under mixed-phase conditions (esp. during cloud event E-10 and E-11). This deviation has been attributed to drop evaporation caused by the Bergeron-Findeisen process (CitationHenning et al. 2004; CitationSchwarzenböck et al. 2001) and is discussed in detail for the CLACE-3 measurements by CitationVerheggen et al. (2007).

Another feature seen in the TOT-INT as well as in the residual number size distributions (Figure 4) is the existence of particles in the size range between 20 and 50 nm. Residual particles in this size range originating from drops in warm clouds have been observed before (CitationSchwarzenböck et al. 2000). They were explained by highly soluble CCN material and high supersaturations (1% or more) occurring through the orographic effect of a mountain site. However, the scavenging fraction at this diameter range was about 0.05 and decreased for smaller sizes, whereas in the actual study F_N(d_p) reaches values between 0.1 and 0.2 and sometimes increases for smaller diameters. A sampling artifact of the Ice-CVI₍₀₎ is unlikely, because the same characteristic is found for the TOT-INT number size distribution. It might be possible that these small residues are related to the collection of ice particles by both the CVI and the total inlet. This hypothesis will be addressed in section 5.5. Since the sampling efficiency of the Ice-CVI₍₀₎ system for ice particles should be higher than for supercooled drops (due to bounce off instead of freezing at the tube walls), the aforementioned up-scaling of the residual particle size distribution might be too large in the particle size range below 50 nm. This suggestion is supported by the fact that the residual particle size distributions in Figure 4 are systematically higher than the difference of total and interstitial particles in this size range.

5.2. Ice-CVI Setup Including PI but Without VI (Ice-CVI₍₁₎)

In this section the impact of the PI on the Ice-CVI sampling characteristics under mixed-phase conditions is studied. In order to examine the pre-segregation of supercooled drops by the PI we concentrate on non or only slightly precipitating cloud events at low and moderate wind velocities (i.e., even less influence of blowing snow) where a CCN mode was observed in the TOT-INT number size distribution and where the VI was not installed. Figure 5 shows four cloud events where the CCN mode is visible (solid grey lines in the upper panels, more pronounced in the lower two examples) with a maximum between 150 and 200 nm similar to the respective plots in Fig. 4. This time the residual particle number size distribution (solid black lines, upper panels) did not follow the shape of this mode. This is a clear indication that the liquid drops did not pass the PI in contrast to the measurements presented in Figure 4 where the PI was not installed. The residual particle size distributions that are now associated with the collection of ice particles only do not possess a distinct shape but slightly increase to smaller sizes.

Since the TOT-INT number size distribution is composed of residuals from liquid drops and ice particles whereas the Ice-CVI residual number size distribution comprises ice particle residuals only, the condition F_N,res (d_p) ≤ F_N,tot-int (d_p) should be met in the lower panel graphs of Figure 5 over the entire particle diameter range. This constraint is only achieved when calculating the residual size distributions and the resultant scavenging fractions (open diamonds, lower panels) in Figure 5 with an assumed Ice-CVI₍₁₎ sampling efficiency of about 1. Due to the fluctuations of the TOT-INT number size distribution at the small and large diameter edge (caused by subtraction of two similar numbers and low counting statistics) it is impossible to infer an Ice-CVI₍₁₎ sampling efficiency for ice particles more precisely.

5.3. Ice-CVI Setup Including VI but Without PI (Ice-CVI₍₂₎)

The segregation capability of the VI is exemplarily demonstrated in Figure 6. The VI was installed and de-installed during two different cloud events at rather constant meteorological conditions (temperature, wind speed and direction) and continuous precipitation which might include windblown snow. Furthermore, the quite constant signals of total and interstitial particle concentration (Figure 6b, black and grey lines) and in-situ measured LWC (Figure 6d, grey line) imply constant cloud microphysical conditions throughout the selected periods. It is obvious from Figure 6 that the installation of the VI strongly reduces N_res from several tens to below 5 cm⁻³ and the sampled CWC from several hundreds to below 20 mg m⁻³. The effect of using the VI is most clearly seen when looking at the time series of the mean volume diameter of the sampled hydrometeors, D_V, calculated from Equation (Equation5) and shown in Figure 6e,

where ρ_hydro denotes the density of the hydrometeors and CWC is here the cloud water content sampled by the Ice-CVI. A value of 0.91 g cm^{− 3} was chosen for ρ_hydro which represents the solid ice density and is used for small, compact ice particles (CitationHeymsfield et al. 2004). During event E-2a where the VI was installed, the D_V values were below the upper size limit of 20 μ m (Figure 6e) given by the VI (and above the lower limit of 5 μ m given by the CVI), whereas without the VI, D_V was mostly above this limit reaching values of 40 μ m. This implies that the higher residual particle concentration and sampled CWC are caused by large ice aggregates that reach the CVI inlet, because they are not pre-segregated by the VI. These agglomerates release several aerosol particles that they have incorporated in the cloud (due to riming and aerosol scavenging) when they become evaporated inside the sampling system, which artificially reduces D_V according to Equation (Equation5) to the inferred range of 40 μ m. Most likely these are precipitating particles (i.e., larger than 1 mm) that do not follow the stream lines at all but could have trajectories depending on the prevailing wind conditions to infrequently reach and pass the omni-directional inlet. A rough estimation assuming spherical ice aggregates with a diameter of 5 mm yields a number concentration of only 6*10^{− 3} particles liter^{− 1} to be sufficient to explain the average measured CWC of 0.4 g m^{− 3} when the VI was not installed. This confirms the very low but still non-zero passing efficiency of the omni-directional inlet for large ice aggregates and the necessity to operate the VI, since already very few collected large particles would cause a large contamination regarding residual number and CWC.

By the way, during the event E-1b the CWC was below the detection limit of the dew point mirror and no Lyman-α hygrometer (which is able to measure CWC below 10 mg m^{− 3}) was operated. Thus reasonable D_V values for this period could not be calculated.

5.4. Ice-CVI Setup Including VI and PI (Ice-CVI₍₃₎)

The findings presented in the previous sections imply that the Ice-CVI operated in a complete setup (VI and PI installed in front of the CVI) only collects small ice particles. For one sampling period of 10 hours (event E-7) the complete Ice-CVI setup was operated under rather constant meteorological and cloud microphysical conditions. Time series of different parameters during this event are shown in Figure 7. In the middle and lower panel, IWC and residual particle concentration (N_res) sampled with the Ice-CVI₍₃₎ is compared to IWC and ice particle concentration (N_ice) time series derived from in-situ measurements by means of the CPI. The CPI provides size resolved concentrations of differently shaped hydrometeors with a size bin width of 10 μ m. In order to compare the Ice-CVI₍₃₎ measurements to the CPI results, only the first two CPI size bins (5–15 μ m and 15-25 μ m) are used to calculate the ice particle concentration N_ice and IWC. However, it should be kept in mind that the CPI counting efficiency for particles smaller than 10 μ m is underdetermined and is difficult to correct for owing to the discrete nature of the CPI images (10 μ m represents 4 pixels across on the image). The counting efficiency of the next size bin is quantitative due to the CPI data analysis carried out according to CitationConnolly et al. (2007). Besides, the upper size bin limit of 25 μ m is somewhat larger than the Ice-CVI cut size.

Only four particle habits as classified by the CPI analysis software are found in the 5–25 μ m size range: spheroid, small irregular, columnar and stellar, where the latter two had a negligible abundance. The spheroid habit is expected to represent liquid drops but it might contain a number of frozen drops that are still almost spherical. Therefore, it is not clear whether IWC (again using a bulk ice density of 0.91 g cm^{− 3}) and N_ice should be inferred by totally or partially ignoring spherical hydrometeors. Thus, two different IWC and N_ice time series are calculated from the CPI measurements and shown in Figure 7. The solid grey lines disregard all spheroids (CPI_1) whereas the dashed grey curves include spheroids of the larger size bin (CPI_2). The latter is chosen for two reasons. Due to the Bergeron-Findeisen process the liquid drops should shrink under mixed-phase conditions, i.e., spheroids in the smaller size bin might be most likely still liquid and spheroids in the larger size bin have a higher probability to be frozen. Moreover, larger drops freeze more easily in comparison to smaller ones, at least in the immersion freezing mode (CitationDiehl et al. 2006). Referring to the IWC and particle number comparison it is obvious that IWC and N_res sampled by the Ice-CVI₍₃₎ (solid black line) agree well with IWC (CPI_1) and N_ice (CPI_1), although both Ice-CVI₍₃₎ parameters have a tendency to be slightly higher. This is most likely a consequence of the reduced counting efficiency of the CPI in this diameter range. Ice particle shattering inside the PI can be ruled out, because this would only increase N_res but not the sampled IWC. Average values are 1.1 mg m^{− 3} and 0.47 cm^{− 3} for IWC (CPI_1) and N_ice (CPI_1) and 1.2 mg m^{− 3} and 0.78 cm^{− 3} for IWC and N_res obtained from the Ice-CVI₍₃₎ sampling, respectively. This means both devices agreed within 10 and 40% with respect to IWC and N_ice (in the sense of a 1:1 relation to N_res). It is remarkable that the Ice-CVI₍₃₎ results never exceed IWC (CPI_2) and N_ice (CPI_2), respectively. This is in agreement with the assumption that some of the larger spheroids are frozen. The CPI detected on average five times more drops (spheroids) in its first size bin so that Ice-CVI₍₃₎ measurements of IWC and N_res higher than the IWC (CPI_2) and N_ice (CPI_2) would indicate liquid drop sampling which is not observed at any time. Thus, the hypothesis is favored that the few collected spheroids are frozen, which at the same time supports the assumption that most of the spheroids detected by the CPI are liquid.

Regarding the CPI measurements as a “quasi quantitative” reference demonstrates that IWC and N_ice are recovered by the Ice-CVI₍₃₎ at very low absolute IWC and N_res values between 1–5 mg m⁻³ and 0.2–1.2 cm⁻³ in the presented cloud event. This implies that ice particles between 5 and 20 μ m are sampled with a high but not quantifiable sampling efficiency. An independent verification that the Ice-CVI₍₃₎ sampling is strongly linked to residuals of ice particles in the intended size range is given by the mean volume diameter D_V (Figure 7, upper panel), which is within the sampling size limits of the Ice-CVI₍₃₎ during the entire cloud period.

5.5. First IN Characterization by Means of the Ice-CVI

The event E-7 is investigated in detail to characterize IN in the lower free troposphere. The averaged sub-micrometer size distribution of the ice particle residuals is shown in Figure 8 (upper panel, d N/dlogd_p multiplied by 1000) in conjunction with the total particle size distribution. The most striking feature is the increase of the ice particle residual size distribution to small diameters that results in a slight increase of the scavenging fraction (Figure 8, lower panel) below 40 nm. This behavior was also observed in Figures 4 and 5 not only for the Ice-CVI sampling but also for the TOT-INT size distributions implying that the same effect occurred when the ice particles were sampled by the heated total inlet. Due to the good agreement of residual and ice particle concentration demonstrated in the preceding paragraph it can be excluded that the small residuals result from ice particle shattering inside the Ice-CVI system or from particle scavenging by riming or impaction after ice particle formation. However, it is not clear whether the small residual particles acted as IN, because they might originate from condensable matter inside those ice particles, that have been produced by secondary ice formation processes. The existing low temperatures and the absence of small columnar crystals in the first two CPI size bins do not support ice multiplication processes like rime splintering (Hallett-Mossop mechanism) and drop fragmentation during freezing but the mechanical fracturing of falling or windblown ice crystals, especially during snow fall, might be possible (CitationRogers and Yau 1991).

The main increase of the scavenging fraction (Figure 8, lower panel) is observed at large particle diameters starting above 280 nm and well separated by a minimum at intermediate sizes from the small increase for diameters below 40 nm. Therefore, these residuals are not expected to be linked to secondary ice particles and are regarded as IN demonstrating that large particles are preferred to nucleate ice. This statement is confirmed by extending the investigated diameter range to super-micrometer sizes. In Figure 9 (upper panel) the total and residual particle number size distributions determined simultaneously by the two OPCs connected to the total and Ice-CVI inlet, respectively, are added to the SMPS results. It is noticeable that the residual particle size distribution decreases less steeply than the size distribution of the total particles (sedimentation losses of super-micrometer particles in the sampling lines have been not accounted for but should be similar for both inlets). Consequently, the scavenging fraction (Figure 9, lower panel) increases from 0.03 at d_p = 0.8 μ m to 0.09 at d_p = 5 μ m. Although the absolute majority of IN were found below 1 μ m, only every thousandth sub-micrometer but already every tenth super-micrometer particle acted as IN. This finding is restricted to particle diameters up to 5 μ m, since larger non-activated particles could enter the CVI.

It is interesting to note that the black curve in Figure 9 (upper panel) looks very similar in shape and number to residual number size distributions measured from aircraft (CitationSeifert et al. 2003) in pure ice clouds (excluding cirrus) above 6 km and at temperatures below −30°C.

The chemical characterization of the ice residuals was carried out in different ways. Mass size distributions of soluble aerosol substances were measured by the AMS. Unfortunately, the AMS was not coupled to the Ice-CVI during event E-7. However, it was measuring downstream of the CVI in a preceding period inside the same cloud (event E-6). The Ice-CVI was operated in an Ice-CVI₍₁₎ setup, i.e., that supercooled drops were pre-segregated but a contribution by precipitating ice aggregates could not be completely ruled out. In Figure 10 the single and summed up AMS mass distributions are shown simultaneously with the total residual mass concentration distribution derived from the SMPS measurements assuming spherical particles and a dry particle density ρ_p. For a direct comparison, the SMPS distributions have been converted from the mobility diameter d_p into a vacuum aerodynamic diameter d_va measured by the AMS using d_va = ρ_p/χ d_p (CitationSchneider et al. 2006) where χ denotes the dynamic shape factor which is assumed to be unity. Two SMPS mass size distributions are shown in Figure 10, which were calculated for a lower and upper limit of ρ_p with 1.7 and 2.5 g cm^{− 3}, whereas the latter is a typical value for pure mineral dust particles (CitationLinke et al. 2006). Mass distributions assuming intermediate dry particle densities (e.g., black carbon with ρ _p≈ 2 g cm^{− 3}) are located in between the two SMPS curves. It is interesting to note that the SMPS distribution assuming the lower ρ_p is located above the second one indicating that the increase in mass due to a higher density is over-compensated by the shift to a larger d_va. However, both SMPS derived total residual mass distributions begin to increase and to significantly deviate from the residual mass distribution determined by the AMS at d_va values between 400 and 500 nm. It is unlikely that this deviation is caused by aerosol mass entering the Ice-CVI due to precipitating ice aggregates, because the aerosol particles that might be incorporated by riming or impaction scavenging should have the chemical composition of the total aerosol. This population is mainly composed of substances that are observable by the AMS (sulfate, nitrate, ammonium, and volatile organic matter). Therefore we conclude that the missing mass is related to IN that consists of refractory compounds not vaporizable at 600°C and thus not detected with the AMS. Indeed, the mass concentration obtained from the SMPS is well above the detection limit of the AMS (CitationWalter et al. 2006). This means, that if the sampled residual mass had mainly consisted of non-refractory particles it would have been detected by the AMS. The conclusion that refractory substances constitute the IN is also supported by the fact that the deviation of SMPS and AMS residual mass occurs in the same diameter range (in terms of d_p) as the increase of the IN scavenging fraction (Figure 8, lower panel), although it should be considered that both observations were not made at the same time. From Figure 10 it might be additionally suggested that organic matter (OM) is present in the residual particles above d_p = 200 nm accounting for about 15% of the sub-micrometer IN mass. However, this absolute OM fraction should be handled with caution, because the inferred total OM mass of the residual particles is close to the theoretical AMS detection limit.

Further indication of the chemical composition of the sampled IN during E-7 (Ice-CVI₍₃₎ setup) are obtained from the results of two PSAPs that were connected to the total and Ice-CVI inlet. The PSAP downstream the Ice-CVI was operated with a maximum integration time of 5 min and a high sample flow rate of 4 liter min⁻¹. In combination with the CVI enrichment by a factor of 7 a detection limit of about 0.2 ng m⁻³ was achieved. Mean BC mass concentrations of 0.7 and 79.1 ng m⁻³ were derived for the residual and total aerosol particles, respectively. Since BC is present in the sub-micrometer particles, the BC mass concentrations can be related to the mean residual and total particle mass of 3.9 and 1850 ng m⁻³ inferred from the integral over the corresponding SMPS mass distributions (assuming ρ_p = 2.5 g cm⁻³). Fractions can be described in two different ways: First, the fraction of BC mass to SMPS derived mass is 18% for the residual while it is only 4% for the total aerosol reservoir, respectively. Second, 0.9% of the total aerosol BC but only 0.2% of the total aerosol SMPS mass is found in the residuals. Both observations suggest that BC is enriched in the residual particles compared to the totally abundant particles, which implies that BC containing particles may have served as IN inside the examined mixed-phase cloud.

Nevertheless, only a fraction (about 30%) of the IN mass inferred from the SMPS results could be explained by OM and BC. Moreover the main IN mass is found in the super-micrometer size range according to the OPC number size distribution (Figure 9, upper panel). Assuming again ρ_p = 2.5 g cm⁻³ and a spherical particle shape yields an IN mass of 11.9 ng m⁻³ for residual particles above the SMPS size range. Although the assumption of sphericity for super-micrometer particles is not very realistic, this estimation reveals a factor of three more mass in the super-micrometer compared to the sub-micrometer residual particle size range. Hence, the postulated existence of refractory particles derived from the AMS results and the observed existence of an excessive fraction of large particles (in relation to the total aerosol population) in the residual particles leads to the suggestion that mineral dust particles are a strong candidate for being IN.

Siliceous particles were indeed found as the dominant component in the super-micrometer (19 of 23 analyzed particles) and as an abundant component in the sub-micrometer size range (19 of 125 analyzed particles) by the single particle analysis with ESEM (Figure 11, all four particles in the upper part), which confirms the importance of mineral dust particles as IN in the lower and middle troposphere. This is in agreement with the glaciation of supercooled altocumulus clouds by Saharan dust layers observed by remote sensing techniques (CitationAnsmann et al. 2005; CitationSassen et al. 2003). In addition, the ESEM analysis (Figure 11, lower part) verifies the dominance of smaller (d_p mainly between 0.3–1 μ m) carbon rich particles (106 of 125 analyzed particles) that are at least non-volatile until 300°C which is consistent with OM and BC found in the residual particles by means of AMS and PSAP. The morphology of the carbon rich particles rules out soot particles as part of the residuals but rather suggests organic particles, which is further supported by the C and O peak in the EDX spectrum shown in Figure 11. The large C and small O signal might imply a contribution of elemental carbon which was most likely seen as BC by the PSAP.

However, Si and C particles are by far the most prominent particle types sampled by the ESEM impactor downstream of the Ice-CVI during event E-7 (Ice-CVI₍₃₎ setup). From the ESEM analysis, it was additionally verified that no significant contamination by abraded particles (stainless steel, brass, or aluminum) from the Ice-CVI itself could be observed for the Ice-CVI₍₃₎ setup. This confirms that abrasion does not occur as already argued in section 3 (sampling issues).