Available online at www.sciencedirect.

com
Proceedings
of the
Combustion
Institute
Proceedings of the Combustion Institute 34 (2013) 2749–2757
www.elsevier.com/locate/proci

Crude glycerol combustion: Particulate, acrolein,

and other volatile organic emissions
Scott A. Steinmetz a,1, Jason S. Herrington b,2,
Christopher K. Winterrowd c, William L. Roberts a,1, Jost O.L. Wendt d,
William P. Linak b,⇑
a
Department of Mechanical and Aerospace Engineering, North Carolina State University, Raleigh, NC 27695, USA
b
National Risk Management Research Laboratory, U.S. Environmental Protection Agency, Research Triangle Park,
NC 27711, USA
c
ARCADIS U.S., Inc., Durham, NC 27713, USA
d
Department of Chemical Engineering, University of Utah, Salt Lake City, UT 84112, USA

Available online 29 August 2012

Abstract

Crude glycerol is an abundant by-product of biodiesel production. As volumes of this potential waste
grow, there is increasing interest in developing new value added uses. One possible use, as a boiler fuel for
process heating, offers added advantages of energy integration and fossil fuel substitution. However, chal-
lenges to the use of crude glycerol as a boiler fuel include its low energy density, high viscosity, and high
autoignition temperature. We have previously shown that a refractory-lined, high swirl burner can over-
come challenges related to flame ignition and stability. However, critical issues related to ash behavior
and the possible formation of acrolein remained. The work presented here indicates that the presence of
dissolved catalysts used during the esterification and transesterification processes results in extremely large
amounts of inorganic species in the crude glycerol. For the fuels examined here, the result is a submicron fly
ash comprised primarily of sodium carbonates, phosphates, and sulfates. These particles report to a well-
developed accumulation mode (0.3–0.7 lm diameter), indicating extensive ash vaporization and particle
formation via nucleation, condensation, and coagulation. Particle mass emissions were between 2 and
4 g/m3. These results indicate that glycerol containing soluble catalyst is not suitable as a boiler fuel. For-
tunately, process improvements are currently addressing this issue. Additionally, acrolein is of concern due
to its toxicity, and is known to be formed from the low temperature thermal decomposition of glycerol.
Currently, there is no known reliable method for measuring acrolein in sources. Acrolein and emissions
of other volatile organic compounds were characterized through the use of a SUMMA canister-based

⇑ Corresponding author. Address: National Risk

Management Research Laboratory, U.S. Environmental
Protection Agency, E305-01, Research Triangle Park,
NC 27711, USA. Fax: +1 919 541 0554.
E-mail address: linak.bill@epa.gov (W.P. Linak).
1
Current address: Clean Combustion Research Cen-
ter, King Abdullah University of Science and Technol-
ogy (KAUST), Thuwal 23955, Saudi Arabia.
2
Current address: Restek Corporation, Bellefonte, PA
16823, USA.

1540-7489/$ - see front matter Published by Elsevier Inc. on behalf of The Combustion Institute.
http://dx.doi.org/10.1016/j.proci.2012.07.050
2750 S.A. Steinmetz et al. / Proceedings of the Combustion Institute 34 (2013) 2749–2757

sampling method followed by GC–MS analysis designed for ambient measurements. Results indicate crude
glycerol combustion produces relatively small amounts of acrolein (15 ppbv) and other volatile organic
compounds, with emissions comparable to those from natural gas combustion.
Published by Elsevier Inc. on behalf of The Combustion Institute.

Keywords: Glycerol combustion; Emission characterization; Bio-fuels; Acrolein measurements; Waste fuels

1. Introduction phosphorous (P) from the catalysts, currently

Glycerol is a major by-product of the transe-
sterification of triglycerides (bio-oils) to produce
biodiesel, and is currently the focus of research
seeking value-added uses for this potentially large
volume waste. Transesterification involves the
reaction of triglycerides (fats and oils) with an
alcohol (typically methanol, in excess) and a base
catalyst (often sodium hydroxide) to form fatty
acid methyl esters (FAME) and glycerol. The pro-
cess may also include the initial use of an acid cat-
alyst (often phosphoric or sulfuric acid) for
esterification of any free fatty acids (FFA), or
addition of an acid after transesterification to neu-
tralize the crude glycerol. Approximately 1 kg of
glycerol is formed per 10 kg of FAME. Once
formed, the FAME product is separated by polar-
ity from the glycerol by-product which also con-
tains unreacted mono-, di-, and triglycerides and
soaps (collectively classified as matiere organique
non-glycerol, MONG), excess alcohol, inorganic
elements from the original bio-material, catalyst,
and water. Excess alcohol is typically recovered
by distilling the “methylated” crude glycerol to
produce “demethylated” crude glycerol. The final
composition of the crude glycerol can vary greatly
depending on the bio-oil feedstock, catalyst, and
alcohol used. Methylated glycerol may contain
10–20% methanol and 5–10% salts, while deme-
thylated glycerol may contain <1% methanol
and 5–15% salts [1]. Crude glycerol can be purified
and utilized in soap, cosmetic, and food and bev-
erage industries at significant cost. However, it is
often treated as waste or stored.
In our previous study [2,3], we examined the
feasibility of using crude glycerol as a boiler fuel
to provide process steam and co-generate electric-
ity. If combined with biodiesel production, this
has the advantages of optimizing energy integra-
tion, eliminating transportation costs, displacing
fossil fuels, and reducing greenhouse gas emis-
sions. We have shown that despite its high viscos-
ity (1 Pa s at 298 K), high autoignition
temperature (643 K), and relatively low heating
value (15–27 MJ/kg), glycerol preheating, proper
atomization, and a high swirl refractory-lined bur-
ner can be used to produce stable 100% glycerol
flames [2]. Unfortunately, this previous study also
identified that high concentrations of residual
inorganic species, primarily sodium (Na) and
makes combustion of crude glycerol in boilers
unattractive. While we previously reported partic-
ulate emissions ranging from 2200 to 3400 mg/m3,
new methods of biodiesel production that utilize
fixed catalyst beds, rather than dissolved catalyst,
could greatly reduce or eliminate the inorganic
constituents. There is also potential to co-fire
crude glycerol in black liquor boilers where the
Na could be recovered and used for the kraft pul-
ping process or in other industrial applications
where high levels of ash are not an issue [4].
Another major concern of glycerol combustion
is the possibility of carbonyl formation and emis-
sion. Acrolein (propenal) is of specific concern, as
glycerol (propane-1,2,3-triol) is known to ther-
mally decompose into acrolein at 563 K [5]. Acro-
lein, the simplest unsaturated aldehyde, has an
acrid smell and is highly toxic. It can cause eye
or throat irritation in concentrations under
0.1 ppmv, and can be fatal in high ppmv concen-
trations [6]. Acrolein is also very unstable, and
unlikely to survive in a combustion environment.
However, determining the presence of acrolein in
furnace stack gases is not trivial, and a reliable
method of doing so is not readily available. Previ-
ously, we used commercially available 2,4-dinitro-
phenylhydrazine (DNPH) coated solid sorbent
cartridges to characterize the presence of selected
carbonyls, and found no evidence of acrolein
above concentrations of 17.5 ppbv [2]. For ambi-
ent samples, DNPH coated solid sorbent car-
tridges and aqueous DNPH solutions in
impingers are common methods of collecting alde-
hydes [7]. Carbonyls present in the sample gas
react with the DNPH to form hydrazone deriva-
tives, which are then extracted, separated by high
performance liquid chromatography, and ana-
lyzed by ultraviolet spectroscopy. Although this
approach and other similar methods have been
used for decades to characterize ambient concen-
trations, it has more recently been shown that
these methods can be unreliable when sampling
unsaturated aldehydes from source emissions [8–
10], and alternative methods are under develop-
ment [11–14]. The presence of nitrogen oxides
has been shown to affect sample recoveries [8].
Additionally, the acrolein–hydrazone derivatives
are unstable, and may further derivatize in the
presence of acidic DNPH [9]. Despite this, DNPH
methods continue to be used, often due to a lack
 S.A. Steinmetz et al. / Proceedings of the Combustion Institute 34 (2013) 2749–2757
of knowledge of these issues or the availability of
alternative approaches.
The objectives of the research presented in this
paper are to better characterize two important
issues previously identified: (1) the composition
and particle size distribution (PSD) of the com-
bustion fly ash, and (2) the emissions of volatile
organic compounds (VOCs), specifically acrolein.
After considering several modifications to avail-
able DNPH methods, we discuss in detail one
non-DNPH method designed for ambient VOC
measurements, for use with source emissions.
2. Materials and methods
2.1. Refractory-lined furnace
Combustion of natural gas, No. 2 fuel oil, two
crude glycerol fuels (methylated and demethylat-
ed), and technical grade glycerol was performed
in a vertical refractory-lined furnace, shown in
Fig. 1. This furnace, rated at 82 kW, utilized an
International Flame Research Foundation
(IFRF) movable-block variable-air swirl burner.
All experiments were performed at its maximum
setting corresponding to a swirl number of 1.8.
Continuous emissions monitors (CEMs) mea-
Fig. 1. Furnace diagram with sampling locations.
2751
sured exhaust concentrations of O2, CO2, CO,
NO, NO2 and total hydrocarbons (THC) similarly
to methods described in 40 CFR 60, Appendix A
[15]. Temperatures were monitored near the bur-
ner and in the stack. PSDs were determined using
a TSI scanning mobility particle sizer (SMPS) and
aerodynamic particle sizer (APS). Together, these
instruments cover particle diameters from 15 nm
to 20 lm. A stainless steel sampling probe was
used to dynamically dilute flue gas. Dilution pre-
vented issues related to the effects of condensed
water on sample conditioning (particle removal),
VOC solubility in the sample containers
(SUMMA canisters), and reduced concentrations
to levels more appropriate for analysis. Com-
pressed nitrogen was used as the diluent, and
was introduced at the probe tip. Furnace and sam-
ple NOx measurements were compared to deter-
mine dilution ratios (70:1 for PSDs, 10–60:1
for VOCs). Three PSDs were averaged for each
fuel. Further details are available elsewhere [16].
2.2. Fuel compositions and delivery
Table 1 presents average heating values and
compositions of the two crude glycerol fuels
(methylated and demethylated) used in this study.
Also presented are compositions for No. 2 fuel oil
and technical grade glycerol. The crude glycerol
samples were received in September 2009 from
Foothills Bio-Energies (Lenoir, NC) and were
produced from the transesterification of low
FFA chicken grease. One drum of technical grade
glycerol was obtained in June 2011 from Redox
Tech, LLC (Cary, NC). These fuels were stored
in sealed drums under cover in outdoor secondary
containment and subjected to ambient North Car-
olina temperatures. Analyses indicate that glyc-
erol fuels have significantly lower carbon and
higher oxygen (determined by difference) contents
compared to No. 2 oil. Heating values are approx-
imately half that of No. 2 oil. Also of note are the
high ash contents of the two crude glycerol fuels,
comprised primarily of Na and P. All the fuels
have minimal nitrogen and sulfur contents, but
the glycerol fuels also have higher calcium, potas-
sium, and chlorine concentrations compared to
No. 2 oil.
All the liquid fuels were fed from drums
through a fuel recirculating delivery system. Bar-
rel heaters were used with the glycerol fuels to
reduce viscosities to facilitate pumping. The fuel
delivery system included a 6 kW in-line heater to
maintain fuel temperatures. The fuel system was
initially preheated with No. 2 oil before using
glycerol fuels. At the end of the day, the system
was switched back to No. 2 oil and the glycerol
purged from the system. The fuel delivery system
utilized a Haldex gear pump to continuously
circulate liquid fuels through the recirculation
loop at 230 L/h. Fuels were circulated in this
2752 S.A. Steinmetz et al. / Proceedings of the Combustion Institute 34 (2013) 2749–2757

Table 1
Fuel analyses.
Units No. 2 fuel oil Methylated glycerol Demethylated glycerol Technical glycerol
HHV MJ/kg 45.9 22.7 26.0 18.6
C % 85.84 45.21 49.46 39.06
H % 12.89 10.40 9.57 8.64
N % <0.5 <0.31 <0.27 <0.008
O % 1.17 41.79 37.72 50.11
S % 0.0032 <0.28 <0.20 <0.05
H2O % 0.02 3.48 2.48 0.37
Ash % <0.05 2.28 3.12 0.18
Ca ppm 53 108 115 –
Na ppm 70 13,700 12,100 –
K ppm 22 580 536 –
Cl ppm 20 153 314 171
Mg ppm 17 30 28 –
P ppm 10 1530 1380 –

way to maintain a constant feed temperature and
pressure (620 kPa). A separate burner supply line,
extending from this loop, allowed the atomization
pressure to be further regulated. The No. 2 oil,
methylated, demethylated, and technical glycerol
fuels were heated to temperatures of 308 K,
318 K, 393 K, and 393 K, respectively. Methyl-
ated glycerol was always maintained below the
boiling point of methanol (338 K).
Experiments carried out by Bohon et al. [2,3]
utilized air atomization for all liquid fuels. Since
pressure atomization is more commonly applied
in package boilers, it was investigated here as an
alternative. Pressure atomization was accom-
plished successfully with No. 2 oil and methylated
glycerol with Delavan model 2.00-80A and 2.75-
80A nozzles, respectively, and produced sprays
superior to those seen previously using air atom-
ization. Two 200-mesh y-strainers were utilized
within the circulation loop to help remove sus-
pended particulate that might otherwise plug the
atomization nozzles. However, air atomization
was utilized for both demethylated and technical
glycerols. The higher viscosity, ash content, and
suspended particulate present in the demethylated
glycerol resulted in plugging, and subsequent
removal of, the y-strainers. Air atomization was
not affected by these factors and plugging was
not an issue. Pressure atomization of technical
glycerol was not attempted. Air atomization of
demethylated and technical glycerols was accom-
plished with a Spraying Systems Co. model Air
Atom 1=4 -JSS nozzle. Compressed air delivered at
393 K and 30 sL/min (standard liters/min) was
used to atomize these fuels.
2.3. VOC measurements
Acrolein and other VOC measurements were
accomplished through collection of diluted flue
gases in SUMMA canisters followed by gas
chromatography–mass spectrometry (GC–MS)
analysis. EPA method TO-15 [17] was used as
published. All samples were collected in 6L
SUMMA passivated canisters. Prior to sample
collection, canisters were cleaned in an oven at
373 K by evacuating them to 133 Pa with a molec-
ular drag pump and then filling them with humid-
ified nitrogen (the nitrogen off-gas from a liquid-
nitrogen Dewar was found to give the cleanest
blanks) to 138 kPa. This cycle was completed five
times before the canisters were brought down to
their final pressure of 1.33 Pa. Canisters were
cleaned in batches of 12 and canister cleanliness
was assessed by batch blanks. The maximum
allowable concentration, after cleaning, of each
species on the TO-15 analyte list was 0.2 ppbv.
Sample values were blank corrected.
Dynamically diluted VOC sampling was per-
formed using the same in-stack dilution probe
described earlier. An acrolein/air mixture, used
for spiking, was generated using a Valco Instru-
ments Co. Inc. (VICI) dynacalibrator permeation
oven. Spiking is the intentional addition of a
known concentration of an analyte in order to
determine analytical recovery. Certified VICI
acrolein permeation tubes were heated to 308 K
to emit 550 ng of acrolein per minute (rate con-
firmed by weight change). Maximum dilution
flows on the VICI oven allowed the production
of an air sample containing 100 ppbv acrolein.
Spiked canisters were humidified and filled with
0.5 sL of spike gas before sampling. Furnace stack
samples were collected from paired sampling
trains over a span of 20 min per sample. Samples
were pulled through commercially available, pre-
ashed, heated (378 K) quartz filters designed for
use in VOC and semi-VOC sampling equipment.
Each of three pairs of samples for glycerol fuels
and two pairs for natural gas contained one pre-
spiked and one non-spiked SUMMA canister.
Comparison of simultaneously sampled canisters
allowed spike recovery rates to be calculated.
One additional SUMMA canister for each fuel
was spiked. This field spike was present near the
furnace during experimentation, and then filled
 S.A. Steinmetz et al. / Proceedings of the Combustion Institute 34 (2013) 2749–2757 2753

Table 2
Summary of operating conditions and exhaust gas emissions.
Units Particle size distributions Volatile organic compounds
No. 2 Methylated Demethylated Natural Methylated Demethylated Technical
fuel oil glycerol glycerol gas glycerol glycerol glycerol
Fuel flow LPM 0.093 0.160 0.160 108 0.104 0.142 0.272
Air flow sLPM 1120 1030 1330 1560 863 1144 1312
Stoich. ratio 1.19 1.18 1.28 1.27 1.53 1.23 1.37
Load kW 60.7 60.7 69.4 67.0 38.8 62.0 100.6
Stack temp. K 826 813 798 791 713 878 1024
Burner temp. K 1425 1396 1243 1345 1267 1339 1318
O2 % 3.6 3.3 4.8 4.8 7.6 4.2 5.4
CO2 % 14.0 15.0 14.8 9.6 9.2 13.7 14.8
CO ppmv 99 85 102 21 31 322 17
NOx ppmv 130 202 185 112 45 119 84
NOx at 0% ppmv 156 240 240 145 70 149 112
O2
THC ppmv 3.2 2.0 5.2 9.7 9.7 9.1 17.3

with nitrogen following the experiments. All can-
isters were filled to a total volume of 5 sL, deter-
mined by pressure. Experimental conditions are
shown in Table 2.
Canister samples were analyzed for VOCs
according to U.S. EPA Compendium Method
TO-15 [17]. Samples were preconcentrated with a
headspace preconcentrator (Entech Instruments
Inc., model 7150) interfaced to an Agilent model
6890 GC and model 5973N MS. The preconcentra-
tor was used to pull a metered volume of sample air
(including internal standards) from the canisters,
cryogenically concentrate the VOCs, and then
transfer the analytes to the GC–MS for analysis.
A Restek fused silica capillary column (RtxÒ-1,
60 m  0.32 mm  1.00 lm) was used with the
GC oven held at 308 K for 5 min; then ramped at
5 K/min to 403 K; then ramped to a final tempera-
ture of 523 K with a 3 min hold. Column flow was
held constant at 2.0 mL/min. The MS source was
set to 503 K; the MS quad was set to 423 K; and
selected ion monitoring was utilized for low ppbv
detection limits. The entire headspace preconcen-
trator and GC–MS system was calibrated with a
5-point calibration curve (2.00, 1.00, 0.50, 0.25,
and 0.125 ppbv) for 82 target organic analytes.
Internal standard calibrations were performed
using bromochloromethane, 1,4-difluorbenzene,
and chlorobenzene-d5. The calibration mixtures
were prepared by blending/diluting three commer-
cially prepared mixtures (Scott Specialty Gas,
Plumsteadville, PA).
3. Results and discussion
Table 2 presents a summary of the operating
conditions and major exhaust products measured
during two sets of experiments to collect PSDs
and VOC emissions. Emissions of CO and THC
are relatively low and comparable to levels mea-
sured for No. 2 oil and natural gas combustion.
NOx emissions (150–240 ppmv corrected to 0%
O2) are comparable to those reported by Bohon
et al. [2,3]. However, NOx emissions of methyl-
ated glycerol were seen much lower (70 ppbv)
in one experiment. This is likely related to the
lower load and higher stoichiometric ratio for this
experiment compared with others. We attempted
to maintain a consistent furnace load of between
60 and 70 kW and nominally 20% excess air for
all experiments. However, this was not always
possible due to periodic difficulties with regard
to fuel feeding and atomization.
3.1. Fly ash characterization
Fly ash elemental analyses presented by Bohon
et al. [3] indicated that the glycerol particulate
emissions contained 42–46% Na, 4–6% P, 3–6%
C, 1–2% K, 1% Cl, and 1% S. Mg, Ca, Fe, Zn,
and other trace elements were measured in con-
centrations less than 1%. Assuming stable oxides,
10–20% of the fly ash mass was undetermined.
Carbon in these samples was determined by a
thermal optical transmittance (TOT) method
[18], and the presence of notable concentrations
of carbon was interesting given the white powdery
consistency of the particulate. TOT methods are
typically used for ambient samples to determine
total carbon and the split between organic and ele-
mental forms. For source samples, this can be
applied to examine unburned carbon.
In an effort to further characterize the carbon
speciation, similar filter media was spiked with
sodium carbonate and sodium bicarbonate [3].
TOT analyses of these samples demonstrated that
the resulting thermogram interprets carbonates as
mixtures of organic and elemental carbon. Fur-
ther analysis using Fourier transform infrared-
attenuated total reflectance (FTIR-ATR) verified
the possible presence of sodium carbonate in the
fly ash samples, as evident by IR responses near
2754 S.A. Steinmetz et al. / Proceedings of the Combustion Institute 34 (2013) 2749–2757
700, 870, and 1450 cm 1. Equilibrium calculations
were also performed using HSC Chemistry to
determine thermodynamically stable species.
These calculations were performed for methylated
and demethylated glycerol corresponding to sys-
tem concentrations of fuel and 20% excess air. A
total of 85 species, comprised of C, H, N, O,
Na, P, S, Cl, and Ca, were considered. Kinetic
and mixing limitations were not considered, so
equilibrium results need be used with care. Fig-
ure 2 presents Na equilibrium predictions for
1 atm and system temperatures from 300 to
2500 K. The equilibrium calculations suggest that
sodium vaporized in the flame forms NaOH,
which could react in the gas phase with SO2,
CO2, and phosphorous compounds. At tempera-
tures below a predicted 1700 K Na dew point,
equilibrium indicates stable condensed-phase mix-
tures of sodium sulfate (Na2SO4), sodium carbon-
ate (Na2CO3), and sodium phosphate (Na3PO4).
These predictions qualitatively agree with the pre-
vious XRF [3] and the FTIR-ATR results, sug-
gesting that Na2CO3 is the primary source of
carbon in the fly ash and that negligible amounts
of unburned carbon are present.
Figure 3a presents the averaged measured vol-
ume PSDs (with standard errors) for each fuel.
The measured PSDs are unimodal indicating
well-defined accumulation modes, with mean vol-
ume diameters of 0.7 and 0.3 lm for the methyl-
ated and demethylated glycerol fuels,
respectively. There was no evidence of a coarse
mode for either fuel. These results indicate that
the ash, composed predominantly of alkali and
alkaline earth elements, are extensively vaporized
during combustion. Once vaporized, the ash
species undergo homogeneous nucleation and par-
ticle growth via coagulation and condensation
processes. The large mean particle diameters indi-
cate extensive particle growth corresponding to
very large particle number concentrations. Num-
Fig. 2. Equilibrium predictions of Na mass fraction (as
indicated species) as a function of temperature for
simulated methylated glycerol combustion (20% excess
air). Dashed lines represent gaseous species, solid lines
represent condensed species.
Fig. 3. Measured and predicted volume PSDs for
methylated and demethylated glycerol fuels. Measured
PSDs (a) present averaged PSDs with error bars for
composite SMPS and APS measurements. Predicted
PSDs (b) are the results of the nucleation of 2 nm nuclei
allowed to grow by coagulation only using the MAER-
OS simulation code. Dashed PSDs correspond to
methylated glycerol; solid PSDs correspond to deme-
thylated glycerol.
ber concentrations measured at the sampling loca-
tion were 3.3  108 and 1.6  108 #/cm3 for the
methylated and demethylated fuels, respectively.
It should be noted that significant ash is lost to
the furnace walls as evident by the persistence of
an accumulation mode measurable during natural
gas combustion weeks after the glycerol experi-
ments. In fact, this furnace contamination made
it impossible to close ash mass balances, and indi-
cates that crude glycerol combustion in boilers is
not likely until new catalyst options are employed.
However, regardless of mass losses, there is no
indication of a coarse mode, or significant particle
growth beyond 1 lm.
Particle coagulation was modeled using the
multicomponent aerosol simulation (MAEROS)
code [19] to determine the extent that this mecha-
nism could be used to describe aerosol growth.
System inputs included an assumed particle den-
sity of 2.2 g/cm3 and mass concentrations of
5050 and 3310 mg/m3 corresponding to the calcu-
lated ash concentrations for methylated and
demethylated fuels, respectively. Residence times
from the burner to sampling location, based on
fuel and air flows and a linear temperature profile,
were calculated to be 3.9 and 3.3 s for methylated
and demethylated glycerols, respectively. Fig-
ure 3b presents the results of these MAEROS cal-
culations for both fuels. Comparisons with Fig. 3a
suggest reasonable agreement between the model
 S.A. Steinmetz et al. / Proceedings of the Combustion Institute 34 (2013) 2749–2757
predictions and the data. The model predictions
suggest higher volume concentrations with slightly
smaller mean diameters. These differences are
likely the result of significant mass losses to the
furnace walls and the added contribution of ash
condensation on existing particles. Model predic-
tions are consistent with the data, with neither
suggesting accumulation above 1 lm.
3.2. Acrolein and other VOCs
Table 3 presents the measured concentrations
of VOCs determined by GC–MS analysis of stack
samples collected in non-spiked SUMMA canis-
ters. Averages and standard errors are presented
for natural gas, methylated, demethylated, and
technical glycerol fuels. Concentrations, in ppbv,
have been dilution and blank corrected, and spe-
cies are presented in elution order. One non-
spiked canister from both the demethylated and
technical glycerol sets was discarded, as the pres-
ence of some compounds was detected in abnor-
mally high concentrations. The paired samples
corresponding to these discarded samples had
concentrations consistent with the remaining four
canisters, indicating the abnormality was not
related to stack emissions. Dilution ratios and
preconcentrator injection volumes, which were
chosen to keep acrolein concentrations within
the GC–MS calibration range, varied by sample.
As a result, most other compounds detected were
in concentrations below the calibration range, and
detection limits for each sample varied.
Acrolein measurements of diluted samples for
all four fuels ranged from non-detect to 1.1 ppbv,
with a method detection limit of 59.6 pptv. This
corresponds to in-stack concentrations ranging
from <7.2 to 25.7 ppbv. Average values are pre-
sented in Table 3. Replicate sample spiking using
the paired train approach described in Section 2.3
yielded average acrolein recoveries of 172%, 96%,
129%, and 140% for natural gas, methylated,
demethylated, and technical glycerols, respec-
tively. Field spikes generated for each fuel test
produced recoveries consistent with each of these
values, suggesting that the variation in recoveries
is more likely due to spiking imprecision rather
than fuel effects. Spike concentrations (10 ppbv)
were chosen before the actual diluted concentra-
tions of acrolein (1 ppbv) were known. Compar-
ing values with large differences in relative size
could lead to further inaccuracies in recovery cal-
culation. These results show that acrolein emis-
sions for crude and technical glycerol are very
small and comparable to emissions from natural
gas.
With the exception of ethanol, acetone, and
isopropyl alcohol, no other VOC was measured
in concentrations higher than 100 ppbv. Natural
gas and technical glycerol typically produced the
lowest and highest VOC concentrations, respec-
2755
tively. However, VOC emission concentrations
for all the glycerol fuels were often comparable,
and at most, no more than 30 times those of nat-
ural gas. It is interesting to note the presence of a
number of halogenated VOCs in very low concen-
trations. These were present in the natural gas
emissions and typically in higher concentrations
in the glycerol emissions. While extremely low,
these measurements are above blank levels and
believed to be real. Table 1 indicates that signifi-
cant quantities of chlorine (100–500 ppm) are
present in all the glycerol fuels, and this is thought
to be the source of the halogen. As indicated in
Table 2, there was some variability in the furnace
loads during the VOC experiments. However, this
variability did not correlate with any observed
trends in the very low VOC emissions.
4. Conclusions
We have previously shown that a refractory-
lined burner designed with high swirl can be used
to successfully burn crude glycerol fuels without
the need to co-fire with other fossil fuels. This bur-
ner provides a thermal environment to promote
fuel ignition, flame stability, and nearly complete
fuel oxidation [2,3]. This was an important first
step in demonstrating a new value added use for
this potential high volume waste produced during
the transesterification of bio-oils to produce bio-
diesel fuels, and if utilized as a boiler fuel during
this process, offers advantages of optimizing
energy integration, reducing waste processing
and transportation costs, fossil fuel requirements,
and greenhouse gas emissions. The initial
research, however, identified two important issues
that could prevent further development of crude
glycerol as a replacement to fossil fuels: very high
particulate emissions, and the potential to pro-
duce acrolein emissions of concern.
A series of experiments were performed to spe-
cifically address these two critical issues. Addi-
tional characterization of the ash behavior
suggests that crude glycerol containing soluble
catalysts (in this case sodium hydroxide and phos-
phoric acid) produces extremely large submicron
particulate emissions that are incompatible with
the operation of package boilers. PSD measure-
ments suggest that these and the other predomi-
nantly alkali and alkaline earth elements present
in the original bio-oil are extensively vaporized
during combustion and form an accumulation
mode aerosol through nucleation, condensation,
and coagulation processes. Equilibrium calcula-
tions and FTIR-ATR analysis for sodium suggest
a mixed composition of carbonates, sulfates, and
phosphates. The size and composition of these
particles are likely to cause extensive fouling and
corrosion problems in boiler applications. New
methods of biodiesel production that may utilize
2756 S.A. Steinmetz et al. / Proceedings of the Combustion Institute 34 (2013) 2749–2757

Table 3
Acrolein and other VOC emissions (ppbv).a,b,c,d
Natural gas Methylated glycerol Demethylated Technical glycerol
glycerol
Average Standard Average Standard Average Standard Average Standard
error error error error
Dichlorodifluoromethane 0.7f,g – 2.4f 0.5 8.9f,g – ND –
Dichlorotetrafluoroethane 3.7 0.6 22.7g 17.2 52.8g 27.0 44.8g 17.8
Ethanol 101.5 14.5 191.7g 131.1 173.0 9.8 260.7 16.0
Acetonitrile 5.9 2.8 45.5 36.5 <37.8e – ND –
Acrolein 13.3 1.8 16.5f 9.2 <18.0e – 20.7f,g –
Acetone 173.6 1.2 190.2f 117.4 <31.0e,f – 42.1f –
Trichlorofluoromethane 2.5 0.0 2.4f 1.2 7.5f,g 2.6 <12.2e –
Isopropyl alcohol 60.5 11.4 99.7f 73.6 <37.8e – 46.5f –
Acrylonitrile 33.5 0.1 36.7g 13.7 31.0g 2.7 54.1g 16.2
tert-Butanol 2.5 0.1 8.3f – ND – ND –
Methylene chloride 2.1g 0.4 7.2f 3.9 <6.0e – <12.3e –
3-Chloro-1-propene 0.6g 0.0 3.3f,g 2.4 11.2f,g – 6.8f,g –
1,1,2-Trichloro-1,2,2- 1.2g 0.0 1.2f,g 0.6 6.7f,g – 6.8f,g –
trifluoroethane
trans-1,2-Dichloroethene ND – 8.1f,g 1.7 ND – ND –
Vinyl acetate 3.2 0.4 8.5f 3.0 ND – <7.8e,f –
2-Butanone 14.0 1.4 20.6f 15.1 <31.0e,f – <33.6e,f –
Ethyl acetate 0.5g 0.0 0.5f,g 0.2 6.8f,g – 7.2f,g –
Tetrahydrofuran 3.1 0.1 10.4g 6.6 13.3f,g 0.8 21.4g 7.9
Ethyl tert-butyl ether 0.5g 0.1 ND – 4.8f,g – 11.6f,g –
1,1,1-Trichloroethane 0.6g 0.0 0.6f,g 0.3 8.4g 0.8 13.5g 5.3
Benzene 1.3g 0.4 1.0f,g 0.5 6.7f,g – <8.7e –
Carbon tetrachloride 2.2 0.2 10.8g 8.8 19.6g 4.4 25.6g 9.8
1,4-Dioxane 1.4g 0.9 4.1g 2.4 4.5g 0.4 7.9g 3.1
4-Methy-2-pentanone 1.2g 0.2 2.0f,g 1.5 7.8f,g – 13.8g 5.0
1,1,2-Trichloroethane 0.9g – 1.4f,g – 18.3f,g – ND –
Toluene 1.7g 0.0 4.8f,g 3.7 1.6f,g – 7.2g 0.7
2-Hexanone 2.6 0.4 3.7f,g 2.6 ND – 20.0f,g –
1,2-Dibromoethane 0.3g – ND – 5.6g 0.6 9.0g 3.6
Ethylbenzene 0.4g 0.0 0.9f,g 0.6 2.6f,g – 5.4g 1.2
m-Xylene 0.7g 0.1 1.7f,g 1.0 3.8g 1.0 12.1g 2.4
Bromoform 0.7g 0.0 9.3g 8.5 13.6g 1.4 21.8g 8.6
Styrene 0.4g 0.0 4.2g 3.7 6.1g 0.6 9.9g 3.7
o-Xylene 0.5g 0.0 1.1f,g 0.7 2.8g 0.8 8.3g 1.7
a
Non-detect (ND) indicates no response in GC–MS analysis.
b
The following compounds were not detected in concentrations above 5 ppbv in any sample: propylene,
chloromethane, 1,3-butadiene, 1,1-dichloroethene, carbon disulfide, 1,1-dichloroethane, methyl-t-butyl-ether, 2-chloro-
prene, cis-1,2-dichloroethene, n-hexane, chloroform, 1,2-dichloroethane, 1,2-dichloropropane, bromodichloromethane,
trichloroethene, isooctane, methyl methacrylate, heptane, dibromochloromethane, tetrachloroethene, chlorobenzene, p-
xylene, 1,1,2,2-tetrachloroethane, naphthalene.
c
The following compounds were ND in all samples: vinyl chloride, bromomethane, chloroethane, vinyl bromide,
cyclohexane, tert-amyl methyl ether, cis-1,3-dichloropropene, trans-1,3-dichloropropene, 1,1,1,2-tetrachloroethane,
bromofluorobenzene, cumene, chlorotoluenes, n-propylbenzene, 1,3,5-trimethylbenzene, 1,2,4-trimethylbenzene, tert-
butyl benzene, 1-ethyl-4-methyl benzene, 1,3-dichlorobenzene, 1,4-dichlorobenzene, sec-butyl benzene, 1,2-dichloro-
benzene, o-cymene, n-butyl benzene, 1,2,4-trichlorobenzene, hexachlorobutadiene.
d
Some sets of triplicate measurements had only one sample that was not ND or BDL. Standard error could not be
calculated.
e
Below detection limits (BDL). Method detection limit defined by multiplying the Student’s t-value (3.143 for six
degrees of freedom) by the standard deviation of seven replicate analyses of a low level calibration standard (50 pptv).
Table values adjusted for dilution.
f
One or more samples were BDL or ND.
g
One or more samples were below the lowest calibration point but above detection limits.

fixed catalyst beds, rather than soluble catalysts, erol in black liquor boilers where the sodium
or a supercritical reaction without catalyst, could could be recovered and used for the kraft pulping
greatly reduce or eliminate the inorganic constitu- process or in other industrial applications where
ents. There is also potential to co-fire crude glyc- high levels of ash are not an issue [4].
 S.A. Steinmetz et al. / Proceedings of the Combustion Institute 34 (2013) 2749–2757
Results of a SUMMA canister collection and
GC–MS analyses method designed to characterize
emissions of acrolein and other VOCs indicate
that combustion of crude glycerol fuels produces
very low acrolein emissions, comparable to those
for natural gas combustion. This indicates that
acrolein emissions are, therefore, not an issue of
concern and this is consistent with the stable
flames and good fuel burnout observed. Analysis
of the same samples for a variety of other VOCs
suggests emissions that are also relatively low
and comparable to natural gas. Although preci-
sion levels could be improved, this approach could
be further developed and refined to provide reli-
able acrolein and VOC emission measurements
from sources.
Acknowledgements
Portions of this work were sponsored under
the NCSU/EPA Cooperative Training Program
in Environmental Sciences Research, Training
Agreement CT833235-01-0 with North Carolina
State University, P.O. EP-09-C00114 with J.O.L.
Wendt, and Contract EP-C-04-023 with ARCA-
DIS U.S., Inc. The authors are grateful to Mr.
Daniel Janek, for his help operating the refrac-
tory-lined furnace, Dr. Simon Lappi, for his help
with FTIR-ATR analyses at NCSU, and Mr.
Myles Bohon, for his help with glycerol combus-
tion issues. The research described in this article
has been reviewed by the U.S. EPA National Risk
Management Research Laboratory and approved
for publication. The contents of this article should
not be construed to represent Agency policy nor
does mention of trade names or commercial prod-
ucts constitute endorsement or recommendation
for use.
References
[1] Glycerol Purification, EET Corp., available at
http://www.eetcorp.com/heepm/biodiesel.pdf.
[2] M.D. Bohon, B.A. Metzger, W.P. Linak, C.J. King,
W.L. Roberts, Proc. Combust. Inst. 33 (2011) 2717–
2724.
[3] M.D. Bohon, Characterization of Glycerin Combus-
tion and Emissions, MS thesis, North Carolina State
University, Raleigh, NC, 2010.
2757
[4] Understanding the Kraft Process in Paper Produc-
tion, Paper Industry, available at http://
www.paperindustry.com/kraft-process.asp
(accessed December, 2011).
[5] Glycerol, MSDS, available at http://www.hvchem-
ical.com/msds/glyc.htm (accessed December, 2011).
[6] U.S. EPA, Toxicology Review of Acrolein, Report
No. EPA/635/R-03/033, available at http://www.e-
pa.gov/iris/toxreviews/0364tr.pdf (accessed Decem-
ber, 2011).
[7] U.S. EPA, Method 0011: Sampling for Selected
Aldehyde and Ketone Emissions from Stationary
Sources, available at http://www.epa.gov/osw/haz-
ard/testmethods/sw846/pdfs/0011.pdf (accessed
December, 2011).
[8] U. Karst, N. Binding, K. Cammann, U. Witting,
Fresenius J. Anal. Chem. 345 (1993) 48–52.
[9] R. Schulte-Ladbeck, R. Lindahl, J.O. Levin, U.
Karst, J. Environ. Monitor. 3 (2001) 306–310.
[10] S.S.H. Ho, K.F. Ho, W.D. Lie, et al., Atmos.
Environ. 45 (2011) 261–265.
[11] S. Uchiyama, Y. Inaba, N. Kunugita, J. Chroma-
togr. A 1217 (2010) 4383–4388.
[12] K. Akiyama, A. Nakayama, SAE Technical Paper,
Report No. 2005-01-2152, available at http://sub-
scriptions.sae.org/content/2005-01-2152/ (accessed
December, 2011).
[13] K. Akiyama, A. Nakayama, SAE Technical
Paper 2010-01-2207, available at <http://subscrip-
tions.sae.org/content/2010-01-2207/> (accessed
December, 2011).
[14] G. Shoeing, Modified impinger method for acrolein
in air, Ashland Chemical Company, available at
http://www.arb.ca.gov/ei/acrolein/modified.doc
(accessed December, 2011).
[15] U.S. EPA Code of Federal Regulations (CFR)
Promulgated Test Methods, available at http://
www.epa.gov/ttn/emc/promgate.html (accessed
December, 2011).
[16] W.P. Linak, R.K. Srivastava, J.O.L. Wendt, Com-
bust. Sci. Technol. 101 (1994) 7–27.
[17] U.S. EPA, Determination of Volatile Organic
Compounds (VOCs) in Air Collected in Specially-
Prepared Canisters and Analyzed by Gas
Chromatography/Mass Spectrometry (GC/MS):
Compendium Method TO-15 in Compendium of
Methods for the Determination of Toxic Organic
Compounds in Ambient Air, Report No. EPA/625/
R-96/010b, available at http://www.epa.gov/
ttnamti1/files/ambient/airtox/to-15r.pdf (accessed
December, 2011.
[18] NIOSH Manual of Analytical Methods, Diesel
Particulate Matter (as Elemental Carbon), available
at http://www.cdc.gov/niosh/docs/2003-154/pdfs/
5040.pdf (accessed December, 2011).
[19] F. Gelbard, J.H. Seinfeld, J. Colloid Interface Sci.
78 (1980) 485–501.