Environ. Sci. Technol.

2006, 40, 155-162

Previous research has shown that fine particle emissions

Effects of Dilution on Fine Particle are sensitive to dilution conditions. A major focus has been
Mass and Partitioning of on nucleation and the particle size distribution, both of which
can be extremely sensitive to dilution conditions (2). However,
Semivolatile Organics in Diesel existing emission and ambient standards are based on fine
particle mass. Fine particle mass emissions depend on the
Exhaust and Wood Smoke phase partitioning of semivolatile compounds in the exhaust.
For sources with significant organic aerosol emissions,
ERIC M. LIPSKY† AND dilution samplers measure higher fine particle mass emission
ALLEN L. ROBINSON* rates compared to filters collected at exhaust temperatures
(1, 3). This well-recognized effect is due to gas-to-particle
Department of Mechanical Engineering, Carnegie Mellon
University, Pittsburgh, Pennsylvania, 15213
conversion of semivolatile species as the exhaust is cooled
during dilution. Cooling reduces the saturation pressures of
the semivolatile compounds in the exhaust.
Dilution samplers are typically operated at dilution ratios
Experiments were conducted to examine the effects of between 20:1 and 200:1. A dilution ratio of 100:1 is generally
high enough to reduce the exhaust temperature to ambient
dilution on fine particle mass emissions from a diesel engine
levels, but the median dilution ratio of vehicle exhaust in an
and wood stove. Filter measurements were made urban atmosphere is around 10 000:1 (20). Therefore, the
simultaneously using three dilution sampling systems concentration of semivolatile species inside a dilution
operating at dilution ratios ranging from 20:1 to 510:1. sampler can be orders of magnitude higher than typical
Denuders and backup filters were used to quantify organic atmopsheric conditions. One also needs to account for effects
sampling artifacts. For the diesel engine operating at of background pollution on the emissions because dilution
low load and wood combustion, large decreases in fine samplers mix exhaust with particle- and organics-free air.
particle mass emissions were observed with increases in Dilution reduces the concentrations of semivolatile spe-
dilution. For example, the PM2.5 mass emission rate cies; partitioning theory indicates that, under isothermal
from a diesel engine operating at low load decreased by conditions, this should reduce the amount of semivolatile
50% when the dilution ratio was increased from 20:1 to material in the particle phase (5, 6). However, the significance
of this effect is not well understood as illustrated by a recent
350:1. Measurements of organic and elemental carbon indicate
review concluding that fine particle mass tends to be
that the changes in fine particle mass with dilution are conserved upon dilution (2). Experiments with diesel exhaust
caused by changes in partitioning of semivolatile organic report modest decreases in fine particle mass emissions at
compounds. At low levels of dilution semivolatile species higher dilution ratios (3, 7). These changes were attributed
largely occur in the particle phase, but increasing dilution to changes in partitioning of semivolatile organics. However,
reduces the concentration of semivolatile species, both studies only considered dilution ratios smaller than
shifting this material to the gas phase in order to maintain 100:1. In addition, one study performed the experiments with
phase equilibrium. Emissions of elemental carbon do not a constant filter temperature of 52 °C (7) while the other only
vary with dilution. Organic sampling artifacts are shown to observed a decrease in emissions at one dilution ratio (3).
vary with dilution because of the combination of changes In contrast, measurements by Hildemann et al. (1) made on
a fuel oil boiler suggest increasing emissions with higher
in partitioning coupled with adsorption of gas-phase
dilution ratio. Interpretation of these data is potentially
organics by quartz filters. The fine particle mass emissions complicated by organic sampling artifacts (8).
from the diesel engine operating at medium load did not This paper examines the effects of dilution sampling on
vary with dilution because of the lower emissions of the fine particle mass emissions from a diesel engine and a
semivolatile material and higher emissions of elemental wood stove. Measurements of PM2.5 mass, organic carbon,
carbon. To measure partitioning of semivolatile materials and elemental carbon emissions were made at dilution ratios
under atmospheric conditions, partitioning theory indicates between 20:1 and 510:1. Backup and denuded filters were
that dilution samplers need to be operated such that the collected to estimate organic sampling artifacts. Partitioning
diluted exhaust achieves atmospheric levels of dilution. Too theory is used to discuss the data in the context of real-world
little dilution can potentially overestimate the fine particle dilution.
mass emissions, and too much dilution (with clean air)
Experimental Setup and Procedure
can underestimate them.
Experiments were performed to measure the effects of
dilution on fine particle mass emissions from a diesel engine
Introduction and a wood stove. Figure 1 shows a schematic of the
experimental setup. Filter samples were collected using three
Many combustion and other high-temperature sources emit
completely independent dilution sampling systems operated
compounds that are semivolatile or volatile at exhaust
simultaneously at different dilution ratios. This approach of
temperatures but undergo gas-to-particle conversion as the
simultaneous sampling with multiple samplers minimizes
combustion products mix with ambient air. Dilution sampling
the effects of temporal variations in emissions on the results.
is a technique developed to simulate these processes in order
Detailed descriptions of the design, characterization, and
to better characterize fine particle emissions (1).
operation of the dilution samplers can be found in Lipsky
and co-workers (9, 10). Briefly, each sampler isokinetically
* Corresponding author phone: (412)268-3657; fax: (412)268-3348;
e-mail: alr@andrew.cmu.edu. collects exhaust through separate, heated inlet lines that are
† Present address: Department of Engineering, Penn State Mc- maintained at a temperature slightly above the exhaust
Keesport, McKeesport, PA 15132. temperature to minimize thermophoretic losses. The sampled
10.1021/es050319p CCC: $33.50  2006 American Chemical Society VOL. 40, NO. 1, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 155
Published on Web 11/30/2005
 Aerosol Characterization. Filter samples were collected
using a sampling train that consists of a sharp-cut PM2.5
cyclone operating at 24 liters per min (lpm) (Figure 1).
Downstream of the cyclone the flow is split and passed
through two filter packs each operating at 12 lpm. One filter
pack contains a single quartz filter (Bare-Q); the second filter
pack contains a Teflon filter followed by a backup quartz
filter (quartz behind Teflon or QBT). The QBT filter is used
as an estimate of the positive sampling artifact from gas
adsorption of semivolatile organic material (8). Identical filter
trains were attached to the end of each dilution tunnel.
The Teflon filter is used for quantifying total PM2.5 mass
emissions. Teflon filters are weighed before and after
FIGURE 1. Experimental setup of simultaneous sampling with three sampling using a microbalance in a temperature and relative
dilution tunnels. Each sampler has a separate heated inlet line. humidity controlled environment (30-40% RH at 21-23 °C).
The filters are equilibrated for 24 h before gravimetric analysis.
exhaust is then rapidly mixed by turbulence with filtered Samples on quartz filters are used to quantify organic and
(HEPA and activated carbon) dilution air inside a dilution elemental carbon (OC and EC) emissions using a Sunset
tunnel. Filter trains are connected to the end of the dilution Laboratories laboratory thermal-optical transmission OC/
tunnel. Each tunnel provides about 2.5 s of residence time EC analyzer. The temperature protocol is a modified version
after mixing but before filter collection. Previous research of the NIOSH 5040 protocol (11). Quartz filters are prepared
has shown that additional residence time beyond that before sampling by baking them at 500 °C in air for at least
provided by the tunnel does not affect filter measurements 6 h to remove any residual carbon. Both quartz and Teflon
under the conditions of these experiments (10). All samplers filters are stored in a freezer (-18 °C) between sample
are constructed out of stainless steel with Teflon gaskets to collection and analysis. All filter measurements are blank
minimize sample contamination. corrected (10).
Two different dilution tunnel designs were used for these A second filter train was occasionally operated in parallel
experiments. To establish the comparability of the different with the first train to further investigate organic sampling
designs, separate experiments were conducted while operat- artifacts. This train consists of a sharp-cut PM2.5 cyclone, a
ing all three samplers simultaneously at the same dilution carbon monolith denuder (MastCarbon Ltd., U.K.), followed
ratio (10). These intercomparison experiments show excellent by a filter pack with a quartz fiber filter and a carbon
agreement between the different samplers across the range impregnated glass-fiber (CIG) filter (47 mm, Schleicher &
of dilution ratios considered here; for example, the average Schuell, GF 3649) (11). The flow rate through the denuded
relative bias in the PM2.5 mass emissions measured using the filter train was 12 lpm to match the filter face velocity of the
two designs is 1% ( 17% (average ( standard deviation). standard train. The CIG filter was analyzed with the Sunset
The dilution ratio within each sampler is determined by OC/EC analyzer in a helium atmosphere using a stepped
simultaneously measuring exhaust and dilution tunnel CO2 temperature profile ramping to 330 °C (11).
levels:
Sources. Emissions tests were performed using a small
(CO2)ex - (CO2)bck diesel generator and a wood stove. The diesel generator is
DR ) (1) a single-cylinder Yanmar L70AE air-cooled diesel engine
(CO2)tun - (CO2)bck connected to a 4.5 kW generator. The engine is EPA and
CARB exhaust emission compliant. Experiments were per-
where (CO2)ex, (CO2)bck, and (CO2)tun are the CO2 mixing ratios formed at constant load: low load (25% of rated capacity)
in the exhaust, dilution air, and dilution tunnel, respectively. or medium load (55% of rated capacity). Exhaust temper-
Separate CO2 monitors are used to continuously monitor atures at the sampling location were in the range of 230-260
(CO2)tun at the end of each tunnel. The dilution air CO2 mixing °C for low-load tests and 260-290 °C for medium-load tests.
ratio is measured both before and after each day of tests; the The engine was operated at the specified load setting for a
small variations in (CO2)bck, are insignificant compared to minimum of 45 min before filter sampling to allow all
the relatively high tunnel and exhaust CO2 levels. No dilution components to achieve a steady-state temperature. Diesel
corresponds to a dilution ratio of one. fuel was purchased at a local gas station.
To compare measurements made at different dilution
The wood stove is an EPA approved Jøtul “602 CB Classic”
ratios, emissions are reported as fuel-based emission factors
wood stove. The wood fuel was a mixture of oak, cherry, and
(e.g., g PM2.5/kg fuel):
some ash. Wood-burning experiments involved starting the
fire with a small amount of wood. After the fire was
[P] established, the stove was loaded to capacity and the wood
EF ) C (2)
[C] f was allowed to burn down for 45-75 min until the combus-
tion stabilized and the exhaust temperature at the sampling
where [P] and [C] are the background-corrected pollutant location was in the range of 130-150 °C. Sampling was then
and carbon concentrations inside the dilution tunnel, started, and stable flaming combustion conditions were
respectively, and Cf is the mass fraction of carbon in the fuel. maintained by adjusting the vents on the door of the stove.
[C] is determined from the measured exhaust gas composi- Exhaust temperatures and O2 and CO levels were reasonably
tion, assuming all of the fuel carbon is emitted as CO and constant during sampling.
CO2. Small amounts of fuel carbon are emitted as either gas-
or condensed-phase organic compounds; however this Results
organic material contributes negligibly to the overall carbon
balance. We use a value of Cf of 0.87 for the diesel fuel and Fuel-based emissions of PM2.5 mass, total carbon (TC), and
0.40 for the wood fuel. Unless otherwise noted, the word organic and elemental carbon (OC & EC) as a function of
“emission” or “emission rate” refers to fuel-based emissions. dilution ratio are shown in Figure 2 for low- and medium-

156 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 1, 2006

FIGURE 2. Fuel-based emissions from low-load diesel (left column), medium-load diesel (middle column), and wood smoke (right column)
experiments. Plots a1-a3 show PM2.5 mass emission rate; plots b1-b3 compare organic carbon emission factors measured with the Bare-Q
and Q - QBT approaches; plots c1-c3 show particulate organic carbon (OC) and elemental carbon (EC) emission factors measured using
the Q - QBT approach. Bare-Q is carbon measured with a quartz filter; Q - QBT is carbon measured with a backup-corrected quartz
filter. Q - QBT is particulate OC, while Bare-Q has substantial positive artifact, as discussed in the text. In plots c1-c3 symbols indicate
total carbon emissions and shading indicates the contribution of OC and EC to the emissions. Plots c1-c3 also show the carbon fractions
from the OC/EC analysis: He1 OC is the OC that evolves at 340 °C, and He2-He4 OC is the OC that evolves at temperatures greater than
340 °C during the OC/EC analysis. The diesel plots combine data from back-to-back experiments conducted on the same day. Lines are
intended as a visual aid. Vertical bars in (a) are experimental uncertainties determined from intercomparison experiments (10).

load diesel operations and a wood smoke experiment. The the changes in partitioning. The changes are not caused by
emissions are presented on a fuel basis in order to compare changes in saturation pressure because the temperature of
measurements made at different dilution ratios on a con- the diluted exhaust at the filter holder was essentially constant
sistent basis; fuel-based emissions of inert pollutants such across this entire set of experiments (27 ( 2 °C). (Note that
as EC should be independent of dilution ratio, appearing as mixing inside the dilution tunnel is not adiabatic; at the lowest
a horizontal line in Figure 2. dilution ratios there was some heat transfer to the sur-
The fuel-based PM2.5 mass emissions decrease with roundings).
increasing dilution ratio during low-load diesel and wood Dilution reduces the concentration of both semivolatile
smoke experiments. For example, Figure 2 shows that species and the sorptive material which, under constant
increasing the dilution ratio from 20:1 to 350:1 decreased the temperature conditions, requires semivolatile species to
PM2.5 mass emissions from the diesel engine operating at transfer from the particle to the gas phase to maintain phase
low load by 55%. Increasing the dilution ratio from 20:1 to equilibrium. This reduces the PM2.5 mass emission rate. A
120:1 decreased the PM2.5 mass emissions from the wood complication is the fact that dilution samplers are operated
stove by over 60%. Within experimental uncertainty, the PM2.5 using cleaned air, while real-world dilution mixes emissions
mass emissions measured during medium-load operation with polluted air. This issue is addressed in the discussion
of the diesel engine are constant (Figure 2). Trends similar section.
to those shown in Figure 2 were observed across the entire If changes in phase partitioning of semivolatile organics
set of experiments; however, combining results from different are responsible for decrease in PM2.5 mass, then these changes
experiments is difficult because of day-to-day variability in should be apparent in measurements of particulate OC. The
engine emissions and experiment-to-experiment variation simplest approach for measuring OC emissions is to use a
in wood smoke emissions. quartz filter (Bare-Q). Figure 2 shows that the trends in Bare-Q
Changes in the phase partitioning of the semivolatile OC with dilution are different than the changes in PM2.5 mass.
organic material with dilution is the likely explanation for For example, at low load the Bare-Q OC increase with dilution
the changes in the PM2.5 mass shown in Figure 2. Although ratio while the PM2.5 mass decreases. However, OC mea-
dilution does not alter the total (gas + particle) emission rate surements can be significantly impacted by sampling arti-
of semivolatile species, dilution does affect the phase facts; these artifacts are carefully considered in the next
partitioning of this material. Gas-particle partitioning occurs section.
via absorption with an organic solution or adsorption to soot Dilution-ratio-dependent losses are another potential
and mineral surfaces (5, 6). Therefore, phase partitioning explanation for the trends shown in Figure 2. Fuel-based EC
depends on the concentration and saturation pressure of emissions measured by the different samplers were within
the semivolatile species and the concentration and com- (12% for a given experiment, indicating consistent collection
position of the sorptive material (5, 6). Under the conditions of a nonvolatile component of PM2.5 (Figure 2c). Furthermore,
of these experiments, changes in concentration of the there were no consistent trends in the EC emission factor
semivolatile and sorptive material with dilution likely cause with dilution ratio.

VOL. 40, NO. 1, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 157

 consistent with a significant positive artifact on the Bare-Q
filter. For the wood smoke experiments (Figure 3c), the TC
measured by a Bare-Q filter is smaller than the Teflon mass.
Assuming wood smoke is dominated by carbonaceous
species, an organic-mass-to-organic-carbon ratio (OM/OC)
of 1.3 is required to close the mass balance between the
Teflon and Bare-Q filters. This value is smaller than expected
given the relatively polar composition of wood smoke (12),
indicating that the wood smoke also creates a net positive
artifact on a quartz filter. Other researchers report that the
positive artifact is dominant when sampling both diesel
exhaust and wood smoke (13-17).
Figure 3 indicates that the artifact-corrected estimates of
particulate OC are less than the PM2.5 mass for all three
experimental conditions. For diesel exhaust, the backup and
denuder approaches measure the same OC and EC emission
rates and the majority of the Bare-Q OC appears to be positive
artifact. On average, particulate OC contributes only 35%
FIGURE 3. Average ratio of total carbon (OC + EC) to PM2.5 mass and 20% of the Bare-Q OC at low and medium loads,
for (a) low-load diesel, (b) medium-load diesel, and (c) wood smoke respectively. In contrast, other papers report that particulate
experiments. Three different estimates of OC are shown: bare quartz OC contributes a majority of Bare-Q OC (13-15). Differences
filter (Bare-Q), backup-corrected quartz filter (Q - QBT), and denuded in engine technology and operating conditions likely influ-
quartz filter (DenQ). There are two different estimates of EC (Q - ence sampling artifacts; in addition, the larger positive
QBT EC and Bare-Q EC are the same). Error bars are standard artifacts reported here may be due, in part, to the effects of
deviations of the ratio of OC + EC to PM2.5 mass across the set of dilution on artifacts discussed below. The CIG filter indicates
experiments. This ratio is expected to be less than one because
that there was little negative artifact from the denuded quartz
of the contribution of elements other than carbon to the PM2.5 mass.
filter.
Organic Sampling Artifacts. Both positive and negative For wood smoke, the EC emissions measured by the
artifacts can significantly affect quartz-filter measurements different approaches are comparable, but OC emissions
of particulate OC (8). Positive artifact is caused by organic measured using the denuded filters are somewhat smaller
vapors adsorbing onto the quartz filter resulting in an than the Q - QBT OC. However, the wood smoke emissions
overestimate in the amount of organic aerosol. Negative are highly variable as indicated by the large error bars in
artifact is caused by particle-phase organic compounds Figure 3. Positive artifact is estimated to contribute between
volatilizing after collection resulting in an underestimate of 25% and 50% of wood smoke Bare-Q OC, which is consistent
the organic aerosol mass. Organic sampling artifacts are with results from Fine et al. (16, 17).
commonly accounted for using a combination of denuders The Teflon mass minus EC divided by the artifact-
and/or backup filters (8). We employed both approaches in corrected, particulate OC can be used as a crude consistency
this study. check for the artifact corrections. Under the assumption that
The OC collected on a backup quartz filter behind a Teflon carbonaceous materials dominate the fine particle mass, this
filter (QBT) is a standard estimate of the positive artifact (8). value is the OM/OC ratio. For wood smoke, the Q - QBT OC
The approach assumes that the inert Teflon filter only collects requires an OM/OC ratio of 1.8 for mass balance closure,
particles, allowing gas-phase organics to pass through to the which is within the range of expected values (12) suggesting
backup quartz filter. The particulate OC emissions are then the Q - QBT approach provides a reasonable estimate of
estimated by subtracting the OC collected on the QBT from particulate OC. However, this approach yields estimated OM/
that collected on the Bare-Q filter (Q - QBT). OC ratios that are much larger than expected for diesel
The denuded filter train provides a second estimate of exhaust; for example, the low-load data require a ratio of 2.6
the particulate OC emissions. The denuder removes most of for mass balance closure versus expected values in the range
the organic vapor from the sample upstream of the quartz of 1.2-1.4 (12). This indicates that, for the diesel exhaust, the
filter, minimizing the positive artifact. However, the depletion Q - QBT (or denuder) approaches are underestimating the
of semivolatile organic vapor from the air stream may cause particulate OC emissions and/or that a significant fraction
volatilization of semivolatile organic material collected on of the emissions are inorganic. Mass balance problems
the quartz filter resulting in a negative artifact (8). This
between sum of aerosol species and Teflon mass are not
negative artifact is estimated using the carbon impregnated
uncommon for measurements of diesel exhaust (13). A linear
glass-fiber (CIG) filter downstream of the denuded quartz
regression of the Teflon mass minus EC versus Q - QBT OC
filter. Therefore, the particulate OC emissions are estimated
yields a slope of 1.4 µg/µg-C and an intercept of 0.5 g/kg fuel
by adding the OC measured on the CIG filter and the denuded
with an R2 value of 0.85 (4). This suggests an OM/OC ratio
quartz filter (DenQ + CIG).
of 1.4, in line with expectations, and that there are significant
Figure 3 compares the average OC emissions measured
inorganic emissions from the diesel engine (water or sulfate
using the three different approaches: Bare-Q OC, backup-
seem to be the most obvious candidates).
corrected (Q - QBT) OC, and denuded OC. To combine data
from different experiments, the OC and EC data are normal- Effects of Dilution on Organic Particulate Emissions.
ized by PM2.5 mass before averaging. The ratio of TC to PM2.5 Figure 2 indicates that the trends in artifact-corrected,
mass should be less than 1 because non-carbon organic particulate OC match those of the PM2.5 mass, which is
components and inorganic species also contribute to PM2.5 consistent with the changes emissions with dilution being
mass. Figure 3 combines data from filters collected at different caused by changes in phase partitioning of semivolatile
dilution ratios. organic compounds. Increasing the dilution ratio from 20:1
When the diesel engine is operated at low and medium to 120:1 decreased the Q - QBT OC by 75% during the wood
loads (Figure 3, parts a and b), the TC measured with a Bare-Q smoke experiment shown in Figure 2. At low load, the diesel
filter is, on average, slightly greater than the PM2.5 mass particulate OC decreased by almost 70% when the dilution

158 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 1, 2006

FIGURE 4. Evolution of particulate carbon during OC/EC analysis
of quartz filters collected at three dilution ratios during a low-load
diesel experiment. Particulate carbon is defined as backup- FIGURE 5. Changes in fuel-based PM2.5 and organic mass emissions
corrected carbon (Q - QBT). The first four groups of bars indicate of low-load diesel exhaust as a function of dilution ratio. Organic
the amount of carbon that evolves from the four different temperature mass is defined as OC multiplied by 1.4. To easily compare the
steps of the He mode of the analysis (340, 500, 700, and 870 °C). The changes in the different measurements with dilution ratio, the quartz
final group of bars indicates the amount of EC that evolves during filter data have been shifted to match the PM2.5 mass data at a
the analysis. Increasing the dilution ratio substantially reduces the dilution ratio of 20. This is done by adding the number indicated
amount of carbon that evolves at the lowest temperature step of in the legend to each measurement. The changes in particulate
the He mode, while the carbon at the higher temperature steps organic mass (Q - QBT) match those of the overall PM2.5 mass. The
remains relatively constant. Backup and Bare-Q levels at the He3- backup-Q filter provides a measure of the gas-phase semivolatile
700C peak were the same at DR ) 200. organic carbon which, on a fuel basis, increases with dilution due
to changes in partitioning.
ratio was increased from 20:1 to 350:1. For the medium-load
diesel experiments, the Q - QBT OC is constant. step of the OC/EC analysis contributes 60-70% of the low-
Dilution also alters the composition of the particle-phase load Bare-Q OC versus only 25-35% at the medium load.
emissions. The ramping temperature protocol used for OC/ Therefore, less semivolatile carbon may be emitted under
EC analysis crudely classifies the OC by volatility; the most medium load.
volatile species evolving in the lowest temperature step of Sampling Artifacts and Dilution. Data presented in Figure
the helium mode (340 °C) and progressively less volatile 5 suggest that the individual quartz filters provide a surpris-
compounds evolving in the higher temperature steps. ingly consistent picture of the changes in phase partitioning
Figure 4 plots a thermogram of particulate OC (Q - QBT) of the semivolatile organics. The measurements are from a
measured during a low-load diesel experiment. Increasing low-load diesel experiment; to make direct comparisons with
the dilution ratio dramatically decreases the amount of OC PM2.5 mass, organic mass and not OC is plotted using an
evolving in the lowest temperature step of the OC/EC analysis, OM/OC ratio of 1.4. Figure 5 indicates that the changes in
while the amount of OC evolving in the higher temperature particulate organic mass, (Q - QBT) × 1.4, with dilution
steps (500 °C or higher) remains relatively constant. At low ratio exactly match the decrease of PM2.5 mass.
dilution ratios, carbon that evolves at 340 °C contributes the Compared to the changes observed with the other filters,
majority of the particulate OC emissions; at high dilution the Bare-Q OC reported on a fuel basis is relatively constant,
ratios the higher temperature (less volatile) fractions domi- varying by less than (25% across the range dilution ratio for
nate the particulate emissions. Figure 2, part c3, indicates the set of filters collected during a given experiment (Figure
that similar trends with dilution are observed in the wood 2b and Figure 5). The Bare-Q OC modestly increased with
smoke particulate OC with dilution. The modest decrease in dilution ratio during the diesel experiments and decreased
particulate wood smoke OC with dilution in the higher modestly during the wood smoke experiments. In contrast,
temperature steps may be related to pyrolysis of semivolatile for experiments in which there was a decreasing trend in
organics during OC/EC analysis. PM2.5 mass with dilution, the OC measured by the QBT filter
Low- versus Medium-Load Diesel Operations. Although increases dramatically with dilution ratio. Figure 5 illustrates
the overall PM2.5 mass emission factors of the diesel engine this trend for a low-load diesel experiment.
at low and medium loads are similar, at low load the PM2.5 One explanation of the Bare-Q and QBT OC data in Figure
mass emission factor is a strong function of dilution ratio, 5 is that the quartz filters provide a consistent measure of
but independent of dilution ratio at medium load. This the gas-phase semivolatile organics. The Bare-Q OC emis-
difference appears related to the composition of the emis- sions are relatively constant because the filter is collecting
sions. both gas- and particle-phase OC and the total OC emissions
At low load the PM2.5 emissions are dominated by OC (gas + particle) do not change with dilution. At low dilution
(average OC/EC ratio of 7) while at medium load emissions ratios the Bare-Q filter collects the majority of the semivolatile
are dominated by EC (average OC/EC ratio of 0.4). This likely organics as particles while at higher dilution ratios it collects
influences the relative importance of the different sorption them as gases. In contrast, the QBT OC emissions increase
mechanisms controlling gas-particle partitioning. At higher with dilution because these filters are only exposed to the
EC loadings adsorption to EC is likely more important than gas-phase compounds whose relative concentration in-
absorption in organic matter. This effect is illustrated by creases at higher dilution ratios due to changes in partitioning.
changes in thermal desorption profiles of diesel nanoparticles This interpretation of the quartz-filter data depends on
with soot loading (18). the filters not being in equilibrium (or saturated) with the
Another difference between the low- and medium-load gas-phase organics (19). Figure 6 plots the mass of OC
diesel exhaust appears to be the composition of the OC collected on the QBT filter as a function of the volume of
emission. Carbon that evolves in the 340 °C temperature exhaust passed through the filter during low-load diesel

VOL. 40, NO. 1, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 159

FIGURE 6. Organic carbon collected on a backup quartz filter behind
a Teflon filter (QBT) as a function of volume of sampled exhaust.
Data are from the low-load diesel experiments. Volume of sampled
exhaust is used as a proxy for the total mass of organics to which
the filter is exposed. The data show that quartz filters do not reach
equilibrium with the gas-phase organics under the conditions of
the experiments. The roll over at high volumes of sample exhaust
may indicate the filters are approaching equilibrium but also may
be caused by changes in gas-to-particle partitioning of semivolatile FIGURE 7. Ratio of total carbon (OC + EC) to PM2.5 mass measured
organics. Higher exhaust volumes correspond to lower dilution using three different filter configurations while sampling from the
ratio experiments. diesel engine operating at (a) low and (b) medium load. The results
show a substantial increase in the positive artifact on a Bare-Q
experiments. The volume of exhaust is a proxy for the mass filter with dilution ratio. Both the denuded quartz and backup-
of gas-phase organics to which the filter is exposed. Except corrected quartz approaches provide a consistent correction for
this positive artifact.
for a couple of low dilution ratio points, the amount of carbon
collected by the backup quartz filter increases with the volume
of sampled exhaust indicating the quartz filters are not in artifact. In contrast, the ratio of Bare-Q TC to PM2.5 mass
equilibrium with the gas phase. At the highest volumes of increases with dilution ratio indicating increasing positive
sampled exhaust, there is evidence of the mass of carbon artifact with dilution ratio. This increase is particularly
collected starting to roll over consistent with the quartz filters dramatic for low-load operations where the Bare-Q TC to
approaching equilibrium. However, Figure 6 is based on PM2.5 mass ratio increases from 0.8 at a dilution ratio of 20:1
measurements made at different dilution ratios. At low to a value of 1.7 at a dilution ratio of 350:1.
dilution ratios a larger fraction of the semivolatile OC exists The trends in Figure 7 can be explained by a combination
in the particle phase; therefore some of the roll over shown of changes in partitioning and effects of the quartz filter
in Figure 6 is likely due to changes in partitioning and not approaching equilibrium. Under low-load conditions, the
equilibrium effects. relative amount of gas-phase semivolatile organics increases
Although Figure 6 indicates that the quartz filters are with dilution due to changes in phase partitioning, which,
clearly not in equilibrium with the exhaust, we attribute the in turn, increases the positive artifact. Under medium-load
increases in dilution-corrected OC measured by the Bare-Q conditions, changes in partitioning were not observed and
filters in the diesel experiments to the effects of equilibrium. the increase in the positive artifact is likely associated with
As a filter approaches equilibrium the collection efficiency the previously discussed equilibrium effects. At high dilution
of gas-phase OC decreases, reducing the amount of gas- ratios, the positive artifact dominates the OC measured with
phase OC collected per unit volume of exhaust. Under the the Bare-Q filter. At lower dilution ratios, the relative amount
conditions of these experiments (variable dilution ratio of positive artifact reported here is consistent with results of
coupled with fixed filter sampling times), this change in previous studies of diesel exhaust (13-15).
collection efficiency will cause an apparent increase in the
fuel-based OC emissions as a function of dilution ratio Discussion
consistent with the diesel Bare-Q OC data shown in Figure The major finding of this paper is that increasing dilution
2. The fact that the ratio of PM2.5 mass to the backup-corrected after the temperature of the exhaust has reached ambient
OC (Q - QBT) remains constant across an experiment levels can dramatically reduce the PM2.5 mass emission rate
indicates that the Bare-Q and QBT filters are approaching from a diesel engine and wood stove. For example, the
equilibrium at the same rate. measured PM2.5 mass emission factor for the diesel engine
An interesting consequence of the changes in phase operating under low load is 2.9 g/kg fuel at a dilution ratio
partitioning with dilution ratio coupled with the efficient of 20:1 versus 1.4 g/kg fuel at a dilution ratio of 350:1. This
collection of gas-phase semivolatile organics by quartz filters decrease is caused by changes in partitioning of semivolatile
is that the positive artifact on a Bare-Q filter increases with organic compounds in the emissions. From a scientific
dilution ratio. This effect is illustrated in Figure 7 which plots perspective, there is not a unique value for the fine particle
the ratio of TC to PM2.5 mass as a function of dilution ratio mass emission rate for sources that emit semivolatile species
for the two diesel load conditions. Across the entire range of because the partitioning of these species varies continuously
dilution ratios, the ratios of Q - QBT and the DenQ TC to with dilution and temperature. One needs to understand
PM2.5 mass are essentially constant indicating that both these effects in order to correctly interpret measurements
approaches provide a consistent correction of the positive made with dilution samplers. For the two sources considered

160 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 1, 2006

here, measurements at dilution ratios less than 100:1 likely atmospheric levels for these sources. The amount of fine
substantially overestimate the emitted fine particle mass particle mass at higher levels of dilution depends on the
relative to the more dilute conditions encountered in the composition of the emissions. Figure 2 indicates that the
real world. This has obvious implications for using emission changes in fine particle mass are most pronounced at low
measurements in air quality models that mix emissions dilution ratios but that the PM2.5 mass emission rate may still
directly into large grid cells or use source profiles with species be decreasing at a dilution ratio of 350:1.
concentrations that are normalized by OC or PM2.5 mass. The fine particle emission rate of our diesel engine
Partitioning theory provides a context for understanding emissions is somewhat higher than other engines (13, 15),
the results of these experiments. Following Pankow (5, 6), while our wood combustion emissions were somewhat lower
the concentration of a semivolatile compound in the aerosol (15-17). Dilution ratios less than 100:1 may be adequate to
phase is as follows: reach atmospheric levels for low-emitting sources such as
natural gas combustion; conversely dilution ratios signifi-
1 cantly larger than 10,000:1 may be needed for smoking cars
Ca,i ) Ctot,i (3)
1 (4). Although not considered in these experiments, the diluted
1+
Kp,iCPM exhaust also needs to be at ambient temperatures. Emission
measurements at typical atmospheric concentrations are
where Kp,i is the partitioning coefficient for compound i, CPM experimentally challenging; a more attractive approach may
is the concentration of particulate matter (a proxy for the be to characterize the composition of the semivolatile species
amount of sorptive material), and Ctot,i is the total (gas + and then use a partitioning model to predict emissions at
particle) concentration of species i in the system. Kp,i depends atmospheric conditions (4).
on the saturation pressure of the species i, the type of sorptive The effects of dilution on partitioning also depend on the
process (adsorptive vs absorptive), and the properties of composition of the emissions. The exhaust must have
sorptive material (5, 6). Under the conditions of these significant levels of semivolatile material in order for
experiments (isothermal dilution), Kp,i will be independent partitioning to be important. For the sources considered here,
of dilution ratio (if one assumes it does not change as the the highest fine particle concentrations were several thousand
composition of the particle phase changes with dilution). micrograms per cubic meter at a dilution ratio of 20. Given
Therefore, CPM and Ctot,i determine the partitioning of the these levels of emissions, eq 3 indicates that compounds
emissions; these concentrations are determined by the overall with partitioning coefficients such that 1/Kp,i is in range of
emission rate, the composition of emissions, and the amount 1 and 5000 µg/m3 at ambient temperature will undergo phase
of dilution. To predict the overall change in semivolatile transition as the exhaust is diluted to typical atmospheric
organic mass one needs to apply eq 3 to each semivolatile concentrations. C-17 through C-29 n-alkanes are compounds
compound in the system. This is a challenging task given the commonly found in combustion exhaust with partitioning
complexity of the emissions. coefficients in this range; however, the vast majority of the
Dilution samplers are typically operated with particle- semivolatile material has never been identified on a com-
and organics-free air. This means that, inside the dilution pound-by-compound basis (13). In order for the changes in
sampler, CPM and Ctot,i will both scale as 1/DR (CPM may partitioning to measurably alter the overall fine particle mass
actually decrease faster than 1/DR due to changes in emissions, the total concentration of these semivolatile
partitioning). Increasing the dilution ratio will reduce the species (ΣCtot,i) must be comparable to the fine particle mass.
semivolatile mass in the particle phase. However, in the real Composition effects likely explain why dilution did not effect
world, emissions are mixed with polluted background air, the fine particle mass emissions of the diesel engine operating
not particle- and organic-free air. This background pollution under medium load even though the concentrations of PM
contributes to CPM and Ctot,i. As the exhaust becomes very in the diluted exhaust were a factor of 10 or more greater
dilute, the background concentrations will strongly influence than typical ambient levels.
the partitioning of the semivolatile material.
To measure the partitioning that occurs under atmo- Acknowledgments
spheric conditions, eq 3 indicates that one needs to operate
The authors acknowledge Emily Weitkamp, Andy Grieshop,
a dilution sampler such that CPM and Ctot,i of the diluted
Mark Prack, Jessica Chiu, Neal Shyam, and R. Subramanian
exhaust are at typical atmospheric levels. If the concentrations
at Carnegie Mellon University for their assistance in con-
inside the dilution sampler are significantly greater than
ducting this research and an anonymous reviewer whose
atmospheric levels, then too much of the semivolatile material
comments greatly improved the manuscript. This research
will partition into the particle phase and the measurements
was supported by the U.S. Department of Energy National
will overestimate fine particle emissions relative to atmo-
Energy Technology Laboratory under Contract DE-FC26-
spheric levels of dilution. Conversely, if the diluted concen-
01NT41017.
trations are significantly smaller than atmospheric levels,
then one can underestimate fine particle emissions. These
ideas are quantitatively explored in a companion paper that Literature Cited
applies absorptive partitioning theory to the low-load diesel (1) Hildemann, L. M.; Cass, G. R.; Markowski, G. R. A dilution stack
and wood smoke data (4). To assist in interpretation of sampler for collection of organic aerosol emissionssDesign,
measurements made with dilution samplers, researchers characterization and field-tests. Aerosol Sci. Technol. 1989, 10,
should report the PM or OC concentration (and temperature) 193-204.
of the diluted exhaust. (2) Kittelson, D. B. Engines and nanoparticles: A review. J. Aerosol
The amount of dilution required to reach atmospherically Sci. 1998, 29, 575-588.
relevant concentrations depends on the emission rate of (3) Frisch, L. E.; Johnson, J. H.; Leddy, D. G. Effect of Fuels and
semivolatile and sorptive material. For the sources considered Dilution Ratio on Diesel Particulate Emissions; SAE Paper 790417;
Society of Automotive Engineers: Warrendale, PA, 1980.
here, the minimum OC concentrations of the diluted exhaust
(4) Shrivastava, M. K.; Stanier, C. O.; Lipsky, E. M.; Robinson, A. L.
were between 100 and 200 µg-C/m3 at a dilution ratio of Modeling semi-volatile organic aerosol mass emissions from
350:1 for the low-load diesel and 510:1 for the wood smoke combustion systems. Environ. Sci. Technol., submitted.
experiments. These concentrations are an order of magnitude (5) Pankow, J. F. An absorption-model of gas-particle partitioning
higher than typical ambient OC levels, indicating that a of organic-compounds in the atmosphere. Atmos. Environ. 1994,
dilution ratio of ∼5000:1 is needed to approach typical 28, 185-188.

VOL. 40, NO. 1, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 161

 (6) Pankow, J. F. Review and comparative-analysis of the theories (15) Hildemann, L. M.; Markowski, G. R.; Jones, M. C.; Cass, G. R.
on partitioning between the gas and aerosol particulate phases Submicrometer aerosol mass distributions of emissions from
in the atmosphere. Atmos. Environ. 1987, 21, 2275-2283. boilers, fireplaces, automobiles, diesel trucks, and meat-cooking
(7) MacDonald, J. S.; Plee, S. L.; D’Arcy, J. B.; Schreck, R. M. operations. Aerosol Sci. Technol. 1991, 14, 138-152.
Experimental Measurements of the Independent Effects of (16) Fine, P. M.; Cass, G. R.; Simoneit, B. R. T. Chemical charac-
Dilution Ratio and Filter Temperature on Diesel Exhaust terization of fine particle emissions from fireplace combustion
Particulate Samples; SAE Paper 800185; Society of Automotive of woods grown in the northeastern United States. Environ. Sci.
Engineers: Warrendale, PA, 1981. Technol. 2001, 35, 2665-2675.
(8) Turpin, B. J.; Saxena, P.; Andrews, E. Measuring and simulating (17) Fine, P. M.; Cass, G. R.; Simoneit, B. R. T. Chemical charac-
particulate organics in the atmosphere: Problems and prospects. terization of fine particle emissions from the fireplace combus-
Atmos. Environ. 2000, 34, 2983-3013. tion of woods grown in the southern United States. Environ.
(9) Lipsky, E.; Stanier, C. O.; Pandis, S. N.; Robinson, A. L. Effects Sci. Technol. 2002, 36, 1442-1451.
of sampling conditions on the size distribution of fine particulate
(18) Sakurai, H.; Tobias, H. J.; Park, K.; Zarling, D.; Docherty, S.;
matter emitted from a pilot-scale pulverized-coal combustor.
Kittelson, D. B.; McMurry, P. H.; Ziemann, P. J. On-line
Energy Fuels 2002, 16, 302-310.
measurements of diesel nanoparticle composition and volatility.
(10) Lipsky, E. M.; Robinson, A. L. Design and evaluation of a portable
Atmos. Environ. 2003, 37, 1199-1210.
dilution sampling system and the effects of residence time on
mass emission rates. Aerosol Sci. Technol. 2005, 39, 542-553. (19) Mader, B. T.; Pankow, J. F. Gas/solid partitioning of semivolatile
(11) Subramanian, R.; Khlystov, A. Y.; Cabada, J. C.; Robinson, A. L. organic compounds (SOCs) to air filters. 3. An analysis of gas
Positive and negative artifacts in particulate organic carbon adsorption artifacts in measurements of atmospheric SOCs and
measurements with denuded and undenuded sampler con- organic carbon (OC) when using Teflon membrane filters and
figurations. Aerosol Sci. Technol. 2004, 38, 27-48. quartz fiber filters. Environ. Sci. Technol. 2001, 35, 3422-
(12) Turpin, B. J.; Lim, H. J. Species contributions to PM2.5 mass 3432.
concentrations: Revisiting common assumptions for estimating (20) Zhang, K. M.; Wexler, A. S. Evolution of particle number
organic mass. Aerosol Sci. Technol. 2001, 35, 602-610. distribution near roadways. Part I: Analysis of aerosol dynamics
(13) Schauer, J. J.; Kleeman, M. J.; Cass, G. R.; Simoneit, B. R. T. and its implications for engine emission measurement. Atmos.
Measurement of emissions from air pollution sources. 2. C-1 Environ. 2004, 38, 6643-6653.
through C-30 organic compounds from medium duty diesel
trucks. Environ. Sci. Technol. 1999, 33, 1578-1587. Received for review February 16, 2005. Revised manuscript
(14) Shah, S. D.; Cocker, D. R.; Miller, J. W.; Norbeck, J. M. Emission
received August 5, 2005. Accepted October 14, 2005.
rates of particulate matter and elemental and organic carbon
from in-use diesel engines. Environ. Sci. Technol. 2004, 38, 2544-
2550. ES050319P

162 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 1, 2006