Introduction
Siberia, a vast region in central Eurasia, has been gaining growing attention
from atmospheric aerosol researchers in the last few decades. The atmospheric
aerosol over Siberia is of particular interest for several reasons. Firstly,
biogenic emissions of volatile organic compounds (VOC) from the vast boreal
taiga forest are thought to lead to the formation of secondary organic
aerosol (SOA) (Tunved et al., 2006). Secondly, Siberia has been documented to
be an important source region of biomass-burning aerosol particles that are
distributed around the globe in the free troposphere (Conard and Ivanova,
1997; Müller et al., 2005; Warneke et al., 2009). Thirdly, Siberia is one
of the few possible background regions in the Northern Hemisphere where
near-pristine conditions prevail for certain periods of the year (Chi et al.,
2013). Such atmospheric observations in remote areas are very important for
providing a reference for evaluating anthropogenic impacts in this and other
regions (Andreae, 2007; Carslaw et al., 2013; Spracklen and Rap, 2013).
Aerosols influence the radiative budget of the Earth's atmosphere in two
different ways. The first is the direct effect, whereby aerosols scatter and
absorb solar and thermal infrared radiation, and thus alter the radiative
balance of the Earth–atmosphere system. Aerosol particles with a diameter
that is comparable to the wavelength of solar radiation (0.2–2 µm)
are the most effective light scatterers (Waggoner et al., 1981), which makes
organic carbon and some inorganic species (e.g., sulfate, nitrate, ammonium)
in the sub-micrometer size range typically the most effective chemical
components of aerosol light scattering. The second is the indirect effect,
whereby aerosols modify the microphysical and hence the radiative properties
and lifetime of clouds (Haywood and Boucher, 2000; Rastak et al., 2014).
These indirect effects of aerosols result from their CCN (cloud condensation
nuclei) and IN (ice nuclei) activity (Twomey, 1997; Ogren and Charlson,
1992).
The hygroscopic properties of atmospheric aerosol particles are vital for a
proper description of these effects, since they describe how the particles
interact with water vapor both under sub- and super-saturated conditions
(e.g., McFiggans et al., 2006; Swietlicki et al., 2008; Rastak et al., 2014).
They are thus of major importance for describing the life cycle of the
aerosol and the related direct and indirect effects on climate.
The hygroscopic properties of aerosol particles in the northern European
boreal forests under sub- and super-saturated conditions have been studied
extensively using the Hygroscopicity Tandem Differential Mobility Analyzer
(HTDMA) and size-resolved CCN counter (Hämeri et al., 2001; Ehn et al.,
2007; Birmili et al., 2009; Sihto et al., 2011; Cerully et al., 2011;
Kerminen et al., 2012; Paramonov et al., 2013; Jaatinen at al., 2014). These
results specifically show that in summer the aerosol particles are enriched
in organic species produced by biomass burning and biogenic emissions, which
overall decrease their hygroscopicity and CCN activity when compared to other
locations in Europe. In addition, due to aerosol aging (coagulation,
condensation, aerosol–cloud interactions, and chemical reactions on the
surface and in the aqueous phase) the growth in aerosol size from Aitken to
accumulation mode leads to an increase in their hygroscopicity (Paramonov et
al., 2013).
An important limitation of the commonly used HTDMA and CCN instruments is
that they are only applicable to small particles due to the restriction of
their particle size range (typically, dry diameter < 300 nm). To
our knowledge, no experimental data for the hygroscopic properties of the
accumulation mode in the size range 0.3–1.0 µm and of the coarse
(> 1 µm) mode in the boreal environment have been
presented up to now. Therefore, we have set out to investigate and
characterize the hygroscopic properties of boreal aerosol particles in the
growing season covering the sub- and super-micron size ranges. The
hygroscopic growth measurements of aerosol particles have been performed by a
filter-based differential analyzer supplemented by chemical and
microstructural studies. The instrumentation and measurement procedures
applied in this study are described below. To characterize the hygroscopic
behavior of the aerosol particles in the 5–99.4 % RH range, we used the
mass-based hygroscopicity parameter interaction model (Mikhailov et al.,
2013). This model was developed to describe and parameterize the hygroscopic
properties of atmospheric particles with poorly defined chemical composition.
Due to the mass-based approach, it can be used to characterize partly
dissolved solutes that may co-exist in metastable equilibrium with amorphous
phases. As a consequence, the model can reproduce both the characteristics of
water uptake under subsaturated conditions and predict their CCN properties
(Mikhailov et al., 2013).
Water uptake modeling and parameterization
The κ-mass interaction model (KIM) was used to describe and
parameterize different regimes of hygroscopicity observed in our FDHA
measurements. Details of the KIM are given elsewhere (Mikhailov et al.,
2013). Briefly, in analogy with the volume-based hygroscopicity parameter
(Petters et al., 2007), we define a mass-based hygroscopicity parameter,
κm:
1aw=1+κmmdmw,
where aw is the activity of water, md is the total
mass of the dry particle material, and mw is the mass of water
in the wet particle (aqueous droplet). By defining the mass growth factor,
Gm, as
Gm=mw+mdmd
and combining Eq. (1) and Eq. (2), we obtain
aw=κmGm-1+1-1.
Based on Eq. (3), an approximate mass-based κm–Köhler
equation can be written as follows:
RH100%=sw≈κmGm-1+1-1exp4σwMwRTρwDdρwρdGm1/3,
where Mw, σw, and ρw are the molar mass, surface
tension, and density of pure water, R is the universal gas constant, T is the
temperature, and Dd and ρd are the volume equivalent diameter
and density of the dry particle. The concentration dependence of κm in the KIM is expressed as follows:
κm=∑iκm,i0cm,i+∑i<∑jαijcm,icm,j+∑iαiicm,i2.
Here κm,i0 is the dilute hygroscopicity, αij and αii are the cross- and self-interaction coefficients,
respectively, and cm,i and cm,j are the mass
concentrations of individual components (i,j) in the aqueous solution. The
mass concentration of each component in the aqueous solution,
cm,i, can be calculated either from the solubility (if
component i is only partially dissolved) or from the dry mass fraction (if
component i is fully dissolved). For mixed organic–inorganic particles,
KIM describes three distinctly different regimes of hygroscopicity: (I) a
quasi-eutonic deliquescence and efflorescence regime at low humidity, where
substances are just partly dissolved and exist also in a non-dissolved phase;
(II) a gradual deliquescence and efflorescence regime at intermediate
humidity, where different solutes undergo gradual dissolution or
solidification in the aqueous phase; and (III) a dilute regime at high
humidity, where the solutes are fully dissolved, approaching their dilute
hygroscopicity. In each of these regimes, the concentration dependence of
κm can be described by simplified model equations:
Reconstructed particulate matter (PM) concentrations in the
accumulation (0.1–1 µm) and coarse (> 1 µm)
size modes. The uncertainty (1 standard deviation) of the ion analysis is
∼ 10 % for anions and cations. For carbonaceous material the
uncertainties arise from standard instrumental error of the SunSet instrument
and the assumptions used, and are estimated overall to be ∼ 30 %.
Aerosol size
PM
Chemical species concentration (ng m-3)
mode (µm)
(ng m-3)
Sea salt
Nss-SO42-
NH4+
Nss-K+
Nss-Ca2+
NO3-
EC
OM
WSOM
WIOM
0.1–1
1600 ± 20
7.7
438
137
20.7
1.2
0.4
17.0
976
830
146
> 1
510 ± 20
33.7
205
57.8
9.7
7.1
n.a.
3.3
196
41.2
155
Regime I:
κm=k1Gm-1
Regime II:
κm=k2+k3Gm-1+k4Gm-1-1+k5Gm-1-2
Regime III:
κm=k5Gm-1-2+k6.
Here k1 to k6 are fit parameters related to the solubility and
interaction coefficients of all involved chemical components (Mikhailov et
al., 2013; Eqs. 39–44). In Eq. (8) the fit parameter k6=κm0 can be regarded as the dilute hygroscopicity parameter of the
investigated sample of particulate matter (Fig. 3b). Its physical meaning is
equivalent to the volume-based parameter, κv, proposed by Petters
and Kreidenweis (2007).
Figure 3a shows an illustrative example of KIM assuming a two-component
mixture (A and B) having deliquescence relative humidities DRHA
and DRHB, respectively (Wexler and Seinfeld, 1991). At
RHeu mixed particles partially deliquesce (I, green line) and the
concentration of each component is given by the eutonic (eu) solubility,
Ceu; as the RH increases further, gradual dissolution of the
solid component A occurs (mode II, yellow area). At the relative humidity of
RHfd both components are fully dissolved (fd) with a
composition equal to the initial particle composition, CA,B. The
subsequent increase in RH leads to dilution of the aqueous solution (III,
blue area). This simplified two-component diagram can be extended to a
multicomponent mixture. Figure 3b illustrates the respective changes of the
κm, which are captured by Eqs. (6)–(8). Note that these
equations can also be used to describe and parameterize mixed particles that
are in a metastable state (Fig. 3b, red dashed line). Specifically, in KIM
the “quasi-eutonic deliquescence and efflorescence” states are considered
to characterize partly dissolved solutes that may co-exist in metastable
equilibrium with amorphous phases (Mikhailov et al., 2013) that undergo
quasi-eutonic deliquescence (RHeu) and quasi-eutonic
efflorescence (RHef), respectively (Fig. 3b).
Results and discussion
Chemical composition
The aerosol chemical mass closure calculations were made in a similar way as
done by Maenhaut et al. (2002). The reconstructed PM mass for accumulation
and coarse mode was obtained as the sum of eight aerosol species, which were
calculated as follows: (1) ammonium (NH4+); (2) nitrate
(NO3-); (3) sea salt estimated as 1.4486 [Na+] + [Cl-],
where 1.4486 is the ratio of the concentration of all elements except
Cl- in sea water to the Na+ concentration; (4) non-sea-salt (nss)
sulfate is obtained as total sulfate minus sea-salt sulfate, whereby the
latter was obtained as 0.252 [Na+], with 0.252 the mass ratio of
SO42- to Na+ in seawater (Riley and Chester,
1971); (5) nss-K+
– as total K+ – 0.0376[Na+]; (6) nss-Ca2+ – as total
Ca2+ – 0.0382[Na+]; (7) elemental carbon (EC); and (8) organic
matter (OM) as estimated as PM – (EC +∑ inorganic_species). The
measured OC data could not be used for this purpose, since the OC measurement
was done on total PM filters, without size fractionation.
An illustrative example of KIM assuming a two-component mixture (A,
B): (a) – phase diagram; (b) – the respective variation
of the hygroscopicity, κm, due to water uptake. The light
colored areas are responsible for the metastable state of the solution.
Explanation and designations are given in the text.
Mineral dust was not considered in the mass closure since the STXM-NEXAFS
results indicate the absence of significant dust-like components in the
samples (Fig. S2.2). In STXM analysis, mineral dust particles are typically a
rather noticeable phenomenon due to their comparably large size, irregularly
shaped morphology, and strong absorption (i.e., at carbon pre-edge and also
at the oxygen edge due to Al, Si, and Fe oxides). Most of the observed
particles differ strongly from dust-like particles in morphology and
absorption properties.
Elemental carbon and organic carbon were measured for total PM; therefore,
several simplifying assumptions were used to estimate EC, water-soluble
(WSOM), and water-insoluble (WIOM) organic matter in the coarse- and
accumulation-mode fractions of the ZOTTO samples. First, EC was divided
between AM and CM as 5:1, which is a typical ratio for smoke particles
(Jaffrezo et al., 2005; Soto-García et al., 2011; Liu et al., 2013).
Second, we assume that the WSOM / OM ratio is equal to 0.85 for AM and 0.21
for CM. Details of the calculation are presented in Supplement S3.
Reconstructed contents of the inorganic and organic species for AM and CM are
reported in Table 2 and shown in Fig. 4. In both size fractions,
SO42- and NH4+ are the dominant ions, with sulfate ions
accounting for 27 and 40 % of particulate matter in AM and CM,
respectively (Fig. 4). Some of the sulfates could have come from forest fires
located to the east of the ZOTTO site, but these must be a minor source,
because the typical sulfate content in aged biomass smoke is only about
5 %. Additionally, sulfates could also have natural sources, namely, the
oxidation of marine-emitted dimethylsulfide (Levasseur, 2013). Again, this
can only account for a minor fraction because biogenic sulfate aerosol
concentrations in the Arctic are typically less than 100 ng m-3 even
in summer, when this source is strongest (Li and Barrie, 1993; Norman et al.,
1999; Quinn et al., 2002; Ström et al., 2003; Gong et al., 2010; Chang et
al., 2011). Most of the sulfate must thus be of anthropogenic origin,
arriving from the north with the Arctic airflow, which passed at
∼ 400 km by Norilsk (Fig. 1), a powerful source of SO2 (Walter et
al., 2002). This is consistent with the back-trajectory analysis (Fig. 1) and
CO levels (Table 1), which indicate a significant influence of Arctic and
eastern air masses at the ZOTTO site.
Sea salt is the next important inorganic component in CM (6.6 %),
suggesting a predominant origin by long-range transport from the Arctic Ocean
(Fig. 1). Nss-K+ was the second-most abundant cation and accounted for
1.3 and 1.9 % of PM in the accumulation and coarse modes. Generally,
K+ is a good indicator of biomass burning and could have come with
eastern air masses from the biomass burning area (Fig. 1). However, given the
lack of intense forest fires (EC/TC ∼ 2 %), the primary emission of
nss-K+ from biogenic sources should not be neglected. Active biota such
as plants and fungi are known to be an additional source of atmospheric
K-rich salts in the air (Pöhlker et al., 2012). A small amount of
nss-Ca2+ was observed in both accumulation (< 0.1 %) and
coarse (1.4 %) mode. Coarse ash particles emitted from wood combustion
generally contain significant amounts of Ca along with Mg, Si, Al, Fe, and K
(Pitman, 2006). Enriched Ca2+ may be also produced by
processes within clouds, which bring sea salt and mineral particles together,
or by the reaction of atmospheric SO2 with marine biogenic CaCO3
particles (coccoliths) (Andreae et al., 1986). More likely, the latter
process dominated since no mineral particles were identified in the CM by
NEXAFS analysis. Overall, as expected, the water-soluble fraction of all
inorganic ions prevails in CM particles, with a mean fraction 61 % vs.
38 % for AM.
Organic compounds account for a large fraction of air particulate mass (Table
2): OM / PM ratios in AM and CM are as high as 61 and 38 %, and the
WSOM fraction was estimated to be ∼ 52 % in AM and ∼ 8 %
in CM, respectively (Fig. 4). The measured concentrations of OC and EC in
total PM were 1000 ± 60 ng m-3 and
20.3 ± 5.6 ng m-3, respectively, in reasonable agreement with
the sum of the reconstructed OM values (1170 ng m-3). These
concentrations are a factor of 2.4 lower for OC and a factor of 8 lower for
EC than those measured by Maenhaut et al. (2011) in the summer season at the
SMEAR II background boreal station (Hyytiälä, Finland). The low
content of elemental carbon (EC / OC ∼ 2 %) suggests that
during our field campaign the effect of forest fires and fossil fuel
combustion on the carbonaceous aerosol fraction was modest. The most likely
sources of the particulate organic carbon are atmospheric oxidation
processes, which convert biogenic volatile organic compounds (BVOCs), such as
monoterpenes and sesquiterpenes emitted by the boreal ecosystem, to secondary
organic aerosol (Kanakidou et al., 2005; Corrigan et al., 2013; Chi et al.,
2013; Mikhailov et al., 2015).
Average mass fraction of organic and inorganic species
in the accumulation and coarse modes of the ZOTTO samples.
The WSOC / OC ratio was estimated at 0.67 ± 0.06. Such a high
water-soluble fraction suggests that the atmospheric conditions in summer
may favor the further oxidation of the secondary organic compounds towards
higher water solubility. Our results are comparable with earlier data
reported for forest environments in the summer season: WSOC / OC =
0.70 ± 0.09 (Timonen et al., 2008) and 0.71 ± 0.05 (Kiss et al.,
2002).
STXM images and elemental maps of representative accumulation mode
particles in the ZOTTO aerosol samples. The particles shown here represent a
volume equivalent diameter range of 0.3–0.6 µm. (Panels a
and b): carbon pre- and post-edge images; panel (c): carbon
elemental map; panel (d): nitrogen map; panel (e): oxygen
map; and panel (f): overlay of C (green), N (blue), and O (red)
maps. Axes in panels (a)–(e) display image dimensions in
µm. Optical density (color code) is displayed for individual maps.
Red boxes and labels in panel (b) exemplify the most common particle
types in STXM samples: (a) and (b) show internally mixed
particles with ammoniated sulfate core and organic coating;
(c)–(g) show mostly purely inorganic ammoniated sulfate
particles; (i) shows ammoniated sulfate with potassium; and
(h) shows a C-rich particle with small inorganic content.
The high level of WIOM (∼ 30 %) in CM can be explained by the
presence of primary biogenic particles, e.g., plant debris, spores, bacteria
and pollen (Pöschl et al., 2010). Internally mixed particles of sea salt
and organic matter also can be produced via aerosol–cloud processing and
fragmentation of organic-rich surface film layers during the bursting of air
bubbles at the sea surface (Andreae and
Rosenfeld, 2008). In general, the mass closure analysis is in agreement with the
results of single-particle elemental composition as observed by STXM-NEXAFS,
which also showed a dominant abundance of organic particulate matter and
ammoniated sulfates (see next section).
STXM images and elemental maps of representative particles in the
coarse mode range with some smaller accumulation mode particles. The
particles shown here represent a volume equivalent diameter range of
0.6–1.9 µm. (a, b) Carbon pre- and post-edge
images, (c) carbon elemental map, (d) nitrogen map,
(e) oxygen map, and (f) overlay of C (green), N (blue), and
O (red) maps. Axes in (a)–(e) display image dimensions in
µm. Optical density (color code) is displayed for individual maps.
Red boxes and labels in (b) highlight internally mixed particles
with ammoniated sulfate and variable amounts of organics.
TEM micrographs of submicron (a, b, c) and
supermicron (d, e, f) particles.
Hygroscopic properties of accumulation (a, b) and coarse
(c, d) modes of ZOTTO aerosol samples: (a, c) mass growth
factors (Gm) observed as a function of relative humidity compared
to KIM; (b, d) mass-based hygroscopicity parameters
(κm) calculated as a function of mass growth factor. The
data points and error bars are from FDHA experiments of hydration (blue
symbols) and dehydration (red symbols). The labels I (Ia, Ib), II (IIa, IIb)
and III indicate different regimes of hygroscopicity (Eqs. 6–8); the borders
of the corresponding fit intervals are indicated by green circles (b, d). Varied symbols in panels (a) and (c) represent
different experimental runs on the same sample.
Aerosol microstructure – STXM-NEXAFS and TEM analysis
Limited particle statistics is an inherent difficulty of most single particle
techniques. Thus, we applied two different single particle approaches in this
study to broaden the statistical basis and to check the consistency of the
independent data sets. The total particle number that has been probed with
STXM is 150, while the total particle number for TEM analysis is 810. The TEM
data allow a classification of the particle ensemble based on morphology,
while the STXM data (though more limited in statistics) provide more detailed
insights into the chemical composition and therefore complement the TEM data
set. The visual TEM analysis of 725 AM and 85 CM aerosol particles has
revealed three main morphology types: (i) homogeneous spheres, which can be
attributed to terpene- and isoprene-based SOA droplets (Kourtchev et al.,
2005; Pöschl et al., 2010), (ii) mixed SOA-inorganic particles with
core–shell morphology, and (iii) irregularly shaped primary biological
aerosol (PBA) particles, such as plant fragments. In the submicron particle
range the balance between uncoated (type i) and coated (type ii) aerosol
particles is 32 and 62 %, respectively, which agrees with the STXM data.
Approximately 5 % can be attributed to PBA particles with an organic
coating fraction of ∼ 50 %. In the coarse particle mode almost all
aerosol particles are internally mixed with surface organic coating
(∼ 80 %). Among them ∼ 20 % have a high-contrast-density
outer core (Fig. 7d–f) presumably formed from sparingly soluble organic
species due to liquid–liquid phase separation upon particle dehydration
(Song et al., 2012). Approximately 13 % are PBA particles. In addition, a
minor fraction (< 1 %) of the fractal-like soot aggregates was
found on the TEM images of the AM and CM aerosol particles.
The x-ray analysis showed that the size range of the particles collected on
the sampling substrates (Si3N4 membranes) is consistent with the
aerosol size distribution in Fig. 2b. The majority of particles is present
in the AM size range (0.3–1 µm), with a small number of larger CM
particles (> 1 µm) (Fig. S2.1).
To characterize the overall elemental abundance in the collected particles,
multiple STXM maps of relatively large areas have been recorded from all
samples and confirm the bulk chemical analysis, showing that
SO42-, NH4+, and OM are the predominant constituents of
the aerosol (example shown in Fig. S2.2). X-ray spectroscopic evidence for
the dominance of SO42-, NH4+, and OM in the particles is
shown in Fig. S2.3. Based on the x-ray absorption spectra, the elemental
ratio N / Osulfate = 0.54 ± 0.12 was calculated (ion
chromatography results in Fig. 4 gave N / Osulfate = 0.41 ± 0.16). This suggests a sulfate salt composition close to ammonium sulfate
[(NH4)2SO4; N / O = 0.5] and/or letovicite
[(NH4)3H(SO4)2; N / O = 0.38]. The individual particles
comprise highly variable amounts of OM with
OM/(NH4)2-xHx(SO4) mass ratios in the range of 0–0.9.
Elemental maps of typical particles are shown in Figs. 5 and 6. Figure 5
displays particles in the AM size range, which exhibit spherical or
elliptical morphologies with variable OM contents and core–shell structure
for a certain fraction of particles. In contrast, Fig. 6 shows comparably
large particles (CM size range) with conspicuous internal structures and
comparably high OM contents. The difference in OM content can also be seen in
Fig. 6. All internally mixed (NH4)2-xHx(SO4)/ OM particles
reveal a clear separation of the inorganic and organic phases.
Complementary to the STXM results, TEM images in Fig. 7 show particles with
core–shell structures. The cores appear as dendritic crystalline-like
material and resemble similar dendritic sulfate salt structures in the STXM
image, Fig. 6 (i.e., particles b, c, e, and f). Moreover, the TEM images also
display OM shells of different thicknesses around the particles (Fig. 7).
Hygroscopic properties and KIM results
Figure 8a and c show the mass growth factors, determined as a function of
relative humidity upon hydration and dehydration for accumulation and coarse
size modes as detailed in Sect. 2.3.5. The different symbols represent
repetitive measurements on the same sample, indicating a good reproducibility
of the water uptake/release results and a negligibly small effect of particle
evaporation due to semivolatile organics. The onset of deliquescence at
∼ 70 % RH is evident for the accumulation (Fig. 8a) and coarse
(Fig. 8c) modes. Upon dehydration these size modes also exhibit an
efflorescence transition at 37 % RH (insert in Fig. 8a) and at 49 %
RH, respectively. Figure 8c indicates a hysteresis for the CM mode starting
at ∼ 95 % RH, i.e., well before the particles deliquescence. The
kinetics and morphology effects of the supermicron size particles might be
responsible for this effect, as will be discussed below.
From the measurement of the Gm(RH) data we derived mass-based
hygroscopicity parameters using Eqs. (3) and (4) and the Kelvin correction
algorithm for submicron particles as described by Mikhailov et al. (2013,
Appendix C). The corresponding plots of κm vs.
Gm are shown in Fig. 8b and d for AM and CM, respectively. In all
size modes the observed dependence of κm on Gm
exhibits three distinctly different sections or regimes of hygroscopicity as
outlined in Sect. 3 and in Fig. 3. The model lines were obtained by inserting
the fit parameters from Table 3 into Eqs. (6)–(8). For the quasi-eutonic
regime (I), the combination of Eq. (3) and Eq. (6) yields a constant water
activity value given by aw=(k1+1)-1. This relation
yields the following quasi-eutonic RH values characterizing the deliquescence
(Ia) and efflorescence (Ib) phase transitions: 74.6 % (36.7 %) for
AM, and 70.0 % (49.8 %) for CM, respectively (Fig. 8b, d). The
gradual deliquescence mode (II) extended up to 96 % RH for the coarse
mode and even further up to 97 % for the accumulation mode, indicating
the presence of sparingly soluble OC in both fractions. In the dilution
regime (III) (Eq. 8), the fit parameter k6=κm0
obtained for submicron aerosol particles is 0.061 ± 0.002 (Table 3).
This value can be compared to the dilute hygroscopicity obtained for PM1
samples during the Amazonian Aerosol Characterization Experiment (AMAZE-08)
in the wet season (Mikhailov et al., 2013), where κm0=0.104± 0.002. The observed ∼ 40 % discrepancy in
κm0 between Amazon and Siberian samples may be caused by
differences in their chemical composition. Hygroscopic ammoniated sulfate is
a good candidate for the observed difference. However, for the AMAZE-08
campaign, the average sulfate loading measured by ion chromatography was only
0.21 µg m-3 (Chen et al., 2009). This concentration is less
than half that determined in the Siberian sample
(0.44 µg m-3, Table 2). One possible explanation is that the
ammoniated sulfate in the Siberian aerosol sample was partly isolated by
sparingly soluble organic species and therefore was not completely involved
in hygroscopic growth. This potential mass transfer limitation effect will be
considered below.
KIM fit parameters for the accumulation (0.1–1 µm) and
coarse (> 1µm) aerosol size modes collected at ZOTTO
(Siberia); n and R2 are the number of data points and the coefficient
of determination of the fit.
Size mode
Regime
Gm range
n
R2
Fit equation
Best fit parameter ±standard error
Accumulation
Quasi-eutonic deliquescence (Ia)
1–1.3
14
0.98
6
k1=0.342 ± 0.005
Gradual deliquescenceand efflorescence (II)
1.3–4.6
32
0.86
7
k2=0.140 ± 0.004; k3=-1.16 × 10-2 ± 8.7 × 10-4 k4=-1.19 × 10-2 ± 1.5 × 10-3; k5(II) = 4.19 × 10-4 ± 8 × 10-5
Dilution (III)
4.6–9.1
14
0.87
8
k5 (III) = 0.45 ± 0.05; k6=0.061 ± 0.002
Quasi-eutonic efflorescence (Ib)
1.02–1
4
0.97
6
k1=1.70 ± 0.07
Coarse
Quasi-eutonic deliquescence (Ia)
1–1.2
4
0.99
6
k1=0.43 ± 0.01
Gradual deliquescence(IIa)
1.2–5.2
19
0.61
7
k2=0.23 ± 0.02; k3=-0.012 ± 0.003 k4=-0.05 ± 0.02; k5(II) = 0.005 ± 0.003
Dilution (III)
5.2–16.6
28
0.56
8
k5 (III) = 0.50 ± 0.08; k6=0.128 ± 0.004
Gradual efflorescence (IIb)
4.50–1.1
14
0.84
8
k2=0.308 ± 0.006; k3=-0.032 ± 0.004 k4=-0.065 ± 0.008; k5(II) = 0.006 ± 0.001
Quasi-eutonicefflorescence (Ib)
1.1–1.0
4
0.99
6
k1=1.01 ± 0.03
Another possible reason for the increased hygroscopicity of the Amazon sample
is that the biogenic SOA produced by tropical rainforests contains more
hygroscopic species than that produced in the Siberian boreal zone. It was
found that the formation of WSOC is closely linked to photosynthetic activity
by the forest ecosystem, which depends on both temperature and solar
radiation (Zhang et al., 2010; Miyazaki et al., 2012). Among these compounds
are the highly water-soluble and hygroscopic isoprene-derived 2-methyltetrols
(Claeys et al., 2004; Ekström at al., 2009; Engelhart et al., 2011).
Because solar radiation and the production of OH radicals are at a maximum in
the tropics, the concentration of 2-methyltetrols in the Amazon Basin is
higher than in boreal forests. The average concentration of 2-methyltetrols
in Amazonian PM1 particles in the wet season is 45 ng m-3
(Decesari et al., 2006), while for the boreal ecosystem in the summertime it
is only ∼ 26 ng m-3 (Kourtchev et al., 2005). Given that the
WSOC / OC ratio for AMAZE and ZOTTO samples is comparable
(63 ± 4 %), it seems possible, therefore, that the water uptake
associated with WSOM in the AMAZE sample was higher than that in the ZOTTO
sample.
Overall, the observed hygroscopicity behavior of the sub- and super-micron
samples is consistent with their chemical composition. A twofold decrease in
the dilute hygroscopicity parameter, k6=κm0
(Table 3), and a reduction of the κm top value from
∼ 0.22 (Fig. 8d) to ∼ 0.12 (Fig. 8b) are consistent with a high
fraction of organic carbon in the accumulation mode compared to the coarse
mode. Likewise, the observed decrease in the quasi-eutonic efflorescence
transition in the dehydration mode by 15 % RH (i.e., the quasi-eutonic
efflorescence transition is ∼ 50 % RH in the coarse mode (Fig. 8c)
vs. ∼ 35 % in the accumulation mode (Fig. 8a)) can also be
associated with a high fraction of organic species in the accumulation mode.
Reconstructed neutral species concentrations (ng m-3) in the
accumulation (0.1–1 µm) and coarse (> 1 µm)
size modes, their mass fraction (fi), volume fraction (εi), density (ρd,i), and CCN-derived hygroscopicity
(κv,i). The κv,i values in brackets
were obtained under subsaturated conditions in the 96–99.4 % RH range.
Uncertainties of all parameters for inorganic and carbonaceous compounds are
estimated to be ∼ 10 and ∼ 30 %, respectively. The subscript
ws stands for parameters related to the water-soluble fraction.
Size mode
Parameters
Chemical compounds
Nss-(NH4)2SO4
Sea salt
NH4NO3
Nss-K2SO4
Nss-CaSO4
EC
WSOM
WIOM
ρd,i, g cm-3
1.77a
2.24a
1.73a
2.66a
2.96a
2.0a
1.4b
1.4b
κv,i
0.61c (0.57)g
0.98d
0.67c
0.52e
0.0016f
0
0.1c (0.01–0.05)h
0
Accumulation
PM (ng m-3)
544
7.7
0.5
45.4
3.9
17
830
146
fi
0.34
4.8 × 10-3
3.2 × 10-4
0.028
2.5 × 10-3
0.011
0.52
0.091
εi
0.30
3.3 × 10-3
2.8 × 10-4
0.016
1.3 × 10-3
8.2 × 10-3
0.57
0.10
fws,i
0.38
5.4 × 10-3
3.6 × 10-4
0.032
–
–
0.58
–
εws,i
0.33
3.7 × 10-3
3.2 × 10-4
0.019
–
–
0.64
–
Coarse
PM (ng m-3)
241
33.3
–
21.5
23.7
3.3
41.2
155
fi
0.47
0.065
–
0.042
0.046
6.5 × 10-3
0.080
0.30
εi
0.44
0.048
–
0.026
0.026
5.3 × 10-3
0.095
0.36
fws,i
0.72
0.10
–
0.064
–
–
0.12
–
εws,i
0.72
0.080
–
0.043
–
–
0.16
–
a Lide (2005), density taken for 25 ∘C,
b Kostenidou et al. (2007), c Petters and
Kreidenweis (2007), d Niedermier et al. (2008), e
Kelly and Wexler (2006),
f Sullivan et al. (2009),g Mikhailov et al. (2013),
h Petters et al. (2009).
In the internally mixed organic/inorganic particles the organic coating was
found to decrease or even suppress the efflorescence of the inorganic salts
(Choi and Chan, 2002; Brooks et al., 2003; Pant et al., 2004; Braban et al.,
2004; Mikhailov et al., 2004; and Zardini et al., 2008), which can be
explained by kinetic limitations caused by an ultra-viscous or gel-like or
semi-solid organic matrix with low molecular diffusivity (Mikhailov et al.,
2009; Shiraiwa et al., 2010) as will be discussed below.
Dilute kappa composition closure
Under the volume additivity assumption, the dilute hygroscopicity parameter
k6=κm0 can be converted to the Petters and
Kreidenweis (2007) volume-based parameter, κv, by the
relation (Mikhailov et al., 2013)
κv=κm0ρdρw.
Based on the ion balance from IC analysis (Table 2), we estimated the mass of
neutral compounds in the accumulation and coarse size modes. In the mass
balance of the neutral salt compounds we first distributed the measured
concentrations of Nss-SO42- and NH4+ ions (Table 2) between
minor compounds: Nss-CaSO4, Nss-K2SO4, and NH4NO3.
The remaining SO42- / NH4+ mass ratio was found to be 2.9
and 3.0 for accumulation and course mode, respectively. This is close to the
sulfate / ammonium mass ratio in ammonium sulfate (AS), which is 2.7. The
higher experimental ratio of ions as compared to the stoichiometric ratio in
AS could be caused by letovicite (Mifflin et al., 2009) with a
SO42- / NH4+ ratio of 3.6 or/and organosulfates
(Hettiyadura et al., 2015), but these species can only account for a minor
fraction because the ion balance shows that sulfate is almost fully
neutralized by ammonium. Based on these calculations, we assume that ammonium
sulfate is the main component among other possible ammoniated sulfate salts
and sulfate containing organic species.
Characteristic parameters obtained for the accumulation and coarse
modes of ZOTTO samples: density (ρd, ρd,ws) (Eq. 10),
KIM-derived (κm0, κm,ws0) and
corresponding volume-based hygroscopicity parameters (κv,t,
κv,ws) (Eq. 9), and mass fraction of water-soluble compounds
(fws). Subscripts t and ws stand for parameters related to
total PM and to the water-soluble fraction, respectively. κv,p is the predicted hygroscopicity based on the ZSR mixing rule
(Eqs. 15, 16). For κm0, κm,ws0
uncertainties result from the goodness of the KIM fit. For the other
parameters the propagated errors were calculated based on the total
differential of a function by neglecting correlations between variables.
Size mode
Parameters
ρd, g cm-3
κm0
κv,t
fws
ρd,ws, g cm-3
κm,ws0
κv,ws
κv,p
Accumulation
1.54 ± 0.09
0.061 ± 0.002
0.094 ± 0.006
0.89 ± 0.11
1.55 ± 0.11
0.098 ± 0.003
0.15 ± 0.01
0.27 ± 0.03
Coarse
1.66 ± 0.07
0.128 ± 0.004
0.21 ± 0.01
0.66 ± 0.08
1.78 ± 0.09
0.20 ± 0.01
0.36 ± 0.03
0.53 ± 0.05
Reconstructed neutral species concentrations are shown in Table 4 with their
mass fraction (fi), volume fraction (εi), density
(ρd,i), and hygroscopicity (κv,i). Using
these parameters, the effective values of ρd and κv,p for the ZOTTO samples have been computed by weighted
averaging of the properties of individual components:
ρd=∑ifiρi-1,
κv,p=∑iεiκv,i.
Equation (11) is the Zdanovskii, Stokes, and Robinson (ZSR) mixing rule
(Petters and Kreidenweis, 2007) where κv,p denotes the
predicted (p) hygroscopicity based on the volume fraction, εi, and the hygroscopicity, κv,i, of the ith component in
the sample. From Eq. (10) and the ρi, fi pairs (Table 4) it
follows that for submicron and supermicron particles the average weighted
bulk densities are 1.54 and 1.66 g cm-3, respectively
(Table 5). Inserting these values and the KIM-derived hygroscopicity
parameters, k6=κm0 (Table 3), into Eq. (9)
yields κv,t=0.094 for the accumulation mode and κv,t=0.21 for the coarse mode (Table 5), where the subscript (t)
denotes the hygroscopicity related to the total dry particle mass.
As water-soluble compounds are the major contributors to hygroscopic growth,
it is useful to further transcribe the original Eq. (2) as follows:
Gm,ws=mw+md,wsmd,ws,
where Gm,ws is the mass growth factor normalized to the mass of
the neutral water-soluble (ws) compounds. Accordingly, the Gm,ws can be calculated from
Gm,ws=1fwsGm-1+1,
where fws=md,ws/md is the mass fraction of
the water-soluble compounds in the PM, which is 0.89 and 0.66 for AM and CM,
respectively. Using Eqs. (3) and (4) and the Kelvin correction algorithm
(Mikhailov et al., 2013), we converted the Gm,ws RH pairs into
κm,ws. The obtained dependencies of κm,ws on Gm,ws in the dilution regime and KIM fit lines (Eq. 8) are
shown in Fig. 9. The corresponding best fit parameter (dilute hygroscopicity)
κm,ws0=k6 for AM and CM and its volume-based
derivative κv,ws are given in Table 5. The volume-based
hygroscopicity κv,ws was calculated from Eq. (9) as
follows:
κv,ws=κm,ws0×ρd,ws/ρw,
where ρd,ws is the average density of the water-soluble
compounds in the dry particles, which was determined by inserting into
Eq. (10) the mass fractions of water-soluble solutes, fws,i and
ρd,i (Table 4).
As expected, the obtained κv,ws is higher than κv,t: the ratio κv,ws/κv,t is
approximately 1.6 for both size modes (Table 5). Since the hygroscopicity
κv,ws accounts only for water-soluble species, the obtained
values of 0.15 for AM and 0.36 for CM (Table 5) can be regarded as an upper
limit characteristic of boreal aerosol particles in Siberia during the
growing season. The κv,ws value of 0.15 obtained here for
the accumulation mode is comparable to the CCN-derived overall median value
of κv=0.15 reported by Gunthe et al. (2009) for tropical
rainforest air during the wet season in central Amazonia and the average
κv value of 0.16 measured by Levin et al. (2014) at a
forested mountain site in Colorado from July to August. We are not aware of
any other field data of κv for the coarse mode at remote
continental sites, but the elevated CM value of 0.36 compared to that for the
accumulation mode of 0.15 is consistent with the chemical analysis results,
indicating a relatively high content of hygroscopic ammonium sulfate in CM
(fws,AS=0.72 vs. of 0.38 for AM) (Table 4).
Mass-based hygroscopicity parameter κm,ws as a
function of mass growth factor, Gm,ws, normalized to
water-soluble compounds (Eq. 12) upon hydration (blue) and dehydration (red).
KIM fit of dilution mode, Eq. (8) – black solid line.
We now compare the hygroscopicity κv,ws, estimated from
FDHA measurements with the ZSR-predicted hygroscopicity, κv,p. Accounting for the fact that AS, WSOM, and sea salt are the main
contributors to the hygroscopic growth (Table 4), Eq. (11) for AM and CM can
be, respectively, reduced to
κv,p=κv,ASεws,AS+κv,WSOMεws,WSOM,κv,p=κv,ASεws,AS+κv,seasaltεws,seasalt+κv,WSOMεws,WSOM,
where εws,i is the volume fraction of water-soluble
compounds scaled to total water-soluble PM (Table 4). The application of
these simplified equations results in a ∼ 5 % underestimation of
κv,p for both size modes.
The simple mixing rule (Eqs. 15, 16) with corresponding pairs of CCN-derived
κv,i and εws,i (Table 4) yields
κv,p=0.27 for AM and κv,p=0.53 for CM.
That is, the estimates from the mixing rule exceed the FDHA-derived κv,ws values (Table 5) by factors of 1.8 and 1.5 for the
accumulation and coarse modes, respectively. The observed discrepancy is too
large to be explained by experimental and PM chemical analysis uncertainties.
The high content of sparingly soluble organic matter produced by oxidation of
biogenic emissions (Mikhailov et al., 2015) can account for the inconsistency
between the ZSR-predicted κv,p and FDHA-derived
κv,ws values. In mixed particles the organic coating can
reduce the water transport across the surface by acting as a physical
barrier. Moreover, at high content of the sparingly soluble compounds this
effect will strongly depend on the water activity range, as these species
exhibit hygroscopic growth at aw close to 1. This particularly
leads to a discrepancy by a factor of 5–10 between SOA hygroscopicity
determined from sub- and super-saturation experiments (Petters and
Kreidenweis, 2007; Prenni et al., 2007; Wex et al., 2009; Petters et al.,
2009; Pajunora et al., 2015). In our FDHA experiment the hygroscopicity, κm,ws0, was obtained in the dilution mode (Eq. 8) in the 96–99.4 % RH range.
Depending on the type of SOA, the κv in a given RH range can
vary from 0.01 to 0.05 (Wex et al., 2009; Petters et al., 2009), but it is
still less than the values of κv=0.1± 0.04 obtained in
CCN experiments (Petters and Kreidenweis, 2007; Gunthe et al., 2009; Wex et
al.,2009; Chang et al., 2010; Engelhart et al., 2011). Thus, due to the
uncertainty of κv, the last term of Eqs. (15) and (16) can
vary by a factor of 10, which translates into 30 and 3 % uncertainty in
κv,p for the accumulation mode and coarse mode,
respectively. Nevertheless, this uncertainty does not cover the observed
difference between the measured κv,ws and predicted
κv,p values. As mentioned above and as will be shown below
for the internally mixed particles, the organic coating can limit the uptake
of water by inorganic species and may thus decrease the hygroscopicity,
especially in sub-saturated water vapor.
Based on the ZSR mixing rule (Eqs. 15 and 16), we can estimate the volume
fraction of ammonium sulfate, εws,AS,p, that is
involved in the hygroscopic growth. Inserting into Eq. (15) the KIM-derived
κv,ws=0.15 and experimental hygroscopicities κv,AS=0.57, κv,WSOM=0.01 and 0.05 obtained for
sub-saturated conditions at 96–99.4 % RH (Table 4, values are in
brackets) yields for the accumulation mode εws,AS,p=0.23± 0.02, while the total volume fraction εws,AS is 0.33 (Table 4). A similar calculation for the coarse mode
using Eq. (16) with the KIM-derived κv,ws=0.36 leads
accordingly to εws,AS,p=0.48± 0.01 vs.
εws,AS=0.72 (Table 4). Based on these estimations, one
can assume that in sub-saturated conditions (96–99.4 % RH) the inorganic
compounds were not completely dissolved: approximately 40 and 50 % of
ammonium sulfate in AM and CM, respectively, are isolated by sparingly
soluble organic species, thereby reducing the hygroscopicity of the ZOTTO
samples.
The water uptake diagram by the supermicron particles as a
function of time during hydration (a, c) and dehydration (b)
experiments. The red lines indicate the relative humidity (RH) change. The
red dashed line in panel (b) helps to guide the eye to see the
changing trend of background amplitude. Explanation of other symbols is
given in the text.
A similar positive difference (up to 77 %) between predicted (from
chemical analysis) and CCN-derived κv values was observed
for pristine tropical rainforest aerosols studied during the AMAZE-08
campaign in central Amazonia (Gunthe et al., 2009). Although in the Amazon
experiment the dry particle size range was 30–300 nm, we do not exclude the
possibility that the observed deviation was also caused by sparingly soluble
SOA, which in the timescale of the CCN measurement (seconds or less) could
impede the water transport to the more hygroscopic species like AS.
Water uptake kinetics and microstructural rearrangements
Figure 10 shows the raw data for hygroscopic cycles performed in the FDHA
with the coarse-mode sample as a function of relative humidity. The areas
under the peaks are proportional to the amount of the water that is absorbed
(negative area) or released (positive area) upon particle hydration or
dehydration. Figure 10a shows that at 60–94 % RH during the first
hydration cycle the water uptake is not a monotonic function of RH. The
insets in panel (a) indicate that the water absorption is accompanied by
partial evaporation. During the second humidifying run (Fig. 10c) the peak
oscillations are observed again, and continue up to ∼ 96 % RH.
Figure 10b shows that upon dehydration some peaks follow the relative
humidity change almost instantaneously, while some other peaks appear with
considerable delay, i.e., already at constant RH (marked with green arrows).
Additionally, the width of these peaks characterizing the timescale of the
dehydration process gradually increases with decreasing RH. Thus, at 79 %
RH the water release takes ∼ 5 min, whereas at 60 % RH it lasts
for ∼ 25 min: first and second peaks in Fig. 10c, respectively. This
time is much more than the inherent response time of the FDHA system, which
is ∼ 10 s. The most plausible explanation for these observations is
kinetic limitation by bulk diffusion in an amorphous (semi-)solid organic
matrix, which inhibits uptake and release of water during hydration and
dehydration cycles (Mikhailov et al., 2009).
As discussed in Sect. 4.2 and shown in Figs. 5 and 6, the internally mixed
particles consist mainly of an inorganic core (mostly ammonium sulfate),
surrounded by organic compounds. The STXM-NEXAFS results are consistent with
the accompanying TEM investigations (Fig. 7), particularly indicating that
the supermicron particles typically have an irregular porosity core embedded
in an organic matrix. As noted above, the core of some of the particles has a
pronounced dendritic structure (Fig. 7e–f). Such dendritic structures are
characteristic of diffusion-limited growth processes (Feder,
1988).
The transport characteristics of water molecules through the organic coating
can be estimated based on the following relation (Atkins, 1998):
x=(4Dwτ/π)12,
where x is the average distance traveled by water molecules diffusing in an
organic shell, τ is the average time to travel over this distance, and
Dw is the bulk diffusion coefficient of water. We chose for our
calculations x=0.2–0.5 µm (i.e., the thickness of the organic
shell of the particles shown in Fig. 7d–f) and τ=1500 s at 60 %
RH and 300 s at 79 % RH, respectively. Using Eq. (10) for x=0.2;
0.5 µm we obtain Dw=2.1×10-13; 1.3×10-12 cm2 s-1 at 60 % RH and 1.0×10-12;
6.5×10-12 cm2 s-1 at 79 % RH, respectively.
These values are in the range of measured Dw in sucrose (Zobrist et al.,
2011) and Dw in α-pinene SOA estimated by percolation theory
(Shiraiwa et al., 2013).
Low bulk diffusion coefficients suggest that organic coatings of the
particles are in amorphous semi-solid states (Mikhailov et al., 2009; Koop et
al., 2011; Shiraiwa et al., 2011). Moreover, the Dw value can be
decreased by solidification and crust formation at the particle surface,
which can strongly inhibit the interfacial transport of gas molecules (Pfrang
et al., 2011). As noted in Sect. 4.2, a significant fraction
(∼ 20 %) of the coarse-mode particles have this type of outer core
(Fig. 7d–f), most likely formed as a result of liquid–liquid phase
separation.
This is in good agreement with the observations that ambient particles in
boreal forests bounce off the smooth hard surface of an impactor (Virtanen et
al., 2010) and that the viscosity of laboratory-generated α-pinene
SOA is reported to be 103–107 Pa s (Renbaum-Wolff et al.,
2013). Additionally, the
water release timescale of 300–1500 s observed in this study (Fig. 7b,
peaks 1 and 2) is in reasonable agreement with model calculations for α-pinene SOA+AS particles with core–shell structures (Shiraiwa et al.,
2013), which predict at 60 % RH an e-folding time within
101–103 s for 1 µm particles and 10-4–100 s for the submicron fraction in the range of 30–100 % RH
(Shiraiwa et al., 2013). These model results also help to explain why in the
timescale of the FDHA experiment the kinetic limitations were observed for
supermicron particles, but have not been detected for the particles in the
accumulation mode.
From the FDHA-derived upper value of Dw∼ 10-12 cm2 s-1 and the typical particle size range of 20–200 nm
used in HTDMA and CCNc instruments, it follows that the characteristic
timescale for water diffusion into aerosol particles is 3–300 s (Eq. 17).
These estimates are consistent with several HTDMA and CCNC studies where
kinetic limitations have been observed due to organic coatings and the
multi-second residence time used in these instruments (CCNC: Abbat et al.,
2005; Henning et al., 2005; VanReken et al., 2005; Shants et al., 2010;
Engelhart et al., 2008; Ruehl et al., 2009; HTDMA: Xiong et al., 1998;
Chuang, 2003; Mikhailov et al., 2004; Sjogren et al., 2007).
As noted above and shown in Fig. 10a (inserts) during hydration cycles in the
60–96 % RH range, the water uptake is followed by a water release
(negative peaks) despite growing RH. These observations can be explained by
the complex morphology of the particles. Most likely at intermediate RHs due
to partial dissolution, the different species become more mobile, which leads
to compaction (restructuring) of the particles and release of excess water.
The fact that in the second hydration cycle the water loss effect is
reproducible (inserts in Fig. 10c) suggests that re-dried particles retained
their irregular microstructure even at a drying rate as low as
∼ 0.4 % RH min-1. The reproducibility of the irregular
particle microstructure in subsequent hydration–dehydration cycles is in
agreement with the earlier results of Zardini et al. (2008), who investigated
laboratory-generated mixed AS–adipic acid particles by the electrodynamic
balance (EDB) technique, and with recent STXM-NEXAFS results for Amazonian
aerosol samples obtained by Pöhlker et al. (2014).
The driving forces of the morphological transformations can be an inverse
Kelvin effect
stimulating capillary condensation in the cracks, veins, and pores (Sjogren
et al., 2007; Zardini et al., 2008; Mikhailov et al., 2004) and their
subsequent collapse or/and Ostwald ripening, i.e., recrystallization that
causes growth of large salt particles and shrinkage of smaller ones
(Pöhlker et al., 2014). In our hydration experiments the water release
peaks were observed up to 96 % RH (inserts in Fig. 10a, c), suggesting
that morphology transformations occurred with sparingly soluble species that
restrain particle restructuring because of kinetic limitations.
In view of the results discussed above we suggest that boreal aerosol samples
have complex morphologies. The organic material tends to be enriched at the
particle surface and forms an envelope with low molecular diffusivity, which
inhibits the access of water vapor to the particle core and leads to kinetic
limitations. In the timescale of our FDHA experiment these kinetic
limitations were clearly exhibited for supermicron particles in the
hydration–dehydration cycles (Fig. 10), which resulted in a pronounced
hysteresis effect (Fig. 8c).
Summary and conclusions
In this study, we have presented the hygroscopic properties of the
accumulation and coarse modes of aerosol particles sampled at the ZOTTO
background station in western Siberia during a summer campaign. The
hygroscopic growth measurements were conducted with a filter-based
differential hygroscopicity analyzer using the range of 5–99.4 % RH in
the hydration and dehydration operation modes. These studies were
complemented with chemical analyses of the samples, focusing on inorganic
ions and organic carbon/elemental carbon. In addition, the microstructure and
chemical composition of aerosol particles were analyzed by x-ray and electron
microscopy techniques. The air-mass history, CO data, and chemical analysis
results indicate background conditions during the sampling campaign.
The mass closure studies show that organic carbon accounted for 61 and
38 % of PM in the accumulation mode (AM) and coarse mode (CM),
respectively. Accordingly, the WSOM was estimated to be ∼ 52 and ∼ 8 %. Sulfate, predominantly in the form of
ammoniated sulfate, was the dominant inorganic component in both size modes:
∼ 27 % in the AM and ∼ 40 % in the CM. Sea
salt was the next abundant inorganic component, responsible for
∼ 6.6 % of PM in the coarse mode. The bulk density of the
dry particles was estimated to be 1.54 ± 0.09 and
1.66 ± 0.07 g cm-3 for the AM and CM size mode, respectively.
The FDHA hygroscopic studies indicate that both AM and CM exhibit pronounced
water uptake approximately at the same RH, starting at ∼ 70 %,
while efflorescence occurred at different humidities, i.e., at
∼ 35 % RH for AM vs. ∼ 50 % RH for CM. This
∼ 15 % RH difference was attributed to a higher content of organic
material in the sub-micron particles, which suppresses water release in the
dehydration experiments. Overall, the observed hygroscopicity behavior of the
sub- and super-micron samples is consistent with their chemical composition
and microscopic structure. The relatively high content of organic carbon and
low content of ammonium sulfate in AM as compared to CM cause at high RH an
approximately twofold decrease in the water uptake. Thus, at 99 % RH the
hygroscopic growth factor, Gm, is ∼ 7.5 and ∼ 13.4
for sub- and super-micron particles, respectively.
The kappa mass interaction model (KIM) was applied to characterize non-ideal
solution behavior and concentration-dependent water uptake by atmospheric
aerosol samples with complex chemical composition. The κm
model reproduces the FDHA mass water uptake results well and reveals three
distinctly different regimes of hygroscopicity: (I) a quasi-eutonic
deliquescence and efflorescence regime at low humidity, where substances are
just partly dissolved and exist also in a non-dissolved phase; (II) a gradual
deliquescence and efflorescence regime at intermediate humidity, where
different solutes undergo gradual dissolution or solidification in the
aqueous phase; and (III) a dilute regime at high humidity, where the solutes
are fully dissolved, approaching their dilute hygroscopicity. The obtained
KIM fit parameters can be used to characterize the hygroscopic behavior of
sub- and super-micron boreal aerosols corresponding to background conditions
in the growing season.
Based on KIM, for dilute aerosol solutions the volume-based hygroscopicity
parameter, κv, was calculated. The κv value
normalized to total PM (κv,t) and to the water-soluble
fraction (κv,ws) was estimated to be 0.098 and 0.15 for the
accumulation mode and 0.21 and 0.36 for the coarse mode, respectively. These
values can be regarded as the lower and upper limits of aerosol
hygroscopicity in the background boreal environment. The measured value of
κv,ws=0.15 for the accumulation mode is in good agreement
with previously reported CCN-derived data for remote sites in the Amazon
forest (κv≈0.15) and the Colorado mountain forest
(κv≈0.16).
The κv,p values predicted from a chemical mass closure study
(ZSR mixing rule) overestimate the FDHA-derived κv,ws values
by factors of 1.8 and 1.5 for the accumulation and coarse modes,
respectively. The observed divergence can be explained by incomplete
dissolution of the hygroscopic inorganic compounds due to kinetic limitations
caused by a sparingly soluble organic coating. Supporting this assumption are
the microstructural and kinetics measurement results. Thus, TEM and x-ray
studies indicate predominantly core–shell structures of the aerosol
particles with an inorganic core (mostly ammoniated sulfate) surrounded by
organic compounds. The kinetic studies of the water release have shown that
boreal aerosol particles enriched in organic species are in a semi-solid
state, which is characterized by the diffusion coefficient Dw∼10-12 cm2 s-1.
In conclusion, the ZOTTO data set obtained in the growing season revealed a
strong influence of biogenic organic carbon on the chemical composition and
microstructure of the ambient aerosols. The sparingly soluble organic coating
controls hygroscopic growth, phase transitions, and microstructural
rearrangement processes in the ambient particles. The observed kinetic
limitations in the water uptake can strongly influence the outcome of
experiments performed on multi-second timescales, such as the commonly used
HTDMA and CCNC instruments.