Semi-volatile and particulate emissions from the combustion of alternative diesel fuels

Article abstract of DOI:10.1016/S0045-6535(00)00242-3

Motor vehicle emissions are a major anthropogenic source of air pollution and contribute to the deterioration of urban air quality. In this paper, we report results of a laboratory investigation of particle formation from four different alternative diesel fuels, namely, compressed natural gas (CNG), dimethyl ether (DME), biodiesel, and diesel, under fuelrich conditions in the temperature range of 800-1200°C at pressures of approximately 24 atm. A single pulse shock tube was used to simulate compression ignition (CI) combustion conditions. Gaseous fuels (CNG and DME) were exposed premixed in air while liquid fuels (diesel and biodiesel) were injected using a high-pressure liquid injector. The results of surface analysis using a scanning electron microscope showed that the particles formed from combustion of all four of the above-mentioned fuels had a mean diameter less than 0.1 μm. From results of gravimetric analysis and fuel injection size it was found that under the test conditions described above the relative particulate yields from CNG, DME, biodiesel, and diesel were 0.30%, 0.026%, 0.52%, and 0.51%, respectively. Chemical analysis of particles showed that DME combustion particles had the highest soluble organic fraction (SOF) at 71%, followed by biodiesel (66%), CNG (38%) and diesel (20%). This illustrates that in case of both gaseous and liquid fuels, oxygenated fuels have a higher SOF than non-oxygenated fuels. Motor vehicle emissions are a major anthropogenic source of air pollution and contribute to the deterioration of urban air quality. In this paper, we report results of a laboratory investigation of particle formation from four different alternative diesel fuels, namely, compressed natural gas (CNG), dimethyl ether (DME), biodiesel, and diesel, under fuelrich conditions in the temperature range of 800-1200°C at pressures of approximately 24 atm. A single pulse shock tube was used to simulate compression ignition (CI) combustion conditions. Gaseous fuels (CNG and DME) were exposed premixed in air while liquid fuels (diesel and biodiesel) were injected using a high-pressure liquid injector. The results of surface analysis using a scanning electron microscope showed that the particles formed from combustion of all four of the above-mentioned fuels had a mean diameter less than 0.1 μm. From results of gravimetric analysis and fuel injection size it was found that under the test conditions described above the relative particulate yields from CNG, DME, biodiesel, and diesel were 0.30%, 0.026%, 0.52%, and 0.51%, respectively. Chemical analysis of particles showed that DME combustion particles had the highest soluble organic fraction (SOF) at 71%, followed by biodiesel (66%), CNG (38%) and diesel (20%). This illustrates that in case of both gaseous and liquid fuels, oxygenated fuels have a higher SOF than non-oxygenated fuels.

Full text of DOI:10.1016/S0045-6535(00)00242-3

Chemosphere 42 (2001) 681±690
Semi-volatile and particulate emissions from the combustion of
alternative diesel fuels
*
Sukh Sidhu , John Graham, Richard Striebich
Environmental Science and Engineering, University of Dayton Research Institute, 300 College Park, Dayton, OH 45469-0132, USA
Abstract
Motor vehicle emissions are a major anthropogenic source of air pollution and contribute to the deterioration of
urban air quality. In this paper, we report results of a laboratory investigation of particle formation from four di€erent
alternative diesel fuels, namely, compressed natural gas (CNG), dimethyl ether (DME), biodiesel, and diesel, under fuel-
rich conditions in the temperature range of 800±1200°C at pressures of approximately 24 atm. A single pulse shock tube
was used to simulate compression ignition (CI) combustion conditions. Gaseous fuels (CNG and DME) were exposed
premixed in air while liquid fuels (diesel and biodiesel) were injected using a high-pressure liquid injector. The results of
surface analysis using a scanning electron microscope showed that the particles formed from combustion of all four of
the above-mentioned fuels had a mean diameter less than 0:1 lm. From results of gravimetric analysis and fuel injection
size it was found that under the test conditions described above the relative particulate yields from CNG, DME,
biodiesel, and diesel were 0.30%. 0.026%, 0.52%, and 0.51%, respectively. Chemical analysis of particles showed that
DME combustion particles had the highest soluble organic fraction (SOF) at 71%, followed by biodiesel (66%), CNG
(
38%) and diesel (20%). This illustrates that in case of both gaseous and liquid fuels, oxygenated fuels have a higher
SOF than non-oxygenated fuels. Ó 2001 Elsevier Science Ltd. All rights reserved.
Keywords: Ignition delays; Biodiesel; Spray combustion; Re¯ected shock
1
. Introduction
particulate matter emissions are particularly noticeable
under high load conditions when the fuel-to-air ratio is
increased and local pockets of poor mixing result in a
pyrolytic reaction that initiates particulate matter for-
mation (Kageyama and Kinehara, 1982). It is estimated
Airborne particulate pollution has emerged as a
global problem. In time-series analyses the concentra-
tion of ®ne particles in the atmosphere has been shown
to correlate with illness and mortality (Dockery et al.,
1
1
that in the US diesel powered vehicles produce 10 g of
soot annually. The main constituents of diesel particles
are elemental (EC) and organic (OC) carbon. Diesel
particles contain several carcinogenic substances, such
as polycyclic aromatic hydrocfarbons, and thus pose
health risks for humans and animals (McClellan, 1986).
The submicron diesel particles also a€ect urban visibility
by dominating the absorption of solar light because of
their high EC to OC emission ratios (Hamilton and
Mans®eld, 1991). These small airborne particles with
diameters less than 2:5 lm have a high probability of
deposition deep in the respiratory tract and are likely
to trigger or exacerbate respiratory diseases (Rantanen
1
993). Motor vehicle emissions are a major anthropo-
genic source of air pollution. Although compression
ignited engines produce less total engine emissions than
spark ignited engines, they form large quantities of ®ne
particulate matter. Field studies in various urban areas
have indicated that diesel-powered vehicles are a major
source of suspended particulate matter (Venkataraman
and Friedlander, 1994; Harrison et al., 1997). These
*
Corresponding author. Fax: +1-937-229-2503.
E-mail address: sidhu@udri.udayton.edu (S. Sidhu).
0045-6535/01/$ - see front matter Ó 2001 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 4 5 - 6 5 3 5 ( 0 0 ) 0 0 2 4 2 - 3

6
82
S. Sidhu et al. / Chemosphere 42 (2001) 681±690
et al., 1993). It is also well documented that ®ne particles
that have a mass median diameter of 2:5 lm can initiate
lung tumors and result in mortality increases in human
population (Kado et al., 1986).
mance for any particular fuel. This procedure is expen-
sive, time consuming, and the comparison of emissions
is complicated by the changes made to the test engine to
accommodate each fuel. Consequently, there is a need
for a relatively simple test protocol to evaluate the
emissions from alternative diesel fuels. Thus, the pur-
pose of this study was to provide comparison of particles
formed from the combustion of conventional and al-
ternative diesel fuels under CI conditions using a simple
testing procedure. In this paper, we present results from
a study where we used a versatile single pulse shock tube
to investigate particle formation from four di€erent
diesel fuels, namely, compressed natural gas (CNG)
dimethyl ether (DME) biodiesel, and diesel. The shock
tube is ideally suited for this task because it can easily
generate the high pressure and temperature environment
of a CI engine, typically 750±1000°C and 20±70 atm
(Armstrong and Hartman, 1959). This allows fuels to be
exposed to combustion conditions representative of the
full range of temperatures and pressures that may be
experienced in a diesel engine.
Although the chemical makeup of the small particles
seems to be responsible for tumors, the identity of the
responsible mutagens is not known. Initially, bacterial-
based mutagenicity essays of particulate matter extracts
indicated that PAHs and nitro-PAHs were responsible
for nearly all the mutagenic activity (Tokiwa et al.,
1
981). However, recent ®ndings have emphasized that
perhaps the semi-polar oxy-PAHs are responsible for a
signi®cant fraction of the mutagenic activity (Durant
et al., 1998). In either case, the actual chemicals
responsible for a signi®cant fraction of the health e€ects
that has led the EPA to propose tough new standards
are not known. These EPA standards have led to
the consideration of mono-ester-diesel fuels and other
alternative diesel fuels as a passive control strategy for
reducing particulate matter emissions (McDonald et al.,
1
995; Purcell et al., 1996). Even though these alternative
fuels have demonstrated decreased total hydrocarbon
emissions, the level of SOF associated with total
particulate matter has increased (Schmidt and Van
Gerpen, 1996). This increase in SOF raises concern
about the overall toxicity of particles formed during the
combustion of alternative diesel fuels.
2. Experimental
All experiments were conducted with a single pulse,
re¯ected-shock tube. The UDRI shock tube, shown in
Fig. 1, includes provisions for studying light fuels as
premixed vapors, or injecting low vapor pressure fuels as
liquids. The shock tube itself is comprised of a 7:6 cm IDÂ
2:74 m long driver section, a 5:08 cm ID Â 2:75 m long
driven section, and a 5:08 cm ID Â 0:9 m long test sec-
tion. The system also includes a 30:5 cm ID Â 61 cm long
dump tank and an evacuation subsystem. The driven and
test sections are connected through a pneumatic ball
The development of alternative diesel fuels is being
driven by the need to reduce the environmental impact
of emissions and not engine performance. However, the
conventional protocol for evaluating new fuels is to
sample the exhaust of a test engine running with the new
fuel. Modi®cations are usually required to the test en-
gines (fuel injection, timing, air/fuel ratio, etc.) for each
fuel and may not necessarily represent optimal perfor-
Fig. 1. General schematic of the UDRI shock tube con®gured for particulate and exhaust gas sampling.

S. Sidhu et al. / Chemosphere 42 (2001) 681±690
683
valve controlled through the systemÕs automatic digital
ring system. The dump tank is connected to the system
for 24 h to provide a solvent blank. Once the solvent
blank was obtained each of the ®lters were extracted in
turn. Prior to being placed in the apparatus, each ®lter
was spiked with 230 ng of the internal standard, then
extracted for 24 h. Following each extraction, the sol-
vent was removed from the system and stored in glass
volatile sample vials. For the GC/FID/MS analysis, 2 ml
of each sample was applied to a special quartz probe and
the solvent allowed to evaporate at room temperature.
This probe was then placed into the inlet of the ana-
lytical system and thermally desorbed at 300°C for
15 min.
®
through a manually operated ball valve. The entire
structure is fabricated from 1.27 cm thick 304 stainless
steel. While the nominal test conditions are set by the
initial state of the system, the actual test conditions are
measured from the shock velocity through the test gas.
This value is measured by two piezoelectric pressure
transducers mounted in the side wall of the test section.
The shock velocity is taken as the distance between the
sensors divided by the arrival time of the shock past them
as measured by a single-sweep digital oscilloscope. The
exposure conditions behind the re¯ected shock are then
calculated using a model on an algorithm developed by
Gardiner et al. and data from the JANAF tables (Gard-
iner et al., 1985).
The thermal desorption was performed using the
System for Thermal Diagnostic Studies (STDS), de-
scribed in detail elsewhere (Striebich and Rubey, 1990).
The system was ®tted with a thermal desorption inlet
consisting of a 0.25 in OD stainless steel tube. For such
analysis, the GFF was sealed in this inlet and a con-
trolled ¯ow of helium was introduced while the tube was
heated in a GC oven to 300°C. E‚uent from the tube
was directed into a second GC oven which was held at
cryogenic temperature ()60°C) to trap the e‚uent at the
head of a capillary chromatographic column (DB-5,
30 m, 0:25 lm J&W Scienti®c). After the thermal de-
sorption was complete, the helium ¯ow was increased to
a ¯ow capable of chromatographic separation of the
e‚uent components. At that point, temperature pro-
grammed GC±MS was used for semiquantitative anal-
ysis of the products eluted from the thermal exposure.
Filters and traps were each treated with the same ther-
mal program (30±300°C at rate of 20°C/min). Quanti-
tation was conducted by spiking each of the ®lter strips
or traps with a known amount of d10-xylene or d10-an-
thracene, respectively. Semiquantitative information
was obtained by direct comparison of the internal
standard to the unknown solute using a unity response
factor for each compound.
Light gaseous fuels (CNG and DME) were prepared
as premixed gases in air within the shock tube's test
section with the pneumatic isolation valve closed. This
valve was opened by the automatic ®re control system
just prior to rupturing the diaphragm and closed ap-
proximately 0.5 s afterwards, sealing the combustion
products in the test section. Liquid fuels (diesel and
biodiesel) were introduced using a Bosch diesel fuel in-
jector ®tted with a single stroke positive displacement
fuel pump. The pump delivers a small volume ꢀ50 ll† of
fuel in a precisely timed manner and in a narrow spray
pattern that does not impinge on the walls of the tube.
The fuel is introduced 500 ls behind the re¯ected shock
which allows for the vibrational relaxation of the bath
gas prior to injecting the fuel. As with the gaseous fuels,
the ®re control system closes the test section isolation
valve approximately 0.5 s afterwards, sealing the com-
bustion products in the test section.
For the collection of combustion products, the test
section was ®tted with two high volume particulate
samplers in series and 20 liter Tedlar sample bags as
shown in Fig. 1. Immediately following each test, the
isolation valve was closed, the exhaust valve to the
sampling system opened, and the test section purged
with 20 l (10 change-volumes) of dry helium. As the gas
was swept from the system, the high volume samplers
collect the particulate fraction on glass ®ber ®lters
Comparison of ®lter Soxhlet extraction results with
the thermal desorption results, shows the same basic set
of products plus several signi®cant contaminants intro-
duced by the Soxhlet procedure. This ®nding is consis-
tent with that reported by Newkirk (1992) who also
compared Soxhlet extraction, direct thermal desorption,
and vacuum oven desorption. Given the speed, sim-
plicity, and sensitivity of the thermal desorption proce-
dure, this is the perfect method for these samples.
(
GFF), while the volatile gases were captured in the
Tedlar bag. The bag was then removed and evacuated
through a sorbent trap packed with 60/80 mesh Tenax
TA, thereby collecting the semi-volatile organic fraction
of the exhaust. The GFFs were weighed before and after
each test to gravimetrically measure the mass of partic-
ulate formed.
3. Results
For chemical analysis, each ®lter was divided in half:
one portion for thermal desorption and the other for
Soxhlet extraction. The Soxhlet system was washed with
solvent (methylene chloride, Aldrich capillary GC grade,
The results from all four fuels are reported below. In
each experiment, the particulate matter was collected on
glass ®ber ®lters and the gas-phase semivolatiles were
captured by an adsorptive trap. Bulk analysis was per-
formed by scanning electron microscopy and total car-
bon analysis. For chemical analysis of glass ®ber ®lters,
99.9+%), charged with 80 ml of fresh solvent plus 230 ng
of d10-anthracene as an internal standard, and operated

6
84
S. Sidhu et al. / Chemosphere 42 (2001) 681±690
(a)
(b)
Fig. 2. GC±MS traces of products obtained from high pressure combustion of CNG at T ˆ 880°C, p ˆ 24 bar, and / ˆ 3. (a) The
chromatogram shows the products captured by the adsorption trap and (b) the chromatogram shows the products desorbed from
particles collected on the glass ®lter.
thermal desorption and Soxhlet extraction with meth-
ylene chloride were used. However, we only show ther-
mal desorption results because they were cleaner and
had greater sensitivity than Soxhlet extraction. The
adsorption traps were also analyzed using thermal
desorption technique. The thermal desorption chroma-
tograms for both traps and ®lters are shown for the
CNG, DME, diesel and biodiesel fuels in Figs. 2±5,
respectively.
results of this series of tests a nominal set of conditions
was selected for CNG that gave a mean exposure tem-
perature of 880°C at 24 atm with an / of 3. These ex-
perimental conditions are similar to fuel/air ratio and
temperature and pressure conditions experienced by fuel
during a diesel engine cycle. These exposure conditions
resulted in good ignition of the test sample with an ig-
nition delay of 510 ls. Under these nominal test con-
ditions, the average mass collected from three tests was
3
90 lg of particulate with a relative standard deviation
3
.1. CNG
of 24%. The relatively high variance is attributed to the
small amount collected from each test (0.28% of the
®lterÕs mass). Assuming the collection of particulate is
100% ecient and given that the test mixture includes
128 mg of CNG, the relative particulate yield will be
0.30%, within the range of 0.25±0.61% obtained during
propanol combustion under similar conditions (Wang
and Cadman, 1998).
To establish an appropriate set of exposure condi-
tions to generate particulate from CNG a series of tests
were performed representing a range of pressures, tem-
peratures, and equivalence ratios ꢀ/†. The temperature
was varied over a range of ꢁ700±1200°C while the
pressure was varied from 20 to 27 atm. Based on the

S. Sidhu et al. / Chemosphere 42 (2001) 681±690
685
(a)
(b)
Fig. 3. GC±MS traces of products obtained from high pressure combustion of DME at T ˆ 840°C, p ˆ 26 bar, and / ˆ 4. (a) The
chromatogram shows the products captured by the adsorption trap and (b) the chromatogram shows the products desorbed from
particles collected on the glass ®lter.
The particle size of the collected material was ob-
tained using a scanning electron microscope. Using a
magni®cation of 10,000Â this analysis showed the
particulate matter consisted of a ¯eece-like material
composed of smaller components estimated to be less
than 0:1 lm in mean diameter. This is in general
agreement with the diesel emission data (Morawska
et al., 1998). A general carbon measurement was con-
ducted on the collected particulate with a Leco carbon
analyzer. The results from this test indicated that a
major fraction (70%) of the collected material was high
molecular weight, mainly elemental carbon. This result
is also consistent with the diesel emission data where the
non-elemental carbon fraction or solvent extractable
organic components of diesel aerosol represent 5±40%
of particulate mass, depending on the fuel and operat-
ing conditions (Pierson and Brachaczek, 1983; Lares-
goiti et al., 1997; Morawska et al., 1998). This analysis
also shows that the relatively lighter organic fraction
could be thermally desorbed at temperatures below
300°C.
The chemical analysis of glass ®ber ®lters and ab-
sorption traps gives a complete picture of the forma-
tion of high molecular weight carbonaceous products
from CNG (see Fig. 2). The adsorbent trap contained
soot precursors from the single ring compounds
(benzene, toluene, styrene, etc.) to the ®rst two and
three ring structures (indene and acenaphthalene),
whereas, the ®rst four ring structures (pyrene and
¯uoranthene) were observed from the particulate ma-
terial on through ®ve and six ring structures (benzo-
pyrene and indenopyrene). We have also found
additional oxygenated materials (methyl furan and
benzaldehyde) in the gas-phase sample, which may
assist in identifying the oxygenated PAHs which we
believe are present in the particulate sample, but have
been dicult to identify based solely on their mass
spectra.

6
86
S. Sidhu et al. / Chemosphere 42 (2001) 681±690
(a)
(b)
Fig. 4. GC±MS traces of products obtained from high pressure combustion of diesel at T ˆ 1130°C and p ˆ 24 bar. (a) The chro-
matogram shows the products captured by the adsorption trap and (b) the chromatogram shows the products desorbed from particles
collected on the glass ®lter.
3
.2. DME
products with small amounts of methane, ethane, and
ethylene as light organic products of incomplete
combustion. This led us to increase the equivalence
ratio to obtain a signi®cant increase in the yield of
products of incomplete combustion and thus soot
formation.
An experimental procedure similar to CNG was
followed to establish an appropriate set of exposure
conditions to generate particulate from DME com-
bustion. The temperature was varied over a range of
ꢁ
640±1200°C while the pressure was varied from 17±
In contrast to the particles collected from CNG,
which were dark gray to black, the particles from DME
were tan at temperatures below 1000°C turning to gray
at temperatures above 1000°C. The tan color of the
DME particulate indicated a lower elemental carbon
(higher organic) content than the CNG particulate.
However, as the temperature was increased, the ele-
mental carbon content in particulate increased as indi-
cated by the color change from tan to gray. Based on the
results of this series of tests, a nominal set of test con-
ditions was selected for DME that gave a mean exposure
temperature of 840°C at 26 atm with an / of 4. This
compares well with the exposure conditions used for
3
2 atm and / from 1±7. The pressure transducer data
showed strong ignition of the DME sample under all
of the exposure conditions with ignition delays varying
from 23 ls with an / of 2 at 1220°C and 17.2 atm to
2
53 ls with an / of 2 at 832°C and 26.8 atm. This
contrasts with the results for CNG, which was rela-
tively dicult to ignite via compression. The light gas
analysis of the exhaust gas also showed that the DME
mixtures were readily ignited under all of the exposure
conditions. It was observed that at / < 3 there is
complete conversion of the DME with only carbon
monoxide and carbon dioxide as the major light

S. Sidhu et al. / Chemosphere 42 (2001) 681±690
687
(a)
(b)
Fig. 5. GC±MS traces of products obtained from high pressure combustion of biodiesel at T ˆ 1130°C and p ˆ 24 bar. (a) The
chromatogram shows the products captured by the adsorption trap and (b) the chromatogram shows the products desorbed from
particles collected on the glass ®lter.
CNG (/ of 3, 880°C, and 24 atm). The nominal expo-
sure conditions described above gave a mean ignition
delay time of 152 ls with a relative standard deviation
of 2.8%.
so light that the particulate material lies below the
surface ®bers of the ®lter where it cannot be readily
imaged.
The results of the overall carbon analysis of the
collected DME particulate matter showed that under
similar combustion conditions, DME produced far less
carbonaceous material than CNG. Speci®cally, the car-
bon analysis indicated that approximately 70% of the
collected material from CNG was high molecular weight
carbonaceous material, while for DME it was on the
order of 30%.
The particulate loading on the high volume samplers
from DME combustion was small in comparison to
CNG experiments. Due to this, samples collected for
analysis were accumulated from three complete cycles
of the shock tube rather than from a single test as was
the case for the CNG analysis. Averaging the mass
collected from three tests (three shots per test) using the
nominal test conditions showed 170 lg of particulate
per test series, or 57 lg per individual test. Assuming
the collection of particulate is 100% ecient, and given
that the each test mixture includes 223 mg of DME, this
gives a relative particulate yield of only 0.025%. Elec-
tron microscopy of the DME ®lters with a magni®ca-
tion of 10,000Â showed that rather than the particulate
matter captured by the ®ber matrix, small particles on
the order of 0:1 lm in diameter are attached to the
individual ®bers. This suggests that the collected ma-
terial is condensed on the ®lter media or the loading is
The chemical analysis of the DME ®lter and adsor-
bent trap shows a striking similarity to CNG samples
(cf. Figs. 2 and 3). The adsorbent trap captured benzene,
toluene, styrene, up to indene and acenaphthalene in the
gas-phase sample. The ®lter collected four ring struc-
tures (pyrene and ¯uoranthene) on through ®ve and six
ring structures (benzopyrene and indenopyrene). There
is a slight shift in retention times of all the products in
the second chromatogram of Fig. 3 when compared to
Fig. 2 and this is due to a change in initial hold time in
DME experiments.

6
88
S. Sidhu et al. / Chemosphere 42 (2001) 681±690
the temperature over a range of ꢁ1100±1750°C while the
3
.3. Diesel
pressure was varied from 16 to 25 atm. Under these
experimental conditions, the ignition delays varied from
1150 ls at 1730°C and 16 atm to 1720 ls at 1130°C and
24.4 atm. These ignition delays are longer than those
observed in diesel combustion, which illustrates that
biodiesel is dicult to combust. Based on the results of
this series of tests, a nominal set of test conditions was
selected for biodiesel that gave a mean exposure tem-
perature of 1130°C at 24.4 atm. This temperature is
similar to that selected for diesel combustion. The
nominal exposure conditions described above gave a
mean ignition delay time of 1720 ls. The exhaust ®lter
visually and gravimetrically also showed a signi®cant
amount of particulate under the selected nominal con-
ditions. Below these conditions, the particulate ®lter was
mainly loaded with unburnt biodiesel fuel.
For commercial diesel experiments, we varied the
temperature over a range of ꢁ850±1400°C while the
pressure was varied from 21±28 atm. As these are liquid
spray combustion experiments, it is not possible to have
a desired equivalence ratio at all points in the test sec-
tion. Liquid spray combusts at locally fuel rich condi-
tions but with presence of excess surrounding air overall
combustion conditions are lean. The ignition delays
varied from 320 ls at 1340°C and 21 atm to 2110 ls at
8
67°C and 27.7 atm under the above-mentioned exper-
imental conditions. Based on the results of this series of
tests, a nominal set of test conditions was selected for
diesel that gave a mean exposure temperature of 1130°C
at 24 atm. This temperature is higher than those selected
for CNG and DME combustion and is also higher than
the temperatures in diesel engine combustion chamber.
We had to select these test conditions because in our
present shock tube con®guration we have only 7 ms time
window for particulate formation. Under the selected
nominal conditions, the exhaust ®lter also showed, vi-
sually and gravimetrically, a signi®cant amount of par-
ticulate. Below these conditions, the particulate ®lter
was mainly loaded with unburnt diesel fuel.
Under these nominal test conditions, the particulate
mass collected was 300 lg. A relative particulate yield of
0.52% was obtained, assuming the collection of partic-
ulate was 100% ecient and that 40 ll of biodiesel was
injected. This particulate yield was almost identical to
that obtained during the combustion of diesel. In case of
biodiesel, the surface analysis using a scanning electron
microscope with a magni®cation of 10,000Â also shows
that the particulate matter consists of a ¯eece-like ma-
terial composed of smaller components estimated to be
less than 0:1 lm in mean diameter. This is in general
agreement with other diesel fuel data. However, the re-
sults from Leco carbon analyzer suggests that a minor
fraction (34%) of the collected material is high molecular
weight material, mainly elemental carbon. This elemen-
tal carbon fraction is lower than that obtained in diesel
combustion.
Under these nominal test conditions, the particulate
mass collected was 300 lg. A relative particulate yield of
.51% was obtained, assuming the collection of partic-
0
ulate was 100% ecient and that 70 ll of diesel was
injected. The surface analysis using a scanning electron
microscope with a magni®cation of 10,000Â showed
that the particulate matter consists of a ¯eece-like ma-
terial composed of smaller components estimated to be
less than 0:1 lm in mean diameter. This is in general
agreement with the diesel emission data. The results
from the Leco carbon analyzer suggests that a major
fraction (80%) of the collected material is high molecular
weight material, mainly elemental carbon.
The chemical analysis of absorption traps (c.f. Fig. 5)
obtained from biodiesel combustion shows single ring
compounds like benzene, toluene, xylene, styrene, phe-
nol along with saturated and unsaturated aliphatic
compounds like hexene, heptane, nonene to decane and
n-C13. The biodiesel adsorption trap also contains oxy-
genated products like benzoic acid, pentenoic acid and
various substituted methyl esters. The adsorption trap
from biodiesel combustion contains less aliphatic prod-
ucts than the adsorption traps from diesel combustion.
The chemical analysis of ®lters obtained from biodiesel
combustion show oxygenated long chain compounds
(c.f. Fig. 5) and some unburnt fuel. Like diesel com-
bustion, PAHs were observed in small amounts; how-
ever, the PAH yields increased with temperature in
contrast to diesel combustion.
The chemical analysis of absorption traps (cf. Fig. 4)
obtained from diesel combustion shows single ring
compounds like benzene, toluene, xylene, styrene, phe-
nol to two ring structures naphthalene along with sat-
urated and unsaturated aliphatic compounds like
hexene, octane, octene to dodecane and n-C17. The
chemical analysis of ®lters obtained from diesel com-
bustion showed oxygenated long chain compounds (c.f.
Fig. 4) and very small amounts of PAHs. In fact, the
PAHs were observed only in ion chromatograms. This
small amount of ®lter products can be attributed to the
large elemental carbon fraction observed during total
carbon analysis.
3
.4. Biodiesel
4. Discussion
To establish an appropriate set of exposure condi-
tions to generate particulate from biodiesel, we varied
In this study we compared particulate formation
from combustion of oxygenated and non-oxygenated

S. Sidhu et al. / Chemosphere 42 (2001) 681±690
689
fuels in both premixed gas-phase and liquid (spray)-
phase. The gaseous fuels (CNG and DME) were
combusted at ꢁ860°C versus 1130°C for liquid fuels
particulate occurs in the gas-phase prior to their arrival
at the ®rst ®lter. From gravimetric analysis results and
fuel injection size, it was found that under the test
conditions described above, the relative particulate
yields from CNG, DME, biodiesel, and diesel were
0.30%, 0.026%, 0.52%, and 0.51%, respectively.
(
diesel and biodiesel) at a pressure of ꢁ24 atm in all
experiments. We had to select higher temperatures for
liquid fuel combustion because in our present shock
tube con®guration we have only 7 ms time window for
particulate formation and the liquid fuels required
higher temperatures to form particles. The gaseous fuels
could not be combusted at higher temperature, as it
would have resulted in complete combustion in case of
DME and thus no particle formation would have taken
place.
On visual inspection, all ®lters appeared to be loaded
with greyish black particles, except for DME where the
particles appeared to be brown in color. The combustion
oxygenate fuels generated particles with higher soluble
organic fraction with DME at 71% followed by biodiesel
(66%), and then non-oxygenate fuels CNG (38%) and
diesel (20%). This is consistent with results of other
studies where the biodiesel particles had higher SOF
than diesel which was in the range 5±30% of particulate
mass depending on operating conditions (McDonald
et al., 1995; Purcell et al., 1996; Schmidt and Van
Gerpen, 1996; Armstrong and Hartman, 1959).
In gas phase experiment, DME was the easier fuel to
burn than CNG, with ignition delay of 152 ls at 840°C
that is factor of three smaller than ignition delay of
CNG under similar conditions. This is in contrast to
liquid fuel experiments in which oxygenate fuel (bio-
diesel) was more dicult fuel to burn with ignition delay
of 1720 ls at 1130°C that is approximately factor of two
larger than ignition delay of diesel under similar condi-
tions. These ignition delay times from liquid fuel ex-
periments at ®rst appear to be reversed because biodiesel
The chemical analysis of soluble organic fraction
and adsorbent trap also indicated di€erences in
chemical composition of SOF and gas-phase semi-
volatiles (adsorbent trap) obtained during the com-
bustion of gaseous and liquid fuels. The combustion of
gaseous fuels showed an orderly progression from the
single ring compounds (benzene, toluene, styrene, etc.)
to the ®rst two-and three-ring structures (indene and
acenaphthalene) in the gas-phase sample (adsorption
trap) to the ®rst four ring structures (pyrene and ¯u-
oranthene) observed from the particulate material on
through ®ve and six ring structures (benzopyrene and
indenopyrene). However, in the case of liquid fuels,
this orderly progression was not as clear. The amount
of two- and three-ringed PAH was much smaller than
that observed during combustion of gaseous fuels.
This could perhaps be due to the fact that under the
combustion conditions investigated in this study, the
reaction time available to the liquid fuels was not long
enough to combust completely as observed by Cad-
man and co-workers in similar experiments (Williams
and Cadman, 1999). The adsorbent traps from liquid
fuels contained larger amounts of aliphatic and oxy-
genated products than those observed from adsorbent
traps of gaseous fuels and the chromatograms ob-
tained were of a highly complex nature. This is
probably due to the complex fuel composition of
diesel and biodiesel in comparison to CNG and DME.
For complete speciation, selected samples should be
subjected to multidimensional gas chromatographic-
mass spectrometric (MDGC±MS) analysis. The
MDGC±MS system utilizes a sequential heart-cutting
technique that should provide further chromatographic
separation and thus identi®cation of previously unre-
solved analytes. This detailed speciation of particulate
matter samples should help provide more information
concerning the health e€ects of these four alternative
diesel fuels.
(
Cetane number: 46±52) has higher Cetane number than
diesel (Cetane number: 40±45) and it is known that
higher Cetane number results in shorter ignition delay
times (Spataru and Romig, 1995). However, biodiesel
(
viscosity @ 40°C, 4.3 cSt) is also almost twice as viscous
as diesel (viscosity @ 40°C, 2.8 cSt) and this leads to
larger biodiesel spray droplets and that results in longer
ignition delays. This is consistent with results of recent
study that showed that as the amount of biodiesel in
biodisel/diesel mix passed 50%, the increase in Cetane
number did not lead to shorter ignition delays but ig-
nition delays increased because of increase in viscosity
(
Zhang and Van Gerpen, 1996). These problems with
Cetane number and ignition delays in alternative fuels
have also been investigated by Hardenberg and Ehnert
(
Hardenberg and Ehnert, 1981). Our results are also
validated by a single-cylinder engine study that showed
that biodiesel had longer ignition delays than diesel at
relatively low powers and speeds and that the di€erence
became less at higher loads and speeds (Crookes et al.,
1
997).
After the particles were formed during combustion of
all four fuels, they were collected and subjected to
gravimetric, surface, total carbon, and chemical analy-
sis. The surface analysis showed that the particle size
was less than 0:1 lm (mean diameter) for both liquid
and gaseous fuels. This observation was in general
agreement with the diesel emission data, which showed
that diesel exhaust particles are mostly submicrometer
agglomerates of carbonaceous particles ranging from 10
to 80 nm (Morawska et al., 1998). The observation that
no particulate matter is collected on the second GFF in
the sample train indicates that agglomeration of ®ne

6
90
S. Sidhu et al. / Chemosphere 42 (2001) 681±690
Dockery, D.W., Pope, A., Xu, X., Spengler, J.D., Ware, J.H.,
5
. Conclusions
Fay, M.E., Ferris, B.G., Speizer, F.E., 1993. N. Engl. J.
Med. 329, 1753±1759.
Based on the results of this study, we can conclude
the following:
Durant, J.L., La¯eur, A.L., Plummer, E.F., Taghizadeh, K.,
Busby Jr., W.F., Thilly, W.G., 1998. Environ. Sci. Technol.
1
. In gaseous fuel experiments, oxygenate fuel (DME)
had lower particulate yields than the non-oxygenate
fuel (CNG). In liquid fuel experiments, both oxy-
genate and non-oxygenate fuel had similar particu-
late yields this could perhaps be due to large
unburnt fuel fraction in both diesel and biodiesel ex-
periments.
32 (13), 1894±1906.
Gardiner Jr., W.C., Walker, B.F., Wake®eld, C.B., 1985. In:
Lifshitzs A., (Ed.), Shockwaves in Chemistry. Marcel-
Dekker, NY (Chapter 7).
Hamilton, R.S., Mans®eld, T.A., 1991. Atmos. Environ. 25A,
715±723.
Hardenberg, H.O., Ehnert, E.R., 1981. SAE Technical Paper
No. 811212, Warrendale, PA.
2
3
. DME combustion particles has the highest soluble
organic fraction followed by biodiesel. This suggests
that oxygenated fuels have higher SOF than non-ox-
ygenated fuels.
Harrison, R.M., Deacon, A.R., Jones, M.R., 1997. Atmos.
Environ. 31, 4103±4117.
Kado, N.Y., Guirguis, G.N., Flessel, C.P., Chan, R.C., Chang,
X.-I., Wesolowski, J.J., 1986. Environ. Mutagenesis 8 (53),
. The combustion of gaseous fuels showed an orderly
progression from the single ring compounds (ben-
zene, toluene, styrene, etc.) to the ®rst two- and-
three-ring structures (indene and acenaphthalene) in
the gas-phase sample (adsorption trap) to the ®rst
four ring structures (pyrene and ¯uoranthene) ob-
served from the particulate material on through ®ve
and six ring structures (benzopyrene and indenopy-
rene).
195.
Kageyama, K., Kinehara, N., 1982. SAE Paper No. 820181.
Laresgoiti, A., Loos, A.C., Springer, A.S., 1997. Environ. Sci.
Technol. 11, 973.
McClellan, R., 1986. Am. Ind. Hyg. Assoc. 47, 1±13.
McDonald, J.F., Purcell, D.L., McClure, B.T., Kittelson, D.B.,
1
995. SAE Technical Paper No. 950400, Warrendale, PA.
Morawska, L., Bo®nger, N.D., Kocis, L., Nwankwoala, A.,
998. Environ. Sci. Technol. 32, 2033.
Newkirk, M.S., 1992. J. Engrg. for Gas Turbines and Power
14, 578±589.
Pierson, W.R., Brachaczek, W.W., 1983. Aerosol Sci. Technol.
, 1.
Purcell, D.L., McClure, B.T., McDonald, J., Bausu, H.N.,
996. Am. Oil Chem. Soc. 73, 381.
1
4
. Despite experiencing vastly di€erent combustion con-
ditions, the surface analysis showed that both liquid
and gaseous fuels generated particles with mean di-
ameter of less than 0:1 lm.
1
2
1
Rantanen, L., Mikkonen, S., Nylund, L., Kociba, P., Lappi,
M., Nylund, N.-O., 1993. SAE Paper No. 932686.
Schmidt, K., Van Gerpen, J., 1996. SAE Technical Paper No.
Acknowledgements
961086, Warrendale, PA.
The partial support of the National Renewable En-
ergy Laboratory (Golden, CO), Dr. Douglas Lawson,
Project Ocer, is gratefully acknowledged.
Spataru, A., Romig, C., 1995. SAE Technical Paper No.
952388, Warrendale, PA.
Striebich, R.C., Rubey, W.A., 1990. American Lab., January.
Tokiwa, H., Nakagawa, R., Morita, K., Onishi, Y., 1981.
Mutation Res. 85, 195.
Venkataraman, C., Friedlander, S.K., 1994. Environ. Sci.
Technol. 28, 563±572.
References
Wang, R., Cadman, P., 1998. Combust. Flame 112, 359.
Williams, R.L., Cadman, P., 1999. In: Proceedings of the 22nd
International Symposium on Shock Waves (paper 3330).
Imperial College, London, UK, 18±23 July.
Armstrong, L.V., Hartman, J.B., 1959. The Diesel Engine: its
Theory, Basic Design, and economics. Macmillan, New
York.
Crookes, R.J., Kainnejad, F., Nazha, M., 1997. Energy
Convers. Mgmt. 38, 1785.
Zhang, Y., Van Gerpen, J.H., 1996. SAE Technical Paper No.
960765, Warrendale, PA.