ARTICLE IN PRESS

Atmospheric Environment 39 (2005) 69096918

www.elsevier.com/locate/atmosenv

Single particle characterization of spring and summer aerosols

in Beijing: Formation of composite sulfate of calcium
and potassium
Xiande Liua, Jia Zhub, P. Van Espenb, F. Adamsb,, Rui Xiaoa,
Shuping Dongc, Yuwu Lic
a
Chinese Research Academy of Environmental Sciences, Beijing 100012, China
b
University of Antwerp, Department of Chemistry, B-2610 Antwerp, Belgium
c
National Research Center for Environmental Analysis and Measurements, Beijing 100029, China
Received 16 December 2004; received in revised form 26 July 2005; accepted 6 August 2005

Abstract

Scanning electron microscopyenergy dispersive X-ray analysis (SEMEDX) was used for the analysis of 2500 single
particles in ve atmospheric aerosol samples collected during the spring and summer of 2000 in Beijing, China. Mineral
dust appeared to be the dominant particles during an Asian dust episode, while in other circumstances mineral dust and S-
containing particles constituted the major particle components. During anthropogenic pollution episodes in the summer, a
large abundance of S-containing particles featured the atmospheric aerosol. Chemical and size distribution characteristics
are discussed for CaS, KS and CaKS particle classes. Formation of CaKS and other S-containing particle classes
with high abundance was closely related to meteorological conditions such as relative humidity and cloud coverage. Simple
and composite sulfate particles with an elongated crystalline morphology were detected which appear to be indicative of
aqueous phase oxidation, such as in-cloud processing for sulfate formation pathway.
r 2005 Elsevier Ltd. All rights reserved.

Keywords: Atmospheric aerosol; Single-particle analysis; Scanning electron microscopyenergy dispersive X-ray analysis; Sulfate
formation; In-cloud processing; Beijing

1. Introduction recent years SEMEDX has been used to study

The application of single particle analysis with
methods such as scanning electron microscopye-
nergy dispersive X-ray analysis (SEMEDX) pro-
vides complementary and more detailed informa-
tion than that available through bulk analysis. In
Corresponding author. Tel.: +32 3 820 2010;
fax: +32 3 820 2376.
E-mail address: freddy.adams@ua.ac.be (F. Adams).
aerosol particles collected over the North Sea (Van
Malderen et al., 1996a; Hoornaert et al., 1996), the
Atlantic Ocean (Posfai et al., 1995; Anderson et al.,
1996), the Antarctic (Artaxo et al., 1992), and
Siberia (Van Malderen et al., 1996b). Urban areas
around the world were also investigated by such
single particle analysis methods as e.g., in Phoenix,
Arizona, USA (Katrinak et al., 1995), Antwerp,
Belgium (Van Borm and Adams, 1989), Seoul and
other cities in Korea (Ro et al., 2002). The study of

1352-2310/$ - see front matter r 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.atmosenv.2005.08.007
 ARTICLE IN PRESS
6910 X. Liu et al. / Atmospheric Environment 39 (2005) 69096918
urban aerosols is important due to their complexity;
many anthropogenic sources are involved. Their
study is also important for health effect studies
tracing the effects of atmospheric pollution. In
general, there appears to be a similarity in aerosol
composition and size distribution for urban areas
around the world but specic compositional char-
acteristics occur according to population size, energy
consumption pattern, industrial structure, geo-
graphic location and topography/economic situation.
The Beijing site is featured by its rapidly
expanding population, increasing trafc density, a
high consumption of coal and ourishing construc-
tion activities. Beijing is situated in a semi-arid
region in North China and surrounded by moun-
tains in the west and north. The weather conditions
usually do not favor dispersion and transport of air
pollutants. At present, the concentration of inhal-
able particles exceeds the national air quality
standards during extended periods of time each
year. The municipal government recently imple-
mented pollution control measures, policies and
various action plans. An in-depth understanding of
air pollution and aerosol chemistry in Beijing is
urgently needed, justifying intensive aerosol re-
search in the area. The urban aerosol in Beijing
has been the topic of several recent studies (He et
al., 2001; Shi et al., 2003; Yao et al., 2002, 2003).
In their study Yao et al. (2002) focused on the
ionic composition of PM2.5 particles. The major
route of sulfate formation in Beijing was identied
as gas-phase oxidation of SO2 in winter and in-
cloud processing in summer, but insufcient ammo-
nium was present to totally neutralize the aerosol.
Based on the size distribution of ionic species and
the mole ratio of sulfate to SO2, more evidence on
the formation mechanisms of secondary aerosol in
Beijing was provided by Yao et al. (2003). During
the summer, sulfate was mainly in the ne particles
with a mass median aerodynamic diameter (MAD)
of 0.770.1 mm. Sulfate formation was attributed to
in-cloud processing. Sulfate formation in the spring
was attributed to non-cloud heterogeneous pro-
cesses, as sulfate had a MAD of 0.4570.05 mm (Yao
et al., 2003).
Shi et al. (2003) examined a variety of particles
collected in Beijing in 2001 by using SEM and image
analysis. These included mineral dust, coal y ash,
soot aggregates as well as sulfates. Some sulfate
particles were crystalline and contained S, Ca, K
and Al and there were indications that two or more
phases of sulfates were present.
In this study, aerosol samples collected in the spring
and summer in Beijing were analyzed with SE-
MEDX. The aerosol composition, pollution char-
acteristics and the formation mechanism of secondary
sulfate aerosol were investigated. For the spring
samples, the difference between Asian dust and normal
urban aerosols was investigated. For the summer
samples, the number variation of major particle classes
and their relation with aerosol pollution situation and
meteorological conditions was observed. The focus of
the study was largely put on the formation of sulfate
particle classes, which showed close relationship with
the urban air quality in the summer.
2. Experimental
2.1. Aerosol samples
The sampling site is located in an urban area,
under the inuence of the trafc, of the north part
of the urban fourth ring of Beijing. PM10 aerosol
samples were collected for 24 h on the roof of the 11
oor main building of the Sino-Japan Friendship
Center for Environmental Protection on 25 April ,
15 May (spring samples) and on 21, 24 and 25 July
(summer samples), 2000. PM10 concentration was
386, 70, 153, 70 and 124 mg m 3, respectively. An
Asian dust event occurred on 25 April, while air
quality was reasonably good on 15 May and 24
July, and showed light pollution on 21 and 25 July,
all this according to National Standard of Ambient
Air Quality and the Air Quality Index System in
China. Nuclepore (polycarbonate) membrane with
0.4 mm pores was used as aerosol sampling lter as it
provides a at surface for SEM observation. The
particulate aerosol sampler with ow rate of
16.7 l min 1 was the stacked lter unit (SFU) of
Gent type recommended by IAEA for a number of
projects (Maenhaut et al., 1994), but only the PM10
size particles (with aerodynamic size less than or
equal to 10 mm) were collected. In this work,
particles with sizes from 2.5 to 10 mm are dened
as coarse particles, those less than 2.5 mm are
dened as ne particles, and those less than 1 mm
are listed as sub-micrometer.
2.2. SEM EDX measurements
Sections of the 47-mm diameter lter sample were
mounted on electron microprobe stubs and vacuum
coated with a carbon layer of about 40 nm. The
individual particle analyses were carried out with a
 ARTICLE IN PRESS
X. Liu et al. / Atmospheric Environment 39 (2005) 69096918
JEOL JXA-733 superprobe (Tokyo, Japan)
equipped with an annular backscattered electron
detector and a TN-2000 EDX detection system
(Tracor Northern, Middleton, USA). The X-ray
detector consisted of a conventional Si(Li) with Be
window that was able to detect elements from Na
upwards in atomic number. An electron accelerating
voltage of 20 kV was used at a beam current of
1.0 nA with magnication of 2000. Automated
analyses were done using the particle recognition
and characterization software package 733B devel-
oped at the University of Antwerp in which a
particle is recognized when its backscattered elec-
tron intensity is larger than a threshold, and its
diameter larger than a preset minimum size of
0.13 mm. For each recognized particle, the diameter,
area and shape factor were computed and an X-ray
spectrum was accumulated for 30 s. A simple, but
fast on-line spectral analysis method was used to
detect the characteristic peaks of the various
elements and to determine their intensity. The
elemental intensities and the size and shape data
of each particle were stored for further off-line
analysis. More details on the automated particle
analysis procedures and software package can be
found elsewhere (Raeymaekers et al., 1988; Van
Borm and Adams, 1989). A preset number of at
least 500 particles were analyzed for each sample.
2.3. Data analysis
Cluster analysis was used for the reduction and
interpretation of the data sets obtained by the
automated SEMEDX measurements. As in previous
studies (Shattuck et al., 1991; Hoornaert et al., 1996),
the particles were classied into particle groups or
particle classes on the basis of their normalized
characteristic X-ray intensities to the sum of all X-ray
intensities expressed as %. Cluster analysis was
performed with the integrated data analysis system
(IDAS) software package (Bondarenko et al., 1996).
Hierarchical cluster analysis was applied rst, after
which the results thus obtained were used as the
initial seed points for non-hierarchical cluster
analysis. The selection of the most appropriate
number of clusters was based on the Akaike
information criterion (Bondarenko et al., 1994).
Cluster analyses were rst applied to the data of
each sample, subsequently the results for all
combined samples were subjected to a second
cluster analysis in which the average normalized
X-ray intensities of the particle classes were used as
6911
input data. A two-dimensional table was made to
sort out different particle classes over the 2500
particles (see Table 1). Many particles in a class
showed similar composition. Excel software was
used for a regression analysis to detect the scatter of
the intensities for the elements.
A program, Quanta2 is adapted for ZAF correc-
tion using rst-order corrections for backscattering,
absorption and secondary uorescence effects,
converting X-ray intensity data of particles into
weight or atomic percentage for elements detected
(Van Borm and Adams, 1991). The method is
applicable to K- and L-lines for the elements Na to
Bi, as the characteristic lines of elements lighter than
Na are absorbed by the beryllium window of the Si
(Li) detector used. As for sulfur-containing parti-
cles, elements detected such as Na, Mg, K, Ca and S
are proposed with stoichiometry as oxide; however,
ammonium ions are not detectable. The method is
implemented in a FORTRAN computer program
written for PC. It is capable of processing the
massive amount of data produced by automated
analysis of large particle collections.
2.4. Position tagged spectra (PTS) measurements
Elemental mapping was carried out at 20 keV and
1.0 nA current with a JEOL JSM-6300 scanning
microscope equipped with a PGT X-ray analysis
system PGT EDX detector and IMIX-PC software
(Princeton Gamma-Tech, USA). The magnication
was 2500 and the total acquisition time was 5 h and
the image size 128  102 pixels. Scanning was
performed with a dwell-time of 1.38 s per pixel
and a step size of 0.47 mm. The elements Al, Si, S, K,
Ca and Fe were selected for mapping on the basis of
the data obtained in single particle analysis.
The results of the X-ray mapping are done via
PTS, resulting in a le containing beam coordinates
(x, y) and X-ray energy (channel number) for each
X-ray photon detected. The data can be displayed
as a spatial map of selected X-ray energies
(elemental maps) (Mott and Friel, 1999). X-ray
spectra of marked areas corresponding to minimum
nine pixels could also be generated.
3. Results and discussion
3.1. Particle classes and their abundance
Particles were classied into particle classes on the
basis of their normalized characteristic X-ray
 ARTICLE IN PRESS
6912 X. Liu et al. / Atmospheric Environment 39 (2005) 69096918

Table 1
Number abundance percentage and size of particle classes in aerosol samples

Particle class Number abundance percentage (%)

25 April 15 May 21 July 24 July 25 July

Minerals
Aluminosilicates 76.7 28.3 12.4 43.6 18.0
(1.3271.45) (0.7971.29) (1.0971.68) (1.2271.58) (1.2571.43)
Si-rich 14.6 0.2 3.8 9.9 3.4
(1.4571.31) (1.4570.31) (0.7571.14) (1.0571.75) (1.9071.64)
Fe-rich 2.0 4.6 2.0 7.4 19.3
(0.8470.48) (0.4170.29) (0.3670.29) (0.4870.67) (0.3970.34)
Ca-rich 5.2 15.4 5.5
(1.5971.88) (1.0971.44) (1.2871.64)

S-containing
S-rich 6.1 3.2
(0.2070.16) (0.2570.35)
NaS 0.9
(1.2471.52)
MgS 4.7
(0.6470.95)
KS 10.8 0.3 0.9 3.4
(0.2070.12) (0.1870.08) (0.5570.98)
CaS 24.3 37.3 5.6 6.8
(0.3270.32) (0.3470.42) (0.3470.28) (0.3670.41)
CaKS 6.1 27.9 41.7
(0.1670.0.04) (0.5170.64) (0.3270.23)
ZnS 2.2
(0.4070.30)
Other S-containing 2.4 2.8

Minor class
Mn-containing 1.3 0.9 2.3
(0.2070.05) (0.4170.21) (0.2670.12)
P-containing 6.4
(0.5070.97)
Zn-containing 1.8
(0.3570.23)
Pb-containing 2.0
(0.4070.30)
Carbonaceous 3.3 0.5 11.0
Others 0.6 3.9 0.5 4.0 1.8

The mean geometric diameter and standard deviation data are given in the parenthesis for a particle class.

intensities using the clustering software IDAS.
Clustering results are listed in Table 1. Particle
classes were sorted into four categories, namely,
mineral dust, sulfur containing, minors and carbo-
naceous.
Mineral dusts are of nature origin and consist of
mainly aluminosilicates and quartz (Si-rich) parti-
cles. It accounted for 97.8% of the particles in the
sample of 25 April 2000, a sample of Asian Dust
event, while the other samples contain a much lower
percentage of mineral dust, but considerable num-
bers of S-containing particles. Iron-rich particles
could originate from iron and steel production and
steel product erosion, but coal combustion can also
not be excluded as a source. As shown in Table 1,
high abundances occur for the 24 and 25 July
samples, when the prevailing wind direction was
southwest and south-southwest. Hence, the source
might be related to an iron and steel manufactory
complex situated in the southwest of the sampling
site. Calcium-rich particles can be attributed to
wind-blown soil and construction dust.
 ARTICLE IN PRESS
X. Liu et al. / Atmospheric Environment 39 (2005) 69096918 6913

Different kinds of sulfur-containing particles can July 21, 2000

be distinguished on the basis of the elements 0.3
S
detected. On the other hand, sulfur-rich particles 0.25 CA
occur with only sulfur detected. In the Asian dust
K
sample of 25 April, sulfur-containing particles were

Atomic percent
0.2
nearly absent, while in the 15 May, 21 and 25 July
samples sulfur-containing particles accounted for 0.15
50% or more of all the particles detected. Appar-
0.1
ently, such particles are an important component of
the urban aerosol in Beijing. Among the three 0.05
samples of July 2000, they appeared with low
abundance on 24 July, but abundance was high on 0
25 July, and even higher on 21 July. There was an 0.1 1 10
increasing trend of sulfur-containing particles when Particle diameter in micron
air pollution was high. There is a positive correla- July 25, 2000
tion between the abundance of sulfur-containing 0.3 0.6
particles and the PM10 concentration in July. S
0.25 CA 0.5
Carbonaceous particles are those detected, which
do not provide any characteristic X-rays. The K
Atomic percent 0.2 0.4
diameter of nearly all those particles is less than
0.3 mm. Such particles contain only low atomic 0.15 0.3
number elements and are carbonaceous in composi-
tion. They are an abundant component of urban 0.1 0.2
aerosols. SEMEDX is not optimal for the detec-
tion of these particles; hence the data for this 0.05 0.1
particle class in Table 1 are very much under-
0 0
estimated. 0.1 1 10
The mean geometric diameter and standard Particle diameter in micron
deviation data are also presented for particle classes
Fig. 1. Scatter plot of atomic percentage of Ca, K, and S over
in Table 1. Mineral dust particles fall in a size range
particle size for individual CaKS particles in aerosol samples of
larger than that of sulfur-containing particles. As 21 July and 25 July 2000. (Note: K data in lower plot to the right
an example, compositional data were illustrated axis.)
against particle size for CaKS particles in Fig. 1.

3.2. Composition of sulfate particles
Sulfates, as sulfur-containing particles, are pre-
sent with high abundance (see Table 1) and are
closely related to anthropogenic emission sources.
Sulfate particles are mainly secondary in nature,
formed by a variety of atmospheric reactions.
Ammonia often accompanies the sulfate formed as
the major neutralizing cation (Yao et al., 2002,
2003). Other sulfates such as calcium and magne-
sium salts are frequently found in Chinese aerosols,
as alkaline dusts are abundant. Potassium sulfate
particles were reported in high numbers in previous
studies (Liu et al., 2002). The atomic percentage of
sulfur and other major elements are listed in Table 2
for sulfate particles, but ammonium ions were not
detectable due to the limitation of the Si (Li)
detector. MgSO4 and CaSO4 have the same
theoretical atomic percentage values, 0.167 for Mg
or Ca, 0.167 for S and 0.667 for O, respectively;
however, when 1:1 mixed with ammonium sulfate,
the atomic percent change to 0.091 for Mg or Ca,
0.182 for S and 0.727 for O, respectively. It means
that mixing with ammonium sulfate leads to lower
percentage for metallic cations, while higher values
for S and O. As shown in Table 2, CaS particles of
15 May 2000 had a composition close to theoretical
CaSO4 values, while those of three aerosol samples
in July 2000 seemed to mix with ammonium sulfate
to some extent. Composition of MgS particles of
21 July 2000 was similar with MgSO4
(NH4)2SO4
rather than MgSO4. CaSO4 was frequently detected
in atmospheric aerosols in many urban areas
around the world (Van Borm and Adams, 1989;
Hoornaert et al., 1996; Liu et al., 1994). CaCO3
particles originated as soil or road dust could react
 ARTICLE IN PRESS
6914 X. Liu et al. / Atmospheric Environment 39 (2005) 69096918

Table 2
The atomic percentage of S-containing particle classes in aerosol samplesa

15 May 21 July 24 July 25 July

Mean7SD N Mean7SD N Mean7SD N Mean7SD N

MgS MgSO4, MgSO4
(NH4)2SO4b

S 0.20370.006 33
Mg 0.07270.021 33
O 0.70470.007 33
KS K2SO4, K2SO4
(NH4)2SO4, K2SO4
2((NH4)2SO4)b
S 0.20370.014 49 0.16870.008 4 0.18270.009 4
K 0.12470.037 49 0.21870.022 4 0.17970.025 4
O 0.67270.020 49 0.61470.012 4 0.63870.013 4
CaS CaSO4, CaSO4
(NH4)2SO4b
S 0.16770.013 77 0.18470.015 358 0.17170.022 4 0.16970.010 42
Ca 0.16570.028 77 0.13070.026 358 0.15870.046 4 0.15270.012 42
O 0.66770.013 77 0.68570.014 358 0.67170.02 4 0.67070.013 42

CaKS CaSO4
K2SO4, CaSO4
K(NH4)SO4b

S 0.14970.021 3 0.18470.011 363 0.17670.011 239
K 0.13570.008 3 0.02570.011 363 0.07570.017 239
Ca 0.10170.007 3 0.10770.017 363 0.08770.016 239
O 0.61570.006 3 0.67970.014 363 0.65970.013 239
a
The atomic percent mean value and standard deviation for each element are presented. N is the number of particles in the particle class.
b
Some possible speciation forms for sulfates.

with SO2 ,or sulfuric acid, or acidic sulfates such as
ammonium sulfate to generate CaSO4 (Hoornaert et
al., 1996). Some CaSO4 particles are primary
particles such as those from desulfurization process
in coal combustion power stations or from soiling of
calcite in building construction materials.
K2SO4 has the theoretical atomic percentage
values, 0.286 for K, 0.143 for S and 0.571 for O,
respectively. However, when K2SO4 (1:1) mixed
with ammonium sulfate, it is 0.167 for K, 0.167 for
S and 0.667 for O, respectively; it seems the case of
KS particles of 25 July 2000. The composition of
the KS particles of 24 July 2000 falls in between the
two cases proposed above. It is possible that K2SO4
was mixed with less (NH4)2SO4 than 1:1 ratio.
Further more, if K2SO4 was 1:2 mixed with
(NH4)2SO4 , it will be 0.118 for K, 0.176 for S and
0.706 for O. This set of values describes better the
composition of KS particles of 15 May 2000. KS
particles were reported as major particle class in
biomass burning aerosols (Liu et al., 2000). It could
be also be possible that they are converted from
K2CO3.
CaSO4 and K2SO4 mixed at 1:1 ratio would lead
to the theoretical atomic percentage values, 0.077
for Ca, 0.154 for K and S, and 0.615 for O,
respectively. It seems comparable with that of
CaKS particles on 15 May 2000. When CaSO4
and K(NH4)SO4 mixing was proposed, it would be
0.084 for Ca and K, 0.167 for S and 0.667 for O,
respectively. It ts well with that of CaKS
particles on 25 July 2000. The composition of
CaKS particles on 21 July 2000 is with low
percents of Ca and K, and high values of S and O,
indicating possible involvement of ammonium
sulfate. In summary, a variety of sulfates were
found in this study based on compositional data of
individual particles. They could be simple sulfates or
composite sulfates as internal mixture of ammo-
nium sulfate and metallic sulfates.
3.3. Formation of composite sulfate particles of
calcium and potassium
CaKS particles are present in high numbers in
two of the summer samples. Fig. 1 shows data with
respect to chemical composition and particle size.
Particles with size below 0.3 mm account for nearly
60% number abundance for the 21 and 25 July
samples. They show, however, a different size
distribution, as particles with size larger than 1 mm
account for 15% in the 21 July sample, while they
are almost absent in the 25 July sample. The
chemical composition of CaKS class in those
 ARTICLE IN PRESS
X. Liu et al. / Atmospheric Environment 39 (2005) 69096918
two summer samples is also different (Table 2).
Sulfate particles are mainly converted from gaseous
SO2 in the atmosphere. According to Seinfeld and
Pandis (1996) aqueous phase oxidation is one of the
major mechanisms for sulfate formation. In-cloud
processes have been postulated which is important
for sulfate formation in certain circumstances
(McHenry and Dennis, 1994; Yao et al., 2002,
2003). CaKS particles, on the other hand, are
composite sulfates and since calcium and potassium
should be generated by different kinds of emission
sources, it is difcult to explain the formation of
CaKS particles by the same simple and straight-
forward atmospheric reactions as proposed for the
CaS and KS particles.
When comparing CaS, KS particles with
CaKS particles, the latter with more complex
composition need both calcium and potassium in
raw materials and an efcient mixing. There is a
negative correlation between the abundance of the
two types of sulfur-containing particles, competing
for common raw materials. As shown in Table 1,
CaKS particles were detected only in the three
samples obtained in typical urban pollution condi-
tions. CaS and KS particles were more abundant
(24% and 11%) than the CaKS particles (6%) in
the 15 May sample. Hence, it seems that the effect of
in-cloud formation was limited, which was ob-
viously related to the dry weather in spring. The
situation was different for the summer samples
when in-cloud processes exerted profound effects.
In the 21 July sample, CaS particle abundance was
37% whereas KS particles were almost absent and
CaKS particles were abundantly present. It
seemed that KS particles were exhausted, restrict-
ing further formation of CaKS particles. In the 25
July sample, both CaS and KS particle abun-
dances were at 3.4% and 6.8%, while the CaKS
particle abundance is up to 42%.
The potassium content of the CaKS particles in
the sample of 25 July is three times higher than on
21 July. The atomic percent of potassium is
2.571.1% on 21 July and 7.571.7% on 25 July
(see Table 2) and it reveals the compositional
dissimilarity between aerosol samples and the
similarity within an aerosol sample. It seems that
the content of potassium in CaKS particles
depends on the availability of potassium containing
particles in air, a high abundance of those particles
leading to CaKS particles with higher potassium
content. The mass transfer and mixing in the in-
cloud processes may lead to homogenization of the
6915
chemical composition of cloud droplets and thus the
particles generated afterwards from those droplets.
The occurrence of CaKS particles in high
numbers was not reported in previous studies. It is
difcult for bulk analytical methods such as ion
chromatography to distinguish CaKS particles
from CaS and KS particles. Its formation is
related to the hot, humid, cloudy weather during
summer in Beijing facilitating the in-cloud forma-
tion processes. The average monthly precipitation in
Beijing is 21 and 34 mm in April and May, while it is
185 mm in July 2000. In Fig. 2, some meteorological
data are presented for July 2000. The total cloud
coverage was 85%, 32% and 65% and low cloud
coverage was 50%, 0% and 15%, respectively, for
21, 24 and 25 July 2000. There appears a positive
correlation between number abundance of sulfate
particles and the cloud coverage, indicating the
importance of the in-cloud process for SO2 oxida-
tion and sulfate aerosol formation.
CaKS particles have consistent composition
within an aerosol sample, but a different composi-
tion between samples (see Table 2 and Fig. 1). From
this, it is inferred that CaKS particles do not
correspond to one specic compound with constant
composition but are a composite of different
sulfates formed as internal mixture in in-cloud
processes.
3.4. X-ray mapping measurements
As a support for the results obtained by single
particle analysis, X-ray maps were collected. As an
example, Fig. 3 presents the secondary electron
image of one eld of the 21 July sample whereas
Fig. 4 shows X-ray maps for selected elements for
the same eld of view. In addition to the 0.4 mm
pores from the polycarbonate membrane, the gure
shows four mineral particles that (see Fig. 4) contain
Si, Al, Fe and K, typical elements for mineral dust.
Three of those particles have an irregular shape, one
spherical particle is typical for coal y ash. Also,
sulfur-containing particles are abundantly present.
Most of these also show up in the Ca map, while
some are also discernable in the K map. The
spectral data here match well with the major particle
classes summarized in Table 1.
Some particles with crystalline morphology were
observed in the summer samples. Three elongated
crystalline particles for the 21 July sample are
shown in Fig. 3 and in the mappings of S and Ca in
Fig. 4. These particles have a width of about 1 mm
 ARTICLE IN PRESS
6916 X. Liu et al. / Atmospheric Environment 39 (2005) 69096918

100 10
90 R.H.(%)
80 8
70 Total cloud (%)
60 6
Low cloud (%)
50
40 4
Temperature (C)
30
20 2 Wind speed (m/s)
10
0 0
19 20 21 22 23 24 25 26 27
Date in July 2000

300 100
250 80
200
60 PM10
150
40 Sulfate particles %
100
50 20
0 0
19 20 21 22 23 24 25 26 27
Date in July 2000

Fig. 2. Meteorological data during late July 2000 (upper plot) and aerosol mass concentration in mg m 3 and number abundance
percentage of sulfate particles for three aerosol samples in July 2000 (lower plot). (Note: Wind speed data in m s 1 in upper plot and
number abundance data of sulfate particles in lower plot to the right axis.)

Fig. 3. Secondary electron image of some aerosol particles collected on 21 July 2000 (magnication of 2500).

and a length from 5 to 10 mm, equivalent to an crystalline morphology implies a mechanism of

average diameter of 2.23.2 mm in the automated aqueous phase formation and supports the hypoth-
measurements. These elongated particles are similar esis of in-cloud processes as a major sulfate
to those illustrated by Shi et al. (2003). The formation route during summer in Beijing. As the
 ARTICLE IN PRESS
X. Liu et al. / Atmospheric Environment 39 (2005) 69096918
Fig. 4. X-ray maps for selected elements (Al, Si, Fe, S, Ca, and K) of the same eld as in Fig. 3 (sample 21 July 2000).
water evaporates, the crystal seeds are then formed
in cloud droplets and nally turn out as airborne
particles.
4. Conclusion
As microscopic analytical method, SEMEDX is
suitable for chemical characterization of mineral
dusts and sulfur-containing particles. Mineral dusts
are dominant in windy days in spring in Beijing, and
become overwhelming during Asian dust events. On
the other hand, mineral dust, as a basic component
of background aerosol in Northern China, is
relatively abundant in periods with improved air
quality. Sulfur-containing particles, mainly sulfates,
feature aerosol pollution in urban Beijing where
coal is the main energy source. The number
abundance of sulfur-containing particles is closely
correlated with the aerosol mass concentration
levels in summer. Although the SO2 emission is
lower in summer than in winter, the conversion of
SO2 into sulfates is fast and efcient in summer due
to the high temperature and humidity. Sulfate to
SO2 ratio as well as size-distribution pattern of
sulfates were used as experimental evidence for the
aqueous phase oxidation of sulfate formation path-
way postulated in previous studies. The occurrences
of a great number particles of composite sulfates of
calcium and potassium as well as the crystalline
morphology of sulfate particles were found as
supporting experimental phenomena in this work.
It is indicative of aqueous phase oxidation in
general and in-cloud processing in particular for
sulfate formation pathway. This explains how
calcium and potassium originated from different
6917
sources turn out as composite sulfates in Beijing
aerosol during the summer. Among possible aqu-
eous phase oxidation pathways, in-cloud processing
is most likely, as mass transfer is more active and
mixing is more efcient in clouds. Aqueous phase
oxidation facilitates not only the formation of
sulfates, but also the transformation of ammonium
sulfate into sulfate of calcium, or potassium, or
internal mixture of both. The observations of this
study justify the pollution control measures that
come into effect in Beijing, including mandatory use
of coal with low content of sulfur, promotion of
clean energy alternatives such as nature gas, the ban
of coal burning boilers in downtown Beijing and the
installation of desulfurication utilities for new
power stations. All these measures reduces SO2
emission and lead to less sulfate aerosol.
Acknowledgments
Financial support from FlemishChinese Bilat-
eral Cooperation Program (BIL01/57), the China
NSFC (20177036, 20477042) and The National
Basic Research Program of China, no. 2003CB415
003 is gratefully acknowledged.
References
Anderson, J.R., Buseck, P.R., Patterson, T.L., Arimoto, R.,
1996. Characterization of the Bermuda tropospheric aerosol
by combined individual particle and bulk-aerosol analysis.
Atmospheric Environment 30, 319338.
Artaxo, P., Rabello, M.L.C., Maenhaut, W., Van Grieken, R.,
1992. Trace elements and individual particle analysis of
atmospheric aerosols from the Antarctic peninsula. Tellus
44B, 318334.
 ARTICLE IN PRESS
6918 X. Liu et al. / Atmospheric Environment 39 (2005) 69096918
Bondarenko, I., Van Malderen, H., Treiger, B., Van Espen, P.,
Van Grieken, R., 1994. Hierarchical cluster analysis with
stopping rules built on Akaikes information criterion for
aerosol particle classication based on electron probe X-ray
microanalysis. Chemometrics and Intelligent Laboratory
Systems 22, 8795.
Bondarenko, I., Treiger, B., Van Grieken, R., et al., 1996. IDAS:
a windows based software package for cluster analysis.
Spectromchimica Acta B 51, 441456.
He, K.B., Yang, F.M., Ma, Y.L., Zhang, Q., Yao, X.H., Chan,
C.K., Cadle, T., Mulawa, P., 2001. The characterization of
PM2.5 in Beijing, China. Atmospheric Environment 35,
49594970.
Hoornaert, S., Van Malderen, H., Van Grieken, R., 1996.
Gypsum and other calcium-rich aerosol particles above the
North Sea. Environmental Science and Technology 30 (5),
15151520.
Katrinak, K.A., Anderson, J.R., Buseck, P.R., 1995. Individual
particle types in the aerosol of Phoenix, Arizona. Environ-
mental Science and Technology 29, 321329.
Liu, X.D., Jia, H., Qi, J.B., Zhang, J.Q., 1994. Characterization
and source identication of Qingdao aerosol with SEMEDX
method. Research of Environmental Sciences 7 (3), 1017
(Chinese with English abstract).
Liu, X.D., Van Espen, P., Adams, F., Cafmeyer, J., Maenhaut,
W., 2000. Biomass burning in southern Africa: individual
particle characterization of atmospheric aerosols and savanna
re sample. Journal of Atmospheric Chemistry 36, 135155.
Liu, X.D., Li, Y.W., Dong, S.P., Van Espen, P., Adams, F., 2002.
Quantitative analysis and elemental speciation of biomass
burning particles. Environmental Chemistry 21 (3), 209217
(Chinese with English abstract).
Maenhaut, W., Franc- ois, F., Cafmeyer, J., 1994. Regional and
global atmospheric aerosol studies using the Gent stacked
lter unit and other aerosol collectors, with multi-elemental
analysis of the samples by nuclear-related analytical techni-
ques. In: Applied Research on Air Pollution using Nuclear-
Related Analytical Techniques, Report NAHRES-19, IAEA,
Vienna, pp. 249263.
McHenry, J.N., Dennis, R.L., 1994. The relative importance of
oxidation pathways and clouds to atmospheric ambient
sulfate production as predicted by the regional acid deposi-
tion model. J. Applied Meteorology 33, 890905.
Mott, R.B., Friel, J.J., 1999. Saving the photons: mapping X-rays
by position-tagged spectrometry. Journal of Microscopy 193
(1), 214.
Posfai, M., Anderson, J.R., Buseck, P.R., Sievering, H., 1995.
Compositional variations of sea-salt-mode aerosol particles
from the North Atlantic. Journal of Geophysical Research
100, 2306323074.
Raeymaekers, B., Demuynck, M., Dorrine, W., Adams, F., 1988.
Characterization of oil y ash particles by electron probe
microanalysis. Internationl Journal of Environmental Analy-
tical Chemistry 32, 291312.
Ro, C.U., Kim, H., Oh, K.Y., Yea, S.K., Lee, C.B., Jang, M.,
Van Grieken, R., 2002. Single-particle characterization of
urban aerosol particles collected in three Korean cities using
low-Z electron probe X-ray microanalysis. Environmental
Science and Technology 36, 47704776.
Seinfeld, J.H., Pandis, S.N., 1996. Atmospheric Chemistry and
Physics: from Air Pollution to Climate Change. Wiley, New
York, pp. 812823.
Shattuck, T.W., Germani, M.S., Buseck, P.R., 1991. Multivariate
statistics for large data sets: applications to individual aerosol
particles. Analytical Chemistry 63, 26462656.
Shi, Z.B., Shao, L.Y., Jones, T.P., Whittaker, A.G., Lu, S.L.,
Berube, K.A., He, T.O., Richards, R.J., 2003. Characteriza-
tion of airborne individual particles collected in an urban
area, a satellite city and a clean air area in Beijing, 2001.
Atmospheric Environment 37, 40974108.
Van Borm, W.A., Adams, F.C., 1989. Characterization of
individual particles in the Antwerp aerosol. Atmospheric
Environment 23 (5), 11391151.
Van Borm, W.A., Adams, F.C., 1991. A standardless ZAF
correction for semi-quantitative electron probe microanalysis
of microscopical particles. X-ray Spectrometry 20, 5162.
Van Malderen, H., Van Grieken, R., Bufetov, N.V., Koutzenogii,
K.P., 1996a. Chemical characterization of individual aerosol
particles in central Siberia. Environmental Science and
Technology 30, 312321.
Van Malderen, H., Hoornaert, S., Van Grieken, R., 1996b.
Identication of individual aerosol particles containing Cr,
Pb, and Zn above the North Sea. Environmental Science and
Technology 30, 489498.
Yao, X.H., Chan, C.K., Fang, M., Cadle, S., Chan, T., Mulawa,
P., He, K.B., Ye, B.M., 2002. The water-soluble ionic
composition of PM2.5 in Shanghai and Beijing, China.
Atmospheric Environment 36, 42234234.
Yao, X.H., Lau, A.P.S., Fang, M., Chan, C.K., Hu, M., 2003.
Size distribution and formation of ionic species in atmo-
spheric particulate pollutants in Beijing, China: inorganic
ions. Atmospheric Environment 37, 29913000.