Separation of particles that play a role in cloud activation and
ice nucleation from interstitial aerosols has become necessary to further
understand aerosol-cloud interactions. The pumped counterflow virtual
impactor (PCVI), which uses a vacuum pump to accelerate the particles and
increase their momentum, provides an accessible option for dynamic and
inertial separation of cloud elements. However, the use of a traditional
PCVI to extract large cloud hydrometeors is difficult mainly due to its
small cut-size diameters (< 5
The glaciation of supercooled clouds through various ice nucleation (IN) processes is an important atmospheric process affecting the formation of precipitation after a series of secondary processes and the Earth's energy budget (Boucher et al., 2013). Currently, the climatic impact of ice-nucleating particles (INPs) is being reassessed due to increasing evidence of their diversity and potential ability to influence cloud properties (e.g., Hoose and Möhler, 2012; Murray et al., 2012; Després et al., 2012; O'Sullivan et al., 2015; Wilson et al., 2015). However, our knowledge of the atmospheric INP abundance and their identities as well as sources remains scarce (e.g., Knopf et al., 2014). The analysis of ice residual particles, which are mainly leftover INPs after evaporating the water content (DeMott et al., 2003), can help to answer this question.
Recently, it has been found that the crystalline structures of Arizona test
dust particles have the innate advantage of nucleating ice over other
aerosols that lack these suitable properties (Kulkarni et al., 2015). Observations of ice
crystal residual particles from a limited time segment of field studies
suggest an apparent increase in the certain types of compositions (i.e.,
biological material), which are known as a good INP (e.g., Pratt et al., 2009; Prenni et al., 2009;
Huffman et al., 2013). Nevertheless, further experimental work and comprehensive analyses
of the INPs' physico-chemical properties are important to develop
representative IN parameterizations and resemble the atmospheric
representation of INPs. In typical mixed-phase clouds over the Arctic,
continental and maritime environment, the average concentration of cloud
particles (i.e., sum of droplets and ice crystals in the volume diameter
range of 2–47
The counterflow virtual impactor (CVI), which was originally developed for airborne sampling by Ogren et al. (1985), has widely been used to virtually isolate cloud elements from trace gas or interstitial particles. In particular, the aircraft CVI enabled the collection of droplet and ice crystals that are imparted into the system as a result of the motion of an aircraft and to subsequently sample their residuals after evaporation of water. In addition to its primary function as an inertial particle separator, the CVI simultaneously functions as a particle concentrator, in which an enhancement factor (EF) is determined as the ratio of input and output flow (Hayden et al., 2008). Further, by coupling the CVI with other online and offline analytical techniques, those residual samples can be analyzed for identifying their abundance and physico-chemical properties (Twohy et al., 1997; Twohy and Anderson, 2008).
Since 1980s, a series of design updates has been applied to improve the overall performance of the CVI; e.g., enhance the transmission efficiency of residuals and minimize artifacts (e.g., Noone et al., 1988; Anderson et al., 1993; Schwarzenböck et al., 2000; Chen et al., 2005; Hayden et al., 2008; Berg et al., 2009; Shingler et al., 2012). These efforts have led to collect a variety of cloud residual samples from various types of atmospheric clouds, such as cirrus clouds (Cziczo and Froyd, 2014 and references therein), marine boundary layer clouds (e.g., Crumeyrolle et al., 2013; Gieray et al., 1997; Twohy et al., 2001; De Bock et al., 2000; Hopkins et al., 2008; Russel et al., 2000; Noone et al., 2000), Arctic clouds (Hiranuma et al., 2013), orographic clouds (Kamphus et al., 2010), lower tropospheric mixed-phase clouds (Mertes et al., 2001), and convective cumulus clouds (Heymsfield et al., 2005; Matsuki et al., 2010).
Based on the CVI principle, Boulter et al. (2006) developed a pumped CVI (PCVI) as an
alternative to the airborne CVI for laboratory applications. Instead of
using the motion of an aircraft or a wind tunnel as in the airborne CVI, the
PCVI utilizes a vacuum pump to bring a particle-laden flow into the system.
This self-governing feature and portability of the PCVI system is especially
advantageous over the regular CVI to provide constant residual-laden flows
to the downstream instruments during studies in a controlled laboratory
setting (China et al., 2015; Hiranuma et al., 2011; Slowick et al., 2011; Crawford et al., 2011; Gallavardin et al., 2008; Kim and Raynor, 2009; Cziczo et al., 2003, 2009;
DeMott et al., 2003) and a ground-based field application (Kupiszewski et al., 2015; Worringen et al., 2015;
Vogel et al., 2013; Corbin et al., 2012; Richardson et al., 2007). In particular, the separation of small ice
crystal residuals (
Recently, Kulkarni et al. (2011) suggested design modifications to the nozzle geometry and
the collector inlet to improve the TE. To date, these design updates have
been reflected in a commercial PCVI. However, a number of limitations
regarding the use of the PCVI had been raised. For example, due to flow
restrictions of the instruments and an associated small cut size in the
current PCVI (< 5
The current study focuses on the engineering development of a new PCVI,
termed the ice-selecting PCVI (IS-PCVI henceforth), to separate cloud
elements with the cut-size diameter larger than
Geometrical parameters of the traditional PCVI and the IS-PCVI derived from the Autodesk Inventor 3-D computer-digitized design.
The schematic of the IS-PCVI is shown in Fig. 1. The IS-PCVI is composed of
five stainless steel parts, and the scale-relevant design of individual
components is depicted in Fig. 2. In order to minimize the impaction losses
of the particles within the individual components, the entire inner surface
of the stainless steel components is electro-polished. The internal volume
is
Schematic description of the IS-PCVI outfit
A unique feature of the IS-PCVI in comparison to a traditional PCVI is its
capability to facilitate large counterflow (CF > 3 L min
Sampling moist air from the AIDA chamber may lead to ice formation within
the IS-PCVI because of the apparent cooling of the sampled air followed by
the nozzle surface cooling. Such cooling spontaneously occurs due to the
flow acceleration and
Structure of individual components of the IS-PCVI (from top to
bottom) is shown in panels
Water from ice crystals and droplets is removed in the heated section
(> 40
Schematic representation of the experimental setup for validation of the IS-PCVI. The grey dashed line represents the heat-insulating wall of AIDA. Different types of aerosol generators used in this work are summarized in Table S1 in the Supplement.
In order to continuously monitor the
Due to the venture-shaped nozzle geometry to induce the jet stream above the
collector inlet (i.e., the region where the IF and CF flows meet), a certain
The IS-PCVI operation is automated, such that the various flows are
controlled by three MFCs (Fig. 1) and
The principle of the IS-PCVI operation is as follows: (i) an inlet nozzle
produces the jet stream of particle-laden air as a downward IF into the
system, (ii) an application of upward ECF creates the stagnation regions
(i.e., a spot where longitudinal velocity is zero; Kulkarni et al., 2011) in the middle
compartment, triggering an inertial separation of cloud particles, (iii) a
In this section, the CFD simulations (Sect. 3.1), methods in experimental studies (Sect. 3.2) and data analysis method (Sects. 3.3–3.4) are described. In our experiments, inertial particle separation using the IS-PCVI and online residual analyses were simultaneously carried out at various temperatures and supersaturations in the AIDA chamber. The performance of the IS-PCVI was characterized using a series of instruments deployed downstream of the impactor.
Summary of the FLUENT settings used for the CFD simulations.
The CFD simulations were performed to evaluate the performance
characteristics of the IS-PCVI. The commercially available CFD software
FLUENT (ANSYS, Inc.) was employed to perform the three-dimensional
simulations to calculate the flow field and various particle size
trajectories. The RNG k-
The GAMBIT software (version 2.4.6) was used to generate the mesh for the
CFD solver. Mesh control, such as multi-blocking, was implemented to obtain
a better density of the mesh in regions of importance such as where velocity
and
The boundaries of the CFD domain were assigned based on the flow conditions
of the IS-PCVI. The boundary condition of “mass flow inlet” was used for
the input, output, and add flow boundary zones of the domain, and “
We now present our experimental approaches validating the newly developed IS-PCVI. The general layout of our experimental setup at the AIDA facility is illustrated in Fig. 3. In this section, we describe the instrumental techniques employed for the (i) aerosol generation (Sect. 3.2.1), (ii) simulated cloud formation (Sect. 3.2.2), and (iii) in situ residual characterization (Sects. 3.2.3–3.2.7).
In this study, we utilized a known composition of aerosols from different
sources to induce nucleation of ice to generate ice crystals and/or droplets
in a controlled manner (summarized in Table S1). As seen in the table, the
generation procedure was based on our previous studies. A custom-built
atomizer, described in Wex et al. (2015), was used to atomize the ammonium sulfate,
sodium chloride, Snomax and PF CGina bacteria suspensions. The suspensions
using the first two materials were prepared with a concentration of 0.12 g solute in 1 L of 18.2 M
The AIDA chamber, consisting of an aluminum cylinder-shaped vessel of 84 m
Throughout the experiment, a mixing fan is activated to quickly homogenize
air in the chamber. The AIDA gas
As illustrated in Fig. 3, a combination of four aerosol particle
instruments, such as a condensation particle counter (CPC, TSI, Model 3010),
a scanning mobility particle sizer (SMPS, TSI, Model 3080 DMA and Model 3010
CPC), an aerosol particle sizer (APS, TSI, Model 3321), and the white light
aerosol spectrometer optical particle counters (welas OPCs, Palas, Sensor
series 2300 and 2500; Benz et al., 2005), was used for particle size and number
concentration measurements. Specifically, the total number of condensation
nuclei (CN) in the AIDA chamber was measured using the CPC 1. The full size
distribution of CN was measured by a combination of the APS and the SMPS
before expansion. In addition to the CN measurement, two OPCs connected to
the bottom of the AIDA vessel (OPC 1 and OPC 2) provided the size
distributions of cloud particles during the expansion. Both welas OPCs were
operated in side-by-side position at individual vertical sampling tubes
directly from the bottom vessel of AIDA (upstream of the IS-PCVI) to keep
track of temporal evolution of particle size distributions in the chamber
during expansion. The detection size ranges (in optical diameters) of the
OPC 1 and the OPC 2 complementally corresponded to 0.7–46 and 5–240
The particle TE of any CVIs is size-dependent. For example, the potential
artifact of the direct impaction of large particles onto the CVI inner wall
is reported in a previous study (Chen et al., 2005). As a result, only particles that
have sizes around the cut size may be transmitted through the CVI. The
transmission of particles larger than 20
We now describe a method to optimize the critical cut size,
Theoretically, a 5–20
We conducted a set of two expansion experiments (INUIT05_09
and_12) to examine if the heated tube sections downstream of
the IS-PCVI (termed as an evaporation section; see Fig. 3) efficiently
evaporates droplets and ice particles. We used 1
The measurement of rejected particle concentrations in the pump flow (Fig. 1b) was conducted to compare it to the residuals concentration in the sample
output flow and the total particle concentration in AIDA. Deploying the CPC 3 in the pump flow line allowed to measure the number concentration of
interstitial particles smaller than a
The Particle Phase Discriminator mark 2, Karlsruhe edition (PPD-2K;
University of Hertfordshire, UK; Kaye et al., 2008) is a laser light
scattering particle spectrometer that acquires high-resolution spatial light
scattering patterns of individual cloud particles in the 7–26
In this study, the PPD-2K was used to examine the phase of inertially
segregated particles and to provide insight into the potential impact of the
CVI-driven artifacts (i.e., particle collision, coalescence, shuttering,
wake capture effect) onto further downstream measurements. Specifically,
coupled with the PPD-2K, a few AIDA expansion measurements were performed to
gain insight in the inertially separated droplets and ice particles
downstream of the IS-PCVI. We conducted a series of two expansion
experiments using NaCl as seed aerosols. NaCl particles were activated to
droplets followed by homogeneous freezing of ice crystals below
Next, we describe the methodology to determine the physical properties
(e.g., size) and chemical composition of the AIDA cloud residuals and
interstitial particles. As described above, sampled air containing high
inertia particles is directed vertically and injected into the heating
tubes. Subsequently, the output flow is guided to the instruments, such as
CPC 2, CPC 3, OPC 3 and miniSPLAT, as illustrated in Fig. 3. The transmitted
residual particles are first counted by the OPC 3 at the outlet of the
IS-PCVI. The residual number was also counted by directing 1.0 volumetric L min
The composition determination as a function of
To obtain the number concentrations of the IS-PCVI residuals that are
comparable with the AIDA droplet and ice particle size distributions
(
CFD-simulated velocity pathlines. The center region, where the downward and upward velocities meet, coincides with the location of two stagnation planes (the first one at the collector inlet and the second one at the CF outlet) and nozzle jet surface distortion.
Comparison of the TE calculated from the CFD studies. The results
of three different IFs (50, 70 and 100 L min
The welas OPC overestimates actual sizes of non-spherical particles as the
estimated size of this OPC represents the particle size inferred by the Mie
theory assuming spherical particles of known refractive index (Benz et al., 2005).
Since we do not know the physical relation between optical size and actual
non-spherical ice size, we estimated the overestimation factor of the OPC
(
Alternatively, we can also optimize this factor by relating the total
condensed water volume (
To verify this notion and to find the conversion factor applied for ice
crystals in our AIDA experiments, maximum optical sizes of ice crystals
measured by the welas OPC during five expansions (INUIT05_55–59; i.e., homogeneous freezing experiments in
The performance validation of the IS-PCVI was carried out at the AIDA facility. The four phases of validation experiment included: (1) CFD simulation study (Sect. 4.1), (2) ice separation, TE and cut-size tests using the AIDA aerosol and cloud particle instruments (Sects. 4.2–4.6), (3) characterization of droplets and ice particles using a combination of AIDA/IS-PCVI/PPD-2K (Sect. 4.7) and (4) physico-chemical characterization of test aerosol particles and ice residuals and potential ice fragments using a single particle mass spectrometer (Sect. 4.8).
An example of the CFD-simulated velocity pathlines under one condition (IF,
CF, output flow
Figure 6 summarizes an AIDA expansion experiment (FIN01_04)
to activate illite NX (CN
Temporal plots of the AIDA freezing experiment. Arrays of
alphabetical panels represent the chamber gas
Lower-bound transmission curves of two PCVIs as a function of
particle size in
Droplet cut-size spectrum as a function of the CF-to-IF ratio. During our validation tests, we systematically varied the IS-PCVI flow conditions (Table S2).
Ice particle cut-size spectrum as a function of the CF-to-IF ratio. During our validation tests, we systematically varied the IS-PCVI flow conditions (Table S2).
Temporal evolution of droplet size distributions during
INUIT05_09. Normalized size distributions of droplets
(d
Comparison of the total particle number concentration in AIDA to
the sum of residuals in the output flow and interstitial particles in the
pump flow. The dashed line represents
Temporal profiles of the static IS-PCVI/PPD-2K coupling
experiment (INUIT05_55)
Figure 7 shows the lower-bound TE spectrum (below 10
Additionally, simultaneous measurements of two identical welas OPCs (OPC 1
and OPC 3; Palas, sensor series 2300) were carried out during
INUIT05_36. We deployed the welas OPCs before and after the
IS-PCVI to estimate the droplet TE
The droplet
Another
Normalized size distribution of ice particles for
INUIT05_55 (static mode)
Temporal plots of the AIDA expansion experiments from
FIN01_38. Panels are arranged to show the chamber gas
The temporal evolution of droplet size distributions measured with the welas
OPCs is shown in Fig. 10. In the first experiment, we switched off all the
IS-PCVI flows (INUIT05_09). The evaporation section
Interstitial particle concentrations were measured in the pump flow in
INUIT05_34, _51 and _58. The CF-to-IF ratio was maintained to be 0.13,
0.13 and 0.16 for INUIT05_34, _51 and _58, respectively, throughout each
experiment. Comparison of the total particle number concentration to the sum
of residuals and interstitial particles is illustrated in Fig. 11. This
figure shows that the total particle number concentration is in good
agreement with the sum of residual concentration and unactivated aerosols.
Two important implications of this figure include (1) the separation of
interstitial components from cloud elements larger than given
Figure 12 shows the time series of the AIDA conditions (panels i and ii),
the welas OPC (panel iii) and properties of droplets and ice particles
detected by the PPD-2K in two expansion experiments (panels iv). The minor
ice peak was observed with the PPD-2K shortly after initiating first
expansion (10:57 in Fig. 12a.iv), presumably due to IN of impurity in AIDA
in the deposition mode. The background aerosol concentration measured by the
CPC 1 in the AIDA chamber prior to expansion was 0.13
In addition,
The results of AIDA experiments to test inadvertent intrusion of unwanted particles are presented below. The time series plots of the AIDA experimental trajectories and residual counts downstream of the IS-PCVI are illustrated in Fig. 14.
In the FIN01_38 experiment, atomized bacteria particles (PF
CGina bacteria; CN
The main goal of this study is to develop a new PCVI device to separate ice
crystals from mixed-phase clouds. The existing traditional PCVI only allows
separation of cloud elements with the upper bound cut-size diameter
Prior to device fabrication, we conducted the CFD analysis. The CFD study
guided us to optimize the design and helped to understand the performance
characteristics of the IS-PCVI across a wide range of experimental
parameters. Various design improvements to minimize the impact of artifacts
and associated particle losses in the system based on the previous study
were also incorporated. After the CFD study, we verified the performance
characteristics of the IS-PCVI in the laboratory setting. A sequential
experiment involving cloud formation, generation of cloud elements such as
supercooled droplets and ice crystals, inertial separation, and residual
analyses were carried out as a function of
The results from a series of experimental validation tests verified the
inertial separation of pristine ice crystals of cut size
Overall, the newly developed IS-PCVI complements the current version of the PCVI by providing an additional capability of large ice residual extraction to elucidate the properties of ice residuals and INPs in mixed-phase clouds. The application of the IS-PCVI to characterize ice crystal residuals can enhance the ongoing INP measurements, and such studies would further guide future IN studies towards a better understanding of INPs.
Aerosol types and associated particle generators used in this work are summarized as part of the Supplement. Detailed characteristics of IS-PCVI properties during INUIT05 and FIN01 campaigns are also summarized as supplements. Other information regarding particle properties (i.e., concentration and size distribution of aerosol, droplets, and/or ice) before and during individual AIDA expansion experiments is available upon request. Temporal profiles of the AIDA cloud simulation experiments (as illustrated in Fig. 6) can also be provided.
N. Hiranuma and O. Möhler oversaw the deployment and operation of instruments in AIDA, led the data analysis and prepared the manuscript, with input from all authors. GK set up, ran and compiled the CFD studies of the IS-PCVI with support of N. Hiranuma. S. Vogt and N. Hiranuma designed and developed the IS-PCVI equipment with contributions from G. Kulkarni, D. J. Cziczo, R. Wagner and engineers at IMK-AAF. E. Järvinen, P. Vochezer, and M. Schnaiter operated the PPD-2K at AIDA during INUIT05 and provided analysis and interpretation of the PPD-2K data. D. M. Bell, J. Wilson, and A. Zelenyuk operated miniSPLAT at AIDA during FIN-1 and provided analysis and interpretation of miniSPLAT data. The INUIT05 campaign was led by N. Hiranuma and O. Möhler. The FIN-1 activity was coordinated by O. Möhler and D. J. Cziczo with assistance of N. Hiranuma.
We thank the Engineering and Infrastructure group at KIT IMK-AAF (Georg Scheurig, Tomasz Chudy and Rainer Buschbacher) for their support in constructing and operating the IS-PCVI. This work was funded by the Helmholtz Association through its research programme “Atmosphere and Climate (ATMO)” and the Deutsche Forschungsgemeinschaft (DFG) through the Research Unit FOR 1525 (INUIT, grant No MO668/4-1 and MO668/4-2). The valuable contributions of the FIN organizers, their institutions, and the FIN-1 Workshop science team are gratefully acknowledged. N. Hiranuma thanks A. Kiselev (KIT IMK-AAF) and K. Rabe (KIT IBG) for their support on the bacteria sample preparation. M. Schnaiter acknowledges the funding by DFG under grant SCHN 1140/2-1. The participation of G. Kulkarni, D. M. Bell, J. Wilson, A. Zelenyuk, and D. J. Cziczo was partially funded by NSF grant#AGS-1461305. Additional support (A. Zelenyuk, D. M. Bell, J. Wilson, and G. Kulkarni) was provided by the US Department of Energy (DOE) Office of Biological and Environmental Research (OBER) Atmospheric Research Systems Program (ASR). The development of miniSPLAT was funded by the DOE Office of Science, Office of Basic Energy Sciences, Division of Chemical Sciences, Geosciences and Biosciences and EMSL User Facility sponsored by the DOE OBER and located at Pacific Northwest National Laboratory. G. Kulkarni acknowledges support from the US DOE ASR. The Pacific Northwest National Laboratory is operated for DOE by Battelle Memorial Institute under contract DE-AC05-76RLO 1830.The article processing charges for this open-access publication were covered by a Research Centre of the Helmholtz Association.Edited by: B. Ervens Reviewed by: two anonymous referees