Académique Documents
Professionnel Documents
Culture Documents
Fuel
journal homepage: www.elsevier.com/locate/fuel
G RA P H I C A L AB S T R A C T
A R T I C L E I N F O A B S T R A C T
Keywords: Various lignin depolymerization methods have been proposed. Nevertheless, the relationship between the
Lignin structure of lignin and its depolymerization behavior has not been widely investigated. Herein, six types of lignin
Lignin structure samples were produced from oakwood (OW, hardwood) and pinewood (PW, softwood) using three different
Depolymerization delignification techniques (ethanolsolv, formasolv, and Klason). The content of ether linkages in the OW-derived
Supercritical ethanol
lignins was approximately three times higher than that in the PW-derived lignins because of the presence of the
Formic acid
sinapyl alcohol unit in the former. The contents of ether linkages in the lignin isolated via the different methods
followed the order: formasolv > ethanolsolv > Klason. The lignin samples were depolymerized in a mixture of
supercritical ethanol (scEtOH) and formic acid at temperatures of 250–350 °C. At 350 °C, regardless of the lignin
type, high conversion (> 95%) and a high bio-oil yield (> 81 wt%) could be achieved, demonstrating that the
combined use of scEtOH-HCOOH was very effective for the depolymerization of various types of lignin. At the
low temperatures of 250–300 °C, the lignin conversion and bio-oil yield were highly dependent on the amount of
ether linkages; for example, at 300 °C, the use of OW-derived formasolv lignin resulted in a high bio-oil yield
(86.2 wt%), whereas the use of OW-derived Klason lignin resulted in a very low bio-oil yield (27.9 wt%). The
properties of the bio-oils produced from the different types of lignin were discussed.
⁎
Corresponding author at: School of Mechanical Engineering and SKKU Advanced Institute of Nanotechnology (SAINT), Sungkyunkwan University, 2066 Seobu-Ro, Jangan-Gu,
Suwon, Gyeong Gi-Do 16419, Republic of Korea.
E-mail address: jaehoonkim@skku.edu (J. Kim).
https://doi.org/10.1016/j.fuel.2017.12.079
Received 15 October 2017; Received in revised form 8 December 2017; Accepted 19 December 2017
Available online 02 January 2018
0016-2361/ © 2017 Elsevier Ltd. All rights reserved.
J. Park et al. Fuel 217 (2018) 202–210
203
J. Park et al. Fuel 217 (2018) 202–210
formasolv lignin. The formasolv lignin samples produced using oak- 2.4. Analysis and characterization
wood and pinewood are designated as OW-F and PW-F, respectively.
Klason lignin was prepared by a two-step acid hydrolysis reaction; first, To quantify the ether linkages in the lignin samples (e.g., β-O-4, α-
the wood sample was hydrolyzed with a concentrated strong acid (72% O-4, 4-O-5), the vanillin and syringaldehyde contents were analyzed
H2SO4) for 2 h at room temperature. The reaction mixture was then using nitrobenzene oxidation, as proposed by Chen [30]. Prior to ni-
diluted to give a 3% H2SO4 aqueous solution by adding DDI water, and trobenzene oxidation, the wood samples were ground and extracted
the mixture was reacted at 110 °C under reflux conditions for 4 h for the using a Soxhlet apparatus with a toluene/ethanol mixture (2:1) at
subsequent weak acid hydrolysis reaction. After the two-step acid hy- 120 °C for 6 h. Fifty milligrams of dried lignin or 200 mg of dried wood,
drolysis reaction, the holocellulose was dissolved in aqueous sulfuric 7 mL of 2 M NaOH, and 0.48 g of nitrobenzene were placed into a
acid and the lignin precipitated at the bottom of the flask. The pre- stainless-steel bomb reactor with an inner volume of 11 mL. The reactor
cipitated lignin was separated from the dissolved holocellulose by was then heated to 170 °C and the oxidation was allowed to proceed for
centrifugation. The recovered lignin was washed with DDI water sev- 2.5 h. After the reaction, the liquid mixture was collected from the re-
eral times and then dried in a drying oven at 80 °C for 12 h. The Klason actor and water and DCM were added to the reaction mixture. The
lignin samples produced using oakwood and pinewood are designated whole solution was kept in a separating funnel for 4 h to separate the
as OW-K and PW-K, respectively. The yields of lignin recovered from oxidized products in the aqueous phase from nitrobenzene in the DCM
each delignification method followed the order: Klason (100 wt phase. The aqueous phase was collected from the funnel and then
%) > ethanolsolv (63–71 wt%) > formasolv (53–54 wt%), as listed in acidified to pH 2–3 using aqueous HCl. The acidified species in the
Table S2. aqueous phase were then extracted using DCM in three liquid-liquid
extraction cycles. The DCM phase was evaporated using a rotary eva-
porator at 40 °C and 0.04 MPa to recover the products for further
2.3. Lignin depolymerization reaction analysis. Thermogravimetric analysis (TGA) was conducted over the
temperature range of 30–800 °C using a TA Instruments Q50 TGA in-
The depolymerization experiments were performed by using a strument at a heating rate of 10 °C min−1. The gas flow rate was fixed at
custom-built, batch reactor made of SUS 316 having an inner volume of 60 mL min−1. Elemental analysis (EA) of the samples was conducted
140 mL, equipped with a magnetically driven stirrer. The reactor was with a Vario EL cube elemental analyzer equipped with a TCD detector
rated to rated 500 °C at 50 MPa. In each lignin depolymerization reac- (Elementar Analysensysteme GmbH, Germany). The combustion tube
tion, 3 g of lignin, 60 g of ethanol and 12 g of formic acid were placed and the reduction tube temperature were maintained at 1150 °C and
into the reactor. The reactor was closed and purged with N2 gas three 850 °C, respectively. The oxygen content was analyzed using a TCD
times through a purge line dipped into the solution to remove dissolved detector in O-mode with a pyrolysis tube at 1170 °C. The higher heating
oxygen in the liquid phase and the oxygen in the reactor head, and then value (HHV) of the raw lignin samples and bio-oil was calculated using
pressurized at 1 MPa with N2. The reactor was heated to the experi- the DIN 51900 standard:
mentally desired temperatures of 250–350 °C at a heating rate of ap-
proximately 20 °C min−1. After reaching the desired reaction tem- HHV (MJ. kg−1) = (34 C+ 124.3 H+ 6.3 N+ 19.3S−9.8O)/100 (5)
perature, the reaction was allowed to proceed for 1 h. After the where C, H, N, S, and O are the weight percentages of carbon, hy-
reaction, the reactor was quickly quenched to 100 °C using a water drogen, nitrogen, sulfur, and oxygen, which were obtained by EA.
basket, in order to prevent further possible reaction during the cooling The molecular weight distribution of bio-oil was characterized by
process, and was then further cooled to room temperature using an using a Bruker UltrafleXtreme™ matrix-assisted laser desorption ioni-
electric fan. The reactor pressure was recorded after complete cooling zation time-of-flight mass spectrometer (MALDI-TOF/MS, USA). The
to calculate amount of gas produced. The produced gas was collected in composition of the produced gases was analyzed by using a
a 0.5-L Tedlar bag for compositional analysis. The reaction mixture in PerkinElmer Clarus 600 GC-Model Arnel 1115PPC Refinery Gas
the reactor was then collected in a beaker by washing with acetone. The Analyzer (RGA, PerkinElmer, CT, USA). The detailed specifications of
liquid and solid products were separated using a Whatman filter paper. the RGA-GC have been reported previously [31]. The bio-oil was
The solid residue was dried in a drying oven at 70 °C for 6 h. The liquid characterized using an Agilent Technologies 7890A GC with a time-of-
product, which contained solvent, bio-oil, and byproduct, was evapo- flight mass spectrometer (TOF/MS) detector for both qualitative and
rated at 50 °C and 0.02 MPa using a rotary evaporator. After the eva- quantitative analysis. A detailed description of the GC-TOF/MS equip-
poration process, the bio-oil was further dried in a vacuum oven at ment is given elsewhere [32]. The functional groups on the sample
60 °C at ∼0 MPa for 6 h to remove the residual solvent. All the ex- surface were characterized by using a NICOLET iS10 Fourier-transform
periments were performed in triplicate and the average values are re- infrared (FT-IR) spectrometer (Thermo Electron Co., USA).
ported.
The conversion and the yields of bio-oil, solid residue, and gas 3. Results and discussion
product were calculated using the following equations:
3.1. Characterization of lignin
Conversion
Weight of dry ash free lignin-Weight of organics in solid residue
= The lignin samples were characterized by proximate and ultimate
Weight of dry ash free lignin
analyses, as listed in Table 1. All the lignin samples produced from the
× 100 (1) different types of wood and by the different delignification methods had
similar contents of volatile matter (49–55 wt%) and fixed carbon
Weight of bio oil (45–49 wt%). The ash content of the Klason lignin was slightly higher
Yield of bio oil (wt%) = × 100
Weight of dry ash free lignin (2) than that of the ethanolsolv and formasolv lignin samples. During the
ethanolsolv and formasolv processes, the lignin was dissolved in the
Weight of solid residue organic solvents and recovered by precipitation in water. On the other
Yield of solid residue (wt%) = × 100 hand, during the Klason process, the lignin remained in the solid form
Weight of dry lignin (3)
and holocellulose was dissolved in the acidic aqueous solvent, and thus
some fraction of ash present in the wood sample could be trapped in the
Weight of produced gas
Yield of gas (wt%) = × 100 solid lignin phase. The sulfur contents of the ethanolsolv and formasolv
Weight of dry ash free lignin (4) lignin samples were under the detection limit of EA (< 0.01 wt%), but
204
J. Park et al. Fuel 217 (2018) 202–210
Table 1
Proximate and ultimate analysis of the lignin feedstocks used in this study.
OW-E 50.6 48.2 0.1 1.1 69.2 5.13 23.9 N.D.f N.D. 0.26 0.92
OW-F 54.4 44.5 N.D. 1.1 63.6 5.15 28.2 N.D. N.D. 0.33 0.97
OW-K 54.2 44.0 0.6 1.2 60.9 4.84 30.1 N.D. 5.85 0.37 0.95
PW-E 52.5 45.7 N.D. 1.8 68.0 5.02 24.7 N.D. N.D. 0.27 0.89
PW-F 53.5 46.2 0.2 0.3 66.4 5.08 26.4 N.D. N.D. 0.30 0.92
PW-K 49.0 49.1 0.6 1.3 60.6 4.57 26.3 N.D. 5.92 0.33 0.90
c,d
VM and FC were determined using TGA in N2.
a
On a dry basis.
b
On a dry, ash-free basis.
c
VM: volatile matter.
d
FC: fixed carbon.
e
Ash content was determined using TGA in air.
f
Not detectable: under the detection limit of EA (0.01 wt%).
the Klason lignin contained approximately 5 wt% sulfur, which may formasolv method was used as compared to the ethanolsolv and Klason
have originated from the sulfuric acid treatment. methods. During the lignin separation reaction, the α-aryl ether and β-
To quantify the ether linkages (e.g., β-O-4, α-O-4, 4-O-5) in the aryl ether bonds of lignin in the wood undergo hydrolysis to produce
lignin samples, nitrobenzene oxidation was employed; during the ni- small soluble fractions, which are detached from the cellulose. Proto-
trobenzene oxidation, vanillin (V) and syringaldehyde (S) were formed nation of the hydroxyl group in the α or β position with acid catalysts
by cleavage of the ether linkages [30], which are the major chemical can convert the α or β carbon into a carbocation, which can break the
bonds in lignin. As listed in Table 2, the amount of V and S produced ether bond. On the other hand, the formed carbocation may be re-
during the nitrobenzene oxidation was highly dependent on the wood condensed in an intramolecular manner to produce cyclic species or
source as well as the type of lignin produced using the various isolation may react with other fragmented lignins [35,36]. The extent of the
methods. When the wood samples were oxidized, the total V + S yield recondensation reaction may depend on the relative amount of guaiacyl
from OW was approximately three times higher than that produced alcohol and sinapyl alcohol units in a given lignin species because the
from PW because of the presence of syringaldehyde in the oxidized sinapyl alcohol, in which the ortho position of the aromatic ring is
mixture. Lignin from hardwood contains both guaiacyl alcohol (G) and occupied by methoxy groups, can effectively suppress self-re-
sinapyl alcohol (S) units with G/S ratios in the range of 1–4, which condensation [36]. Therefore, the OW-derived lignin samples gave rise
leads to a high ether linkage content in the range of 64–86%. On the to higher V + S yields than the PW-derived lignin samples. In addition
other hand, softwood lignin consists of almost 95% guaiacyl alcohol to the wood source, the degree of recondensation is also known to be
units, and thus the amount of ether linkages (59–74%) in softwood is dependent on the delignification method and conditions [12]. As listed
lower than that in hardwood [33,34]. The V + S content of the lignin in Table 2, the formasolv method resulted in consistently higher V + S
samples was an order of magnitude smaller than that of the “intact” yields than achieved with the ethanolsolv method, regardless of the
lignin present in the corresponding woods. This indicates that the wood type. The lower reaction temperature (120 °C) of the formasolv
structure of lignin changed significantly to form CeC condensed lin- method than that of the ethanolsolv method (190 °C) may account for
kages at the expense of the ether linkages during the lignin separation the lower degree of condensation in the former case. In addition, when
process. Notably, the V + S yields from the lignin samples were highly the Klason method was used, much lower V + S yields were obtained
dependent on the delignification methods. For example, when OW was than achieved with the formasolv and ethanolsolv methods because of
used, the V+S yield followed the order: OW-F the use of a large quantity of strong acid [35]. Because the CeC bond
(129.3 µmol g−1) > OW-E > (116.4 µmol g−1) > OW-K dissociation energy (BDE) (234.9–483.4 kJ mol−1) is larger than the
(83.4 µmol g−1). A similar trend in the V yield was observed for the PW- CeO BDE (182.7–311.6 kJ mol−1) [37], the high degree of re-
derived lignin samples. This indicates that less condensation to form condensation can make lignin depolymerization more difficult.
condensed linkages at the expense of ether linkages occurred when the The functional groups in the lignin samples were analyzed using FT-
IR, and the results are shown in Fig. 1. Signals of hydroxyl (νeCeOH at
∼3430 cm−1, including vibrations from COOH and H2O), carbonyl (ν-
Table 2 C]O at 1710 cm−1), skeletal aromatic ring vibrations (νeC]C at 1605
Nitrobenzene oxidation data.
and 1510 cm−1), and CeH (νeCeH at 1460 and 1365 cm−1) groups
Feedstock Vanillin (V) Syringaldehyde (S) (µmol Total (V + S) were clearly detectable in the profile of the lignin samples, which
(µmol g−1) g−1) (µmol g−1) agrees well with previous observations [28,38]. The signals of the CeO
groups at 1320 and 1120 cm−1 only appeared in the profile of the OW-
OWa 520.1 788.2 1308.3
derived lignin samples, indicating that the corresponding ether bonds
OW-Eb 49.8 66.6 116.4
OW-Fc 35.6 93.7 129.3 originated from the syringol group. On the other hand, the signals of the
OW-Kd 40.6 42.8 83.4 CeO groups at 1270 and 1030 cm−1 (which correspond to the guaiacol
PWe 459.3 0 459.3 group) in the profile of the PW-derived lignin samples had a stronger
PW-E 35.4 0 35.4 transmittance than observed for the OW-derived counterparts. Com-
PW-F 68.9 0 68.9
PW-K 23.0 0 23.0
pared to the ethanolsolv and formasolv lignin samples, the Klason
lignin exhibited weak peaks at 1320 and 1270 cm−1, indicating that
a
Oakwood. fewer CeO bonds of the guaiacol and syringol groups were present.
b
Ethanolsolv lignin from oakwood. This agrees well with the nitrobenzene oxidation results.
c
Formasolv lignin from oakwood. The lignin samples were not completely soluble in acetone or formic
d
Klason lignin from oakwood.
e acid, and thus only the molecular weights of the soluble fraction were
Pinewood.
205
J. Park et al. Fuel 217 (2018) 202–210
Fig. 2. (a) Yields of bio-oil and solid reside produced from the depolymerization of OW-E,
Fig. 1. FT-IR spectra of the lignin samples (a) wavenumber range from 4000 to 800 cm−1, OW-F, and OW-K at temperatures of 250–350 °C. Reaction conditions: 9–10 MPa (250 °C),
(b) wavenumber range from 1400 to 1100 cm−1. 19–23 MPa (300 °C), 30–46 MPa (350 °C), 3 g lignin, 60 g ethanol, 12 g formic acid,
60 min. (b) Correlation between the amount of ether linkages and bio-oil yield.
206
J. Park et al. Fuel 217 (2018) 202–210
Klason process. Overall, as shown in Fig. 2b, there is a strong correla- achieved with the scEtOH + HCOOH mixture above 350 °C, regardless
tion between the amount of ether linkages (determined by nitrobenzene of the type of lignin used.
oxidation) and the bio-oil yield at the low-to-medium temperature The molecular weights of the bio-oil produced from the OW-derived
range of 250–300 °C, whereas a high-yield bio-oil (> 81 wt%) was lignin samples are shown in Fig. S3a. After depolymerization, the peak
207
J. Park et al. Fuel 217 (2018) 202–210
of the bio-oil produced from OW-E shifted from 900 m/z (Fig. S1) to comparable to that observed in the depolymerization of OW-F and OW-
500 m/z. A similar peak shift to the low-molecular-weight range was E. Therefore, despite the low content of labile ether linkages in the PW-
observed for the bio-oil produced from OW-F. In addition, the signals of derived lignins produced by the formasolv and ethanolsolv methods,
the high-molecular-weight species (> 2000 m/z) observed in the profile the high depolymerization activity associated with scEtOH + formic
of the OW-F sample (Fig. S1) disappeared in the profile of the bio-oil. acid is not significantly affected by the structure of lignin derived from
The molecular weight of the bio-oil produced from OW-K was similar to different wood sources. The lowest conversion (95.1%) and bio-oil yield
that from OW-E. This indicates that scEtOH + HCOOH is a highly (81.7 wt%) were achieved with PW-K because of the high content of
versatile medium for the effective depolymerization of various types of CeC linkages. However, even when the highly recalcitrant PW-K was
lignins with different chemical bonds. used as the feedstock, the bio-oil yield was higher than that achieved
The chemical species in the bio-oils produced from OW-E, OW-F, via the acid-catalyzed depolymerization of dealkaline lignin at 250 °C
and OW-K at 350 °C were analyzed using GC-TOF/MS and the chro- for 30 min (60 wt% THF soluble) [25], base-catalyzed depolymerization
matograms are shown in Fig. 3. The chemical species in the bio-oils of organosolv lignin at 300 °C (52 wt%) for 40 min [23], and metal-
were categorized into three groups: linear and branched short-chain catalyzed depolymerization of alkali lignin at 265 °C (45.8 wt%) for 1 h
oxygenated species in the short retention times (< 10 min), mono- [61] (see Table S6 for a comparison of the bio-oil yield). This indicates
aromatic species in the middle retention time (10–20 min), and long- that scEtOH + formic acid is a highly versatile and effective medium
chain fatty acid alkyl esters in the long retention time (> 20 min) zone. for the depolymerization of various types of lignin.
The linear and branched short-chain oxygenated species (see Table S4 The molecular weight distributions of the bio-oils produced from
for detailed information) may have originated from the decomposition the PW-derived lignins are shown in Fig. S3b. As in the case of the OW-
of ethanol and formic acid at 350 °C [39]. To confirm the decomposi- derived lignins and their corresponding bio-oils, the peaks were shifted
tion of ethanol and formic acid, a mixture of 60 g of ethanol and 12 g of to the low-molecular-weight direction for the bio-oils produced from
formic acid (without lignin) was treated at 350 °C and 29–35 MPa for PW-derived lignin. The bio-oils produced from the different lignin
30 min and the product was analyzed using GC-TOF/MS, as shown in samples had similar molecular weight distributions. Again, this clearly
Fig. S4. Similar types of oxygenated species were formed during the indicates the effective depolymerization of the various types of lignin in
blank reaction. At 250–300 °C, relatively less of the short-chain oxy- the scEtOH + formic acid medium.
genated species was observed (Fig. S5), indicating less ethanol de- The GC-TOF/MS chromatograms of the bio-oils produced from the
composition at the low liquefaction temperatures. Among the mono- PW-derived lignins are shown in Fig. 4. Again, the main chemicals were
aromatic species, the most abundant compounds were guaiacol, syr- mono-aromatic species in the middle retention time and short-chain
ingol, and their alkylation compounds (e.g., creosol and 3,5-dimethoxy- oxygenated species were also present in the short reaction time zone, as
4-hydroxy toluene). The long-chain fatty acid ethyl esters (e.g., octa- in the case of the bio-oils produced from the OW-derived lignins. The
decanedioic acid and E-11-hexadecenoic acid ethyl ester) in the bio-oil absence of syringol in the bio-oil produced from the PW-derived lignin
produced from the OW-derived lignins may have originated from the samples is quite straightforward because the lignins were derived from
esterification of woody fat with ethanol [60]. To confirm this, Soxhlet softwood. However, the bio-oil produced from PW-derived lignin did
extraction of OW in hexane for 6 h was conducted and the recovered not contain long-chain fatty acid esters and acid. This could be because
liquid extract is shown in Fig. S6; a substantial amount of woody fat of the absence of fatty acid species in the PW.
could be extracted. Similarly, long-chain fatty acids and their ester
species were detected in the OW-derived lignins (Fig. S7). 4. Conclusion
As shown in Fig. S5, the low-temperature liquefaction resulted in a
relatively large amount of carbonyl-containing or double bond-con- Six different types of lignin samples were depolymerized in a mix-
taining branched mono-aromatics such as 1-(4-hydroxy-3-methox- ture of supercritical ethanol (scEtOH) and formic acid at temperatures
yphenyl)-2-propanone (8, Fig. S5a), 2,6-dimethoxy-4-(2-propenyl)- of 250–350 °C. The content of ether linkages in the samples was highly
phenol (10, Fig. S5a), and 3,5-dimethoxy-4-hydroxyphenylacetic acid dependent on the wood source (i.e., whether it was derived from oak-
(15, Fig. S5a). This suggests that the deoxygenation and hydrogenation wood, OW or pinewood, PW) and the lignin isolation method (etha-
ability associated with scEtOH may be lower at the lower reaction nolsolv, E; formasolv, F; Klason, K). This may be because the formation
temperatures. This agrees well with the EA results, as listed in Table S5. of condensed linkages at the expense of ether linkages is dependent on
As the reaction temperature increased, the O/C ratios of the bio-oil the presence of the sinapyl alcohol unit and the isolation conditions.
produced from the OW-derived lignins decreased and the HHV of the The amount of ether linkages in the lignin samples played an important
bio-oil increased. For example, when the reaction temperature in- role in determining the bio-oil yield in the low-to-medium temperature
creased from 250 to 350 °C, the O/C ratio of the bio-oil produced from range; for example, the content of ether linkages followed the order:
OW-K decreased from 0.31 to 0.25 and the HHV increased from 25.8 to OW-F (129.3 µmol g−1) > OW-E (116.4 µmol g−1) > OW-K
30.4 MJ kg−1. −1
(83.4 µmol g ), and the bio-oil yield followed the order: OW-F
(77.5 wt%) > OW-E (68.0 wt%) > OW-K (18.7 wt%) at 250 °C. On
3.3. Comparison of the reactivity of the lignin samples derived from OW and the other hand, at the high temperature of 350 °C, all the lignin sam-
PW ples, even PW-K with an extremely low amount of ether linkages
(23.0 µmol g−1), could be converted effectively into high-yield bio-oil
As discussed in Section 3.1, the lignin samples produced from PW (> 81 wt%). The compositional analysis suggested that during the de-
had lower V + S contents than those produced from OW, suggesting polymerization, hydrodeoxygenation and hydrogenation occurred to
that the PW-derived lignin samples may have lower depolymerization some degree.
activity. To examine the difference in the depolymerization behavior
depending on the wood sources, the PW-derived lignin samples were Acknowledgements
depolymerized in a mixture of scEtOH and formic acid at 350 °C for 1 h.
During the course of liquefaction, the pressure in the reactor increased This work was supported by the New and Renewable Energy Core
from 30 to 46 MPa. As listed in Table S3, when PW-F was used, a high Technology Program (No. 20143030090940) of the Korea Institute of
conversion of 99.4% with a high bio-oil yield of 90.7 wt% and a low Energy Technology Evaluation and Planning (KETEP) financed by the
solid residue yield of 0.6 wt% were achieved. The use of PW-E resulted Ministry of Trade, Industry and Energy, Republic of Korea. Additional
in a slightly lower conversion (97.1%) and a lower bio-oil yield (88.1 wt support from the Technology Development Program to Solve Climate
%) as compared to those of PW-F. This depolymerization behavior is Changes of the National Research Foundation (NRF) funded by the
208
J. Park et al. Fuel 217 (2018) 202–210
209
J. Park et al. Fuel 217 (2018) 202–210
210