Introduction
Black carbon (BC) aerosols are a strong light-absorbing component in the
atmosphere and a major contributor to positive radiative forcing (Bond et
al., 2013; Change, 2015; Kondo, 2015). They are mainly generated from
incomplete combustion of fossil fuels and biomass. Fresh BC emissions from
traffic are hydrophobic (Lammel et al., 1995; Dusek et al., 2006). However,
the particles can be enhanced by internally mixing with secondary materials
through atmospheric aging processes (Schneider et al., 2005; Shiraiwa et al.,
2007; Matsui et al., 2013; Pratt et al., 2011). Hygroscopicity of BC aerosol
significantly affects its removal rate through wet deposition, absorption
in the human respiratory tract (Löndahl et al., 2007), optical properties
(Chen et al., 2012), and their surface reactivity (Mogili et al., 2006).
Therefore, it is critical to understand the relations between the atmospheric
aging processes and the hygroscopicity of BC aerosols. Atmospheric aerosols,
including BC particles, always have various sizes with distinct chemical
compositions. To better understand or describe an aerosol population,
mixing state is often used. The definition of mixing state, provided by
Winkler (1973), refers to both internal and external mixtures in aerosols. In
an external mixture, individual particles in a given size range consist of
different chemical species. Chemical composition of particulate mass in that
size range will be determined by the relative contributions of the chemically
distinct particles. In an internal mixture, however, all particles
in a given size range are composed of the same mixture of two or more
chemical compounds (Heintzenberg et al., 1990). The mixing state of
atmospheric BC particles is closely linked to their sources and aging
processes (Weingartner et al., 1997; Gysel et al., 2003; Petzold et al.,
2005; Chirico et al., 2010; Heringa et al., 2011). Aerosol hygroscopicity is
determined by the chemical composition of each individual particle
(Heintzenberg et al., 1990; Gysel et al., 2003; McMeeking et al., 2011; Liu
et al., 2013). Thus, aerosol mixing states play a critical role in
determining the hygroscopicity of BC particles.
Many studies have reported the mixing state of atmospheric BC particles. For
example, aerosol's mixing state can be determined using a hygroscopic tandem differential mobility analyzer (HTDMA) (Swietlicki
et al., 2008) or a combined volatility–hygroscopicity TDMA (VHTDMA) (e.g.,
Johnson, 2005). Shiraiwa et al. (2007) investigated the evolution of mixing
state of BC using a SP2 in the polluted air transported from the Tokyo area
in summer. The fraction of thickly coated BC with a core diameter
(DC) of 180 nm increased at a rate of 1.9 % h-1. The
increase rates were lower for larger DC values (Shiraiwa et al., 2007).
Healy et al. (2012) used an aerosol time-of-flight mass spectrometer (ATOFMS)
to study the mixing state of BC particles in Paris. The smaller BC particles
(Dva ≤ 400 nm) were mainly externally mixed, indicating
they were from local or regional sources, while bigger BC particles
(Dva ≥ 400 nm) were mainly internally mixed with nitrate,
indicating they were from medium to long-range transport. Kuwata and
Kondo (2008) conducted volatility TDMA (VTDMA) measurements and showed that
the aerosol was often an external mixture of less and more-volatile
particles.
According to our knowledge, there are only a few direct measurements of BC
particles' hygroscopic properties in the atmosphere. One of the previously
used techniques was coupling hygroscopic measurements with a VTDMA system. It
was found that the less volatile aerosol components were mainly composed of
BC at close proximity to urban environments (Kuwata et al., 2007; Rose et
al., 2011). The relationship between hygroscopicity and mixing state of BC
aerosols had been studied. BC particles exposed to subsaturated sulfuric acid
vapor exhibit a large change in morphology. These particles are very
hygroscopic and act as efficient cloud condensation nuclei. Coating with
sulfuric acid and subsequent hygroscopic growth increase their light-scattering coefficient by 10-fold and light absorption coefficient by nearly
2-fold at relative humidity (RH) = 80 % compared to uncoated BC particles (Zhang et al.,
2008). Herich et al. (2008) combined an ATOFMS with a HTDMA to investigate
the mixing state and hygroscopicity of BC-containing particles at an urban
site in Zurich, Switzerland. The result shows that most BC-containing
particles internally mixed with organics and combustion species
(-26CN- and -42CNO-). They have lower hygroscopicity
compared with mixed sulfate and nitrate (Herich et al., 2008). With a similar
setup, our previous study finds that condensation of amine and secondary
inorganic species would enhance the hygroscopicity of submicron particles,
including BC particles (Wang et al., 2014). Laborde et al. connected a SP2 downstream of a HTDMA (RH = 90 %) and show that the majority of
urban aerosol particles with high hygroscopicity (growth factor, GF ≈1.6) do not
contain a detectable refractory BC (rBC) core, while hydrophobic or less
hygroscopic particles (1.1 ≤ GF ≤ 1.2) have a BC core with no or
little coating of soluble species (Laborde et al., 2013). Similarly, by
coupling a SP2 with a HTDMA, McMeeking et al. (2011) introduced a method for
measuring the hygroscopicity of externally and internally mixed BC particles.
They tested this technique using uncoated and coated laboratory-generated
model BC compounds. The obtained information is compared to the
hygroscopicity distribution of ambient BC aerosols. Their results suggest
that the dominant fraction of the BC particles does not readily act as cloud
condensation nuclei at 0.2 % supersaturation in an urban area. In
addition, Liu et al. (2013) deployed a similar instrument setup and
investigated the relation between the hygroscopic properties and mixing state
of BC particles (Liu et al., 2013). It shows that the GF of BC particles was
influenced by the composition of soluble materials.
These previous studies have relied on real-time hygroscopicity measurements
but usually without detailed temporal information on changes of mixing state
of BC particles. Time-resolved information on aerosol mixing state would be
very useful to identify their sources and aging processes. During summertime in heavily polluted areas, atmospheric aging processes could be
much more complex due to higher temperature, higher pollutant
concentrations, and stronger sunlight radiation. However, according to our
knowledge, few studies have reported time-resolved analysis on the mixing
state and hygroscopicity of BC particles of particular GFs during summertime. Therefore, with high-time-resolution single-particle analysis, this
field study has aimed to determine the relations between BC particle
hygroscopicity and real time mixing state measurement in Shanghai, a heavily
polluted megacity area. It would provide more insights on the effects of
atmospheric processes on hygroscopicity of BC particles.
Experimental section
HTDMA–SP2 system
Temporal variation in BC aerosol mixing state and hygroscopic property were
measured using a custom-built HTDMA–SP2 system (Fig. 1). Similar to systems
that couple an HTDMA with another instrument, such as those used by Herich et
al. (2008), Zelenyuk et al. (2008), and Wang et al. (2014), our system used SP2 (DMT, Boulder,
CO, USA) downstream of a HTDMA to measure rBC content as a function
of hygroscopicity. The first DMA in the HTDMA system selects monodisperse
dried particles. Then the selected aerosols are humidified at a specified
RH (RH = 85 % in this study). The size distribution of
humidified (wet) particles is measured with a scanning mobility particle
sizer
(SMPS), which includes another DMA and a condensation particle counter (CPC,
model 3771, TSI Inc.). GF is the ratio between particle wet size and dry
size. The two DMAs were operated with recirculating sheath flows and a
sheath-to-sample flow ratio of 10:1. The HTDMA is encapsulated in a
thermostatted box to reduce temperature fluctuations. The DMA housing
temperature was controlled at 20 ∘C. Aerosol flow was set at
0.43 L min-1 (the sum of the CPC (0.4 L min-1) and the SP2
(0.03 L min-1) flow rates). The hygroscopicity measurement was
calibrated using (NH4)2SO4 particles. The SP2 identifies
BC-containing particles at each selected GF. The water uptake properties of
BC particles can be linked directly to the mixing state measured by SP2.
Schematic diagram of experimental setup.
Temporal profiles of gaseous pollutants (O3,
SO2, CO, and NO2), temperature, relative humidity
(RH), and PM2.5 and PM10 mass concentrations.
SP2 can measure number and mass size distribution of rBC-containing particles
(Baumgardner et al., 2004; Schwarz et al., 2006). Briefly, SP2 detects
incandescence and scattering light signals of rBC-containing particles
induced by a 1064 nm Nd:YAG intra-cavity laser. The mass of rBC is
proportional to the intensity of the incandescence signal. A particle with an
incandescence signal (above a threshold) is treated as an rBC particle, while
a particle that only exhibits scattering signal is considered a non-rBC
particle. SP2 detection efficiency was close to unity for larger rBC
particles. The minimum rBC mass that could be observed with near-unity
detection efficiency was ∼ 0.7 fg rBC, corresponding to a 90 nm mass-equivalent diameter; the detection efficiency declined rapidly at lower sizes
(Gong et al., 2016). The total ambient mass concentrations of rBC were
possibly underestimated because of the reduced detection efficiency for small
rBC particles (Schwarz et al., 2006; McMeeking et al., 2010).
The conversion from rBC mass to the effective rBC core diameter requires us to
assume an effective density for rBC cores in the particles. In this study, an
effective density of 1.8 g cm-3 was used to convert the ambient rBC
mass to the mass-equivalent diameter. This value was recommended by many
previous studies (Bond and Bergstrom, 2006; Gong et al., 2016).
The scattering properties of externally and internally mixed rBC particles
may be distorted due to particle mass loss induced by laser heating in SP2.
Thus, scattered light from an rBC particle may not yield a full Gaussian
waveform. The Gaussian scattering function was reconstructed from the leading
edge of the scattering signal (before particle is heated by the laser), which
was measured with a two-element avalanche photodiode (APD). This
method (leading-edge-only fit or LEO fit) can determine the scattering
properties of individual rBC particles more accurately
(Gao et al.,
2007). Optical diameter of an rBC particle (Dp) was derived from
Mie theory with the LEO fitted scattering signal and rBC core size
(Dc) (Moteki et al., 2010; Liu et al., 2014; Laborde et al.,
2013). The absolute coating thickness of an rBC particle was calculated as
(Dp – Dc) / 2, based on the assumption of a
concentric core–shell morphology. However, rBC aging in the atmosphere may
result in an imperfect core–shell structure (Matsui et al., 2013).
Averaged hygroscopic growth distributions of (a) total
ambient particles and (b) BC particles at RH = 85 % for
three selected sizes (D0=120, 240, and 360 nm).
Diurnal variations in BC (a) number concentration and
(b) number fraction (the ratio of the number concentration of BC
particles to that of total particles at a certain D0 and GF) for three
selected sizes.
In this study, ambient particles with three electrical mobility sizes (120,
240, and 360 nm in dry diameter) were selected first by a DMA, then humidified
at a RH = 85 %. For the measurement of the overall hygroscopic
distribution of total ambient particles (such as Fig. 3a), the second DMA was
operated in a continuous scanning mode. For the measurement of temporal
trends in the hygroscopic distribution, the second DMA was operated in a
stepped mode by sending particles with fixed GFs of 1.0, 1.2, and 1.4
(representing hydrophobic, transition, and hydrophilic modes, respectively) to
the CPC and SP2.
Single-particle aerosol mass spectrometer (SPAMS)
A single-particle aerosol mass spectrometer (SPAMS Hexin Analytical
Instrument Co., Ltd., China) was used in parallel to the HTDMA–SP2 system.
The SPAMS first measures the size of a single aerosol particle. Then, it uses
a 266 nm laser to disintegrate the particle and ionize its chemical
compounds, of which mass-to-charge ratios (m/z) and concentration are
determined with a bipolar time-of-flight mass spectrometer. Detailed
information on the SPAMS has been described elsewhere (Li et al., 2011).
In this work, a total of 158 410 individual particle mass spectra were
collected, accounting for about 56 % of all the particles that were sized
in the SPAMS. Elemental carbon (EC) ion clusters have been considered an
important marker for BC aerosols (Gong et al., 2016). Using
Cn+ (n = 1, 2, 3…) as the BC marker, a total of
64 368 BC-containing particles were identified, accounting for about
40.1 % of sampled particles. The mass spectra of BC-containing particles
were classified into several types based on their similarities using a
clustering algorithm called adaptive resonance theory (ART-2a) (Song et al.,
1999). Similar to previous studies (Huang et al., 2013; Zhai et al., 2017;
Gong et al., 2016; Spencer et al., 2007), the vigilance factor, learning
rate, and iterations for the ART-2a algorithm were set to 0.85, 0.05, and 20,
respectively. Finally, five particle types were manually combined based on
the similarity of their chemical nature.
Noticeably, the particles with a mobility size at 120 nm cannot be detected by
the SPAMS, as they were smaller than the lower limit (200 nm) of the size range
of the SPAMS. According to Slowik's study, the vacuum aerodynamic diameter
(dva) of compact aggregated BC particles was linearly
proportional to mobility diameter (dm), specifically,
dva = 1.3⋅dm (Slowik et al., 2004). Here we
assume that most detected BC particles follow this relation
(dva = 1.3⋅dm). In this study, only
particles with D0 = 120, 240, and 360 nm were studied with the
HTDMA–SP2 system. These mobility sizes correspond to
dva = 150, 312 and 468 nm, respectively. Thus, the SPAMS
cannot provide mixing state information for particles with
dva = 150 nm (dm = 120 nm), which was out
of the SPAMS detection range (200 to 2000 nm).
Other instruments
OC / EC analyzer
Hourly mass concentrations of EC and organic carbon (OC)
were measured using a semicontinuous OC / EC analyzer (model 4, Sunset
Laboratory Inc., Portland, USA) based on the National Institute of
Occupational Safety and Health thermal–optical transmittance measurement
protocol (NIOSH 5040), with a PM2.5 impactor inlet. Detailed information
can be found in a previous publication (Wang et al., 2016a). Concentration
of secondary organic carbon (SOC) was estimated using the method of the minimum
ratio of OC / EC (Chou et al., 2010), which is calculated by the
following equation.
SOC=OCtotal-EC⋅(OC/EC)pri,
where OC and EC are the measured hourly mass concentrations of OC
and EC. (OC / EC)pri is the OC / EC ratio
in primarily emitted combustion aerosols. At urban locations, the
(OC / EC)pri was assumed to be the minimum value of the
OC / EC ratio throughout the whole study sampling period (Cao et al.,
2013). The OC vs. EC plot is displayed in Fig. S1 in the Supplement. We use
the minimum value 2.2 as the (OC / EC)pri in this work.
Monitor for AeRosols and Gases in Air (MARGA)
A Monitor for Aerosols and Gases in Air (MARGA, Applikon Analytical B. B.
Corp., ADI 2080, the Netherlands), with a PM2.5 cyclone impactor, was
deployed to measure the concentrations of inorganic ionic species (i.e.,
SO42-, NO3-, and NH4+) in
PM2.5. Detailed description about the MARGA is available in the previous
publication (Du et al., 2010).
Sampling period and site
The measurements of the relationship between mixing state and hygroscopicity
of BC particles varying with time were carried out from 4 to 16 July 2017
using a SPAMS and a HTDMA–SP2 system. The sampling site is located at the
Department of Environmental Science and Engineering on the main campus of
Fudan University (31.30∘ N, 121.5∘ E), and it is surrounded
by residential and commercial areas. An elevated road (the Middle Ring Line)
with heavy traffic is ∼ 400 m away from the sampling location.
Meteorology
The meteorology and air quality information were obtained at a nearby air
quality monitoring station, which is operated by the Shanghai Environmental
Monitoring Center (Yangpu site)
(http://www.semc.com.cn/aqi/home/Index.aspx, last access: 29 May 2018).
The station was 3.3 km from the sampling site. Temporal profiles of measured
gaseous pollutants (O3, SO2, CO, and NO2),
temperature, RH, PM2.5, and hourly PM10 mass concentrations from 4
to 16 July 2017 are shown in Fig. 2. The temperature and RH varied between
24.1 and 38.0 ∘C and 46 % and 100 %, with an average of
30.4 ∘C and 74.1 %, respectively, during the sampling period.
Figure 2 shows that the temperature was negatively correlated with RH but
positively correlated with O3 mass concentration. Hourly
O3 concentration usually peaked in the afternoon during this
period. Its maximum value reached 316 µg m-3 at 14:00 LT on
12 July, showing extremely active photochemical activities in this afternoon.
The maximum value of PM2.5 reached 72 µg m-3 at
18:00 LT on 13 July with an average of 29.2 µg m-3.
Meanwhile, PM10 varied from 15 to 141 µg m-3, with an
average of 57.9 µg m-3.
Results and discussion
The focus of this paper is temporal variation in hygroscopicity and mixing
states of ambient BC particles. Due to the slow scanning rate of DMA voltage
and sampling time requirement of SP2, obtaining temporal information of BC
concentrations requires the DMA size selection to be fixed for a certain
amount of time. Therefore, for the HTDMA–SP2 system, only one size (D0)
was studied for each sampling period. Specifically, D0 = 120, 240,
and 360 nm were measured during 4 July 2017 06:00:00 LT–7 July 2017
05:00:00 LT, 7 July 2017 06:00:00 LT–10 July 2017 05:00:00 LT, and
10 July 2017 06:00:00 LT–16 July 2017 05:00:00 LT, respectively.
We studied three GFs (GF = 1.0, 1.2, and 1.4) for each D0. The GF
selection was based on the GF size distribution of BC particles. A general
picture of hygroscopicity of total sampled ambient particles is shown in
Fig. 3a, which illustrates averaged hygroscopic GF distributions
at three selected sizes (D0 = 120, 240, and 360 nm). All three GF
curves featured a bimodal distribution, which contained a hydrophobic mode
peak at GF = ∼ 1.0 and a hydrophilic mode peak at larger GFs
(1.3–1.6). Clearly, the hygroscopic particles were typically more abundant
than the hydrophobic ones. These hygroscopic particles featured a size-dependent hygroscopic growth, significantly shifting to a larger GF with
increasing particle size. This feature was conventionally attributed to a
size-dependent chemical composition (Swietlicki et al., 2008; Ye et al.,
2013). In contrast, hygroscopic GF distributions of BC particles
only show one mode (Fig. 3b). The GF curve of BC particles peaked at
GF = ∼ 1.0. The BC-containing number fraction decreased sharply for
larger GFs. When GF was greater than 1.4, few BC particles were detected by
the
SP2.
GFs of 1.0, 1.2, and 1.4 were selected to represent hydrophobic, transition,
and hydrophilic modes for BC particles, respectively. Higher GFs
(GF > 1.4) were not selected due to the low BC-containing number
fractions at these GFs. For each day, the sampling was divided into eight 3 h
sampling periods. Three GFs (GF = 1.0, 1.2, and 1.4) were set sequentially
for 1 h during each sampling period.
Here, the diurnal trend in (in Sect. 3.1) and classification of (in Sect. 3.2)
BC particles are described and discussed in detail. Then, the relations
between mixing state and hygroscopicity of BC particles are elucidated (in
Sect. 3.3).
Diurnal variations in BC particles with different
hygroscopicities
Diurnal variation in hydrophobic-mode BC particles
The averaged diurnal patterns of the number concentration and number fraction
of BC particles are shown in Fig. 4a and b, respectively. The number fraction
of BC particles is defined as the ratio of the number concentration of
sampled rBC particles to that of total samples, including
rBC and non-BC particles, at a certain D0 and GF. The majority
of hydrophobic-mode (GF = 1.0) BC particles typically exhibited two peaks
for all three D0 values (Fig. 4a). They were likely to be freshly emitted from
combustion sources (McMeeking et al., 2011). The number concentrations of
hydrophobic BC particles reached their first peak in the morning around 06:00
to 09:00 local time (LT) and were then followed by a dip in the afternoon around
12:00 to 17:00 LT. The hydrophobic BC particles reached the second peak in
the evening and then slowly decreased during the night. This trend is similar
to that found in some field studies in other city areas, such as Shenzhen and Xiamen in
China (Huang et al., 2012; Wang et al., 2016b). The elevated BC particle
number concentration in the morning and early evening can be explained by
increases in local anthropogenic emissions, especially those from rush hour
traffic (Dreher et al., 1998; Allen et al., 1999; Bhugwant et al., 2000).
This trend presumably was also intensified by lower boundary layer heights at
those times.
As shown in Fig. 4b, hydrophobic BC particles accounted for the largest
percentage among three GFs for all D0 values. However, the number fractions of
hydrophobic BC particles decreased with the increased D0. For
D0 = 120, 240, and 360 nm, the maximum number fractions of the
hydrophobic BC particles were ∼ 80 %, 70 %, and 60 %,
respectively. One possible reason is that the majority of fresh BC particles'
diameters are smaller than 200 nm (Kondo et al., 2006), corresponding to our
finding that relatively lower fractions of BC particles were detected at
larger sizes. In addition, factors affecting the BC number fractions at
GF = 1.0 may also be related to the behavior of non-BC-containing
particles and their size dependence.
Diurnal variations in rBC (a) core size and
(b) coating thickness at different GFs.
Diurnal variations in transition and hydrophilic-mode BC
particles
The transition and hydrophilic-mode BC particles likely originated from aged
particles. Condensation of hydrophilic secondary materials (e.g., sulfate,
nitrate, and secondary organic compounds) would significantly enhance the
water uptake ability of BC particles. As shown in Fig. 4a, the number
concentrations of some transition and hydrophilic-mode BC particles (like
transition mode for 120 and 240 nm BC particles) showed a clear daily
maximum during 12:00–15:00 LT (Fig. 4a). This trend could be explained by
the intense aging processes during this time when sunlight intensity and
atmospheric oxidants' concentration reach their peak values. However, the
number concentration of transition and hydrophilic-mode BC particles could be
affected by other atmospheric aging processes. For example, nitrate formation
could be significantly enhanced during nighttime due to the hydrolysis of
N2O5 (Mozurkewich et al., 1988; Wang et al., 2009, 2016a). This
process would make more transition and hydrophilic-mode BC particles during
nighttime. To understand their diurnal trends, the measurement of their
chemical compositions and mixing states is essential, which will be discussed
in Sect. 3.3.
Figure 4b also shows the diurnal variations in number fractions of transition
and hydrophilic-mode BC particles. Unlike the hydrophobic mode BC particles,
the transition and hydrophilic-mode (GF = 1.2 and 1.4) BC particles with
larger sizes tended to contribute higher number fractions of total particles.
This trend was more pronounced for the hydrophilic mode (GF = 1.4): the
maximum number fractions of BC particles were ∼ 10 %, 10 %, and
20 % for D0 = 120, 240, and 360 nm, respectively. One possible
reason is that the sizes of fresh BC particles are likely to be small. The
median diameter of BC particles originating from traffic emissions is usually
< 200 nm (Harris et al., 2001; Zervas et al., 2006; Xue et al.,
2015). They have to grow to larger sizes (e.g., 360 nm) through
aging/coating, which also increases their hygroscopicity. Another possible
reason is that these hydrophilic BC particles were from a different source. A
candidate is biomass burning (BB) aerosols, which have slightly higher
hygroscopicity than those from traffic emissions (Laborde et al., 2013).
Detailed discussion of BB aerosol will be shown in Sect. 3.2.
In addition, the diurnal variations in the BC particle number fraction showed
that during nighttime, a larger proportion of BC particles were in
hygroscopic mode compared to during daytime, indicating that hygroscopicity of BC at
night was much stronger than that in the daytime. The main reason will be
discussed in Sect. 3.3.
Diurnal variations in rBC core diameter and coating thickness
The rBC core diameter (Dc) can be obtained if a single particle
BC mass is known and assuming that the core is spherical and BC density is
1.8 g cm-3. The corresponding coating thickness is (Dp=Dc) / 2 (Gong et al., 2016). Diurnal variations in average rBC
core diameter and coating thickness at different GFs are displayed in Fig. 5a
and b, respectively. For a certain BC particle size, a larger core size and a
thinner coating thickness corresponded to BC particles with lower
hygroscopicity (e.g., GF = 1.0). When BC particles became more hygroscopic
(i.e., GF increases), the coating thickness increased.
It is interesting to note that the core sizes for the hygroscopic-mode BC
particles increased during nighttime (21:00 to 06:00 LT), while coating
thickness decreased (since the entire electron mobility diameter was fixed).
This observation suggests that the coating material on BC particles might be
different between daytime and nighttime. To achieve the same GF, an increased
coating thickness is required for less hygroscopic coating materials, as the
hygroscopicity of a rBC core is always constant (GF = 1.0).
In this study, the main uncertainty associated with the HTDMA-measured GF of
soot particles was influenced by particle morphology. For fresh BC particles
with an aggregate structure, the mobility diameter (Dmob)
measured by a DMA is normally larger than its geometric volume / mass
equivalent diameter (Dve) (DeCarlo et al., 2004). However,
coating on soot aggregates can modify its morphology (Weingartner et al.,
1997; Lewis et al., 2009; Pagels et al., 2009) by making soot aggregate more
compact. Change of particle morphology affects Dmob measurement.
It has been reported that more compact BC particles tend to exhibit smaller
mobility diameter and higher effective density (Zhang et al., 2008; Pagels et
al., 2009). In HTDMA measurements, if BC-containing particles' shape is
significantly fractal, the water adsorption process in HTDMA would likely
make them more compact, and therefore their GFs soot particles would be
underestimated. These effects are less pronounced for particles that are
less fractal. Due to the limitation of HTDMA, the complex morphology or
ρBC of BC-containing particles cannot be explicitly determined
in this study. Thus, a conventional core–shell model for a BC-containing
particle has to be assumed. The GF = 1.0 results show average coating
thicknesses of ∼ 20–35 nm, suggesting the presence of non-BC
materials or the effect of a nonspherical shape on size measurement.
Diurnal variations in the distribution of BC particle growth
factors
Fresh BC particles usually have a low GF. Through aging processes, the GF of BC
particles increases. It would be interesting to see how the GF distributions
of BC particles change during the day and obtain a rough estimate of how fast the
aging process occurred. Figure 6 shows the diurnal variations in the
distribution of BC particle growth factors. Note only three GFs were measured
by the HTDMA–SP2 system. Here, the BC aerosol number fraction for each GF is
defined as (BC aerosol number concentration for this GF) / (sum of BC
aerosol number concentration for GF = 1.0, 1.2, and 1.4). It is found that
the BC aerosol number fraction for GF = 1.0 reached two maxima at around
09:00 and 18:00 LT, probably due to rush hour traffic. Only 3 h after
09:00 or 18:00 LT, the BC aerosol number fractions for GF = 1.0 dropped
significantly (from 0.44 to 0.26 and from 0.40 to 0.36, respectively).
Meanwhile, the BC aerosol number fractions for GF = 1.4 increased (from
0.34 at 09:00 LT to 0.49 at 12:00 LT and from 0.44 at 18:00 LT to 0.50 at
21:00 LT). Evidently, the GF distribution of BC particles changed rapidly
even in just 3 h. This change was probably due to BC particle
aging, and the aging timescale is around several hours.
Diurnal variations in the GF distribution for BC particles. The BC
aerosol number fraction for each GF is defined as (BC aerosol number
concentration for this GF) / (sum of BC aerosol number concentration for
GF = 1.0, 1.2, and 1.4).
The coating thickness of BC particles from this study (20 ∼ 80 nm for
BC particles with an electrical mobility diameter of 120–360 nm) was in the
range of previous measurements from other cities. For example, Laborde et
al., measured BC coating thickness in Paris during wintertime (Laborde et
al., 2013). They found coating thickness was approximately 33 nm on average
for a rBC core size from 180 to 280 nm. Liu et al. report an average coating
thickness of ∼ 40 nm for BC particles with an electrical mobility
diameter of 163 nm (Liu et al., 2013). A field study in London during
wintertime shows that the average coating thickness for BC particles
137, 143, and 169 nm in diameter were ∼ 15, 22, and 33 nm, respectively
(Liu et al., 2014).
The rapid change in BC particle coating thickness suggests that the aging
timescale was around several hours. This timescale is consistent with a
previous modeling study (Riemer et al., 2004).
BC-containing particle types identified by SPAMS
The mixing state and aging degree of BC-containing particles have been
studied using a SPAMS. Noticeably, the particle size range for the SPAMS is
from 200 to 2000 nm. The detection efficiency drops rapidly below 400 nm
and above 1200 nm (Li et al., 2011). However, most pure BC particles are
smaller than 200 nm (Kondo et al., 2006), which is close to the lower size
limit for the SPAMS and can only be detected at a low efficiency. Based on
SPAMS mass spectra patterns, BC particles were classified into five types:
EC, NaKEC (sodium- and potassium-rich EC), ECOC (EC and OC), KEC (potassium-rich EC), and others. Their relative
contributions are shown in Table 1. The average mass spectra for each
particle type are shown in Fig. S2 in the Supplement.
Number counts and fractions of the five types of BC-containing
particles detected by the SPAMS.
Number count
Fraction of
Type
of particles
particles
Pure EC
5191
8.1 %
KEC
21 456
33.3 %
NaKEC
11 001
17.1 %
ECOC
21 225
33.0 %
Others
5495
8.5 %
Total BC-containing
64 368
100 %
particles
Pure EC particles only presented BC fragment ions (Cn+ and
Cn-) in both positive and negative ion mass spectra. There
were low signals of secondary species, such as sulfate or nitrate, indicating
that pure EC had not gone through significant aging in the atmosphere; thus
the EC type was made up of freshly emitted BC particles.
NaKEC particles exhibited strong signals for BC fragment ions in both
positive and negative mass spectra, additionally potassium (+39K+)
and sodium (+23Na+) in positive ion mass spectra, and nitrate
(-46NO2- and -62NO3-) and sulfate
(-97HSO4-) in the negative ion mass spectra. Hydrocarbon-like
organic aerosol (HOA) is dominated by alkyl fragment signatures as well as
the CnH2n+1+ (m/z = 29, 43, 57) and CnH2n-1+
(m/z = 27, 41, 55) ions. The time series of HOA correlated well with
those of NO2 and CO, two tracers of vehicle emissions (Fig. S3a in
the Supplement). The diurnal pattern of HOA ion intensity further suggests
the association of HOA with traffic activities, as it showed two obvious
peaks during the morning and evening rush hours (Li et al., 2017). HOA as a
tracer of traffic emission correlated reasonably well with the NaKEC particle
numbers (R2=0.560), as shown in Fig. S3b in the Supplement.
The ECOC particles internally mixed with many OC signals, including
+37C3H, +43CH3CO+,
+50C4H2+,
+51C4H+,
+61CH3C(OH) = OH+,
+62(CH3)2NHOH+, and +23Na+ and
BC fragment ions (Cn+). The presence of a high signal intensity for
sulfate (-97HSO4-) and a relatively low signal intensity for
nitrate (-46NO2-, -62NO3-) suggests that they
were aged BC particles. BC particles with various intensities of OC, nitrate,
and sulfate were commonly detected in ambient measurements with an ATOFMS
(Moffet et al., 2008; Ault et al., 2009; Dall'Osto and Harrison, 2006) and
were also attributed to aged traffic emissions (Healy et al., 2012).
KEC particles were characterized by an intense +39K+ signal in
the positive ion mass spectra and strong signals for -26CN- and
-42CNO- in the negative ion mass spectra. Significant
intensities of ion fragments of levoglucosan, such as
-71C3H3O2- and -73C3H5O2-, were
also observed. Typical BC fragments Cn- appeared in the
negative ion mass spectra. Similar to ECOC, the presence of a high signal
intensity of -97HSO4-, -46NO2-, and
-62NO3- signals indicates significant particle aging in the
atmosphere (Leskinen et al., 2007; Reid et al., 2005). These characteristics
suggest that their sources are BB or coal combustion or both (Andreae, 1983;
Soto-Garcia et al., 2011; Wang et al., 2013; Gong et al., 2016). Particles
with similar mass spectral patterns previously observed in several urban
field studies were also assigned to the sources of combustion of biomass or
coal (Moffet et al., 2008; Healy et al., 2012; Bi et al., 2011; Wang et al.,
2013; Gong et al., 2016).
The others particle type was not grouped to any of the previous four
types and it accounts for only 8.5 % in total BC particle number
concentration. The average mass spectrum of this particle type is displayed in
Fig. S2e in the Supplement. However, this is not the focus of this study.
The relative fractions of aerosol types as a function of particle size were
plotted in Fig. S4 in the Supplement. Generally, the number fraction for each
particle type is highly dependent on particle size. Sharp changes in BC
particle mixing states have been found between the size ranges of 200–400 nm
and 400–800 nm. EC and NaKEC are the major fraction types in the
200–400 nm size range. In contrast, the larger size range (400–800 nm)
was dominated by ECOC and KEC types. For convenience of discussion, we
separated particles into two groups based on their dva, namely
G200-400 (200 nm < dva < 400 nm) and
G400-800 (400 nm < dva < 800 nm). In
this work, dva = 150 nm (dm = 120 nm) was
out of the SPAMS detection range, while dva = 312 nm
(dm = 240 nm) and dva = 468 nm
(dm = 360 nm) fell in the range of G200-400 and
G400-800, respectively.
Diurnal variations in number fraction of each classified particle
type:
(a) G200-400:200 < dva < 400 nm
and
(b) G400-800:400 < dva < 800 nm.
Averaged diurnal variations in Ox, SOC,
SO42-, and NO3- mass concentration in
(a), (b), (c), and (d), respectively. The
blue line in (d) is daily variation in relative peak area of
NO3- measured by SPAMS.
The diurnal variations in number fraction of each particle type in
G200-400 were calculated and shown in Fig. 7a. It is found that number
fraction of the EC and NaKEC types displayed pronounced diurnal patterns with
two major peaks in the early morning (06:00–09:00 LT) and in the evening
(18:00–21:00 LT), possibly relating to traffic. From the hygroscopicity
measurement (D0 = 240 and 360 nm in Fig. 4a), the elevated BC
particle number concentration at GF = 1.0 during these two time frames
suggests that the elevated concentrations of hydrophobic BC particle aerosol
were probably associated with EC and NaKEC types, which are produced from
traffic sources.
A different pattern has been observed for G400-800 (shown in Fig. 7b).
The ECOC and KEC types accounted for the major number fractions in the
400 nm ∼ 800 nm range. The diurnal variations in these two particle
types share a similar trend, while the other two types (EC and NaKEC) showed no
significant variation. The EC and NaKEC types only attributed to small
proportions of G400-800 particles. Interestingly, the number fraction of
ECOC in G400-800 also showed two major peaks in the morning and evening,
suggesting that ECOC was aged BC particles from traffic emissions. The KEC
peak in the evening was much more pronounced than that in the morning, and
this might be due to BB, which is still widely used for domestic
cooking (in the evening) in the countryside around Shanghai. These BB
aerosols were then transported to the sampling site.
The relations between the mixing state and hygroscopicity of BC
particles
To elucidate the relation between mixing state and hygroscopicity of BC
particles, the detailed chemical composition and mixing state information
from the EC / OC, MARGA, and SPAMS was compared to the HTDMA–SP2
hygroscopicity measurements. As discussed in Sect. 3.1, we found that the rBC
core sizes for the hygroscopic-mode particles increased during nighttime
(21:00–06:00 LT) while coating thickness decreased, indicating the BC
particle coating compositions were different between daytime and nighttime.
The major secondary aerosol coating materials in the polluted boundary layer
could be SOC, sulfate, and nitrate. Therefore, the
diurnal trends in these species have been investigated and compared to BC
particles' hygroscopicity.
Major secondary ionic species
The dominant ionic species in urban aerosols in Shanghai are sulfate,
nitrate,
and ammonium (Ye et al., 2013). To study the chemical composition dependence
on hygroscopicity, mass concentrations of SO42-,NO3-, and
NH4+ were measured using a MARGA during this field study. As shown
in Fig. 8c, the average sulfate concentration varied in a small range from
∼ 4.8 to 6.1 µg m-3. Its concentration in the daytime
was only slightly higher than that of nighttime. The average mass
concentration of NO3- varied between
1.1 µg m-3 and 4.4 µg m-3 with an average of
2.3 µg m-3. Similar to our previous study (Wang et al.,
2016a), the nitrate concentrations at night were clearly elevated (Fig. 8d).
The relative peak area (RPA) of NO3- in BC particles
measured by SPAMS is also consistent with the MARGA measurement (Fig. 8d).
During the nighttime in summer, lower temperature, higher RH,
and a high concentration of NO3 (N2O5) favor the formation
of nitrate in the particle phase (Wang et al., 2009, 2016a).
Elevated SOC concentrations in the daytime
Photochemical reactions are a major formation pathway of SOC (Kroll and
Seinfeld, 2008; Zhang et al., 2018). Odd oxygen
(Ox = O3 + NO2) was often used
as an indicator of photochemical oxidant concentration in the atmosphere
(Herndon et al., 2008; Hu et al., 2016; Wood et al., 2010). The diurnal
variations in SOC and Ox mass concentrations at a 1 h
resolution were plotted in Fig. 8. In this work, average SOC and
Ox varied between 1.8 and 8.8 µgC m-3 and
58 and 214 µg m-3, respectively. The correlation coefficient
(R) between SOC and Ox was 0.772 (shown in Fig. S5 in
the Supplement), indicating that the SOC formation was associated with the
photochemical oxidant concentration during this study.
Single-particle mass spectrometry was also used to further investigate the
mixing state and possible formation pathways of SOC. The RPA
of +43[CH3CO+ / CHNO+] during the daytime is a tracer of
SOC formation (Qin et al., 2012; Zhang et al., 2014, 2018). Time series of
hourly-averaged RPAs of m/z +43 in ECOC particles was shown in
Fig. 9. Overall, the m/z +43 curve peaked in the afternoon, which was
consistent with the trend in Ox. This result indicates
that SOC (m/z = +43) produced by photochemical reactions condensed on
BC particles. The average ECOC particle size versus time is also shown in
Fig. 9. It peaked between 13:00 to 15:00 LT in the afternoon. Since the
concentration of sulfate in the daytime was only slightly higher than that in
nighttime (Fig. 8c), the increase in ECOC particle size was mainly caused by
the condensation of SOC rather than secondary inorganic species. Therefore,
in the afternoon, the intense photochemical process resulted in BC particles
coated with more organic materials, leading to an increased coating thickness.
Diurnal variations in relative peak area (RPA) of m/z+43 from
ECOC particles and the average aerodynamic particle size of ECOC
particles.
Relations between the coating thickness of 360 nm for BC particles with a GF of 1.4 and the concentration of (a) SOC,
(b) nitrate, and (c) sulfate; relations between the number
fraction of BC particles in 360 nm aerosol with a GF of 1.4 and the
concentration of (d) SOC, (e) nitrate, and
(f) sulfate.
Hygroscopicity and mixing state (coating material)
As discussed above, the chemical composition measurement clearly shows BC
particles were coated with more SOC in the daytime and with more nitrate in
the nighttime. Sulfate concentration did not change much between daytime and
nighttime. Meanwhile, at a given GF, the coating for hygroscopic-mode BC
particles was thicker in the daytime and thinner in the nighttime. The water
uptake ability of nitrate is much stronger than secondary organics. Thus,
compared to SOC, less nitrate coating is needed for a given hygroscopicity
or GF.
To better understand this finding, we estimated volumes of different coating
materials required for a BC particle with a given hygroscopicity using the
Zdanovskii–Stokes–Robinson (ZSR) mixing rule (Stokes and Robinson, 1966),
GFZSR(RH,Dp)=∑iGFi(RH,Dp)3εi.1/3.
The εi is the volume fraction of rBC, nitrate, sulfate, or
organic coating in BC particles. For simplicity, we assume rBC is covered by
a
mixture containing either SOC / (NH4)2SO4 or
(NH4)2SO4 / NH4NO3, representing the
mixing state of BC in daytime and nighttime, respectively. The GF for SOC is
set to be 1.2 since Sjogren et al. reported a uniform growth factor
GFSOA = 1.2 (RH = 90 %) according to the ZSR
modeling results and field measurements (Sjogren et al., 2008). This value
is at the high end of a previously measured SOA hygroscopicity range, thereby
representing highly aged and oxidized SOA (Varutbangkul et al., 2006;
Baltensperger et al., 2005). The GFs of pure BC,
(NH4)2SO4, and NH4NO3 aerosol with a dry size of
163 nm at RH = 90 % are calculated using the Aerosol
Diameter-Dependent Equilibrium Model (ADDEM) model. Their values are 1.0,
1.7,
and 1.8, respectively (Topping et al., 2005a, b).
For a coated BC particle with a dry diameter of 163 nm and a GF of 1.4, the
following relations would hold for a SOC / (NH4)2SO4
coating (Eq. 3) and (NH4)2SO4 / NH4NO3
coating (Eq. 4).
1.4=3εBC⋅GFBC3+εSOC⋅GFSOC3+ε(NH4)2SO4⋅GF(NH4)2SO431.4=3εBC′⋅GFBC3+εNH4NO3⋅GFNH4NO33+ε(NH4)2SO4⋅GF(NH4)2SO43
εBC and εBC′, representing the
volume fractions of the rBC core in two mixing states, are given by
εBC=1-εSOC-ε(NH4)2SO4 and
εBC′=1-εNH4NO3-ε(NH4)2SO4, respectively.
Here, we assume
ε(NH4)2SO4 is
constant in both mixing states.
Combining the equations above, the ratio of volume fraction of SOC to
NH4NO3 is
εSOCεNH4NO3=6.6.
This calculation shows that a higher volume fraction of
SOC / (NH4)2SO4 is needed for a BC particle to achieve
the same GF as the one covered by
(NH4)2SO4 / NH4NO3.
This result confirms that different atmospheric aging pathways lead to
changes in aerosol mixing state with distinct hygroscopicities: during
nighttime with low temperature and high RH, formation or condensation of
nitrates on BC particles enhanced the hygroscopicity of BC particles and
resulted in the thinner coating of BC particles in the hygroscopic mode for each
selected size. During daytime, condensation of photochemically generated SOC
on BC particles was associated with the thicker coating of BC particles with less
enhancement of hygroscopicity. The sulfate coating can enhance hygroscopicity
of BC particles. However, unlike nitrate and SOC, its formation did not show
a significant difference between day and night. This finding is consistent
with our previous measurement of secondary species formation in Shanghai
during the summertime. Shanghai is a typical Chinese megacity with heavy air
pollution. In summer, high NOx emission and ozone
concentration lead to enhanced nitrate formation via the N2O5 pathway
during nighttime (Wang et al., 2009; Pathak et al., 2009). Secondary organic
aerosol and particulate sulfate are usually formed through photo-oxidation of
organic vapor and SO2 during daytime (Kroll and Seinfeld, 2008;
Wang et al., 2016). The differences in aging pathways between daytime and
nighttime result in different coating materials and thickness on rBC cores,
which can further impact their hygroscopicity. Noticeably, Liu et al.
report that the hygroscopicity of BC particles was largely driven by the
coating of ammonium nitrate (Liu et al., 2013). However, this may not be the
case in Shanghai during summertime, as most particulate ammonium in
Shanghai has been found to be in the form of ammonium sulfate (Wang et al.,
2009; Pathak et al., 2009). Indeed, our SPAMS spectra for BC particles do not
show the presence of large ammonium peaks (Fig. S2 in the Supplement)
because most ammonium was in the form of ammonium sulfate, which is
difficult to be ionized and detected in SPAMS (Wang et al., 2009).
Figure 10a–c compare the coating thickness of 360 nm BC particles
(GF = 1.4) with other chemical indicators, such as SOC, nitrate, and
sulfate. It shows that there was a positive correlation between the coating
thickness and the SOC concentration and a negative correlation between the
coating thickness and the nitrate concentration, which is consistent with
their diurnal trends. There was a positive correlation between the coating
thickness and sulfate concentration. Figure 5b shows the coating thickness
for 360 nm BC particles whose GF = 1.4 peaked in the afternoon.
Formation of sulfate was also slightly enhanced during the afternoon due to
stronger solar irradiation, resulting in this correlation.
Figure 10d–f show the number fraction of BC particles in 360 nm aerosol with a GF of 1.4 versus the concentrations of SOC, nitrate, and sulfate. It is
found that the SOC had a negative correlation with the number fraction of BC
particles in 360 nm aerosol with a GF of 1.4, while nitrate and sulfate
had a positive correlation with the number fraction. This finding is
consistent with the fact that SOC has a lower GF than nitrate and sulfate
salts. When more SOC is formed and more likely covers BC particles' surfaces,
fewer BC particles' GFs can reach 1.4. In contrast, when more nitrate or
sulfate is formed and condenses on BC particles, more BC particles' GFs can
reach 1.4.
Conclusions
In this study, a HTDMA–SP2 system along with SPAMS were used to measure BC
particles' hygroscopic properties in Shanghai during the summer of 2017. Three
hygroscopic modes, namely the hydrophobic mode, transition mode, and
hydrophilic mode with GFs at 1.0, 1.2, and 1.4, respectively, were selected
to study the diurnal variations in rBC core and coating thickness as a
function of time.
Our results reveal that the hygroscopicity of BC particles is determined by
the coating layer thickness and materials, both of which are affected by
atmospheric aging processes. For a specific BC particle size, a thin coating
layer corresponded to freshly emitted BC particles with low hygroscopicity
(e.g., GF = 1.0). When BC particles became more hygroscopic (i.e., GF
increases), the coating thickness increased. A high yield of particulate
nitrate during nighttime was observed, and the nitrate coating greatly enhanced
the hygroscopicity of BC particles. During daytime, strong SOC formation from
photochemical oxidation played an important role in the evolution of the BC
mixing state. A thinner layer of nitrate coating could convert fresh BC
particles to aged hygroscopic ones while a thicker coating layer of SOC and
sulfate was required to reach the same overall hygroscopicity.
This study shows that atmospheric aging processes in a polluted city area
play critical roles in the fast change of aerosol mixing state during
summertime. Time-resolved information on particle hygroscopicity is
necessary to evaluate the aging process, wet removal, and climate
effects of BC aerosols.