Chemical characterization of individual particles and residuals of cloud droplets and ice crystals collected on board research aircraft in the ISDAC 2008 study

Authors


Abstract

[1] Ambient particles and the dry residuals of mixed-phase cloud droplets and ice crystals were collected during the Indirect and Semi-Direct Aerosol Campaign (ISDAC) near Barrow, Alaska, in spring of 2008. The collected particles were analyzed using Computer Controlled Scanning Electron Microscopy with Energy Dispersive X-ray analysis and Scanning Transmission X-ray Microscopy coupled with Near Edge X-ray Absorption Fine Structure spectroscopy to identify physico-chemical properties that differentiate cloud-nucleating particles from the total aerosol population. A wide range of individually mixed components was identified in the ambient particles and residuals including organic carbon compounds, inorganics, carbonates, and black carbon. Our results show that cloud droplet residuals differ from the ambient particles in both size and composition, suggesting that both properties may impact the cloud-nucleating ability of aerosols in mixed-phase clouds. The percentage of residual particles which contained carbonates (47%) was almost four times higher than those in ambient samples. Residual populations were also enhanced in sea salt and black carbon and reduced in organic compounds relative to the ambient particles. Further, our measurements suggest that chemical processing of aerosols may improve their cloud-nucleating ability. Comparison of results for various time periods within ISDAC suggests that the number and composition of cloud-nucleating particles over Alaska can be influenced by episodic events bringing aerosols from both the local vicinity and as far away as Siberia.

1 Introduction

[2] Current knowledge of aerosol-cloud interactions in the Arctic is inadequate for accurate climate predictions, in part due to limited data on the composition of airborne particles that serve as cloud condensation nuclei (CCN) and ice nuclei (IN). Seasonal variability of cloud properties in the Arctic is so uncertain that the role of cloud feedbacks cannot be determined [Morrison et al., 2005; Solomon et al., 2009]. Warming trends observed at high latitudes are a result of the increased anthropogenic emissions and air pollution by aerosols, greenhouse gases, and ozone [Law and Stohl, 2007]. Aerosols reaching the North Slope of Alaska contain a variety of chemical constituents, including soot, biomass burning particulates, mineral dust, sulfates, and organic compounds [Quinn et al., 2009; Warneke et al., 2010].

[3] Many of the Arctic haze pollutants found near Barrow, Alaska are transported from anthropogenic sources and boreal fires in Eurasia [Brock et al., 2011; Earle et al., 2011; Fisher et al., 2010; Garrett and Zhao, 2006; McFarquhar et al., 2011]. Specifically, during the springtime of 2008, pollution plumes observed over Alaska were traced to biomass burning events in Siberia, Russia [Warneke et al., 2009]. In addition, large quantities of dust transported from Asia have frequently been observed in the Arctic [Avramov et al., 2011; Brock et al., 2011; Cahill, 2003; Earle et al., 2011; Fisher et al., 2010; McFarquhar et al., 2011; Stone et al., 2007; Tomasi et al., 2007]. During winter and early spring, particle mass concentrations greater than 2 µg m−3 have been measured in the Arctic and attributed to anthropogenic emissions in Europe and Asia [Barrie, 1986; Brock et al., 2011; Law and Stohl, 2007]. Substantially lower concentrations of fine particles are typically observed in the late spring to early summer due to slow transport and efficient removal processes [Law and Stohl, 2007]. However, as the ice breaks up in spring and the open ocean is exposed, the decrease in transported anthropogenic aerosol particles is counterbalanced by an increase in new particles formed as a result of dimethyl sulfide and biological emissions from the ocean [Ferek et al., 1995; Leck and Bigg, 2005a, 2005b].

[4] Cloud-nucleating ability depends on properties of the available aerosol [e.g., Quinn et al., 2008; Fornea et al., 2009; Dusek et al., 2006; Hiranuma et al., 2011; Kamphus et al., 2009]. In the Arctic, mixed-phase clouds containing both droplets and ice crystals are frequently present in addition to liquid and glaciated clouds [Jackson et al., 2012; McFarquhar et al., 2007, 2011; McFarquhar and Cober, 2004; Avramov et al., 2011; Prenni et al., 2007; Rogers et al., 2001; Shupe et al., 2005, 2006; Verlinde et al., 2007]. In a year-long period of remotely sensed data collected at the Barrow research site on the North Slope of Alaska operated by the Atmospheric Radiation Measurements program of the U.S. Department of Energy, mixed-phase clouds were observed ~40% of the time [Shupe et al., 2006]. Mixed-phase clouds of varying degrees of glaciation were also observed during the Indirect and Semi-Direct Aerosol Campaign (ISDAC) in spring 2008 [Avramov et al., 2011; Earle et al., 2011; McFarquhar et al., 2011]. Although mixed-phase clouds are frequently observed, the microphysical processes underlying their formation in the Arctic are not well understood [Liu et al., 2011].

[5] For aerosols of a single-component composition and known size, well-established Köhler theory predicts their cloud-nucleating ability [Köhler, 1936]. However, the complex and evolving properties of atmospheric aerosols provide a challenge in predicting cloud formation [Gunthe et al., 2011; Moore et al., 2011; Y. Ma et al., Rapid modification of cloud-nucleating ability of aerosols by biogenic emissions, submitted to G. Res. Letts., 2013]. Recent field measurements reveal apparently conflicting evidence regarding the relative importance of particle size and composition in predicting CCN activation. For example, measurements from recent campaigns suggested that CCN concentrations are mainly determined by the aerosol size distribution [Dusek et al., 2006; Hudson, 2007; Zelenyuk et al., 2010; Earle et al., 2011]. In contrast, additional measurements collected in a wide variety of locations indicated that the chemical composition is important in determining the cloud-nucleating ability of aerosols [Furutani et al., 2008; Quinn, et al., 2008; Twohy and Anderson, 2008; Topping and McFiggans, 2012]. Laboratory measurements in which both particle size and chemistry are monitored independently illustrate the requirement that a particle must meet both the minimal size and composition criteria in order to activate as CCN and further suggest that the dominant factor (size or composition) changes throughout an aerosol's atmospheric lifetime and depends on aerosol aging processes [Moffet et al., 2010c; Ma et al., submitted manuscript, 2013]. Hence, apparent conflict in field results cited above may be a result of variations in sampled air masses from project to project and even flight to flight. During the ISDAC project, single particle mass spectrometry measurements indicated that while particle size was the dominant factor in cloud activation for the submicron particle size range, a slight enhancement of sulfate in cloud droplet residuals was also observed [Zelenyuk et al., 2010]. The measurement techniques utilized in this study provide a means to analyze possible composition dependence in great detail and more qualitative insight on size dependence. Thus, we focus on the composition effects on CCN potential. The relative importance of aerosol size and chemistry in predicting ice-nucleating properties is also a topic of recent discussion [DeMott et al., 2010; Wang et al., 2012]. However, field observations show that the concentration of liquid droplets exceed ice crystal concentration by 4 orders of magnitude or more in the mixed-phase clouds observed during the Arctic springtime [Avramov et al., 2011; Rogers et al., 2001]. Thus, under most conditions, the majority of collected cloud particle residuals arise from liquid droplets rather than ice crystals. Nevertheless, recent modeling studies of Arctic mixed-phase clouds show sensitivity to assumed concentrations of IN and CCN, and availability of such data is crucial to improve the performance of models [Fridlind et al., 2007; Klein et al., 2009].

[6] In this study, we report the chemical composition, mixing state, and size of individual airborne particles and residuals from mixed-phase clouds sampled during ISDAC using instrumentation onboard the National Research Council Canada Convair 580 research aircraft. The physico-chemical properties of ambient particles and residuals are compared to seek their plausible relevance to cloud-nucleating aerosols [Andreae and Rosenfeld, 2008; Kolb and Worsnop, 2012]. Chemical analysis of field-collected individual particles was performed using Computer Controlled Scanning Electron Microscopy and Energy Dispersive X-ray analysis (CCSEM/EDX) and Scanning Transmission X-ray Microscopy coupled with Near Edge X-ray Absorption Fine Structure spectroscopy (STXM/NEXAFS). A STXM/NEXAFS mapping approach was used to investigate the internal composition (mixing state) of individual particles [Hopkins et al., 2008; Moffet et al., 2010a; Takahama et al., 2010]. The results are interpreted in the context of ambient conditions and field data collected from other onboard probes, including Condensation Particle Counters (CPCs) which provide particle number concentrations, a combination of the 2 Dimensional Cloud probe (2DC), Cloud Imaging Probe (CIP), 2 Precipitation Probe (2DP), and a 2 Dimensional-Stereo probe (2D-S) that measure cloud particle concentrations [Lawson et al., 2006], and a Continuous Flow Diffusion Chamber (CFDC) that measures number concentration of particles acting as ice nuclei under controlled CFDC chamber conditions

2 Experimental Procedure

2.1 Particle Collection and In Situ Measurements of Particle and Cloud Properties

[7] During ISDAC, the Convair 580 aircraft conducted 27 flights over the northern slope of Alaska [McFarquhar et al., 2011]. During the flights, the source of the sampled air varied between an ambient isokinetic inlet and a counterflow virtual impactor (CVI) inlet. To provide insight into the effect of aerosol composition on cloud properties, we selected representative particle samples from cloud-free air collected directly through the isokinetic inlet and cloud residuals collected through a CVI during time periods when the aircraft was flying through mixed-phase clouds. Use of the CVI allowed sampling of cloud droplets or crystals larger than the optimized cut size of ~11 µm diameter [Hayden et al., 2008]. Prior to entering a cloud, the sampling of ambient air through the CVI was routinely performed as a background test of CVI performance. During the sample periods included here, ambient particle removal was near optimal, with counts of ambient aerosol through the CVI, measured prior to entering into clouds, 10 cm−3 or less. Particles larger than 100 µm diameter were efficiently removed by impaction on a 90° bended drain line after the CVI. Downstream of the CVI inlet, the sample line was heated to approximately 40°C to remove liquid water and ice, and the residual-laden sample was distributed to a suite of instruments, including the Time-Resolved Aerosol Collector (TRAC) used for particle collection [Laskin et al., 2006]. Reported number concentrations of particles passing through the CVI have been corrected using an enhancement factor (EF) determined as

display math

where v is the true aircraft airspeed (m s−1), D is the CVI tip internal diameter (mm), PAmb and P0 are the ambient and standard pressures, respectively, TAmb and T0 are the ambient and standard temperatures, respectively, Q is the CVI sample flow (L min−1). In this study, an EF factor of 5.6 was applied to correct ice nucleus and aerosol concentrations sampled through the CVI [Hayden et al., 2008].

[8] A schematic of the Convair sampling system and all instruments referred to in this work is shown in Figure 1. The TRAC was used to deposit consecutive particle samples onto prearranged microscopy substrates (Carbon type B, Cu 400 mesh grids. Ted Pella, Inc.) with a time resolution of 5 min. The selected sampling periods include characteristic episodes of ambient particles collected directly through the isokinetic inlet (Flight 30-Substrate 6) and droplets and crystal residuals collected through the CVI during flights in mixed-phase clouds (Flight 30-Substrate 19 and Flight 31-Substrate 62) as shown in Table 1. These particular time periods were chosen for this study based on the availability of in situ aerosol and cloud probes as well as cloud phase data [McFarquhar et al., 2011]. Samples are identified according to Flight Number-Substrate Number in Table 1 and throughout this manuscript. Since diverse sampling conditions were encountered during ISDAC, two additional samples were selected from two unique episodes. The first of these episodes was a flight through a mixed-phase cloud which contained aerosol with ice nucleation efficiency far exceeding the campaign average (Flight 34 on 29 April) according to in situ observations. The second episode was a flight through a biomass burning plume that originated in Siberia (Flight 25 on 19 April).

Figure 1.

Sampling instruments layout for ambient and in-cloud measurements used in this study.

Table 1. Particle Samples Used for CCSEM/EDX and STXM/NEXAFS Analyses
      # of Particles Analyzed by
  1. a

    Samples are identified according to Flight Number-Substrate Number.

Particle PopulationFlight #Date in 2008TRAC Time, UTCSample IDaInlet ModeCCSEM/EDXSTXM/NEXAFS
Biomass Burning2519 April21:57:26–22:02:33F25-S20Ambient950124
Cloud-Free Ambient3026 April20:15:34–20:20:41F30-S6Ambient588071
In-Cloud Residual3026 April21:22:19–21:27:26F30-S19Residual185045
 3127 April1:03:11–1:08:19F31-S62Residual53028
High IN (Residuals)3429 April2:28:14–2:38:30F34-S66 to S67Residual153091

[9] Detailed characterization of the ambient conditions and aerosol properties during collection of TRAC samples was provided by instruments onboard the Convair 580, as summarized in Table 2. Two CPC instruments measured number concentrations of fine mode aerosol. The first CPC (TSI, Inc. 3775) had an independent inlet that continuously sampled ambient aerosol. The second CPC (Particle Measurement Systems, Inc., PMS 7610) was connected to the shared inlet with the TRAC, the CFDC, and other instruments, and the inlet was operated in ambient (isokinetic) and CVI modes. The PMS 7610 undercounted particles by a factor of ~2.5 throughout the project, likely due to an apparent flow problem. We also note that the limit of detection is 4 nm for the TSI CPC and 10 nm for the PMS probe, which could contribute to some of the observed differences in concentration. Thus, the TSI CPC measurements provided reliable determination of ambient aerosol concentrations as well as a means to correct the PMS CPC data. The PMS CPC data were used to estimate residual concentrations (corrected for the CVI enhancement and flow correction factors) and to calculate IN fractions. According to the PMS CPC measurements, typical concentrations of cloud droplet residues collected through the CVI were almost an order of magnitude lower than concentrations of ambient particles.

Table 2. In Situ Measurement Averages During TRAC Sampling
Particle PopulationdBiomass BurningCloud-Free AmbientIn-Cloud ResidualHigh IN (Residuals)
  1. Reported values represent average ± standard deviation (Avg ± Dev) and 1 min averaged min/max range (Min/Max).

  2. a

    Reported concentrations are corrected by the CVI enhancement and flow correction factors.

  3. b

    The number of ice crystals (Nice) is based on combined cloud probe measurements according to the method described in Jackson et al. [2012].

  4. c

    Percentage activated as IN (time average of CFDC concentration/total particle concentration based on the corrected CPC 2).

  5. d

    Reported conditions and measurements are derived from adjacent period in the same biomass burning plume episode (21:36:54–21:42:01). Filter measurement was conducted on CFDC from 21:49:22 to 22:03:57. Note very similar ambient conditions, CPC measurements, and elemental compositions for both intervals.

Sampling Inlet CategoryAmbientAmbientResidualResidual
Ambient ConditionsAvg ± DevMin/MaxAvg ± DevMin/MaxAvg ± DevMin/MaxAvg ± DevMin/Max
Temp. (°C)−14.6 ± 3.3−19.9/−10.2−37.5 ± 1.5−38.9/−34.9−11.3 ± 1.5−13.2/−8.6−10.8 ± 1.1−11.8/−8.5
Supersaturationwater (%)−68.9 ± 9.8−79.1/−49.2−50.6 ± 11.6−60.8/−34.9−2.6 ± 4.4−6.7/1.1−9.8 ± 9.1−29.7/0.4
Supersaturationice (%)−64.0 ± 11.1−75.6/−43.9−27.1 ± 15.9−41.0/−5.28.9 ± 5.61.4/12.40.3 ± 10.9−23.6/12.5
Particulate Measurements        
CPC 1 (TSI 3775), ×105 L−19.0 ± 6.44.6/20.43.0 ± 0.42.7/3.32.4 ± 0.81.5/3.13.4 ± 2.01.9/6.1
aCPC 2 (PMS 7610), ×105 L−15.9 ± 2.84.0/10.23.3 ± 0.42.9/3.70.11 ± 0.160.01/0.350.21 ± 0.40/1.0
bNice, L−10 0 0.39 ± 0.360.15/0.6611.4 ± 20.40/54.3
In Situ Ice Nucleation Experiments        
CFDC Operating Conditions        
Chamber Temp., (°C)−29.6 ± 0.2−30.0/−29.4−31.0 ± 0.1−31.2/−30.9−24.1 ± 3.8−29.1/−18.0−21.8 ± 0.8−22.5/−19.7
Supersaturationwater, (%)−2.1 ± 1.1−3.2/−0.2−1.2 ± 0.8−2.0/0.2−8.1 ± 4.6−17.9/−0.32.4 ± 1.6−1.6/3.6
Supersaturationice, (%)31.0 ± 1.529.3/33.734.1 ± 1.133.1/36.217.0 ± 10.10/29.426.9 ± 2.919.4/29.0
Ice Nucleus Concentration        
aIN, L−11.2 ± 1.30.3/2.10.08 ± 0.100.01/0.150.62 ± 0.350.03/1.145.6 ± 3.53.1/8.1
c% Activation2.0 × 10−4 2.0 × 10−5 6.3 × 10−3 2.6 × 10−2 

[10] The CFDC and Single Particle Laser Ablation Mass Spectrometer (SPLAT) were also operated on the shared inlet. Full descriptions of the CFDC and SPLAT are available in Rogers et al. [2001] and Zelenyuk et al. [2009, 2010], respectively. The CFDC provides the concentration of all particles that activate as IN under the chosen temperature and relative humidity within the CFDC chamber. While the CFDC method provides no direct information regarding the nucleation mechanism, it can be inferred that the concentration of IN measured under conditions below water saturation has most likely nucleated by deposition freezing, whereas those sampled under conditions above water saturation may include those nucleated by immersion, condensation, and depositional freezing [McFarquhar et al., 2011]. The SPLAT provides single particle size for particles in the range of 0.05 to 3 µm diameter and composition of particles in the narrower range of 0.1 to 1 µm diameter. In addition, we report the number of ice crystals estimated by a combination of multiple cloud probes, including the 2DS (10 < D < 1280 mm), 2DC (25 < D < 800 mm), CIP (15 < D < 960 mm or 25 < D < 1600 mm depending on operating mode), and 2DP (200 < D < 6400 mm), mounted on the wing of the aircraft, adapted from Jackson et al. [2012]. A Droplet Measurement Technologies 3-laser Photo-Acoustic Soot Spectrometer (PASS-3) was also deployed on the aircraft for the measurement of aerosol light absorption and scattering at 405, 532, and 781 nm wavelengths [Flowers et al., 2010].

2.2 Computer Controlled Scanning Electron Microscopy and Energy Dispersive X-ray Analysis

[11] Particle samples selected for analysis were first imaged by CCSEM to assess the general coverage of particles on the substrate and their overall visual appearance. Typical images of particle residuals collected through the CVI and ambient particles collected in the ambient sample inlet are shown in Figures 2a and 2b, respectively. Substantially higher loadings of particles on substrates were observed in the ambient samples. This was consistent with variations in number concentrations detected by the in situ particulate and cloud probes. CCSEM/EDX particle analysis was used to determine the elemental composition of hundreds to thousands of individual particles in each sample. A detailed description of the method is published elsewhere [Laskin et al., 2006] and is described briefly here.

Figure 2.

SEM images of two samples: (a) Residuals collected through the CVI inlet. (b) Ambient particles collected through the ambient sampling inlet during the flight of 26 April 2008. Images are taken at 7000X magnification.

[12] An FEI XL30 digital field emission gun environmental scanning electron microscope was used in this work. The instrument is equipped with an EDX spectrometer (EDAX, Inc.), which allows X-ray detection of elements with atomic numbers higher than beryllium (Z > 4), and two solid-state detectors positioned to collect transmitted and backscattered electrons, respectively. During image processing, the transmitted electron signal was mixed with the backscattered signal to create high contrast images of particles. A working distance of 10 mm, an electron accelerating voltage of 20 keV, and a beam current of 400 pA were used for X-ray microanalysis of all particle samples. During each CCSEM/EDX measurement, a designated area of ~380 µm × ~330 µm on the substrate was analyzed. This relatively large area was chosen to account for spatial inhomogeneities of particle samples where larger particles are concentrated in the central region of the sample, and smaller particles scattered over a wider area. [Laskin et al., 2003; Hopkins et al., 2008]. X-ray spectra were acquired for 10 s from each detected particle larger than 0.3 µm. The relative atomic concentrations of 16 elements commonly present in atmospheric particles, i.e., C, N, O, Na, Mg Al, Si, P, S, Cl, K, Ca, Mn, Fe, Zn, and Pb, were analyzed.

2.3 Scanning Transmission X-ray Microscopy / Near Edge X-ray Absorption Fine Structure Spectroscopy

[13] Spectromicroscopic analysis of individual particles was performed on a subset of the selected samples using the STXM instrument at the Advanced Light Source of the Lawrence Berkeley National Laboratory (Berkeley, CA). The STXM instrument has been described in detail in the supporting information and elsewhere [Kilcoyne et al., 2003; Moffet et al., 2010b], and only a brief description is included here.

[14] The STXM technique provides spatially resolved NEXAFS spectra and information regarding spectral components within single particles [Moffet et al., 2010a, 2010b]. A representative NEXAFS spectrum of a single particle collected during the episode of high concentrations of in situ ice nuclei is shown in Figure 3. The spectrum contains discernible peaks at 285.4 eV, 288.5 eV, and 290.4 eV. These peaks arise from the carbon 1s → π*R(C*[DOUBLE BOND]C)R transition for materials containing sp2 hybridized carbon-carbon double bonds (e.g., black carbon (BC)), the carbon 1s → π*R(C*[DOUBLE BOND]O)OH transition in carboxylic groups, and the carbon 1s → π*C*O3 transition in carbonates (CO3), respectively [Hopkins et al., 2007]. To further probe the internal structure of particles, carbon K-edge spectral maps were generated from the STXM/NEXFAS analysis. Maps include four major components identified by STXM/NEXAFS: organic carbon (OC), BC, CO3, and a broadly defined inorganic species (IO). Details on classification of STXM/NEXAFS spectra into these categories are discussed in Moffet et al. [2010a].

Figure 3.

NEXAFS K-edge spectrum of a particle collected during the High IN episode. The solid bold black line is the experimental data, and the solid blue line is the fit using individual peaks labeled in the legend. A list of peak assignments and relative spectral areas is given in Table S3.

3 Results and Discussion

3.1 Sampling Periods and Ambient Conditions

[15] Using CCSEM/EDX and STXM/NEXFAS, we examined the size and composition of mixed residuals of cloud drops and ice crystals from mixed-phase clouds collected during several flights (Flight 25, 30, 31, and 34). Time periods and characteristics of the six samples from four flights included in this study are shown in Table 1. Our analysis extends the work of Zelenyuk et al. [2010], who reported the results of in situ single particle mass spectrometry measurements conducted during Flight 31 of ISDAC. A cloud sorting algorithm was used to select samples corresponding to clear air and within clouds (liquid, mixed, or ice phase) environments [McFarquhar et al., 2007; Jackson et al., 2012]. Classifications of cloud phase for samples used in this study are reported in Table S1 of the supporting information. Throughout the project, in-cloud data were most commonly collected in mixed-phase clouds. Such were the conditions during the collection of residuals on 26, 27, and 29 April. For example, for the majority of time during F30-S19 and F31-S62, the clouds were classified as mixed phase, with ~20% of the time designated as liquid only. A notable exception was the episode of High IN residuals observed on 29 April 2008 (i.e., F34-S66 and S67), which was classified as mixed-phase cloud with patchy areas of clear sky (Table S1). Samples containing mixed-phase residuals were analyzed to infer chemical composition of those particles that facilitate nucleation of droplets and ice crystals. During the High IN period, observed concentration of IN are several orders of magnitude higher than the campaign average. We chose to include analysis of this sample to look for any compositional differences which coincide with elevated IN levels. Nevertheless, it should be noted that the sampling method employed here did not provide segregation between nucleated particles and ambient ones, and thus plausible causality between particle composition and nucleation ability may be inferred, but cannot be directly measured. Additional samples (i.e., F25-S20 and F30-S6) were chosen from those collected on the ambient inlet in time periods corresponding to 100% clear sky cloud-free conditions, to allow comparison between the composition of ambient aerosols and the subset of those aerosols which act as cloud nuclei. Previous studies have shown that when ice crystals impact aircraft sampling surfaces, tiny pieces of the inlet surface may chip off causing contamination to the sample [Cziczo et al., 2004]. To test for this potential problem, we analyzed an additional sample collected under ice cloud conditions (not shown). In that sample, only a very small fraction (< 0.5%) of the particles were metallic, suggesting that the majority of particles sampled are actual residuals, not artifacts.

[16] To assess the origin and trajectory of sampled air masses, we compared 5 days backward trajectories for the four research flights discussed in this manuscript using the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model (Figure 4) [Draxler and Rolph, 2003]. For each flight, results are presented for different flight altitudes corresponding to the mean, highest, and lowest altitudes traveled during the sample collection periods, as shown in Figure 4. Locations of the sample collection periods used for backward trajectory calculations are summarized in Table S2 of the supporting information. On 19 April, the Convair 580 flew through a biomass burning plume originating from Siberia [Brock et al., 2011]. The air mass sampled on this day passed directly over Russia a few days prior to sampling (Figure 4e). Both forest fires in northern Kazakhstan and agricultural burns in southern Russia were occurring during that time [Warneke et al., 2009].

Figure 4.

HYSPLIT 5 day backward trajectory for the sample collection periods: (a) F30-S19: residual samples in mixed-phase clouds on 26 April, (b) F31-S62: residual samples in mixed-phase clouds on 27 April, (c) F30-S6: cloud-free air on 26 April, (d) F34-S66 and 67: High IN episode on 29 April, and (e) F25-S20: the biomass burning period on 19 April. Red lines indicate the backward trajectory of the air mass for the average flight altitude of each flight. Blue and green lines represent the trajectory of air mass from the highest and lowest altitudes flown, respectively. Altitude refers to pressure altitude. Each data point in the time series is a 6 h average.

[17] Additional details regarding ambient conditions, and in situ particle, and ice nucleus measurements performed during the sample collection periods are summarized in Table 2. During flight within the biomass burning plume on 19 April, particle concentrations measured by the TSI CPC were much higher than those in the typical ambient air. The measurements of the SPLAT and PASS-3 instruments also detected the presence of the biomass plume [McFarquhar et al., 2011]. During the collection of cloud-free ambient particles on 26 April, total ambient particle concentrations detected by the TSI CPC were very similar to those collected during the collection of other in-cloud residual samples. As can be seen in Table 2, the concentrations of cloud droplet residues measured by the PMS CPC were an order magnitude lower than ambient particle concentrations due to the inlet operation in CVI mode.

[18] Overall, during the sample periods selected for this study, observed IN concentrations (1 min averages) varied widely, with frequent values of less than 0.01 L−1 to as high as 8.1 L−1, according to the CFDC measurements. As expected, the percentage of particles that activated as IN was higher during the in-cloud samplings (> 6 × 10−3%) than ambient samplings (≤ 2 × 10−4%). Similarly, ice crystal concentrations detected by the cloud probes varied greatly [Jackson et al., 2012]. Under ambient sampling periods including the cloud-free ambient periods and the biomass burning event, no ambient ice crystals were observed. In contrast, during the in-cloud residual sampling on 26–27 April, the clouds contained an average of 0.39 ice crystals per liter. During the unusual conditions encountered in the High IN episode on 29 April 2008 (Flight 34), the CFDC measured an average IN concentration of 5.6 L−1. The highest 1 min average IN concentration, 8.1 L−1, was recorded at 2:29 GMT. In addition, strikingly high concentrations of ambient ice crystals (up to > 50 L−1) were observed during the High IN episode.

3.2 Composition of Ambient Particles and Cloud Droplet Residuals

[19] We first discuss the composition data from ambient particles and cloud droplet residuals from mixed-phase clouds. Samples from cloud-free ambient air (F30-S6) and mixed-phase clouds (F30-S19 and F31-S62) were analyzed by CCSEM/EDX to determine the elemental composition of 5880 and 2380 particles, respectively (Table 1). After data collection, particles were classified using a rule-based classification scheme into five distinct groups: CNO, CNOS, NaMg, AlSiCa, and Other (Figure 5). The first group, CNO, included particles containing only carbon, nitrogen, and oxygen. Oxygenated carbonaceous particles often observed in the troposphere are included in this group [Middlebrook et al., 1998; Hand et al., 2005; Day et al., 2009]. In addition to carbon, nitrogen, and oxygen, many of the particles also contained small amounts of sulfur, possibly from internal mixing with sulfates. These particles were placed in the CNOS group. When other elements (Na, Mg, Al, Si, P, Cl, K, Ca, Mn, Fe, Zn, or Pb) were detected, the particle composition was classified as either sea salt, NaMg, for particles in which [Na] + [Mg] > [Al] + [Si] + [K] + [Ca] + [Mn] + [Fe], or mineral dust, AlSiCa, for particles in which [Na] + [Mg] < [Al] + [Si] + [K] + [Ca] + [Mn] + [Fe]. All remaining particles were categorized as Other. The NaMg group was comprised of particles containing sodium and magnesium salts typically indicative of sea salt [Hopkins et al., 2008; Laskin et al., 2002, 2005, 2012; Moffet et al., 2012].

Figure 5.

Classification scheme to define particle types in the CCSEM/EDX analysis.

[20] In Figure 6, the first two columns show the percentages of particle groups derived from CCSEM/EDX analysis in each particle population. In ambient samples, the majority of particles were classified as CNOS or CNO. Less than 3% of the ambient particles were identified as NaMg. In contrast, 28% of the residuals were NaMg, suggesting that sea salt particles acted as CCN or possibly IN. An elevated fraction of NaMg particles in samples of cloud residuals is consistent with the calculated backward trajectories traveling near the ocean surface (Figures 4a and 4b). Sea salt is well known to enhance a particle's ability to act as CCN [e.g., Cruz and Pandis, 1998].

Figure 6.

Stacked column diagram of particle groups for the studied particle population, including cloud-free ambient particles (F30-S6) and in-cloud residual particles (F30-S19 and F31-S62). The first two columns represent percentages of particle types identified by the CCSEM/EDX analysis. The third and fourth columns represent percentages of particle type identified by STXM/NEXFAS. The colors indicate particle classes distinguished by different combinations of organic carbon (OC) internally mixed with other components: inorganics (IO), black carbon (BC), and carbonates (CO3). The numbers in parenthesis represent the total number of particles.

[21] Many particles classified as NaMg, AlSiCa, and Other also contained substantial amounts of organics (up to 90%) in internally mixed particles. This observed complexity in composition implies that the sea salts and minerals found in our study were aged and mixed with other components. A population dominated by highly processed aerosol during ISDAC was also reported in Zelenyuk et al. [2010]. The majority of particles found in that study were oxygenated organics mixed with sulfates, biomass burning particles, and processed sea salts.

[22] Most importantly, the elemental composition of cloud residuals was different from that of the population of ambient aerosols, indicating that a subset of the ambient population had a composition that enhanced their cloud-nucleating ability. CNOS particles comprised a larger fraction of ambient population than the cloud droplet residues. This result differs from the findings of Zelenyuk et al. [2010], who reported higher sulfate content in cloud droplets residuals of very small size (i.e., 200–400 nm). However, SPLAT measurements report composition of particles in a size range of 0.100–0.8 µm [Zelenyuk et al., 2009] whereas CCSEM/EDX analyses cover particles in a size range of 0.3–2.5 µm [Laskin et al., 2003]. In both ambient particles and cloud droplet residuals, the combined fractions of the two remaining particle classes, AlSiCa and Other, never exceeded 10% of the total particles. This indicates that AlSiCa and Other particles may not contribute substantially to cloud formation.

[23] Following elemental analysis, a total of 71 particles from a cloud-free air (F30-S6) and 73 residuals collected in mixed-phase clouds (F30-S19 and F31-S62) were analyzed by STXM/NEXAFS (Table 1). The spectrum for each sample was deconvoluted to determine the contributions of the various functional groups. Determining relative concentrations from the observed optical depths requires a peak fitting procedure. The functions used for the deconvolution and the associated fitting procedure are described in Moffet et al. [2010a, 2010c] and only briefly presented here. A compilation of the organic functional groups detected by STXM/NEXAFS and the average peak area (for the sample average spectra) are summarized in Table 3. The table includes the energies of specific electronic transitions associated with each functional group. The peak area for the functional group is reported in OD × eV units per particle. Here, we assume that peak area is proportional to mass, which requires that the absorption cross sections do not change significantly with the presence of different functional groups. We note that the Cloud-Free Ambient and In-Cloud Residual cases are examined in this section, and the High IN and Biomass Burning cases are discussed in the following section. The estimated total carbon mass in residuals was smaller than that in ambient particles, suggesting that organics do not enhance cloud nucleation, though it is not clear from our data whether or not organics inhibit nucleation. More interestingly, variations in organic functional groups were also detected. The largest organic particles in the ambient particle population showed a substantial presence of aliphatic and aromatic carbons (292.2 eV). Ambient samples also contained a relatively large amount of carboxylic groups (288.5 eV) and carbonate (290.4 eV) that might be relevant to the ice nucleation during that period. Black or elemental carbon components, characterized by a peak at 285 eV, attributed to sp2 hybridized C[DOUBLE BOND]C bonds were seen in some of the particles [Hopkins et al., 2007]. Unlike the droplet residuals discussed in the following section which were coated, coatings were not observed on these ambient particles. The black carbon and carbonate peaks were more pronounced and better resolved in the residual samples.

Table 3. Peak Areas of Each Spectral Feature Given by the Average Single Particle Carbon K-edge Spectrum
   Peak Area (Optical Depth × eV) ± Uncertainty (%)b
  1. a

    Excited state transition. Chemical bonding information are referred from Moffet et al. [2010a], and references therein.

  2. b

    The peak area is normalized to the total area within individual particles. The uncertainty values are reported as the relative standard uncertainty estimated by the sample standard deviation divided by the sample mean and the square root of the recorded spectra per sample.

  3. c

    This is the area from the step function peaked at 294.5 eV that includes the energy to 320 eV. This area is related to the total carbon present within the sample.

Energy, eVTransitionaFunctionalityBiomass BurningCloud-Free AmbientIn-Cloud ResidualHigh IN (Residuals)
285.41s → π*C*[DOUBLE BOND]C0.17 ± 6.7%0.22 ± 12.4%0.21 ± 12.0%0.22 ± 11.0%
286.5K 1s → π*R(C*[DOUBLE BOND]O)R or C*OH0.19 ± 7.2%0.31 ± 9.3%0.21 ± 8.7%0.28 ± 6.0%
287.7K 1s → C-H*C*H, C*H2, C*H30.54 ± 1.9%0.25 ± 6.0%0.30 ± 8.5%0.35 ± 4.7%
288.5K 1s → π*R(C*[DOUBLE BOND]O)OH0.80 ± 1.5%1.08 ± 3.6%0.76 ± 7.3%0.84 ± 4.2%
289.5K 1s → 3pσ*R-OC*H2-R0.91 ± 1.2%0.68 ± 3.7%0.55 ± 7.7%0.59 ± 3.7%
290.4K 1s → π*C*O30.10 ± 8.4%0.13 ± 13.3%0.14 ± 15.2%0.22 ± 12.8%
297.1 and 299.7L2 2p1/2 → & L3 2p3/2Potassium0.10 ± 13.5%0.29 ± 15.9%0.32 ± 18.1%0.10 ± 13.5%
292.2K 1s → σ*C*-C, C*-O1.47 ± 3.0%1.87 ± 6.8%1.08 ± 10.7%1.37 ± 8.3%
3001s → σ*C*[DOUBLE BOND]C, C*[DOUBLE BOND]O1.45 ± 4.4%2.28 ± 6.5%1.99 ± 6.8%2.11 ± 6.0%
cEdge stepTotal Carbon19.32 ± 0.6%11.57 ± 2.5%7.04 ± 3.8%11.55 ± 2.2%

[24] Particles detected in all samples were sorted into eight major groups based upon the combination of the four major components described above including organic carbon (OC) compounds, inorganic (IO) compounds, carbonate (CO3), and black carbon (BC), which the STXM detects in individual particles, following the procedure described in Moffet et al. [2010a]. The percentage of particles classified into each molecular composition group is shown in Figure 6 (i.e., last two columns). In the ambient particle population, there was a high percentage of particles containing only organic constituents (21%), and mixed OC and IO components (58%). In residual samples, the majority of particles had more complex internal mixing of different components than ambient particles. The fraction of residuals containing more than three components was much larger (55%) than that in the ambient particles (15%), indicating higher compositional heterogeneity of droplet residuals. We note that the composition of analyzed residues may be affected by irreversible changes in drying particles such as formation of organic salts [Laskin et al., 2012], oligomers, organo-nitrates, and organo-sulfates [Nguyen et al., 2012]. Soluble organics can be taken up by activated cloud droplets and modified through in-cloud aqueous chemistry and subsequent drying [He et al., 2013]. Furthermore, unactivated airborne particles can be captured by cloud droplets and then be present in the dry residues [Pekour and Cziczo, 2011]. Regardless, the sum of all groups containing BC and CO3 in residuals (25 and 47%, respectively) increase by more than a factor of two over the ambient particles (11 and 13%, respectively). Further, it is possible to roughly estimate the mass of carbonate relative to other components by taking the area of the carbonate peak and dividing it by the sum of the areas for the other transitions (excluding the transitions at 292.2 and 300 eV because those contain contributions from carbonate). According to this approximation, we find that the mass fraction of carbonate is significantly higher in some of the CVI samples than in the ambient samples. Overall, the estimated mass percentages carbonate are 1.6 and 1.5% in the High IN and in-cloud residuals, respectively, compared to 0.9% in the cloud free ambient sample and 0.5% in the biomass burning episode. These results suggest that these components particularly carbonate, may be relevant to cloud formation in the geographical area of the study. Carbonate ions are present at some level in aerosols collected above in the Arctic Ocean throughout the year [Bates et al., 2009]. Further, calcium carbonate is predicted to be enhanced in Arctic sea salt [e.g., Sander et al., 2006]. Other local sources of carbonate include soil minerals and freshwater streams in Arctic [Keller et al., 2007]. In summary, both the elemental and molecular composition analyses indicate differences between ambient and droplet residual particles.

3.3 Size-Dependent Composition

[25] Because differences in particle composition may be correlated with particle size, it is difficult to determine which characteristic fundamentally drives cloud-nucleating ability. The single particle techniques employed in this study provide a means to survey both size and composition. To distinguish between possible effects of size and composition, particles detected by each technique were sorted into the following approximate size ranges: < 0.5 µm, 0.5 to 0.75 µm, and > 0.75 µm based on an equivalent circle diameter derived from the observed 2-D particle projection area. For each size range, the fractions of particles in each CCSEM/EDX composition category are shown in Figure 7. For the ambient aerosols, notable differences in elemental composition between various size ranges were observed (Figure 7a). First, the fractions of particles classified as AlSiCa and NaMg were slightly higher than average in the largest particles (> 0.75 µm diameter), indicating the presence of larger mineral dust particles and sea salts, respectively. Second, the CNOS fraction was higher in smaller particles (< 0.75 µm) consistent with the characteristic size of sulfate particles [Zelenyuk et al., 2010]. In the residual particles, an increase in the NaMg fraction and a reduction in the CNO fraction above 0.5 µm diameter were observed (Figure 7b). Comparing relative contributions of the NaMg fractions in the cloud residuals and the ambient particles, we see that NaMg particles comprise a larger fraction of residuals than ambient particles, with the greatest enhancement in the largest size bin. This indicates that large NaMg containing particles likely play a selective role in mixed-phase cloud formation. As the relative fractions of AlSiCa are not significantly different between the ambient and the residual particle samples, their potential role in cloud formation is unclear.

Figure 7.

Percentages of particle types identified by the CCSEM/EDX analysis for the following: (a) cloud-free ambient particles (F30-S6) and (b) in-cloud residual particles (F30-S19 and F31-S62). The total population of ambient particles (5880) is subdivided into three particle size ranges (equivalent circle diameter): < 0.5 µm (5210), 0.5–0.75 µm (640), and > 0.75 µm (30). Similarly, the population of residuals (2380) is subdivided into < 0.5 µm (2030), 0.5–0.75 µm (305), and > 0.75 µm (45). The numbers in parenthesis represent the total number of particles in each size range.

[26] STXM/NEXFAS data for ambient and residual particles for each particle size range is shown in Figure 8. The OC and mixed OC IO classes account for the majority of ambient particles in all size ranges (Figure 8a). The OC and OC IO classes comprised more than 85% of the < 0.5 µm diameter particles, and 60% of the > 0.5 µm particles. An enhancement in the percentage of particles containing CO3 was observed in all sizes of residuals compared to ambient aerosols. As seen in Figure 8b, for the residuals, internal mixtures containing the CO3 component were 26% (< 0.5 µm), 67% (0.5 to 0.75 µm), and 60% (> 0.75 µm). These fractions were larger than the fractions in ambient particles in the same particle size ranges. The sum of all BC containing particles in the residuals (25%) is more than a factor of two higher than detected among the ambient particles (11%). Enhancement in BC containing particles in residuals was more pronounced in the two smallest size ranges. The fraction of residuals containing more than three classes incrementally increased with size, indicating that mixing complexity increases with particle size. The presence of three or more classes was more common in the residuals than in the ambient particles. This suggests that internal mixing may facilitate cloud-nucleating ability [Möhler et al., 2008].

Figure 8.

Stacked column diagrams of particle classes identified by the STXM/NEXAFS analysis. Fractions of the particle classes within each size range of the (a) cloud-free ambient particles (F30-S6) and (b) in-cloud residual particles (F30-S19 and F31-S62), respectively. Particle classes are distinguished based on the particle mixing state of organic carbon (OC), inorganics (IO), black carbon (BC), and carbonate (CO3) components. The total population of ambient particles (71) is subdivided into < 0.5 µm (50), 0.5–0.75 µm (18), and > 0.75 µm (3). Likewise, residuals (73) are subdivided into < 0.5 µm (34), 0.5–0.75 µm (24), and > 0.75 µm (15). The numbers in parenthesis represent the total number of particles in each size range.

[27] Overall, our results suggest that both size and chemistry of cloud droplet residuals are different from the ambient particles. Notably, sea salt particles were prominent in the residual samples. Our results also suggest that the particles containing CO3 and BC are common constituents in droplet residuals. No clear correlation between the fraction of BC and CO3 containing residuals and particle size was observed. Nevertheless, the observed differences in composition of cloud residuals and ambient particles may be attributed to different CCN and IN propensities of different particle types and to the physico-chemical processes occurring in the aqueous phase of cloud droplets. Perhaps, the significant mixing observed in residuals may be a reflection of their wet chemistry with organic acids [Laskin et al., 2012; Chou et al., 2013].

3.4 Special Cases: High IN Activity Episode and Biomass Burning Plume

[28] The CCSEM/EDX and STXM/NEXAFS classifications of residuals collected through the CVI during High IN episodes separated into three size ranges are shown in Figure 9. The CCSEM/EDX data (Figure 9a) shows that the fraction of large (> 0.75 µm) particles is greater in High IN residuals (8%, i.e., 120 out of 1530 particles) than any other particle population used in our study (< 2%, i.e., Figures 7 and 11), suggesting that the presence of larger particles might be a unique feature of the High IN episode. Also, the percentage of particles containing sodium and magnesium salts was higher during the High IN episode than those during any other analyzed period. The abundance of sea salt particles may trigger immersion freezing, resulting in the high concentrations of ice nuclei observed during this period [Wise et al., 2012; Wagner and Möhler, 2013]. We note that, even in this highest IN case, INs are only a very small subset of the sampled particles, accounting for ~0.026% of the total particles (Table 2). High ice nucleation efficiency of sea salt particles has been observed by others, though it has yet to be determined whether observed nucleation efficiency is due to the presence of sea salt or other marine components such as sulfates or dissolved organic carbon present in the lofted particles during periods of marine biological activity [Schnell, 1977; Rosinski et al., 1986, 1987; O'Dowd et al., 1997; Leck and Bigg, 2005a, 2005b; Wise et al., 2012]. In this study, the relatively large size of sea salt particles cannot be ruled out as a factor contributing to high cloud-nucleating activity (Figure 9a). Additional differences were observed between the cloud particle residuals in the High IN episode and residuals in other periods (Figure 9b). Notably, the black carbon fraction in the High IN samples was larger than any other residual samples and ambient particles. The High IN residuals also contained a larger fraction of the carbonate containing particles (total average fraction ~30%) than the ambient sample (13%) and the biomass sample (16%). As in other residual samples, the percentage of residuals during the High IN episode containing more than three classes incrementally increased with size. During the High IN episode, the fraction of particles containing CO3 also increased with size. In contrast, the fraction of High IN residuals containing BC component was similar among size bins, and particles containing only organic species (OC) were smaller than 0.5 µm in size. All particles larger than 0.5 µm were composed of internal mixtures having at least two components.

Figure 9.

Percentages of particle types identified by (a) CCSEM/EDX and (b) STXM/NEXAFS analysis for the samples collected at High IN episodes (F34-S66 and F34-S67), subdivided into three particle size ranges. The total population of particles analyzed by the CCSEM/EDX (1530) is subdivided into < 0.5 µm (1060), 0.5–0.75 µm (350), and > 0.75 µm (120). Likewise, particles analyzed by the STXM/NEXAFS (91) are subdivided into < 0.5 µm (43), 0.5–0.75 µm (30), and > 0.75 µm (18). The numbers in parenthesis represent the total number of particles in each size range.

[29] Figure 10 illustrates examples of STXM/NEXAFS maps of particles collected in the High IN residual samples. As shown in the figure, the residuals had characteristic organic coatings and are relatively large. These particles contained BC, IO, and CO3 components mixed with organic constituents. All particles contained carboxylic groups as shown by the green areas surrounding the cores in the figure. These results suggest that inorganic species and/or coated black carbon may be essential components of effective cloud nuclei. None of the maps of residuals collected in our samples contained evidence of uncoated black carbon or pure carbonate particles. One caveat of these measurements is that the spatial distribution of components within these residuals could be modified as a result of particle processing or through drying of the original droplet or ice crystal in the heated part of the CVI sample line. Thus, we cannot say for certain that the particle morphology observed here, i.e., inorganic cores surrounded by organic coatings, is identical to that in the atmosphere. Nevertheless, these results suggest that atmospherically aged black carbon (soot) and carbonate particles may have potential relevance to cloud formation in the Arctic.

Figure 10.

Characteristic STXM/NEXAFS maps of particles in High IN residual samples. Colors in the plot represent the dominant component at each pixel. Residues contain cores (composed of inorganic, black carbon, and carbonate) coated with organic material.

[30] Figure 11 shows the size-resolved CCSEM/EDX and STXM/NEXAFS analysis for ambient particles collected during the biomass burning episode. As can be seen in CCSEM/EDX results (Figure 11a), particles collected during the biomass episode had a high fraction of sulfur containing particles, 57%, which is similar to that of ambient particles, 51% (See Figure 7a). The STXM/NEXAFS data (Figure 11b) shows that particles with dominant organic content accounted for the vast majority of particles found in the biomass burning samples. Organic particles comprised up to 85% of the < 0.5 µm diameter particles, and 60% of the 0.5 to 0.75 µm particles. Overall, the presence of particles with three or more components was somewhat lower in the biomass particles than in the ambient particles, and much lower than in the cloud droplet residuals. This trend is true especially for the particles in the first two smallest size ranges, which comprise > 90% of the total population of biomass particles.

Figure 11.

Stacked column diagrams of particle classes identified by the (a) CCSEM/EDX and (b) STXM/NEXAFS analysis for the biomass burning particles (F25-S20). The total population of particles analyzed by the CCSEM/EDX (950) is subdivided into < 0.5 µm (790), 0.5–0.75 µm (150), and > 0.75 µm (10). Likewise, particles analyzed by the STXM/NEXAFS (124) are subdivided into < 0.5 µm (73), 0.5–0.75 µm (40), and > 0.75 µm (11). The numbers in parenthesis represent the total number of particles in each size range.

[31] Figure 12 shows representative STXM/NEXAFS maps of particles from the biomass burning plume. Despite the wide range of bonding and particle morphology previously observed in biomass particulates [e.g., Hopkins et al., 2007; Zelenyuk et al., 2010], here we observe that the majority of particles contain only organic material and are much smaller in size than the High IN residuals discussed above. The larger fraction of smaller particles (< 0.5 µm diameter) without identifiable cores is similar to the ambient particles described in Figure 8a. Although biomass burning particles have been reported to act as effective IN [Stohl et al., 2006; Prenni et al., 2009], our in situ CFDC measurements of IN number concentrations did not substantiate this (Table 2). While the concentration of IN during this episode was similar to that during the High IN episode, this can be attributed to the much larger aerosol concentrations. In fact, according to CFDC measurements, the percentage of particles activated as IN is 2 orders of magnitude lower than during the High IN period, though higher than during the cloud-free ambient case.

Figure 12.

Characteristic STXM/NEXAFS maps of particles in biomass burning samples. Colors in the plot represent the dominant component at each pixel. The representative particles have smaller diameter compared to the residuals and no identifiable cores.

[32] Table 3 includes an assessment of broad differences between particles collected during biomass burning and High IN episodes. The total carbon content was highest in the biomass samples, and lower in other samples. In the biomass plume, elevated levels of ethers (289.5 eV) were also found, presumably due to the presence of sugars in the biomass burning particles [Hays et al., 2005]. The presence of sp2 hybridized C[DOUBLE BOND]C bonds and carbonate were observed in all samples. However, the intensity of these peaks relative to the total area is slightly larger during the High IN episode based on comparison of peak heights in the NEXAFS spectra. This is likely indicative of coated black carbon and carbonate that may have relevance to Arctic cloud formation. However, we note that these chemical properties may not conclusively provide an indication of the particle properties that cause ice nucleation because IN were not segregated from the ambient particles. As described above, all residuals were collected through the CVI inlet and thus represent both CCN and IN present in the air, not only particles that are effective IN. The spatial distribution of clouds was patchy, and liquid droplets and ice crystals may vary depending on apparent supersaturations at time, even within a single cloud. It is also possible that our analyses were insensitive to small changes in total composition that may abruptly enhance ice nucleation.

4 Conclusions

[33] Selected samples of airborne particles collected during the ISDAC 2008 field study were analyzed to probe the chemical composition of aerosols in ambient air and the contents of dry residuals of cloud droplets and ice crystals in mixed-phase clouds, including a high ice nucleation episode, as well as composition of particles during a biomass burning plume episode. The CCSEM/EDX analysis provided with electron microscopy imaging and elemental composition of thousands of individual particles, while the STXM/NEXAFS analysis allowed mapping of major molecular features within single particles and analysis of internal mixing of hundreds of individual particles.

[34] The results from our microscopy analyses should be treated with caution because of the modest number of samples analyzed. This said, our results suggest that, on average, cloud droplet residuals are larger than ambient particle residuals. This suggests larger sizes may increase the propensity of aerosols to act as CCN or IN. Our collected data shows clear compositional differences between ambient aerosols and cloud droplet residuals. The frequencies of carbonate and black carbon were ~four and two times higher in residuals than in the out-of-cloud ambient particles. This suggests that these components may be relevant to cloud formation in this geographical area of the study. In addition, significant mixing was observed in residuals, which may be a reflection of products formed through wet chemistry with organic acids [Laskin et al., 2012]. This strong composition dependence to those aerosols acting as cloud nuclei is interesting, given that the majority particles of most compositions in the size range sampled here (> 0.25 µm diameter) would be expected to activate, unless they contain significant concentrations of either insoluble or hydrophobic components.

[35] During the High IN episode, the percentage of sodium and magnesium salts in the mixed-phase cloud residuals was substantially larger (49% of population) than other episodes (28% or less in all cases), which implies that these salts may improve the ice nucleation efficiency. The fraction of cloud residuals containing BC and carbonate components observed during that episode was similar to other cloud droplet residual samples. In addition, all High IN residuals analyzed by the STXM/NEXAFS were characterized by cores containing an insoluble or low solubility component (i.e., black carbon or inorganic compounds) and coated by carboxylic organics. This suggests that aerosol aging processes contribute to the Arctic cloud formation. During a flight through a biomass burning plume that originated from Siberia, the collected particles were composed largely of organic material without identifiable cores. Analyzed particles from the biomass sample contained fewer black carbon and carbonate content than the residual samples. Fractions of particles containing BC and carbonate in biomass burning plume samples (4% and 16%, respectively) were similar to ambient particles and substantially lower than those characteristic of cloud droplet residuals. Our measurements provide no indication that biomass burning particles have ice nucleation ability superior to ambient aerosol sampled during other time periods.

[36] In summary, our results suggest two things. (1) The number and composition of the cloud-nucleating particles in the Arctic can be significantly altered by episodic events. (2) The composition of cloud droplet residuals differs from ambient particle residuals, which suggests that chemical composition plays a key role in determining which subset of the aerosol population serves as cloud nuclei. Our data imply that carbonate provides a clear chemical advantage for the Arctic mixed-phase cloud nucleation over other aerosol components, and that sodium and magnesium salts facilitate cloud nucleation due to either chemical or physical characteristics for the episodes analyzed in this study. Future laboratory studies, in which only one variable (either aerosol size or composition) is varied at a time, are needed to provide further insight on the fundamental mechanism of ice nucleation and the Arctic cloud formation (Ma et al., submitted manuscript, 2013). In addition, our measurements provide evidence that residuals of cloud droplets have characteristic organic coatings and are relatively large, suggesting that aerosol aging and a higher degree in internal mixing state enhances cloud-nucleating ability. Accurate predictions of aerosol nucleation ability in the Arctic region require consideration of aerosol aging effects on particle composition and mixing state in addition to differences in aerosol size.

Acknowledgments

[37] The authors gratefully acknowledge financial support provided by the Atmospheric System Research program of the Department of Energy's office of Biological and Environmental Research. Data were obtained from the Atmospheric Radiation Measurement Program sponsored by the U.S. Department of Energy, Office of Science, Office of Biological and Environmental Research, Climate and Environmental Sciences Division. S. Brooks acknowledges National Science Foundation NSF-CAREER program, Award 054875. N. Hiranuma acknowledges a Summer Research Institute Fellow in Interfacial and Condensed Phase Chemical Physics of the Pacific Northwest National Laboratory. R. C. Moffet acknowledges additional financial support from a Lawrence Berkeley National Laboratory Seaborg Fellowship. The work of G. McFarquhar was supported by BER, DOE under grants DE-SC0001279 and DE-SC0008500. The STXM/NEXAFS particle analysis was performed at beamlines 11.0.2 and 5.3.2 at the Advanced Light Source at Lawrence Berkeley National Laboratory. The expertise of A.L. Kilcoyne and T. Tyliszczak for the STXM work is gratefully acknowledged. The work at the Advanced Light Source was supported by the Director, Office of Science, Office of Basic Energy Sciences, of the U.S. Department of Energy under Contract DE-AC02-05CH11231. The CCSEM/EDX particle analysis was performed in the Environmental Molecular Sciences Laboratory, a national scientific user facility sponsored by the Department of Energy's Office of Biological and Environmental Research at Pacific Northwest National Laboratory. PNNL is operated by the U.S. Department of Energy by Battelle Memorial Institute under Contract DE-AC06-76RL0.

Ancillary