Insoluble aerosol particles trapped in glacial ice provide insight into past
climates, but analysis requires information on climatically relevant particle
properties, such as size, abundance, and internal mixing. We present a new
analytical method using a time-of-flight single-particle mass spectrometer
(SPMS) to determine the composition and size of insoluble particles in
glacial ice over an aerodynamic size range of
Aerosol particles play a significant role in global climate both directly, through regulation of regional and global radiation budgets and atmospheric photochemical cycles, and indirectly, through their ability to modulate cloud microphysical and precipitation processes. In both cases, the abundance, size, morphology, and composition of the aerosols dictate their ability to affect climate (Seinfeld and Pandis, 2006). As such, the ideal atmospheric particle measurement has been described as one that is able to “count each particle and report its size, chemical composition, and morphology rapidly” (Wexler and Johnston, 2011).
Inferences of past climates have been achieved using ice core records, whose chronologies provide direct and high-resolution records of past atmospheric composition (Legrand and Mayewski, 1997). A large number of methodologies have been developed to determine properties of aerosols in snow and ice. However, all have important limitations.
So-called “offline” particle retrieval methods, requiring a precursor step to first concentrate particles prior to analysis, have been widely employed for ice core studies. Offline studies typically use a pre-existing analytical instrument and are therefore of low expense, requiring only that the sample be transported to the instrument. When paired with quantitative analytical techniques such as Raman (Sakurai et al., 2011) and sublimation energy dispersive X-ray spectroscopy (Drab et al., 2002; Iizuka et al., 2009; Oyabu et al., 2015), as well as optical techniques such as transmission and scanning electron microscopy (Murr et al., 2004; Ellis et al., 2015), offline methods can achieve determinations of particle size, composition, and morphology. There are limitations of offline techniques, however, including the time and resources required for sample recovery and transport, and these limitations often render such methods insufficient for acquiring contiguous, high-resolution particle records from ice cores. Further, offline methods can increase the chance of sample contamination and/or particle losses occurring during handling, transport, and storage (Ohata et al., 2011).
“Online” instrumentation, in contrast, allows for rapid, real-time
delineations of ice core particulate properties, necessary for
high-resolution measurements of particle mass-concentrations and (or)
size distributions along an ice core. For example, Coulter counter (CC) or
laser particle detection (LPD) based instrumentation is typically used for
determining insoluble particle (“dust”) size distributions in glacial ice
(Hamilton and Langway, 1967; Knipperz and Stuut, 2014). However, neither CC
nor LPD techniques provide associated information on particle composition or
morphology; broad assumptions of both parameters are thus required to infer
ice core dust concentrations from these instruments. Alternate online
techniques, such as inductively coupled plasma mass spectrometry (ICP-MS)
and ion chromatography (IC), are also routinely used to determine elemental
(e.g., Cd, Ce, Zn, Pb) and soluble ion (e.g., Na
The recent advent of a commercially available intracavity laser-induced incandescence photometer, the Single Particle Soot Photometer (SP2; Droplet Measurement Technologies, Inc., Boulder, CO) has allowed for a new class of online measurements where both concentrations and size distributions of (refractory) carbonaceous particles in snow and ice can be recovered on a per-particle basis (McConnell et al., 2007; Kaspari et al., 2011; Ginot et al., 2014; Lim et al., 2014). McConnell et al. (2007) were the first to measure past black carbon deposition continuously along a Greenlandic ice core using SP2 instrumentation. More recent SP2 studies have highlighted the necessity of quantifying instrumental errors and systematic uncertainties when extracting trace amounts of highly heterogeneous material from snow and ice (Schwarz et al., 2012; Ohata et al., 2013; Wendl et al., 2014). Specifically, these studies have shown that correction factors, primarily a result of the efficacy achieved in nebulizing insoluble particles from a melted ice core matrix and aerodynamically delivering them to the SP2 for analysis, can be large and vary significantly between laboratory setups (Lack et al., 2014; Katich et al., 2017).
To date no single online method has proven capable of determining the size
and chemical composition for all particle types situated in glacial ice.
Here we expand upon the advantages of the SP2-based methodology by using a
time-of-flight (ToF) single-particle mass spectrometer (SPMS), allowing for
analyses of small aerosol masses (
Positive spectra classification.
Negative spectra classification.
The PALMS instrument has previously been used for in situ measurements of airborne
particles in both laboratory settings and airborne and ground-based
field campaigns. The PALMS instrument has been described previously (Murphy
and Thomson, 1995; Thomson et al., 2000; Cziczo et al., 2006,
2013). In brief, accumulation- to coarse-mode (
Since PALMS has a single ToF mass spectrometer, spectra are limited to
either positive or negative ions during a given sampling period (Murphy and
Thompson et al., 1997a, b). The classification of a particle in either the
positive or negative ion mode depends on that particle's chemical makeup;
e.g., a mineral dust particle identified in the positive ion mode is
typically associated with metal-oxide (Na
The ice core samples used in this study are from two distinct locations in
west-central Greenland. DS14 is situated on the Disko Island ice cap
(69
Core processing involved discretization of samples via stainless steel band
saw cuts followed by standard sample preparation procedures for firn samples
(Osterberg et al., 2006), which included manually shaving the outer 4–5 mm
of firn/ice from each sample using a pre-cleaned ceramic ZrO microtome blade
under a laminar flow clean bench. All samples remained chilled at
To delineate ZrO ceramic or metal (stainless steel) artifacts derived from
core processing, band saw cuts were made on frozen ultrapure water controls
(18.2 M
Nebulization was used to aerosolize particles after melting the ice samples.
The condensed-phase water was then evaporated in the low-humidity
nebulization flow, thereby releasing particles. In order to quantitatively
analyze the production of aerosol, the liquid solution was nebulized and
transported to PALMS at known flow rates. Figure 1a shows a schematic of the
experimental setup. The melted sample was placed into a glass container attached
to a custom Collison-type atomizer. Dry, inert carrier gas (N
An important aspect of analyzing particles re-aerosolized from a liquid suspension, as described in the last section, is the homogenous dilution of soluble components of the original aerosol in the liquid water matrix. Following evaporation of the condensed water during nebulization and transport (Fig. 1a), this material is formed from two distinct processes during nebulization and transport to PALMS. First, nebulized droplets not containing an insoluble residual particle evaporate to form small particles comprised solely of the solute ions in the original droplet. Second, droplets containing insoluble residual particles are “coated” with the soluble ions following evaporation of the droplet's condensed liquid. This may be visualized as a slurry of insoluble mineral dust particles in a water matrix containing a dissolved salt. Nebulized droplets of the slurry may or may not contain a mineral dust particle with the frequency dependent on the concentration of the dust; however, all droplets will contain salt ions with the amount dependent on the concentration in the solution. In this simple system, nebulization would produce either pure salt particles or mineral dust particles coated by salt. We refer to the latter as “post-aqueous coating”. This process is illustrated in Fig. 1b.
Residual particles composed purely of soluble material after removal of
condensed-phase water were not typically observed. This is likely because
the amount of soluble material in
A particle extraction efficiency curve was developed to quantitatively
determine the number of particles analyzed by PALMS versus the initial
number concentration of particles in solution. The extraction efficiency,
The determination of
A mass concentration can be determined with a statistically representative number of particles analyzed by PALMS. Here we define a statistically representative number of particles as the minimum number required to develop a particle size distribution such that outliers, particularly large particles, do not apply erroneous weighting to subsequent mass determinations. In the context of incandescence-based single-particle methods (i.e., SP2), Schwarz et al. (2012) have suggested 10 000 particles per sample.
As defined by Wendl et al. (2014), an “external” calibration approach, as
commonly employed in SP2-based measurements of refractory black carbon in
glacial snow and ice, assumes that
The normalized logarithmic size distribution of retrieved particles, i.e.,
those measured by PALMS, for the
The size-dependent extraction efficiency curve determined from the
experimental data is illustrated in Fig. 2. The 657 nm diameter particle
size showed the greatest efficiency, determined via Eq. (1) to be
4.0
Most particle losses are assumed to occur via “dumping” of excess flow
(Fig. 1a). Since the PALMS aerodynamic inlet allows a maximum inlet flow
rate of 0.44 L min
Provided information on
The SP2 represents the most common online, single-particle method currently
used for measuring particles in glacial snow and ice, providing a benchmark
for methodological comparisons to SPMS. The methodology presented here was
constructed to allow for a broader size distribution of aerosols sent to
PALMS compared to SP2 nebulization techniques used in most prior ice core
studies (McConnell et al., 2007; Ohata et al., 2011, 2013; Schwarz et al.,
2012; Wendl et al., 2014) but with much lower efficiency. For example, Ohata
et al. (2011) achieved a max efficiency of
More recent studies have achieved high nebulization efficiencies
(
4931 spectra (2362 positive and 2569 negative) were analyzed from DS14 (DS14-01 to DS14-05) and 553 spectra (233 positive and 320 negative) from GW14 (GW14-01 to GW14-04). The main discrepancy between the number of particles measured between the two sites is not particle loading but is due to 60 min of sampling time for DS14-05 (as opposed to 10 min for all other samples) in the positive and negative ion modes. Categorization is broadly divided into natural and anthropogenic sources, as previously noted in modern-era particles from Greenland (Drab et al., 2002; VanCuren et al., 2012), but with some overlap. For example, biomass burning and mineral dust now come from both source types. Representative particle spectra for all classes, in both the positive and negative ion modes, are provided in Appendix 2.
Particle class abundances at the DS14 and GW14 sites and their relative locations in Greenland (lower left). The Ca-rich and processed particle sub-classes discussed in the text are associated here with the mineral/metallic classes. Likewise, biological and inorganic P-rich sub-classes are associated with the sulfate/organic class. Particle classes below the 1 % abundance level are not shown and contaminant particles from ice core processing were eliminated before data analysis. See Sect. 3.2 for additional details.
In positive ion mode, mineral dust is distinguished using primary alkaline
markers, Na
The biomass-burning category contains signatures of carbon-cluster isomers,
elevated K
Log-normalized 3 pt smoothed size distributions for particle
classes containing > 100 particles at DS14
Soot is distinguished in both the positive and negative ion modes by the
presence of elemental carbon (e.g., C
The inclusion of sea salt in both the positive and negative ion modes suggests an artifact arising from post-aqueous processing. This class was observed at both GW14 and DS14 but at low abundance (< 1–2 %). Sea-salt aerosol would have dissolved upon melting of the ice samples prior to analyses. This classification may therefore represent evaporation of an uncommonly large droplet or a coating where the underlying particle composition is not observed. In addition, a non-negligible percentage of the positive ion mode particles from both sites were not readily characterized, with “unclassified” particles comprising 8 and 6 % of particles at DS14 and GW14, respectively. Manual inspection of spectra from this class indicates particles containing key markers from numerous classes (Cziczo et al., 2013), thereby leading to confusion of the classification method. This may represent particles undergoing severe post-aqueous processing or coagulation within the solution or after nebulization and led to the work discussed in more detail in the next section.
We note that a single-particle mass spectrometer similar to PALMS was recently used to measure chemical composition of insoluble particles in precipitation samples collected in Sierra Nevada (Creamean, et al., 2013, 2014, 2016). In those studies, post-aqueous processing was not investigated in detail and single-particle classifications were not adjusted based on the possibility of processing.
Atmospheric particles are not chemically homogenous and often contain both
soluble and insoluble components (Murphy, 2005). Several particle classes
presented here, specifically the sea-salt and sulfate/organic classes and,
to a lesser extent, the P-rich and Ca-rich classes, were inferred to contain
an appreciable proportion of soluble components, though of indeterminate
origin (i.e., of atmospheric or post-aqueous processing origin). To test the
degree of post-aqueous coating to these four particle classes, four negative
ionic species were chosen as class-representative markers: the ions
PO
Median normalized peak intensity for PO
For nearly all species the observed peak areas decreased for the filtered
(insoluble particle-free) experiments to the DI and filtered particle tests
(Fig. 5). Wilcoxon rank sum tests were used to determine whether the difference in
the median peak area between the two experiments (i.e., the DS14-06 sample run
versus the DI and the PSL-doped DS14-06 sample run) was statistically significant. In both the
PO
We found the number of particles measured during the 10 min sample runs
described above were insufficient to quantitatively discern mass
concentration within the ice. To test the outcome of longer sample runs,
sample DS14-05 was run for an hour in both the positive and negative ion
modes. This resulted in multiple classes containing > 500 particles, the threshold number of particles employed here for
class-dependent mass concentration measurements, but still an order of
magnitude smaller than the number suggested by Schwarz et al. (2012) for
laser-induced incandescent methods. Sulfate/organic-rich particles were not
considered due to the uncertainty regarding post-aqueous processing (see
prior section). Representative class densities were taken as 0.8 g cm
DS14-05 particle relative abundance (RA) and calculated mass
concentration values. The values of
Direct comparison of these concentrations to previously published values is
complicated by differences in the particle size range retrieved by various
analytical techniques as well as site-to-site differences. For example,
CC techniques can extend the size range of insoluble particle
retrieval up to 50
A more direct comparison of SP2 vs. SPMS methodologies is available
utilizing black carbon measurements made recently on DS15, a firn core
collected in 2015 adjacent to the DS14 site. Using the SP2 method (following
the methodology of McConnell et al., 2007; see also Wendl et al., 2014), a
mean black carbon concentration in the upper 5 m of DS15 was found to be
0.78
While our results show potential exists for using SPMS to determine insoluble mass concentrations of particles in snow and ice, they also identify areas where more work is needed before SPMS can be used as a quantitative tool. These include (i) executing multiple extraction efficiency calculations as a function of particle class (in addition to size), (ii) incorporating regularized tests for drifts in SPMS extraction efficiency and (or) regularly employing “blank” tests between sample measurements in order to improve delineation to changes in background particulate levels, (iii) achieving a greater number of particle measurements, either through improvements in particle extraction/PALMS transmission or longer sample integration times, and (iv) comparing SPMS-derived particle concentrations with results from well-founded alternate high-precision instrumentation (e.g., an Ultra-High Sensitivity Aerosol Spectrometer (UHSAS; Droplet Measurement Technologies Inc., Boulder, CO) or CC techniques).
In this study we develop and apply a new methodology utilizing SPMS to characterize particulates trapped in snow and ice at the single-particle level. We show that a single instrument, in this case PALMS, can be used to discern the aerodynamic size, composition, and concentration of insoluble ice core particulates. This online method reduces preparation time and resources required for filter-based particle retrieval methods. Based on compositional differences in particles found in two Greenlandic firn cores, we define eight distinct particle classes in the positive ion mode and six in the negative ion mode (Tables 1 and 2; Appendix 2). These classes are a combination of common atmospheric types previously described in the literature and types relatively rare in the atmosphere but common in the ice samples. Differences in the relative abundances of classes found between the two sites are consistent with the sites' unique geography and climatology, notably marine versus high-altitude/inland locations. Furthermore, this study demonstrates the feasibility of using PALMS to infer a sample's mass concentration using an external calibration of a size-dependent extraction efficiency which parameterizes the nebulization, delivery, and ablation or ionization of particles from concentration slurries (Wendl et al., 2014).
We find that two classes, sulfate/organic and sea-salt particulates, appear to be artifacts of post-aqueous processing, highlighting the need for further evaluation of these particle classes in future studies. More broadly, this phenomenon is a critical feature of particles analyzed from a liquid suspension, which has not been fully appreciated in previous studies (e.g. Ault et al., 2011; Creamean et al., 2014): completely soluble aerosols or soluble fractions of particles are homogenized into the melt solution from which insoluble components are extracted. These components are then present as small, solute-only particles and as a surface coating on insoluble particles. The former were typically smaller than the PALMS threshold size but this need not be the case in future studies using different instruments. While the concentration of soluble material can be reduced by dilution, at this time there is no solution to quantifying the original soluble aerosol, finding the original association of soluble with insoluble components, or completely removing this material from the analysis step.
We group the future application of SPMS to ice core studies into qualitative
and quantitative uses. We show that SPMS can be readily used for qualitative
analysis, as is done in atmospheric studies (Murphy, 2005); future
application may include particle classification and source allocation using
mass-spectral compositional markers. Quantitatively, we show that mass
concentration may be feasible but with long sample periods required for
statistically relevant conclusions at current instrument rates. Measurement
frequency during an hour-long period registered
Data used to generate the results figures are included in a Harvard Dataverse dataset with the same name as this paper
(Osman, 2017,
Here, we derive the determination of nebulization efficiency (
In this section, we provide exemplary particle spectra for each class, including the traditional particulate classes using the algorithm described in Cziczo et al. (2013), as well as the additional particle classes used in this study, in both the positive and negative ion mode. All major peaks corresponding to known ionic fragments have been labeled (see, e.g., Murphy and Thompson, 1997a, b; and Cziczo et al., 2013, for details).
Exemplary spectra taken from the DS14 and GW14 samples for the traditional positive ion mode particle classes using the classification algorithm described in Cziczo et al. (2013). From top to bottom: mineral/metallic, organic, biomass-burning, soot, and heavy oil combustion/vanadium-rich particles.
Exemplary spectra taken from the DS14 and GW14 samples for three positive ion mode classes commonly observed in the ice core data. From top to bottom: processed, Ca-rich, and contamination/heavy-metal particles. These classes were originally categorized in the mineral/metallic positive ion category using the algorithm of Cziczo et al. (2013).
Exemplary spectra taken from the DS14 and GW14 samples for the traditional negative ion mode particle classes using the classification algorithm described in Cziczo et al. (2013). From top to bottom: mineral, sulfate/organic, sea salt, and soot particles.
Exemplary spectra taken from the DS14 and GW14 samples for two negative ion mode classes commonly observed in the ice core data. From top to bottom: biological and P-rich inorganic particles. Both classes were originally categorized in the sulfate/organic negative ion category using the algorithm of Cziczo et al. (2013).
MO helped design the experiment, prepared the ice core samples, analyzed the samples, contributed to the data analyses, and wrote the manuscript. MZ helped design the experiment and contributed to the data analyses. SBD collected the DS14 and GW14 ice cores and assisted with ice core sample preparation. DJC helped design the experiment and supervised the laboratory work. All authors participated in writing and editing the manuscript and contributed to the interpretation of the results.
The authors declare that they have no conflict of interest.
This work was supported by an internal Reed Grant from MIT and National Science Foundation award PLR-1205196 to Sarah B. Das. Matthew Osman acknowledges government support awarded by DoD, Air Force Office of Scientific Research, National Defense Science and Engineering Graduate (NDSEG) Fellowship, 32 CFR 168a. Maria A. Zawadowicz acknowledges the support of NASA Earth and Space Science Fellowship. Daniel J. Cziczo acknowledges the support of the Victor P. Starr Career Development Chair at MIT. We thank two anonymous reviewers, whose comments greatly improved the content of this manuscript. Edited by: Joachim Curtius Reviewed by: two anonymous referees