Introduction
Ice formation in clouds influences the life-time of clouds and precipitation as well as solar and thermal radiation (Lohmann and
Feichter, 2005). There are two major pathways by which ice is formed in the atmosphere: homogeneous and heterogeneous
freezing. The spontaneous formation of ice inside droplets at temperatures of lower -37 ∘C and without any catalysts is
called homogenous freezing (Cantrell and Heymsfield, 2005) and occurs in high clouds at a saturation similar to that of liquid
water (Koop et al., 2000). Heterogeneous freezing needs particulate catalysts to initiate freezing at temperatures higher than
-37 ∘C in mixed-phase clouds. These catalysts are called ice nucleating particles (INP). There are four different
heterogeneous ice formation modes (Pruppacher and Klett, 2010): deposition nucleation, condensation freezing, immersion freezing
and contact freezing. Because heterogeneous freezing starts at higher temperatures than homogeneous freezing, ice crystals, which
are formed by ice nucleating particles grow faster (by diffusion) and perhaps precipitate first (DeMott et al., 1998).
Ice formation is very selective because of the low number concentration of particles that can act as INP. Roughly about one in
105 atmospheric particles can act as INP (Rogers et al., 1998; DeMott et al., 2010). The ice nucleation is induced at some
point on the surface of the nucleating particle and the ice nucleation ability of different types if aerosol particles is usually
described either by an ice active-site density approach or by classical nucleation theory involving a contact angle (see
e.g. Steinke et al., 2014, and references therein). Several laboratory and field studies have shown that certain mineral dusts
belong to the most important ice nucleators in the atmosphere (e.g. DeMott, 2003; DeMott et al., 2003; Cantrell and Heymsfield,
2005; Kamphus et al., 2010; Hartmann et al., 2011; Murray et al., 2011, 2012; Hoose and Möhler, 2012; Atkinson
et al., 2013; Diehl et al., 2014). Also biological particles like bacteria, pollen and spores (e.g. Diehl et al., 2001; von Blohn
et al., 2005; Möhler et al., 2007; Pratt et al., 2009; Prenni et al., 2009; Diehl and Wurzler, 2010), soot (e.g. Cozic et al.,
2008a; Pratt and Prather, 2010; Pratt et al., 2010) and effloresced salts (Abbatt et al., 2006; Wise et al., 2012) have good ice
nucleation capabilities. There is also some evidence that glassy organics (Froyd et al., 2010; Murray et al., 2010), porous
particles (Adler et al., 2013) and lead-containing particles (Cziczo et al., 2009; Ebert et al., 2011) have favorable ice
nucleation qualities.
Direct measurements of the chemical composition of single INP or IPR (Ice particle residuals = particles remaining after the ice
of cloud ice particles is evaporated) are important to better understand ice formation in mixed-phase clouds, to improve the
prediction of precipitation, and to evaluate the influence of anthropogenic aerosol on these processes. Measurements of INP/IPR
require instrumentation that can separate the INP/IPR from the background or interstitial aerosol particles and supercooled drops
and measure the chemical composition on-line. In several field studies a combination of a CFDC (Continuous Flow Diffusion Chamber;
Chen et al., 1998), a CVI (Counterflow Virtual Impactor; Cziczo et al., 2003) which separates the ice
crystals from background aerosol and a Single Particle Mass Spectrometer (SMPS; DeMott et al., 2003; Cziczo et al., 2003, 2004;
Cziczo, 2004; Pratt et al., 2009; Corbin et al., 2012) to measure ice nucleating particles were deployed. Previous measurements of
INP and IPR (on-line and off-line) conducted at the Jungfraujoch showed that particles consisting of mineral components dominate
the ice particles number (Kamphus et al., 2010), but also particles containing black carbon (Mertes et al., 2007; Cozic et al.,
2008a) and lead (Cziczo et al., 2009; Ebert et al., 2011) were found to be enriched in IPR.
Here we report on measurements conducted with the single particle instrument ALABAMA (Aircraft-based Laser Ablation Aerosol Mass
Spectrometer; Brands et al., 2011) to analyze the chemical composition of IPR sampled by two unique different ice sampling
inlets. The goal of these measurements was to investigate the chemical composition of IPR in ambient mixed-phase clouds in upward
transported boundary layer air and free tropospheric air at the high-alpine research station Jungfraujoch (JFJ;
3580 ma.s.l.) in Central Europe. Additionally, background aerosol was measured by the use of a total inlet during cloud
free conditions.
Experimental
Measurement location and meteorological conditions
The intensive field campaign INUIT-JFJ/CLACE 2013 in January/February 2013 at the High Alpine Research Station Jungfraujoch (JFJ,
Swiss Alps; Sphinx Laboratory, 3580 ma.s.l.; 7∘59′2′′ E,
46∘32′53′′ N) took place in the framework of the DFG (Deutsche Forschungsgemeinschaft)-funded
research unit INUIT (Ice Nuclei research UnIT) and the Swiss National Science Foundation-funded Interaction of aerosols with Clouds and Radiation project (see also overview paper by Schneider et al., this issue). The IPR were sampled out of
mixed-phase clouds by the Ice-CVI (Ice Counterflow Virtual Impactor; Mertes et al., 2007) and the ISI (Ice Selective Inlet;
Kupiszewski et al., 2014) and fed into the ALABAMA single particle mass spectrometer (Brands et al., 2011). During cloud free
times the ALABAMA was connected to the heated total-inlet to sample the background aerosol particles. The Ice-CVI and the
total-inlet were located in a central position on the roof of the Sphinx Laboratory and the ISI was located at the railing at the
eastern edge of the roof. A short description of the instruments is given in Sects. 2.1 to 2.4. The meteorological conditions
during the campaign along with basic information on the encountered cloud types are summarized in Table 1.
ALABAMA (Aircraft-based Laser ABlation Aerosol MAss spectrometer)
The ALABAMA (Brands et al., 2011) consists of three sections: aerosol inlet, sizing region and desorption/ionization region. The
aerosol inlet is an aerodynamic lens (Liu-type; Liu et al., 1995a, b, 2007; Kamphus et al., 2008) which focuses aerosol
particles to a narrow beam by utilizing 6 apertures with decreasing diameter. The particles are separated from ambient air and
accelerated to 50–100 ms-1, while entering the vacuum-system. For optimal sampling conditions the pressure in front
of the aerodynamic lens should be between 0.5 and 5 hPa and for this a critical nozzle is used, which limits the flow to
80 cm3min-1 under standard working conditions and which reduces the pre-pressure of the lens to
3.8 hPa. During this campaign, the critical orifice was constructed by a constricted O-ring such that also under
Jungfraujoch conditions (ambient pressure around 630 hPa) the lens pressure was regulated to best operating
conditions. A skimmer separates the first and the second pumping stages. Two continuous wave detection lasers, which are
orthogonal to the particle beam, are located in the second pumping stage. As a modification of the instrument described in Brands
et al. (2011), two UV laser diodes (Blu-Ray laser; InGaN, 405 nm) are now used. The particles pass through the laser beams
and the scattered light is reflected by an elliptical mirror and detected by a photomultiplier tube (PMT). The vacuum aerodynamic
particle size (DeCarlo et al., 2004) can be determined from the velocity of the particles by calibration with particles of known
size. In this section the ablation laser (pulsed Nd-YAG-Laser, 266 nm; 6–8 mJ per pulse, 5.2 ns per pulse,
max. 21 Hz) is triggered. If one particle passes through both laser beams, the in-house designed and built electronic
control system sends out a trigger signal to the ablation laser, the laser shoots and ionizes the particles. The ions are
separated in the Z-shaped bipolar time-of-flight mass spectrometer (TOFWERK AG, Switzerland) by their mass-to-charge ratio. In
principle both signals (for positive and negative ions) can be detected using a multichannel plate (MCP). However, for this study
only positive ions were available due to technical issues. The average mass resolution m/Δm of the Z-ToF mass
spectrometer ranges between 100 for low m/z values and 200 for larger ions (Brands et al., 2011).
Additionally the particle size distribution was measured by
a Sky-OPC (Optical Particle
Counter; Grimm 1.129) integrated
into the ALABAMA instrument upstream of the aerodynamic lens covering a size
diameter range from 0.25 to 52 µm. Due to the size limitation by the inlet system
a maximal particle size of approximately 3 µm was
measured. Both instruments sampled through
1/4′′ stainless steel tubes with different lengths (ISI =370 cm; Ice-CVI =126 cm; total =261 cm).
Particle losses inside the sampling tube were calculated with the Particle
Loss Calculator (von der Weiden et al., 2009). In
the size range between 200 and 500 nm the transmission efficiency is
about 99 % for all three inlets. The larger the
particles and the longer the tubes, the higher are the
losses inside the tubes. For a size of 3 µm the
transmission efficiency is between 90 and 45 %.
The switching between the three different inlets was performed
manually.
Ice-CVI (Ice Counterflow Virtual Impactor)
The Ice-CVI is designed to sample IPR from mixed-phase clouds. A detailed description and instrumental characterization is
provided in Mertes et al. (2007), thus it is only briefly described here. The Ice-CVI uses an omnidirectional inlet, which removes
particles larger than 20 µm from the aspired air. The separation of the particles larger than 20 µm out of
the sample flow is done with a virtual impactor (VI) downstream of the inlet. After that the supercooled droplets (>5 µm) are removed by a pre-impactor (PI). The PI consists of impaction plates colder than 0 ∘C. The
supercooled droplets freeze on these plates upon contact while the ice crystals bounce off and remain in the sample
flow. Subsequently, a counterflow virtual impactor (CVI, Mertes et al., 2005) rejects the interstitial aerosol particles so that
only small ice crystals (i.e. those with aerodynamic size diameters between 5 and 20 µm) remain in the sample flow. It
is assumed that such small ice crystals have very recently formed in the close vicinity to the measurement site. For complete
sublimation, the small ice particles are injected into particle-free and dry carrier air inside the CVI. The aerosol particles
that remain after evaporation/sublimation of the ice are termed IPR. The walls of the evaporation tube as well as the carrier air
are at room temperature. The IPR can be transferred to different aerosol instruments for physico-chemical characterization. The
sampling principle of the CVI leads to an enrichment of the collected particles. This enrichment (factors between 5 and 10) is
given by the ratio of the velocity upstream and downstream the CVI inlet tip.
A CPC (Condensation Particle Counter, model 3010, TSI Inc.) was connected downstream of the Ice-CVI at a suitable position for
measuring to determine the total particle number concentration of the IPR. Significant non-zero particle number concentrations
indicate the presence of ice particles in mixed-phase clouds.
ISI (Ice Selective Inlet)
Extracting IPR from mixed-phase clouds was also possible with the ISI (detailed description in Kupiszewski et al., 2014). Small
ice crystals (the IPR within are assumed to be representative of the original ice nuclei) are segregated from larger ice crystals,
supercooled droplets and interstitial aerosol particles by a combination of four inlet components. The first component is the
omnidirectional inlet, which is shielded from above in order prevent large precipitating particles entering the
inlet. Subsequently, a custom-made cyclone (BGI Inc.) removes ice crystals, with aerodynamic diameters of ca. 20 µm
(D50 cut size, i.e., the particle diameter at which 50 % of incoming particles are removed from the sample flow) and
larger. Further downstream the removal of supercooled droplets takes place in a custom built ice-coated droplet evaporation
chamber using the Wegener–Bergeron–Findeisen process. The key difference to the Ice-CVI is that here the separation of
supercooled droplets from ice crystals is conducted in the airborne state via evaporation. Finally, the interstitial aerosol and
the cloud condensation nuclei (CCN) released during droplet evaporation are removed by a pumped counterflow virtual impactor
(D50≈ 5 µm). The extracted ice crystals with aerodynamic diameter between 5 and 20 µm are
heated and the released IPR are transferred into the laboratory for physical and chemical characterization using state of the art
single particle aerosol instrumentation (single particle soot photometer (SP2), ALABAMA single particle mass spectrometer,
Wideband Integrated Bioaerosol Sensor (WIBS) and a Grimm optical particle counter). The ISI additionally allows for simultaneous
counting, sizing and imaging of the hydrometers contained in the cloud with the use of two WELAS (white light aerosol
spectrometer) aerosol sensor system (Palas GmbH) and a Particle Phase Discriminator (PPD). These are placed behind the cyclone
(first WELAS sensor) as well as behind the droplet evaporation unit (second WELAS sensor and PPD). The ISI, together with the
downstream aerosol instrumentation, thus provides number size distribution of sampled hydrometers and aerosol particles,
information on ice crystal shape and surface roughness, as well as number size distribution and chemical composition of ice
residual particles.
The ISI also enriches the collected particles inside the pumped counterflow virtual impactor with an enrichment factor between 1
and 7.
Further instruments
Particle Volume Monitor (PVM; Gerber, 1996): The PVM measured the liquid water content (LWC = total liquid water
present in a cloud) and the surface area of cloud droplets (e.g., Wendisch, 1998). Cloud droplets cross a laser beam and the
scattered light is measured. Elevated LWC values are used to identify cloud events, using a threshold value of
20 mg m-3. During the measurement campaign two PVMs were operated by PSI (Paul Scherrer Institute) and the University of
Manchester.
Cloud Droplet Probe (CDP): The CDP is a forward-scattering spectrometer for determination of particles sizes and
concentrations of cloud droplets. Counting and sizing of the particles is done by detecting light scattered by the particles (for
more details see for example Lance et al., 2010).
Scanning electron microscopy (SEM): Ice particle residuals (IPR) were collected by a two stage impactor system (50 %
cut-off aerodynamic diameters 1.0 and 0.1 µm, respectively) on transmission electron microscopy grids and elemental
boron substrates and analyzed by scanning electron microscopy (SEM, FEI Quanta 200 FEG, FEI Eindhoven, the Netherlands) and
energy-dispersive X-ray fluorescence (EDX, EDAX, Tilburg, the Netherlands) to characterize the particles with regard to their
chemical composition, morphology, size, internal mixing state and electron beam stability (volatility). (For further details see
Kandler et al., 2011). Here we only present the size distribution of IPR measured by SEM and compare them to the IPR sizes
obtained by the ALABAMA and the sky-OPC. Detailed results on the chemical composition analysis of IPR, INP, and background aerosol
by SEM are given in Worringen et al. (2014).
ALABAMA data evaluation
The data evaluation was done with the software package CRISP (Concise Retrieval of Information from Single Particles; Klimach,
2012) based on the software IGOR Pro (Version 6, WaveMetrics) and following the procedure described in Roth (2014). This software
package includes mass calibration, conversion of the mass spectra into “stick” spectra by integration over the peak width, and
various procedures to sort the mass spectra into groups or clusters of similar spectra. This sorting can be done manually, by
means of cluster algorithms (fuzzy c-means or k-means), or by searching for certain parameters (e.g., ion
signals at a certain m/z ratio).
The Clustering depends to some degree on the interstitial settings of the algorithm and its parameters. The start parameters are
Normalization type and time, Preprocessing type, Initialisation type, the number of clusters
and the start cluster difference. Using the fuzzy c-means algorithm (e.g. Bezdek et al., 1984; Hinz et al.,
1999; Huang et al., 2013) the parameters fuzzifier and fuzzy abort have to be set accordingly. The cluster
parameters are explained in detail in the Appendix.
As only the cations were detected during the INUIT-JFJ/CLACE 2013 campaign the grouping of the clusters into different particle
types by using certain marker peaks was difficult in some cases. At the end, seven different particles types were extracted: black
carbon (BC), organic carbon (OC), internal mixture of black carbon and organic carbon (BC/OC), lead containing particles,
industrial metals, BioMinSal (containing possible bioparticles, minerals or salts), and minerals. The average mass spectra of each
particle type are shown in Fig. 1. The association of the mass spectra to the individual particle types is explained in detail
below. The error limits of the number of mass spectra per particle types was estimated using counting statistics.
Organic fraction: black carbon (BC) and organic carbon (OC)
The clusters categorized as BC are due to the presence of the typical fragmentation of Cn+ (Dall'Osto, 2004; Pratt and
Prather, 2010) of higher mass-to-charge ratios (see Fig. 1a). In comparison to the black carbon, the organic carbon cluster
(Fig. 1b) shows different fragmentations of aromatics ([C4H3]+ (m/z 51); [C5H3]+ (m/z 63);
[C6H5]+ (m/z 77); [C7H7]+ (m/z 91), Dall'Osto and Harrison, 2006; Pratt and Prather,
2010), amines ([CHN]+ (m/z 27); [C3H8N]+ (m/z 58), Pratt and Prather, 2010) or other organic material
([C2H3]+ (m/z 27); [C3H]+ (m/z 37); [CH3O/CHNO]+ (m/z 43);
[C4H2/C3N]+ (m/z 50); [C3H7O]+ (m/z 59);
[C11H10/C8H16NO]+ (m/z 142), Pratt and Prather, 2010). Black carbon internally mixed with
organic (BC/OC) shows both fragmentation types of [Cn]+ and [CnHmNO]+ (Corbin
et al., 2012; Fig. 1c).
Lead containing particles
Previous measurements at the Jungfraujoch showed that lead-containing particles can be found in IPR (Cziczo et al., 2009; Ebert
et al., 2011). The cluster of lead containing particles shows the typical isotope pattern of lead (m/z 208, 207, 206, 204;
Fig. 1d).
Industrial metals
The cluster classified as “industrial metals” (Fig. 1e) shows metal signals from anthropogenic (urban/industrial) emissions like
[Cr]+ (m/z 52, 53, 54, 50), [Co]+ (m/z 59), [V]/[VO]+ (m/z 51/67), [Fe]+ (m/z 54, 56), [Mn]+
(m/z 55), [Ni]+ (m/z 58, 60, 61, 62) and [Zn]+ (m/z 64, 66, 68) (De Foy et al., 2012). Chromium and nickel
containing particles might also be due to contamination from stainless steel tubes. However, due to no clear evidence of
contamination these particles were not excluded from the further analysis.
BioMinSal cluster
The spectra in this cluster type contain only metal cations and can therefore not be unambiguous be assigned to a certain
particle type. The spectra in the BioMinSal cluster show a mixture or single signals of [Na]+ (m/z 23), [Al]+
(m/z 27), [K]+ (m/z 39, 41), [Ca]+ (m/z 40) and/or [Fe]+ (m/z 54, 56). An example of an average spectrum of
this particle type is shown in Fig. 1f.
The interpretation of this particle cluster is difficult. The signals are most likely cations of different kinds of salts
(Ka2SO4 or KNO3 are more conceivable than sea salt), minerals or primary biological particles, but without
corresponding anion an unambiguously assignment to one particle type is not possible. Previous investigations of biological
aerosol have shown that metal signals like [Na]+ (m/z 23), [K]+ (m/z 39) or [Ca]+ (m/z 40) dominate the
cation spectra (e.g. Fergenson et al., 2004; Pratt and Prather, 2010; Frank et al., 2011). During laboratory studies with test
aerosol we also observed mainly single metal signals of sodium and potassium in biological particles, but as well in mineral dust
particles. Also iron (Fe+ m/z 54, 56), which is a typical marker for mineral dust, was observed to occur in some biological
particles (maybe due to hemoglobin or due to other ions with an m/z of 56). For these reasons the cluster “BioMinSal” cannot
be further differentiated in this study.
Minerals
Typically marker peaks of minerals are: [Li]+ (m/z 6, 7), [Mg]+ (m/z 24, 25), [Al]+ (m/z 27),
[Si]/[SiO]+ (m/z 28, 29, 30/44), [Ca]+ (m/z 40), [Ti]/[TiO]+ (m/z 48, 46, 47, 49, 50/64) and/or
[Fe]+ (m/z 54, 56) (Trimborn, 2002; Dall'Osto, 2004; Dall'Osto et al., 2010) Clusters containing at least three of the
above marker peaks were classified as minerals (Fig. 1g).
Others
The cluster others includes all particle mass spectra, which are very noisy or could not be unambiguously assigned to any
of the above cluster types.
Results and discussion
During the INUIT-JFJ/CLACE 2013 campaign we measured 71064 background aerosol particles during a measurement time of 217 h
and 1808 ice particle residuals during a measurement time of 256 h: 1664 spectra behind the Ice-CVI during 221 h
and 144 spectra behind the ISI during 35 h.
Ice particle residual composition and air mass trajectories
The ALABAMA was used to analyze IPR extracted from mixed-phase clouds by two different inlets: ISI and Ice-CVI. The times
during which cloud particles were sampled by the ISI were identified using the cloud liquid water content (LWC)-data measured with
two PVM instruments (PSI and University of Manchester) and a CDP (University of Manchester) while the times when the Ice-CVI
sampled cloud particles were identified using the ice residual concentration measured with the CPC behind the Ice-CVI. The Ice-CVI
cloud events were classified according to the sample efficiency and properties of the Ice-CVI. This classification resulted in
(++)-events, where mainly IN were measured; (+-)-events, where the measurements were possibly influenced by secondary ice or
contamination by fragments of incoming snowflakes; and (- -)-events, where hardly any ice residuals were detected. The (- -)-events
are not included in the following analysis. Table 1 shows the identified cloud events with measurement time, number of acquired
mass spectra, event classification and weather conditions and cloud type. The meteorological situation during the whole campaign
was mainly influenced by low pressure systems with low temperatures.
The occurrence of artifacts during the measurements using the Ice-CVI as well as the ISI is described in Worringen
et al. (2014). Aluminum oxide is an artifact of the measurements with the Ice-CVI and silicon oxide is an artifact of the
measurements with the ISI. The impaction plates of the Ice-CVI contain aluminum. When the ice crystals bounce on the impaction
plates of the Ice-CVI, some aluminum oxide particles are removed from these plates and can be carried with the air stream to the
analysis instrument. The ISI was calibrated with silicon oxide particles, which remained in the system and occurred during the
measurements as artifacts. In our dataset it is difficult to distinguish between artifacts and a real component of a particle.
Therefore the peaks of m/z 27 (Al+) and m/z 43 (AlO+) with respect to the Ice-CVI measurements and the peak of m/z
28 (Si+) with respect to the ISI measurements were ignored.
In total, we measured 329 spectra during the (++)-events (61 h measurement time) and 1335 spectra during the
(+-)-events (160 h measurement time) by using the Ice-CVI. Behind the ISI we measured 144 spectra during 35 h
measurement time.
Figure 2 shows the results of the clustering algorithm for all ice particle residual measurements. The Ice-CVI data were analyzed
separately for all (++)-events (left pie chart) and for all (+-)-events (middle pie chart). The right pie chart shows the
results for all ISI measurements. This analysis shows that the minerals group is one of the dominating fractions in IPR
composition measured by using the Ice-CVI during the (+-)-events (21 %) and the ISI (33 %). Interestingly, during the
(++)-events minerals represent only a smaller fraction (6 %). We also detected lead containing particles, but only while
sampling through the Ice-CVI (12 % in the (++) cases and 11 % in the (+-) cases). During the (++)-events measured
behind the Ice-CVI, black carbon internally mixed with organic carbon (31 %) is one of the main fractions. Together with OC
(also 31 %), these two particle types represent 62 % of all IPR. It is worth noting that there are more similarities
between the Ice-CVI and ISI during (+-)-events than the (++)-events. A reason for that can be that the ISI measurements are
also influenced by secondary ice or incoming snowflakes. During the (+-)-events as well as during the ISI-events precipitation
occurred. On the other hand, the measurements at the ISI and the Ice-CVI were not done simultaneously, such that also differences
in the air mass origin could play a role here.
Backward trajectories were calculated with CRISP (Klimach, 2012) with access to HYSPLIT (Hybrid Single Particle Lagrangian
Integrated Trajectory Model, National Oceanic and Atmospheric Administration; Draxler and Rolph, 2012, 2014), which was run using
GDAS (Global Data Assimilation System) meteorological dataset. The geographical origin of the air masses and change of altitude
during the transport can be used to infer whether the air masses are influenced by surface emissions or mainly characterized by
free tropospheric air and long distance transport. During all measurement episodes (total, Ice-CVI, ISI) the air masses approached
the measurement station from slightly different directions (Fig. 3). The air masses during the (++)-events came primarily from
north-western directions from North America over France and during the (+-)-events additionally more from the North from
Scandinavia and Germany. The origin of the air masses during the measurements using the ISI are similar to those of the
(++)-events but with a tendency towards southern directions (Fig. 3). During non-cloud episodes air masses approached the
measurement location from all directions.
Previous measurements at the JFJ showed that there is an enrichment of mineral dust and black carbon in IPR composition (Mertes
et al., 2007; Cozic et al., 2008b; Kamphus et al., 2010). An enrichment of black carbon is not observed in our
measurements. Cziczo et al. (2009) found that lead-containing particles are also enriched in the IPR compared to the background
aerosol. Our measurements show also that mineral dust (measured behind the ISI and during the (+-)-events) and lead (only
measured behind the Ice-CVI) is one of the dominating fractions in the IPR composition. Worringen et al. (2014) report on SEM
analyses of particles sampled during the same project behind the ISI and the Ice-CVI. They found lead in two states, namely
homogeneous lead-rich particles and small lead inclusions in particles of other material. While the small lead inclusions are not
considered as potential artifacts, the large lead-rich particles might be artifacts from mechanical re-suspension of lead
containing particles from the surface of the impaction plates of the Ice-CVI. In our case the lead-containing particles are
internally mixed with organics, minerals or other metals, therefore we interpret these particles as real atmospheric particles and
not as artifacts.
Event-to-event variability
Figure 4 shows the IPR chemical composition measured during five different events. Although the air masses sampled during all five
events have very similar origin and history of altitude (Fig. 5a and b) the chemical composition for each event is different from
the others. All air masses during these events arrived rising from a western direction to the measurement platform, such that the
chemical composition of these five events was probably influenced by local emissions. During Event 3 and 4 the air masses arrived
from a south-western direction. This is reflected in the chemical composition: both events show organic carbon as the main
fraction and the amount of lead containing particles is comparable (13 % Event 3 and 15 % Event 4). The differences
between Event 3 and 4 are the absence of BC in Event 4 and of minerals in Event 3. The chemical composition of Event 12 is also
comparable to Event 4. In this case also the dominating fractions are OC, BioMinSal and minerals. The absence of lead containing
particles and presence of 3 % BC in Event 12 are the main differences between Event 12 and Event 4.
Another reason for similarities and differences between these 5 events can be the temperature (Fig. 6). Events 3 and 4 have the
highest fraction of OC (nearly 50 % of the particles) and the highest temperatures (Event 3: average temperature
T‾=-7.8 ∘C; Event 4: T‾=-8.6 ∘C). In contrast to that, Event 12 has the highest amount
of minerals and the BioMinSal-type (64 %) and has an average temperature of about T‾=-17 ∘C. Event 4 has
also a high amount of “metal” components (minerals and BioMinSal 38 %) but by closer consideration of the BioMinSal-type it
can be shown that this type consists mostly of metal ions together with organic ions in Event 4 and only metal ion signals during
Event 12. Therefore the data indicate, that organic material dominates the IPR composition at higher temperatures and metal
components from mineral or salty origin dominate at lower temperatures. This finding is surprising, because previous studies found
that organics are ice active only at temperatures below -30 ∘C (Hoose and Möhler, 2012). However, the organic
signal in the mass spectra might also originate from bioparticles, which would explain the ice formation at higher temperatures,
because bioparticles have been reported in the literature by many authors to be ice active at higher temperatures as previous
measurements show (e.g. Möhler et al., 2008; Hoose and Möhler, 2012). An exception is Event 1. During this event the
lowest temperatures (T‾=-19.2 ∘C) of all events prevailed, but a high relative amount of organic components
was measured (80 % BC/OC and OC). A reason for this high relative amount of organic material can be local emissions. At this
time the total particle data show a short period with high concentrations (>2000 cm-3, see Fig. 7), which points to
local emissions.
Comparison between background aerosol particle composition and IPR
During non-cloud phases the ALABAMA was connected to the total inlet to measure background aerosol particles. Figure 8 shows the
averaged composition of all background aerosol particles along with the IPR composition measured behind the Ice-CVI during the
(++)-events and behind the ISI. In comparison to the IPR composition, the measurements show a high fraction of black carbon
(18 %; Fig. 8) and only a minor fraction of BC/OC (4 %). The fraction of the particle type “BC/OC” in the background
aerosol is comparable to that in the IPR sampled by the ISI. However, also the aerosol particles are dominated by the particle
types “organic carbon” (42 %) and “BioMinSal” (30 %). Previous high mountain-top measurements at the Storm Peak
Laboratory (SPL; 3200 ma.s.l.) in northern Colorado show also organic as the major component in the background aerosol
(DeMott et al., 2003; Cziczo, 2004). The differences as compared to the IPR compositions are the negligibly small amount (<0.1 %) of lead-containing particles and the much smaller amount of minerals (about 1 %) than observed in the IPR. This is
partly in contrast to other measurements. Previous measurements with the ATOFMS at the Jungfraujoch as well as airborne
measurements over North America show 5 and 10 % lead-containing particles in the background aerosol (Murphy et al., 2007) but
also only a small amount of mineral dust or fly ash measured at the Storm Peak Laboratory (DeMott et al., 2003).
Previous investigations show that the presence of a high potassium peak in the mass spectra in combination with organic carbon is
a marker for biomass burning (e.g. Pratt and Prather, 2010; Twohy et al., 2010; Corbin et al., 2012) or biogenic aerosol (e.g.
Trimborn, 2002; Pratt and Prather, 2010). Closer consideration of the three different carbon clusters (BC, BC/OC, OC) shows that
potassium is present in nearly every mass spectrum. In the background aerosol 45282 mass spectra including black or organic carbon
were measured: 92 % of all these mass spectra show potassium. The BC-type includes the mass spectra without potassium
(20 %; only 1 % in the OC/BC-type and 6 % in the OC-type). The same observation holds for the IPR composition: the
particle types including organic carbon measured by the use of the Ice-CVI as well as the ISI show more spectra including
potassium (94 % of all OC and OC/BC spectra while using the Ice-CVI and 98 % while using the ISI). Because of these results
spectra with a potassium peak cannot clearly be assigned to biomass burning or biological particles.
Influence of the air mass trajectories on the aerosol composition
The chemical composition of the background (out-of-cloud) aerosol has been shown in Fig. 8. The high variability of the measured
IPR might be due to the different aerosol properties when air masses arrive from different geographic origin and/or different
altitudes at the Jungfraujoch. Therefore, we investigate here how air masses origin influences the aerosol properties. Because of
the saddle position of the JFJ, there are only two possibilities of how the air masses can approach the measurement platform
(Fig. 6). Predominantly, the air masses arrive from northwestern direction over the Swiss plateau (wind direction of
approx. 315∘). Less frequently, the approach was from a southeastern direction from the inner Alps (approx. 135∘,
via the Aletsch Glacier). According to the direction from which air masses approach the measurement platform, the chemical
composition is influenced predominantly by anthropogenic and organic emissions or by mineral dust sources. To assess the influence
of origin and the average altitude of the air masses during their transport to the Jungfraujoch, we compare two time periods with
aerosol measurements using the total inlet (11 February and 19 February 2013). The chosen time periods exhibit nearly the same
relative humidity, potential wet bulb temperature and temperature (Fig. 6).
Figure 9 shows the chemical composition of the aerosol particles for both time periods, along with corresponding
back-trajectories. The comparison shows that more particles containing BC (internally mixed with ammonium or the BioMinSal-type)
and the BioMinSal-type are observed in the air masses arriving from a northwestern direction (19 February 2013), while the
measurements during the SE-event show more OC containing particles (internally mixed with ammonium or the BioMinSal-type). The
fraction of particles measured containing only potassium and pure organic carbon is very similar for both events.
The air masses not only approached locally from different directions, but also had different long-range origins (see back
trajectories Fig. 9). The air masses measured on 19 February 2013 (NW-event) approached from the north over Germany. The chemical
composition of the measured particles is dominated by black carbon internally mixed with ammonium and the BioMinSal-type. In
contrast to the 19 February 2013, the air masses on 11 February 2013 that also originated mostly from the North but approached
over southern France, and from southern Europe (western Po Valley) contain higher amounts of OC and BioMinSal containing
particles. The vertical transport pathway of the air masses could be an indicator for a local, anthropogenic influence on chemical
composition. In contrast to the air masses from the north-western direction the air masses from the southeast have been lifted
from low altitudes, thereby bringing anthropogenic emissions (e.g., from the Po Valley) up to the altitude of the measurement
station. The air masses from the north-western direction stayed nearly at the same altitude for the last three days prior to the
measurement.
Size distribution of IPR and background aerosol
Ice-CVI and ISI both extract small ice particles in the size range from 5 to 20 µm from mixed-phase clouds (Mertes
et al., 2007; Kupiszewski et al., 2014) while the upper cut-off of the total inlet was 40 µm for wind speeds up to
20 ms-1 (Weingartner et al., 1999). Figure 10 shows the size distributions of IPR measured with the ALABAMA (a, c,
e) and the Sky-OPC (b, d, f) compared with ESEM measurements (a–d). Figure 10a shows the size distributions of IPR measured by
ALABAMA and ESEM applying the Ice-CVI. Figure 10c the size distributions measured by ALABAMA and ESEM using the ISI. In Fig. 10b
and d, the corresponding size distributions measured with the Sky-OPC are shown. Figure 10e and f shows the size distribution
measured by ALABAMA and the Sky-OPC using the total inlet. The measurement times for Ice-CVI and ISI are given in Table 1; the
measurements using the total inlet were done during all cloud-free periods. For all three inlets (total, Ice-CVI, ISI) the maximum
of the detected IPR measured with the ALABAMA and also with the Sky-OPC is in the size range of 300–650 nm. The IPR size
distributions measured by ALABAMA and SEM behind the Ice-CVI are in very good agreement (Fig. 10a). In contrast, the IPR size
distribution measured behind the ISI are somewhat different. (Fig. 10c): both distributions show two maxima, but the positions of
the maxima are not the same. The ALABAMA IPR distribution has a primary maximum in the size bin 0.4–0.5 µm and
a secondary maximum in the size bin 1.3–1.6 µm. The SEM IPR distribution both maxima are shifted to larger sizes. The
first maximum is in the 1–1.5 µm interval and the second maximum is in the 2–2.5 µm interval. However,
total count rates are very low, so this difference should not be overemphasized. Furthermore, SEM measures geometric diameters,
while the sizes determined by ALABAMA refer to the vacuum aerodynamic diameter. In contrast to the size distribution measured with
the ALABAMA and ESEM, the size distribution measured with Sky-OPC does not show a maximum but only a decreasing number
concentration of particles with increasing particle diameter from 100 to 500 nm. This difference is a result of the
detection efficiency of the ALABAMA and ESEM, which is the highest around 400 nm. Figure 10f shows that the background
aerosol particles have a size range of up to approximately 20 µm with a higher number of smaller particles
(<1 µm). The relative amount of larger particles is (d>1 µm) clearly elevated for the ice residuals,
both behind the ISI and Ice-CVI (Fig. 10b and d). Note that the measurements of background particles (out of cloud) and IPR were
conducted by definition at different times, such that the absolute amount of particles cannot be compared.
As mentioned above, relatively more IPR larger than 1 µm were observed compared to the background aerosol
particles. The percentage of the background aerosol particles larger than 1 µm (vacuum aerodynamic diameter) measured
by ALABAMA was only 3 %, while 10–19 % of the IPR behind the Ice-CVI and ISI were larger than
1 µm. Therefore, the chemical composition of these particles has been analyzed separately:
Figure 11 shows the chemical composition of the IPR and background aerosol particles larger than 1 µm. The IPR larger
than 1 µm measured behind the Ice-CVI are dominated by minerals, OC (also mixed with BC) and the BioMinSal-type, but
lead containing particles are found to a similar percentage (15 %) as in all IPR (12 %, see Figs. 2 and 8). OC is the
main fraction of the particles larger than 1 µm measured behind the ISI. In the background aerosol supermicron
particles are dominated by the BioMinSal-type. The finding that the BioMinSal-cluster is one of the major components the IPR as
well as the background aerosol for d>1 µm strengthens the assumption that the metal ions detected in these mass
spectra are likely metal cations of primary aerosols like sea salt, minerals or biological particles, which are mainly found in
the coarse mode (Seinfeld and Pandis, 2006).
The size distribution allow also for an estimation of the ice active site (e.g., Vali, 2008) density of particles larger than
1 µm compared to those smaller than 1 µm: assuming that each IPR corresponds to at least one ice active
site on the original aerosol particle (otherwise it would not have acted as an INP), we can estimate the ice active site density
for the particles analyzed by ALABAMA, separated for particles smaller than 1 µm and larger than 1 µm. The
number of ice active sites per IPR surface (ice active site density) is 2.0×10-7 nm-2 for
d>1 µm and 1.6×10-6 nm-2 for particles with d<1 µm. Relating the number of ice
active sites to the total aerosol surface yields 7.1×10-10 nm-2 for d>1 µm and 4.3×10-9 nm-2 for d<1 µm. In both cases the ice active site density is larger for particles smaller than
1 µm. This is an important finding and explains the high number of IPR found in the size range below
1 µm. However, it should be noted that the assumption that each IPR corresponds to one ice active site may
underestimate the number of ice active sites. In cases where more than one ice active site is found on the particles, also only
one ice crystal will form and one IPR will remain. Thus the estimated number yields a lower limit for the ice active sites.
Summary and conclusions
This paper presents results of the single particle measurements during the INUIT-JFJ/CLACE 2013 campaign conducted at the
Jungfraujoch in January/February 2013. In agreement with previous measurements at the Jungfraujoch (Chou et al., 2011; Kamphus
et al., 2010; Mertes et al., 2007), it was found that larger particles are more efficient ice nuclei. Specific analysis of only
those particles >1 µm showed that the chemical composition of these larger particles is dominated by organic
material and the “BioMinSal” type, both in the IPR as well as in the background aerosol. Besides, the “minerals” group was
indicated as one of the main fractions in the IPR composition measured behind the Ice-CVI. From the finding that IPR have a higher
relative number of larger particles than the background aerosol particles we conclude and confirm a better ice nucleation ability
for the larger, primary aerosol particles such as minerals, salt and biological particles.
In general, IPR are enriched in organic material, partly also internally mixed with BC. Additionally, minerals and lead were found
in IPR as has already been observed in previous measurements conducted at the JFJ (Mertes et al., 2007; Cozic et al., 2008b;
Cziczo et al., 2009; Kamphus et al., 2010; Ebert et al., 2011). In comparison to the composition of the IPR measured using the
Ice-CVI, the measured composition using the ISI is dominated by minerals, and no lead containing particles were found, while
12 % of the IPR sampled using the Ice-CVI included lead. In contrast to Cozic et al. (2008) we found no enrichment of pure
black carbon in the IPR composition. Only the chemical composition of the background aerosol particles shows a high fraction of
black carbon. These results comply with previous investigations (DeMott et al., 1999; Dymarska et al., 2006; Hoose and Möhler,
2012) describing black carbon as a good ice nuclei but only at lower temperatures and its ice nucleating ability appears to be
reduced if coatings by organic material or sulfuric acid are present. During the INUIT-JFJ/CLACE 2013 campaign the temperatures
were never lower than -30 ∘C; see Fig. 6. In the background aerosol, the main particle fraction is also organic
carbon. The high amount of organic material in the background aerosol as well as in the IPR motivates further investigation of
different organic material by single particle mass spectrometry to better characterize the organic fraction. While there have been
previous observations that organic matter is of general importance for ice nucleation ability (DeMott et al., 2003; Cziczo et al.,
2004), recent laboratory studies have shown the importance of porous and glassy organic matter for ice nucleation (Murray et al.,
2010; Adler et al., 2013).
The investigation of the influence of the temperature and the origin and altitude of the air masses on the chemical composition of
IPR and aerosol particles indicate that organic material dominates the IPR composition at higher temperatures and metal components
from mineral or salt at lower temperatures which is in contrast to previous findings (e.g. Möhler et al., 2008; Hoose and
Möhler, 2012). It was illustrated that the origin (geographic and altitude) of the air masses influenced the chemical
composition of the aerosol particles. For the non-cloud aerosol, the particles sampled in air masses characterized by free
tropospheric air and long distance transport were found to be dominated by BC and single metal ions from salt, mineral or
biological sources (BioMinSal-type). These particle types were probably partly mixed with secondary inorganic components, because
NH4+ was detected in some spectra. In contrast to that, the particles in air masses which are influenced by surface
emissions are dominated by organic carbon and potassium.