ARTICLE IN PRESS

Atmospheric Environment 40 (2006) 212–224

www.elsevier.com/locate/atmosenv

Development and evaluation of personal respirable particulate

sampler (PRPS)
Seung Joo Leea, Philip Demokritoub,, Petros Koutrakisb,
Juana M. Delgado-Saboritc
a
Occupational Health and Environmental Safety Division, APAC Laboratory, 3M Korea Yoido P.O. Box 93, Seoul, Korea
b
Exposure, Epidemiology and Risk Program, Department of Environmental Health, School of Public Health, Harvard University,
401 Park Drive, P.O. Box 15677, Landmark Center West, Room 404M, Boston, MA 02215, USA
c
Inorganic and Organic Department, Universitat Jaume I, Castellon 12071, Spain
Received 1 November 2004; received in revised form 16 August 2005; accepted 16 August 2005

Abstract

This paper presents the development, laboratory evaluation, and field tests of a personal respirable particulate sampler
(PRPS). The PRPS is designed as a personal sampling system to collect particulate matter (PM0.5, PM1.0, PM2.5, PM4.5,
and PM10) and gaseous pollutants, including O3, SO2, and NO2. It operates at a flow rate of 5.0 LPM and consists of five
selectable impaction stages (with cutpoints of 10, 4.5, 2.5, 1.0, and 0.5 mm), a backup filter, and two diffusion passive
samplers. In each impaction stage, particles are collected onto a polyurethane foam (PUF) substrate. This substrate, using
no adhesive, was shown to have minimum particle bounce and re-entrainment. A backup 37 mm Teflon membrane filter is
used downstream to collect particles smaller than the cutoff diameter of the final impaction stage. The impaction stage
cutpoints were characterized in the laboratory using artificially generated polydisperse aerosols. Particle losses for each
stage were found to be acceptably low. The performance of the PRPS was also compared with that of a collocated micro-
orifice cascade impactor (MOI) and real-time particle sizing instruments (SMPS/APS) in laboratory experiments using
artificially generated particles. The size distributions measured by the PRPS were found to be much closer to those
measured by the real-time particle sizing instruments than to those measured by the MOI. A field PM intercomparison
study was also conducted using the PRPS and three reference samplers, the Harvard Impactor (HI), the USEPA PM2.5
Well Impactor Ninety Six (WINS), and the Harvard Personal Environmental Monitor (Harvard PEM) sampler. The
PM10, PM2.5, and sulfate concentrations measured by PRPS were in a very good agreement with those obtained from the
reference samplers.
r 2005 Elsevier Ltd. All rights reserved.

Keywords: Personal sampling system; Particulate matter; Impaction; PUF

1. Introduction

There are numerous epidemiological studies

Corresponding author. Tel.: +617 384 8847. showing associations between increased airborne
E-mail address: [email protected] particulate matter (PM) concentrations and adverse
(P. Demokritou). health effects, which include the respiratory and

1352-2310/$ - see front matter r 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.atmosenv.2005.08.041
 ARTICLE IN PRESS
S.J. Lee et al. / Atmospheric Environment 40 (2006) 212–224 213

cardiovascular diseases (Castillejos et al., 2000; Cheng, 1995). Recently, polyurethane foam (PUF)
Dockery and Pope, 1994; Dockery et al., 1993; has been used and evaluated as an impaction
Hoek et al., 2001; Moolgavkar, 2000; Pope et al., substrate to avoid the use of adhesives and to
1995; Schwartz and Dockery, 1992; Schwartz and minimize particle bounce/re-entrainment. For the
Morris, 1995). Total suspended particulates (TSP), same sampling flow, for the PUF impaction
particles smaller than 10 mm (PM10), and particles substrate, the cutoff diameter is much lower than
smaller than 2.5 mm (PM2.5) have all been correlated for the conventionally used flat, rigid substrates
with health effects in both epidemiological and (Demokritou et al., 2002a, 2004a, b; Kavouras and
toxicological studies. Koutrakis, 2001). Since the pressure drop is
The relation between outdoor ambient particulate determined by the geometry of impaction nozzle
concentrations and personal exposure to PM is and the flow rate, a lower pressure drop can be
important in order to evaluate human health effects. achieved for PUF. It was shown that PUF can also
However, these issues have been difficult to address collect large quantities of particles without changing
since personal exposure is affected by not only the particle collection characteristics, and the collected
outdoor particle concentration but also by human particles can be easily extracted for toxicological,
time–activity patterns and indoor sources. It has biological, and chemical characterization studies
been shown in many studies that outdoor PM (Salonen et al., 2000).
concentration is a poor indicator of personal In this paper, we present the development,
exposure because concentrations in the breathing laboratory characterization, and field evaluation of
zone can be different for different microenviron- a personal respirable particulate sampler (PRPS)
ments (Chang et al., 1999; Clayton et al., 1993; that utilizes PUF as the particle collection medium.
Lioy et al., 1990; Sexton et al., 1984). Therefore, It is suitably compact and lightweight to mini-
the most accurate way to determine personal mally interfere with the normal activities of the
exposure is to use personal samplers for monitoring individual subject. This sampler includes five
directly pollutants in the breathing zone of the independently selectable impaction stages (10, 4.5,
individuals. 2.5, 1.0, and 0.5 mm, respectively). In addition,
Many forms of size-selective sampling devices, two passive diffusion samplers can be used for
i.e., conventional inertial impactors, impingers, collection of gaseous copollutants such as O3, NO2,
electrostatic precipitator, and cyclones, were devel- and SO2.
oped and evaluated to measure the personal
exposure of PM. Among those sampling devices, 2. Methods
the impactors are widely used to classify the
particles into different particle size ranges (Marple, 2.1. Impactor design considerations
1970). Several personal monitoring systems have
been developed to classify particles from 0.2 to According to conventional impaction theory (for
20 mm (Demokritou et al., 2001, 2002a, b; Misra flat, rigid substrates), the dimensionless Stokes
et al., 2002; Rubow et al., 1987). number, Stk, is the principal factor for impaction
A major problem associated with the application and is given by
of impactors to classify and collect particles is
bounce-off and re-entrainment of particles from the rp d 2p UC c
impaction substrates. In order to minimize particle Stk ¼ , (1)
9mW
bounce, impaction substrates are usually coated
with adhesive agents such as grease or mineral oil where m is the air dynamic viscosity, W nozzle
(Huang and Tsai, 2003; John et al., 1991; Sehmel, width, rp particle density, dp particle diameter, U air
1980; Wall et al., 1990). However, these materials velocity in the impactor nozzle, and Cc Cunningham
may induce chemical interference during analysis of slip correction factor. The slip correction factor is
the collected particles. Also, particles collected on specified by (Hinds, 1999)
adhesive-coated impaction substrates may not be
suitable for physiological/toxicological studies. dp
Conventional impactors have also shown relatively Cc ¼ 1 þ ð2:34 þ 1:05e
0:390l=d p Þ, (2)
l
low collection capacities due to their flat/rigid
impaction surfaces (Lee et al., 2005; Tsai and where l is the mean air molecule free path.
 ARTICLE IN PRESS
214 S.J. Lee et al. / Atmospheric Environment 40 (2006) 212–224

The dimensionless Reynolds number (Re) was application. The PUF substrates are easily inserted
calculated using Eq. (3) for the rectangular impac- and detached from the substrate holder, and are
tion nozzle: suitable for gravimetric and chemical analysis.
Dimensions for the five different nozzles and
2Qrg
Re ¼ , (3) impaction substrates are shown in Table 1. The
ZðL þ W Þ thickness of the PUF for all stages is 0.64 cm, which
where Q is the flow rate, rg air density, L length of is adequate to insure that accelerated particles do
the nozzle, and W width of the nozzle. It is not penetrate through the PUF onto the substrate
recommended that the Re should be below 3000 to holder.
prevent highly turbulent flow and increased particle Particles smaller than the cutoff size of the final
losses (Marple et al., 1993). The same authors also impaction stage are collected on a Teflon membrane
showed that the efficiency curve is sharpest for Re backup filter, held in a suitable holder (37 mm
values between 500 and 3000. diameter, 2 mm pore size, No. R2PJ037, Pall Life
Sciences, Ann Arbor, MI). Other filter materials
such as polypropylene fiber may be used as a
2.2. Description of the PRPS backup filter.
Two passive samplers are attached on the side of
The assembled PRPS and its internal parts are an elutriator for the measurement of gaseous
shown in Fig. 1 (with only the stages for 10, 2.5, and copollutants, much the same as in an earlier
1.0 mm cutpoints). The sampling apparatus consists personal sampling system developed in our research
of three basic components: (1) a set of impactor group (Demokritou et al., 2001). As the sample air
stages, (2) a backup filter, and (3) a set of two passes through the elutriator, the pollutant gases
passive samplers. The PRPS is compact and is pass through diffusion-controlling inlets and are
relatively lightweight (0.3 kg). Each impaction trapped on coated filters. The first passive sampler
stage consists of a slit-shaped acceleration nozzle uses the method previously developed and validated
and a circular PUF impaction substrate. Individual in our laboratory (Koutrakis et al., 1993), which is a
impaction stages can be added or removed, depend- nitrite-coated glass fiber filter to collect ozone. The
ing on the size ranges needed for a particular second passive sampler measures gaseous sulfur
dioxide and nitrogen dioxide (Ogawa & Company
USA Inc., FL). This passive sampler uses a
triethanolamine-coated cellulose filter.

2.3. Laboratory characterization of PRPS impaction

stages

The experimental apparatus used for the labora-

tory performance of the system is shown schemati-
cally in Fig. 2. The equipment consists of two major
components: (1) the polydisperse particle generation
system, and (2) the particle sampling and monitor-
ing system. Due to the wide range of PRPS
cutpoints, two different artificial particle generators
were employed to generate and measure the poly-
disperse particles for the calibration of the various
stages.
For the 0.01–1.0 mm (aerodynamic) size range, an
aqueous solution of sodium chloride (NaCl) was
atomized using a Collision type atomizer (Model
3076, TSI Inc., MN). All the generated particles
were mixed in the dilution chamber (volume of
1.15  10
2 m3) with dry, particle-free room air.
Fig. 1. The personal respirable particulate sampler (PRPS). After dilution, the mixed stream was passed though
 ARTICLE IN PRESS
S.J. Lee et al. / Atmospheric Environment 40 (2006) 212–224 215

Table 1
Physical characteristics, flow parameters, and experimental results of PRPS

Stage no.

1 2 3 4 5

Physical characteristics
Acceleration nozzle
W (cm) 0.432 0.196 0.178 0.102 0.102
L (cm) 1.219 0.965 0.787 0.457 0.380
S/W 1.0 1.5 1.6 1.6 1.6
Substrate
PUF diameter (cm) 1.91 1.91 0.95 0.95 0.95
Material PUF PUF PUF PUF PUF
Flow parameters
Re 910 1155 1409 2427 2912
U (cm s–1) 158 441 595 1787 2150
Experimental results
d50 (mm) 10.0 4.55 2.65 1.05 0.58
pffiffiffiffiffiffiffi
Stk 0.48 0.54 0.39 0.37 0.22
s 1.34 1.38 1.40 1.44 1.69
DP (kPa) 0.01 0.02 0.03 0.27 0.34

Note:
pffiffiffiffiffiffiffi W, nozzle width; L, nozzle length; S, substrate-to-nozzle distance; Re, Reynolds number; U, nozzle air velocity; d50, 50% cutpoint;
Stk, square root of Stokes number; s, collection efficiency curve sharpness; DP, pressure drop.

Fig. 2. Experimental setup for laboratory characterization of PRPS.

a diffusion drier (Model 3062, TSI Inc., MN) and electrostatic charge distribution. The aerosol gen-
then passed through a Kr-85 neutralizer (Model erator for the 1.0–20.0 mm size range is also shown
3012, TSI Inc., MN) to stabilize the particle in Fig. 2. It uses an aqueous suspension of solid
 ARTICLE IN PRESS
216 S.J. Lee et al. / Atmospheric Environment 40 (2006) 212–224

glass spheres (nominal size 3–10 mm, density ¼ The alternating measurements were repeated at
2.48 g cm
3, Polysciences, PA). A suspension of least five times to ensure adequate precision for both
approximately 1.0 g of glass beads per 200 ml of the SMPS and the APS. The reproducibility of the
distilled water was continuously mixed with a measurements was 75% in upstream particle
magnetic stirrer. For particle sizes of more than concentration. Particle losses for each impaction
1 mm diameter, the charge has insignificant effects stage were measured using the impaction nozzle
on the measurement of collection efficiency (Fuchs, without the impaction substrate.
1963; Lee et al., 2005; Liu and Pui, 1974; For the SMPS data, each particle size interval was
Wiedensohler, 1988). Therefore, the neutralizer converted from mobility equivalent diameter to
was not used to stabilize the particle electrostatic aerodynamic diameter. Assuming that the mobility
charge (Boltzmann) distribution for generating equivalent diameter is equal to the equivalent
1.0–20 mm solid glass bead spheres. The suspension volume diameter (Kasper, 1982), the conversion
was aerosolized using a nebulizer (Retec X 70, Retec was made as (Peters et al., 1993)
Development Laboratory, OR) and then diluted sffiffiffiffiffiffiffiffiffiffiffiffi
using dry, particle-free room air to evaporate the pffiffiffiffiffiffi rp C m
water from the particles. Cad a ¼ dm, (4)
r0 w
The output of the aerosol generator passed into
the top of a cylindrical anodized aluminum dilution
where da is aerodynamic diameter, dm the mobility
tube (120 cm H  5.08 cm O.D.). Supplemental air
equivalent diameter, Ca the slip correction factor for
was drawn in through a HEPA filter at the top of
the aerodynamic diameter, Cm the slip correction
the dilution tube. To insure adequate aerosol mixing
factor for the mobility equivalent diameter, w the
inside the tube, a round plate was placed inside the
dynamic shape factor, rp the density of the particle,
tube just downstream of the input flows. Each
and r0 the unit density (1 g cm
3). For sodium
impaction stage was mounted inside the dilution
chloride particles, the dynamic shape factor and
tube and evaluated separately, as shown in Fig. 2.
particle density were assumed to be 1.08 (that of a
To make sure the laminar flow inside the duct, the
cube) and 2.2 g cm
3, respectively.
upstream sampling port was placed approximately
The collection efficiency for a given particle size,
larger than eight-duct diameters after particle
Eff, for each impaction stage, was calculated as
generation port. The downstream sampling port
was also placed eight-duct diameter downstream of C upstream
C downstream C upstream
the outlet of the impaction stage. Eff ¼ ¼1
,
C upstream C downstream
A scanning mobility particle sizer (SMPS) (Model
3080, TSI Inc., MN) and an aerodynamic particle (5)
sizer (APS) (Model 3310, TSI Inc., MN) were used to
where, Cupstream and Cdownstream are upstream and
quantify particle number as a function of particle size
downstream particle number concentrations, re-
with aerodynamic diameters from 0.01 to 1.0 mm and
spectively. The collection efficiency data were fitted
from 0.6 to 20 mm, respectively. The flow through
using the Weibull sigmoidal algorithm (Sigma Plot,
each stage was maintained at a constant 5 LPM.
SPSS Inc., IL):
Sampling flow of SMPS was 0.3 LPM. For the
SMPS, isokinetic sampling probes were used, 1=c
=bÞc
alternating between upstream and downstream of y ¼ y0 þ a½1 þ e
ðx
x0 þb ln 2 , (6)
the sampler, sampling for 135 s duration. The
where a, b, c, x0 , and y0 are the coefficients
upstream sampling port was placed approximately
determined by the algorithm. The sharpness (s) of
eight-duct diameters upstream from the impaction
the collection efficiency curve was calculated by
stage. The downstream sampling port was similarly
placed on a second port connected to the outlet of sffiffiffiffiffiffiffiffiffi
the impaction system. d 84:1
s¼ , (7)
For the APS, the sampling flow was 5 LPM. The d 15:9
impactor stage was placed inside the dilution tube.
Up/downstream particle concentrations were tested where d84.1 and d15.9 are the sizes of particles having
with/without the impactor stage. Particle concen- collection efficiencies of 84.1% and 15.9%, respec-
trations were measured for 1 min duration. tively (Hinds, 1999).
 ARTICLE IN PRESS
S.J. Lee et al. / Atmospheric Environment 40 (2006) 212–224 217

2.4. Laboratory intercomparison tests dilution tube. Four isokinetic probes were placed at
a distance of more than 10 times the tube diameter
Laboratory intercomparison experiments were below the top of the dilution tube to sample for the
conducted to compare the size distribution mea- PRPS, the MOI, and the real-time instruments
sured by the PRPS and the micro-orifice impactor (SMPS/APS).
(MOI) (MSP Corporation, Minneapolis, MN). As in the laboratory performance characterization,
SMPS and APS were also collocated to obtain the for the SMPS data, each particle size interval was
real-time particle concentration. The experimental converted from mobility equivalent diameter to
setup for these experiments is shown in Fig. 3. In aerodynamic diameter. For NaCl particles, the
order to simulate the typical bi-modal size distribu- dynamic shape factor was assumed to be 1.08 (that
tion of atmospheric aerosols, the mixture of the of a cube) and the particle density is 2.2 g cm
3. For
Coarse Test Dust (CTD) (Powder Technology Inc., the CTD, the dynamic shape factor was assumed to
Burnsville, MN) and NaCl was used as the test be 1.36 (that of quartz microcrystals) and the particle
aerosol. Polydisperse aerosols were generated in the density is 2.65 g cm
3. It should be noted that for the
0.01–20 mm particle size range. CTD was dried over laboratory intercomparison tests, the generated sub-
24 h prior its use at a temperature of 50 1C and micron aerosol was a mixture of NaCl and CTD. An
aerosolized using the fluidized-bed aerosol genera- additional laboratory experiment was conducted to
tor (Model 3400A, TSI Inc., MN). An aqueous determine the relative amounts (fractions) of NaCl
solution of NaCl was atomized using an atomizer and CTD. Based on the fractions of NaCl and CTD
(Model 3076, TSI Inc., MN), as described above for used for the SMPS measurements, the aerodynamic
the impactor characterization experiments. The diameters were calculated. The particle number
mixture of aerosolized CTD and NaCl was then concentration was measured with the SMPS for
passed through the Kr-85 neutralizer to reduce the 135 s duration using isokinetic sampling probes. First,
particle charge close to the Boltzmann equilibrium only NaCl aerosol was generated; secondly, both
distribution. The output of the aerosol generator NaCl and CTD aerosol were used; and finally, only
passed into the top of a cylindrical dilution tube CTD aerosol was generated. Measurements were
(150 cm H  12.82 cm O.D.). Supplemental air was repeated at least five times to ensure adequate
drawn through a HEPA filter at the top of the precision. The reproducibility of the measurements

Fig. 3. Laboratory intercomparison experimental setup.

 ARTICLE IN PRESS
218
Fraction of NaCl number conc. in generation (%) S.J. Lee et al. / Atmospheric Environment 40 (2006) 212–224

100 The sample duration was about 180 min at

90 approximate average particle mass concentration
of 200 mg m
3 and 360 min at 100 mg m
3. At the end
80
of each run, PUF substrates, the 37 mm Teflon
70 backup of PRPS, and 37 mm Teflon filers of MOI
60 were equilibrated for 48 h in a temperature and
50 relative humidity-controlled room (4075% RH and
2172.8 1C). Just prior to gravimetric analysis, both
40
the filters and the PUF substrates were exposed to
30 210
Po ionizing units (Staticmaster Model 2U500,
20 NRD Inc., Grand Island, NY) for more than 10 s to
10
minimize effects of electrostatic charge. For all
samples, 100% replicate weighing was performed
0
using an electronic microbalance (Mettler Toledo
0.0 0.1 0.2 0.3 0.4 0.5 0.6
(a) Mobility diameter (µm)
MT5, Columbus, OH) both before and after
sampling, with a third weighing done when the first
two were not within 75 mg. This procedure was
Fraction of NaCl number conc. in generation (%)

100

90
implemented to improve the precision of the
gravimetric analysis and to exclude outliers (Allen
80
et al., 1999).
70

60 2.5. Field study

50
Two collocated PRPS systems, each of which is
40 made up with PM10 and PM2.5 stages including a
30 backup filter, were used for the field study. For the
20
intercomparison study, an HI for PM10 (Marple
et al., 1987)), a PM2.5 HI, a PM10 Harvard PEM
10
(Demokritou et al., 2001), a USEPA PM2.5 WINS
0 Impactor (Rupprecht & Patashnick Co., Inc.), and a
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 PM2.5 Harvard PEM were collocated on the roof of
(b) Aerodynamic diameter (µm)
Harvard Countway Library in Boston, Massachu-
Fig. 4. Fraction of number concentration of NaCl in generated setts. Samples were collected during November 2003
aerosols: (a) as a function of mobility diameter and (b) as a (for the intercomparison with HIs and Harvard
function of aerodynamic diameters. PEMs) and May 2004 (for the intercomparison with
the WINS) for sampling durations of 48 h. Initial
was 78%. Figs. 4a and b show the fraction of NaCl and final sampling flows for each sampler were
(in %) in the generated aerosol as a function of measured with calibrated rotameters (Matheson,
mobility diameter and consequent aerodynamic East Rutherford, NJ). The HI and the Harvard
diameter, respectively. The aerodynamic diameter PEM collect particles on 37 mm Teflon membrane
was calculated from the proportionally adjusted filters (Pall Life Sciences, Ann Arbor, MI) whereas
dynamic shape factor and particle density in each WINS impactor collects particles on 47 mm Teflon
size bin. membrane (Pall Life Sciences, Ann Arbor, MI). HI,
The MOI operates at 30 LPM and was used to Harvard PEM, WINS, and PRPS operated at 10, 4,
classify particles in the following size intervals 16.7, and 5 LPM, respectively. Final filter/PUF
p0.56, 0.56–1.00, 1.00–3.2, 3.2–10, and X10 mm. weighing was performed within 10 days after field
Teflon membrane filters of 37 mm (Cat. No. sampling, following EPA guidelines, using the
R2PJ037, Pall Life Sciences, Ann Arbor, MI) were technique described above.
used as impaction substrates. Average particle PUF is a porous material, and is known to absorb
number concentrations were also calculated using a small variable quantity of water vapor from the
the real-time data obtained by the collocated APS air with increasing relative humidity (Demokritou
and SMPS. et al., 2004b). To correct field test substrates for the
 ARTICLE IN PRESS
S.J. Lee et al. / Atmospheric Environment 40 (2006) 212–224 219
pffiffiffiffiffiffiffi
effects of relative humidity, the laboratory blanks square root of Stokes number ð StkÞ at the cutoff
were used. The difference in mass for each field test diameter, the collection efficiency curve sharpness
substrate, before and after sampling, was adjusted (s), and the pressure drop (DP) across the stage. The
to take into account changes in the corresponding cutoff diameters in each stage are 10.0, 4.50, 2.65,
blanks. No humidity effects were necessary for the 1.05, and 0.58 mm, respectively. For all stages, the s
Teflon filters. Teflon membrane filters were used for values are sufficient to provide adequate separation
the impaction stages of the MOI, and backup filters of particles larger than the cutpoint from the
of Harvard PEM, HI, WINS, and PRPS (Allen airstream.
et al., 1999; Demokritou et al., 2004a). In previous studies, when the porous PUF
Quality assurance and quality control (QA/QC) impaction substrate was used, a decrease of cutoff
procedure included the determination of the limit of diameter was observed for the same acceleration
detection (LOD), the precision, and the coefficient nozzle configuration, compared to flat and rigid
of variation (CV) for the PRPS. The LOD was substrates (Demokritou et al., 2004a, b; Huang and
calculated based on three times of the standard Tsai, 2003; Kavouras and Koutrakis, 2001; Lee
deviation of mass for the field blanks. Precision and et al., 2005) The decrease of cutoff diameter results
CV were obtained from each stage of the PRPS. in a decrease in the dimensionless Stk values. The
Precision was calculated as the root mean square theoretically calculated Skt value is approximately
(RMS) difference for the
pffiffiffi paired sample values for 0.7 for a rectangular impaction nozzle configuration
the PRPS, divided by 2. The CV was determined and a rigid, flat impaction surface. In this study
as the precision divided by the sample mean using the PUF impaction substrate, the Stk value
concentration. has the range 0.2–0.5, which is significantly smaller
than 0.7. As a result, for a given cutpoint, a
3. Results and discussions lower acceleration is required, with a correspond-
ingly lower pressure drop. This feature can reduce
3.1. Impactor characterization experiments the vaporization losses of semi-volatile particle
components.
The collection efficiency curves for the five Fig. 6 illustrates the particle losses for all five
impactor stages are shown in Fig. 5. Also, Table 1 stages as a function of the particle aerodynamic
presents the critical parameters for each stage of the diameter. Particle losses inside each stage were
PRPS. These include the cutoff diameter (d50), the determined by measuring particles up- and down-
stream of each stage without the impaction sub-
strate. For the coarse particle size range (aerody-
100 namic diameter 2.5–10 mm), the particle losses vary
from 5% to 15%. For fine particles (p2.5 mm),
90

80
Collection efficiency (%)

100
70
PM0.5
90
PM2.5 PM10
60 80
70
Particle loss (%)

50
PM1.0 PM4.5
60
40
50
30 PM.5
40 PM1.0 PM2.5 PM4.5 PM10

20 30
20
10
10
0 0
1 10 0 1 2 3 4 5 6 7 8 9 10 11
Aerodynamic diameter (µm) Aerodynamic diameter (µm)

Fig. 5. Collection efficiencies as a function of aerodynamic Fig. 6. Interstage particle losses as a function of aerodynamic
diameter for each stage of PRPS. diameter.
 ARTICLE IN PRESS
220 S.J. Lee et al. / Atmospheric Environment 40 (2006) 212–224

particle losses are less than 10%. Comparable 3.2. Laboratory comparison of the PRPS, the MOI,
results were previously reported for other impactors and real-time instruments
(Demokritou et al., 2002b; Kavouras et al., 2000;
Marple et al., 1991). Particle losses within the PRPS Fig. 7 shows the mass size distributions measured
are caused by the turbulent flow in the acceleration with the MOI, the PRPS, and the SMPS/APS. It is
nozzle area and the wall surface between stages. It apparent that the size distribution obtained from
should be noted that the collection efficiencies the PRPS is in good agreement with that of the real-
determined for each stage (see Fig. 5) include time instrumentation (SMPS/APS), while the MOI
particles collected on the impaction substrate as measurements do not show as good an agreement.
well as the particles lost on the acceleration nozzle The MOI measurements indicate lower coarse
and wall. particle concentrations than the other two methods.
It should be noted that the dry particles can easily
bounce off from the flat and non-adhesive impac-
tion surface such as Teflon membrane filters (Lee
et al., 2005; Marple et al., 1991). Systematic
450 laboratory tests have been conducted to determine
∆Mass Concentration/∆log(Dp), µg/m3

400 particle separation characteristics of uncoated,

SMPS/APS data
350 coated, and PUF impaction substrates under
PRPS various particle-loading conditions (Lee et al.,
300
2005).The particles bounce and re-entrainment from
250 MOI the upper MOI stages, result in particle collection
200 on next stage and/or higher particle loss on the wall.
150
Also, particles are more likely to be deposited on the
multiple nozzle of the MOI than the single-slit
100
nozzles of the PRPS. Indeed, visual examination
50 indicated that there was a large particle deposition
0 on the wall and the multiple nozzles of the MOI
0.1 1 10 after the laboratory intercomparison test. Similar
Aerodynamic diameter (Dp), µm results between the MOI and other cascade im-
Fig. 7. Laboratory particulate mass concentration measured with pactor were previously reported (Demokritou et al.,
the MOI, the PRPS, and the SMPS/APS at high loading particle 2004a). Table 2 summarizes the results from the
concentration. laboratory comparison of the PRPS, the MOI, and

Table 2
Laboratory comparison of the particulate mass collected with the MOI, the PRPS and SMPS/APS

MOI stage size range (mm) and collected mass (mg)

MOI test o0.56 0.56–1.0 1.0–3.2 3.2–10 410 Total collected mass (mg)

High concentration 129.5 55.2 82.0 26.8 0.5 294.0

Low concentration 56.5 35.2 30.0 6.9 1.2 129.8
PRPS stage size range (mm) and collected mass (mg)

PRPS test o0.50 0.50–1.0 1.0–2.5 2.5–10 410 Total collected mass (mg)

High concentration 148.3 50.7 147.6 186.6 -0.5 532.8

Low concentration 62.8 18.8 35.0 24.9 -2.9 138.7
SMPS/APS size range (mm) and collected mass (mg)

SMPS/APS test o0.50 0.50–1.0 1.0–2.5 2.5–10 410 Total collected mass (mg)

High concentration 85.2 84.9 146.8 105.5 20.2 442.6

Low concentration 30.9 29.8 39.2 26.9 6.9 133.7
 ARTICLE IN PRESS
S.J. Lee et al. / Atmospheric Environment 40 (2006) 212–224 221

Table 3
Sample mean concentration, precision, coefficient of variation (CV), limit of detection (LOD) for PRPS

Stage Description Number of Mean conc. Precision CV LOD

sample (mg m–3) (mg m–3) (mg m–3)

1st Particle collection above 10 mm 12 8.50 0.45 0.05 1.29

2nd Particle collection from 2.5 to 10 mm 12 5.52 0.26 0.05 0.72
Backup Particle collection below 2.5 mm 12 12.15 0.42 0.03 0.47

the real-time instruments (SMPS/APS) for both 35

relatively high (about 500 mg) and relatively low
(about 130 mg) total collected mass. The ratio of

PRPS PM10 mass concentration (µg/m3)

30
MOI to PRPS mean total collected mass was 0.55 Y = 0.99X - 1.17
r2 = 0.99
and 0.93 for high and low concentration tests, 25
respectively.
20
3.3. Field study results
15
Table 3 gives the result of the PRPS field
evaluation study including number of tests, mean 10 Data
concentrations, precision, CV, and LOD. Six pairs 1:1
Regression
of PRPS samples were obtained, with the first stage
5
(collecting the particles above 10 mm), the second
stage (collecting the particles from 2.5 to 10 mm),
0
and the backup filter (collecting the particles below
0 5 10 15 20 25 30 35
2.5 mm). For these samples of ambient Boston air,
PEM PM10 mass concentration (µg/m3)
the mean ambient concentrations of each stage were
(a) Harvard PEM vs. PRPS
all considerably higher than the LOD values. The
results of CV values for each stage were below 5%. 35
Figs. 8a and b show that PM10 mass concentra-
tions measured by the PRPS agree fairly well with
PRPS PM10 mass concentration (µg/m3)

30
the Harvard PEM and the HI, respectively. PM10 Y = 0.94X + 2.98
for the PRPS is overall about 1.2 mg m
3 or 6.2% r2 = 0.97
25
lower than for the Harvard PEM, and about
2.6 mg m
3 or 16.5% higher than for the HI PM10.
20
The PM10 concentration of PRPS is obtained from
the sum of the mass of the second impaction
substrate (collecting the particles from 2.5 to 15
10 mm) and the backup filter (collecting the particles
below 2.5 mm) divided by the total sample volume. 10
This calculation is appropriate because the particle Data
1:1
loss for the PM2.5 stage was quite small (less 5 Regression
than 5%).
Figs. 9a–c show that the PM2.5 mass concentra- 0
tions measured by the PRPS are also fairly in good 0 5 10 15 20 25 30 35
agreement with the Harvard PEM, the HI PM2.5, HI PM10 mass concentration (µg/m3)
and the USEPA PM2.5 WINS. PM2.5 mass concen- (b) HI vs. PRPS
tration of PRPS was obtained from the mass of Fig. 8. The PM10 mass concentration measured with the PRPS in
Teflon backup filter divided by sampling volume. comparison with the (a) Harvard PEM and (b) HI in the ambient
PM2.5 for the PRPS is overall about 0.8 mg m
3 or air around Boston.
 ARTICLE IN PRESS
222 S.J. Lee et al. / Atmospheric Environment 40 (2006) 212–224

30 30
PRPS PM2.5 mass concentration (µg/m3)

PRPS PM2.5 mass concentration (µg/m3)

25 25
Y = 0.94X - 0.83 Y = 1.00X + 2.96
r2 = 0.96 r2 = 0.95
20 20

15 15

10 10
Data
1:1 Data
Regression 1:1
5 5
Regression

0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
3 3
(a) PEM PM2.5 mass concentration (µg/m ) (b) HI PM2.5 mass concentration (µg/m )

20
PRPS PM2.5 mass concentration (µg/m3)

Y = 1.00X + 1.15
15 r2 = 0.98

10

5 Data
1:1
Regression

0
0 5 10 15 20
3
(c) WINS PM2.5 massconcentration (µg/m )

Fig. 9. The PM2.5 mass concentration measured with the PRPS in comparison with the (a) Harvard PEM, (b) HI, and (c) WINS in the
ambient air around Boston.

6% lower than for the Harvard PEM, and about 3.0 4. Summary and conclusions
and 1.5 mg m
3 higher than for the HI and WINS,
respectively. The personal respirable particulate sampler
Finally, Figs. 10a and b show that the PM2.5 (PRPS) was developed and evaluated. It operates
sulfate mass concentrations measured by the PRPS at 5 LPM with five independently selectable impac-
agree quite well with the Harvard PEM and USEPA tion stages. This system can be used to collect
PM2.5 WINS impactor. PM2.5 sulfate mass concen- particles with sizes between 10, 4.5, 2.5, 1.0, and
tration for the PRPS is about 0.1 mg m
3 (2.1%) 0.5 mm on PUF substrates. Particles smaller than
higher than for the Harvard PEM and about 0.5 mm are collected on a Teflon membrane backup
0.03 mg m
3 higher than for the WINS impactor. filter. It has a relatively low total pressure drop of
 ARTICLE IN PRESS
S.J. Lee et al. / Atmospheric Environment 40 (2006) 212–224 223

5 study was conducted with the Harvard personal

environmental monitor (PEM), Harvard impactor
(HI), and USEPA PM2.5 WINS impactor. The size
PRPS PM2.5 sulfate concentration (µg/m3)

Y = 0.90X + 0.19 distribution obtained by PRPS agrees well with that

4
r2 = 0.99
of the real-time instrumentation (SMPS/APS). The
average collected mass of PRPS was found to be
3 higher than that of the MOI at laboratory
intercomparison test. The difference is most likely
due to the particle bounce-off from the uncoated
impaction substrates and the increased wall and
2
nozzle particle losses inside the MOI. The field study
Data results indicate fairly good agreement between the
1:1 PRPS and the three reference methods (HI,
1 Regression
Harvard PEM, and WINS) for PM10 and PM2.5
mass and sulfate concentrations.
Consequently, the personal sampler can be used
0 to provide the estimation of human exposure and to
0 1 2 3 4 5
improve our understanding of the chemical and
(a) PEM PM2.5 sulfateconcentration (µg/m3)
toxicological characteristic of particles.
7
Acknowledgments
PRPSPM2.5 sulfate concentration (µg/m3)

6
This work was supported by US Environmental
5 Y = 1.01X + 0.03 Protection Agency/Harvard Particle Health Effects
r2 = 0.99
Center (EPA Grant No. R827353) and the NIEHS
Harvard Center (NIEHS Grant No. ES00002).
4

3 References

2 Allen, G.A., Oh, J.A., Koutrakis, P., Sioutas, C., 1999.

Data Technique for high-quality ambient coarse particle mass
1:1
measurements. Journal of the Air & Waste Management
Regression
1 Association 49, 133–141.
Castillejos, M., Borja-Aburto, V.H., Dockery, D.W., Gold, D.R.,
Loomis, D., 2000. Airborne coarse particles and mortality.
0 Inhalation Toxicology 12, 61–72.
0 1 2 3 4 5 6 7 Chang, L.T., Sarnat, J., Wolfson, J.M., Rojas-Bracho, L., Suh,
(b) WINS PM2.5 sulfate concentration (µg/m3) H.H., Koutrakis, P., 1999. Development of a personal multi-
pollutant exposure sampler for particulate matter and criteria
Fig. 10. The PM2.5 sulfate mass concentration measured with the gases. In: Pollution Atmospherique, Numero Special 40e
PRPS in comparison with the (a) Harvard PEM and (b) WINS. Anniversaire de L’APPA, pp. 31–39.
Clayton, C.A., Perritt, R.L., Pellizzari, E.D., Thomas, K.W.,
Whitmore, R.W., Wallace, L.A., Ozkayak, H., Spengler, J.D.,
about 2.0 kPa. The PUF used for the impaction 1993. Particle Total Exposure Assessment (PTEAM) study;
substrates results in insignificant particle bounce-off distribution of aerosols and elemental concentrations in
personal, indoor, and outdoor samples in a southern
and re-entrainment, without using adhesives such as California community. Journal of Exposure Analysis and
grease or mineral oil as coatings. Interstage losses Environmental Epidemiology 3, 227–250.
were less than 10% and 20% for fine and coarse Demokritou, P., Kavouras, I.G., Ferguson, S.T., Koutrakis, P.,
particles, respectively. 2001. Development and laboratory performance evaluation of
a personal multipollutant sampler for simultaneous measure-
The PRPS was also evaluated in a laboratory
ments of particulate and gaseous pollutants. Aerosol Science
study using real-time measurement instruments and Technology 35, 741–752.
(SMPS/APS) and compared with the micro-orifice Demokritou, P., Gupta, T., Ferguson, S.T., Koutrakis, P., 2002a.
impactor (MOI). Also, a field intercomparison Development and laboratory performance evaluation of a
 ARTICLE IN PRESS
224 S.J. Lee et al. / Atmospheric Environment 40 (2006) 212–224

personal cascade impactor. Journal of the Air & Waste Marple, V.A., 1970. A fundamental study of inertial impactors.
Management Association 52, 1230–1237. Ph.D. Thesis, University of Minnesota, MN.
Demokritou, P., Kavouras, I.G., Ferguson, S.T., Koutrakis, P., Marple, V.A., Rubow, K.L., Turner, W., Spengler, J.D., 1987.
2002b. Development and experimental evaluation of a Low flow rate sharp cut impactors for indoor air sampling:
personal environmental monitoring system for simultaneous design and calibration. Journal of Air Pollution Control
measurements of particulate matter and criteria gases. Association 37, 1303–1307.
Aerosol Science and Technology 35, 741–752. Marple, V.A., Rubow, K.L., Behm, S., 1991. A microorifice
Demokritou, P., Lee, S.J., Ferguson, S.T., Koutrakis, P., 2004a. uniform deposit impactor (MOUDI): description, calibration
A compact multistage (cascade) impactor for the character- and use. Aerosol Science and Technology 14, 434–446.
ization of atmospheric aerosols. Journal of Aerosol Science Marple, V.A., Rubow, K.L., Olson, B.A., 1993. Inertial,
35, 281–299. gravitational, centrifugal and thermal collection techniques.
Demokritou, P., Lee, S.J., Koutrakis, P., 2004b. Development In: Aerosol Measurement. Van Nostrand Reinhold, New
and evaluation of a high loading PM2.5 speciation sampler. York.
Aerosol Science and Technology 38, 111–119. Misra, C., Singh, M., Shen, S., Sioutas, C., Hall, P.M., 2002.
Dockery, D.W., Pope, C.A., 1994. Acute effects of particulate air Development and evaluation of a personal cascade impactor
pollution. Annual Review of Public Health 15, 107–132. sampler (PCIS). Journal of Aerosol Science 33, 1027–1047.
Dockery, D.W., Pope, C.A., Xu, X., Spengler, J.D., Ware, J.H., Moolgavkar, S.H., 2000. Air pollution and daily mortality in
Fay, M.E., Ferris, B.G., Speizer, F.E., 1993. An association three US counties. Environmental Health Perspective 108,
between air pollution and mortality in six US cities. New 777–784.
England Journal of Medicine 329, 1753–1759. Peters, T.M., Chein, H., Lundgren, D.A., 1993. Comparison and
Fuchs, N.A., 1963. On the stationary charge distribution of combination of aerosol size distributions measured with a low
aerosol particles in a bipolar ionic atmosphere. Geofisica Pura pressure impactor, differential mobility particle sizer, elec-
e Applicata 56, 185–193. trical aerosol analyzer, and aerodynamic particle sizer.
Hinds, W.C., 1999. Aerosol Technology. Wiley, New York. Aerosol Science and Technology 19, 396–405.
Hoek, G., Brunekreef, B., Fischer, P., van Wijnen, J., 2001. The Pope, C.A., Dockery, D.W., Schwartz, J., 1995. Review of
association between air pollution and heart failure, arrhyth- epidemiological evidence of health-effect of particulate air-
mia, embolism, thrombosis, and other cardiovascular causes pollution. Inhalation Toxicology 7, 1–18.
of death in a time series study. Epidemiology 12, 355–357. Rubow, K.L., Marple, V.A., Olin, J., 1987. A personal cascade
Huang, C.-H., Tsai, C.-J., 2003. Mechanism of particle impaction impactor: design, evaluation, and calibration. American
and filtration by the dry porous metal substrates of an inertial Industrial Hygiene Association Journal 48, 532–538.
impactor. Aerosol Science and Technology 37, 486–493. Salonen, R.O., Pennanen, A.S., Halinen, A.I., Hirvonen, M.-R.,
John, W., Fritter, D.N., Winklmayr, W., 1991. Resuspension Sillanpaa, M., Hillamo, R.V.K., Koskentalo, T., Aarnio, P.,
induced by impacting particles. Journal of Aerosol Science 22, Ferguson, S.T., Koutrakis, P., 2000. A chemical and
723–736. toxicological comparison of urban air PM10 collected during
Kasper, G., 1982. Dynamics and measurement of smokes I—size winter and spring in Finland. Inhalation Toxicology 12,
characterization of nonspherical particles. Aerosol Science 95–104.
and Technology 1, 187–199. Schwartz, J., Dockery, D.W., 1992. Increased mortality in
Kavouras, I.G., Koutrakis, P., 2001. Use of polyurethane foam Philadelphia associated with daily air pollution concentra-
as the impaction substrate/collection medium in conventional tions. American Journal of Epidemiology 135, 12–19.
inertial impactors. Aerosol Science and Technology 34, 46–56. Schwartz, J., Morris, R., 1995. Air-pollution and hospital
Kavouras, I.G., Ferguson, S.T., Wolfson, J.M., Koutrakis, P., admissions for cardiovascular-disease in Detroit, Michigan.
2000. Development and validation of a high volume low cut- American Journal of Epidemiology 142, 23–35.
off inertial impactor (HVLI). Inhalation Toxicology 12, 35–50. Sehmel, G.A., 1980. Particle resuspension: a review. Environment
Koutrakis, P., Wolfson, J.M., Bunyaviroch, A., Froehlich, S.E., International 4, 107–127.
Hirano, K., Mulik, J.D., 1993. Measurement of ambient Sexton, K., Spengler, J.D., Treitman, R.D., 1984. Personal
ozone using a nitrite-coated filter. Analytical Chemistry 65, exposure to respirable particles: a case study in Waterbury,
209–214. Vermont. Atmospheric Environment 18, 1385–1398.
Lee, S.J., Demokritou, P., Koutrakis, P., 2005. Performance Tsai, C.-J., Cheng, Y.-H., 1995. Solid particle collection
evaluation of commonly used impaction substrates under various characterization on impaction surface of different designs.
loading conditions. Journal of Aerosol Science 36, 881–895. Aerosol Science and Technology 23, 96–106.
Lioy, P.J., Waldman, J.M., Buckley, T.J., Butler, J., Pietarinen, Wall, S.M., John, W., Wang, H.C., Coren, S., 1990. Measure-
C., 1990. The personal, indoor and outdoor concentrations of ment of kinetic energy loss for particles impacting surfaces.
PM10 measured in an industrial community during the winter. Aerosol Science and Technology 12, 926–946.
Atmospheric Environment 24, 57–66. Wiedensohler, A., 1988. An approximation of the bipolar charge
Liu, B.Y.H., Pui, D.Y.H., 1974. Electrical neutralization of distribution for particles in the submicron size range. Journal
aerosols. Journal of Aerosol Science 5, 465–472. of Aerosol Science 19, 387–389.