Production of cellulosic gasoline: Via levulinic ester self-condensation

Article abstract of DOI:10.1039/c8gc01432a

Most biomass to biofuel processes are limited to the production of linear or minimally branched hydrocarbons. Motor gasoline, however, consists of highly branched linear and/or cyclic alkanes. This work describes the optimization of the levulinic ester self-condensation reaction and the efficient conversion of its products, which are highly branched cyclopentadienes, into a mixture of substituted cyclopentanes with high octane ratings and excellent density and flow properties.

Full text of DOI:10.1039/c8gc01432a

Green Chemistry
View Article Online
View Journal
PAPER
Production of cellulosic gasoline via levulinic ester
self-condensation†
Cite this: DOI: 10.1039/c8gc01432a
a
Zheng Li, a,b Andrew L. Otsuki a and Mark Mascal
*
Most biomass to biofuel processes are limited to the production of linear or minimally branched hydro-
carbons. Motor gasoline, however, consists of highly branched linear and/or cyclic alkanes. This work
describes the optimization of the levulinic ester self-condensation reaction and the efficient conversion
of its products, which are highly branched cyclopentadienes, into a mixture of substituted cyclopentanes
with high octane ratings and excellent density and flow properties.
Received 8th May 2018,
Accepted 16th July 2018
DOI: 10.1039/c8gc01432a
rsc.li/greenchem
carbons with RON values between 70 and 80.4,5 Compound
3 can be derived in high yield from levulinic acid 1 (RvH), a
Introduction
The production of bio-based fuels by the condensation of bio- platform molecule of unrivaled potential which can be gen-
genic aldehydes or ketones with carbohydrate-derived platform erated directly from raw, cellulosic biomass.6 We have also
molecules, often furfurals, followed by hydrodeoxygenation reported the synthesis of branched cycloalkanes via ketone
generally leads to C9+ products. The application of this intermediate 5, the product of an intramolecular aldol con-
approach has given rise to dozens of papers describing routes densation of 2,7-octanedione 4, a Kolbe dimer of levulinic
to diesel-type fuels, which consist mainly of linear alkanes in acid (Scheme 1).7
the C10–C20 range.1–3 The paraffinic fraction of motor gaso-
In yet another approach to levulinic acid oligomerization,
line, on the other hand, is comprised of hydrocarbons up to we recently introduced cyclopentadiene- and cyclopentane-
about C12, a key characteristic of which is the extent of based diesters 6 and 7 in the context of novel, renewable
branching in the carbon chains. In general, the more monomers with branched, carbocyclic cores of interest for
highly branched the molecule, the higher the performance the preparation of low surface energy materials.8 Although 6
of the fuel, as measured by the antiknock index, expressed had been previously synthesized by base-induced self-con-
in terms of research octane number or “RON”. Thus, densation of ethyl levulinate, yields were very low and the
while 2,2,4-trimethylpentane (RON = 100) performs well in structure was found to have been misassigned. Below, we
spark-ignition engines, its singly branched isomer 3-ethyl- describe the optimization of the dimerization reaction that
hexane has RON = 34, while the unbranched n-octane has leads to 6, as well as the conversion of a mixture of cyclic
RON = −20.
condensates of levulinic ester into C8–C13 branched hydro-
To date, few reports of cellulosic biomass-derived alkanes carbons that, like those in Scheme 1, are characteristic of
sufficiently branched to serve as biogasoline or even mid- motor gasoline.
performance gasoline blendstocks have surfaced. Those that
do emerge generally involve the cracking, pyrolysis, or gasifi-
cation of biomass or its derivatives in combination with
catalytic upgrading or reforming methods commonly
employed in the petroleum industry, i.e. essentially using
biomass as a feed for refinery practices. Recently, we have
shown that the intrinsic branching in the angelica lactone
dimer 3 made it possible to produce branched C7–C10 hydro-
aDepartment of Chemistry, University of California Davis, 1 Shields Avenue, Davis,
CA, 95616, USA. E-mail: mjmascal@ucdavis.edu
bCollege of Energy, Xiamen University, Xiangan South Road, Xiamen, Fujian,
Scheme 1 Reagents and conditions: a. K10 montmorillonite, 165 °C,
92%; b. K2CO3, 70 °C, 94%; c. H2, Ir-ReOx/SiO2, 220 °C, 88%; d. −e−,
KOH/MeOH, Pt–Pt, undivided cell, 65%; e. PTSA, H2, Pd/C, hexane,
80 °C; f. H2, Pd/C, Al(OTf)3, 220 °C, 85% over two steps.
361102, China
†Electronic supplementary information (ESI) available. See DOI: 10.1039/
c8gc01432a
This journal is © The Royal Society of Chemistry 2018
Green Chem.

View Article Online
Paper
Green Chemistry
which was 9.0 ppm too far downfield. We therefore unequivo-
cally assign structure 6 to the base-induced self-condensation
Results and discussion
The sodium ethoxide catalyzed aldol condensation of ethyl product of ethyl levulinate, and conclude the anomalous shift
levulinate 1 (RvEt) was first reported by Duden and Freydag value for C4 in the computational model is the result of a
as early as 1903,9 and revisited by Shimizu in 1950.10 limitation in the method.
The structure of the dimer was (wrongly) assigned as the
The synthesis of 6 was initially reproduced exactly as
2-(2-carboxyethyl)-4-methylcyclopenta-1,3-diene-1-carboxylic acid described in the literature (Scheme 2),9 which involved the
system 8, and the yield of the product was <10%, due mainly reaction of 1 (RvEt) with an equimolar quantity of sodium
to the competing hydrolysis of the starting material to sodium ethoxide in cold ethanol followed by two days of standing at
levulinate in the course of the condensation reaction. We room temperature. Filtration of the reaction mixture, evapor-
repeated the procedure and determined that the correct struc- ation of the solvent, and neutralization with aq H2SO4 led to
ture was in fact the 3-(carboxymethyl)-2,4-dimethylcyclopenta- the separation of an oil which proved to be mainly a monoester
1,3-diene-1-carboxylic acid system 6, the hydrogenation of of 6. The crude product was therefore dissolved in ethanol, a
which proceeded smoothly to cyclopentane 7.8 Considering few drops of H2SO4 were added, and the mixture was heated at
the potential usefulness of 6 and 7 as renewable platforms for 50 °C overnight prior to analysis. The yield of 6 that we
materials and biofuel production, we undertook to examine obtained by this method (5%) was consistent with that in the
the chemistry in greater detail and optimize the condensation literature.9,10 As the condensation reaction proceeds, the pro-
reaction to a standard of preparative usefulness.
duction of H2O leads to hydrolysis of the starting ester, and
The rationale for assigning structure 8 to the self-conden- accumulation of the sodium levulinate salt was evident as the
sation of ethyl levulinate was reasonable, even if it ultimately reaction gradually thickened to an unstirrable paste. It was
led to the wrong conclusion. Duden and Freydag considered therefore proposed that a drying agent be included in the
both structures 6 and 8, but inclined toward the latter based mixture, and to this end Na2SO4, CaO, 3 Å molecular sieve,
on the notion that the kinetic product (enolization to the and triethyl orthoformate were evaluated. In all cases, signifi-
methyl group) should dominate the condensation chemistry.9 cant improvements in yield were noted. Bases other than
With the benefit of modern analytical techniques, we quickly sodium ethoxide were also tried, including sodium carbonate,
ascertained that the NMR data supported structure 6. In par- triethylamine, and 1,8-diazabicyclo[5.4.0]undec-7-ene (DBU),
ticular, 13C-NMR using a DEPT (Distortionless Enhancement however no production of 6 was observed in any of these cases.
by Polarization Transfer) pulse sequence showed peaks for It was found that the reaction time could be reduced from
four methyl and four methylene groups, rather than the 48 hours to only 5 hours if the temperature was raised to
three methyl and five methylene groups that would be 50 °C, and this also had a positive effect on yield (Table 1).
expected for 8. Computational modeling of the NMR shielding Concentration was determined to be another key parameter,
tensors of 6 using the gauge-independent atomic orbital where increasing dilution improved the yield of 6 to a
(GIAO) ab initio method gave a satisfactory agreement for all maximum of 40% (Table 1, entry 8). The remaining mass
shifts apart from the 13C peak for the carbon at the 4-position. balance of reactions at full conversion was found to be mainly
Thus, while the absolute average deviation from the experi- a complex mixture of trimeric products.
mental shifts for all carbons apart from C4 was 0.95
The varying degrees to which self-condensates of ethyl levu-
0.72 ppm, the calculated value for C4 was 8.1 ppm downfield linate could be reduced was explored. Simple hydrogenation of
of its observed shift. The computation of shielding tensors for 6 over Pd/C gave a high yield of 7 (92%), which is formally an
alternative conformers led to only minor variations in the annulated adipic ester, as mixture of stereoisomers for which
data. This anomaly led us into the literature to determine no separation was attempted. We have previously noted the
whether (1) there was a potential issue with our structural
assignment, or (2) the applied method systematically overesti-
mates 13C shifts in related systems. A search of the Reaxys
database turned up a single example of a 1-carboxy-4-methyl-
substituted cyclopentadiene with reported 13C NMR data, i.e.
2,3,4,5-tetramethylcyclopenta-1,3-diene-1-carboxylic acid 9.11
Interestingly, the modeled 13C NMR shifts show an absolute
Scheme 2 Self-condensation of ethyl levulinate. Reagents a. NaOEt,
average deviation from the reported values of only 1.08
Na2SO4, EtOH. For conditions, see Table 1; b. EtOH, H2SO4, 50 °C,
0.56 ppm with the exception of C4, the calculated shift for overnight.
Green Chem.
This journal is © The Royal Society of Chemistry 2018

View Article Online
Green Chemistry
Paper
Table 1 Conversion of 1 (RvEt) to 6 under various conditionsa
Entry
Volume of EtOH (mL)
T (°C)
t (h)
Conversion of 1 (%)
Yield of 6 (%)
1
2
3
4
5
6
7
8
12
20
30
30
30
40
40
40
60
60
60
60
50
50
50
50
50
50
50
70
50
50
70
35
5
5
1.25
2.5
5
5
7.5
5
100
100
88
94
100
84
100
100
75
87
95
22
28
17
26
28
36
33
40
38
34
33
35
9
5
10
11
12
7.5
7.5
24
81
a Reaction conditions: 5.0 g of ethyl levulinate, 1.5 g of Na2SO4, 2.4 g NaOEt in the stated volume of ethanol at the stated temperature. After the
stated reaction period, the mixture was neutralized to pH 2, filtered, and subjected to esterification conditions (see ESI).
potential application of 7 as a novel monomer.8 For the pur- ing 7 alongside alcohol and ester products, but little hydro-
poses of producing a biofuel, the complete deoxygenation of carbon (Table 2, entry 1). Our follow-up approach was to
dimer 6 was undertaken. We have established in previous work include a Lewis acid, in this case γ-Al2O3, to promote the de-
that the most hydrogen-economic approach to generate hydro- carboxylation reaction. Adjustment of conditions eventually
carbons from carboxylic acids is thermal decarboxylation, led to a 94% yield of C8–C10 branched cycloalkanes, quantified
which in effect removes three C–O bonds at the cost of no by GC-MS against an internal standard (see ESI†).
external hydrogen.5 This is typically accomplished by heating
As shown in Scheme 3, diester 6 is first saturated to cyclo-
at a high temperature in the presence of a hydrogenation cata- pentane 7. The expectation was that 7 would decarboxylate to
lyst. A test reaction with 6 and Pd/C under 4 MPa H2 at 320 °C produce the C8 product 10, and this was indeed observed
in tetradecane solvent gave a partially reduced mixture includ- (42%). However, we were not able to fully suppress reduction
Table 2 Conversion of 6 to cycloalkanes
Entry
Catalyst
Atmosphere
T (oC)
t (h)
Total cycloalkane yield (%)
1
2
3
Pd/C
Pd/Al2O3
Pd/Al2O3
4 MPa H2
2 MPa H2
4 MPa H2
320
300
320
5
3
5
<5%
34
94
Conditions: 0.80 g of 6, 0.15 g of catalyst, 10 mL of tetradecane, temperature and time as indicated.
Scheme 3 Conversion of 6 to branched cycloalkanes. Observed products are bolded. Reagents and conditions: a. 0.80 g of 6, 0.15 g of Pd–Al2O3,
10 mL of tetradecane, 4 MPa of H2, 320 °C, 5 h.
This journal is © The Royal Society of Chemistry 2018
Green Chem.

View Article Online
Paper
Green Chemistry
of the carbonyl groups, and a mixture of C9 (assigned as 12 + allowed an approximation of the yield to be made based on a
14, 44%) and C10 (16, 8%) products was also seen. Estimation weighted average C10.5H21 product. The reaction was repeated
of the RON of these hydrocarbons using the recently published twice using 15 g of the crude dimer–trimer mixture and the
artificial neural network-based group contribution method of products combined to give an isolated yield of 12.70 g (83%) of
Kubic et al.12 gives values of 101 (10), 91 (12), 92 (14) and 85 hydrocarbons. No solvents were used to aid removal of the
(16), with a product distribution weighted average of RON = product from the reaction vessel, but rather the mixture was
95, which is higher than standard grades of commercial immediately distilled and the fractions analyzed by GC-MS to
gasoline.
provide an approximate product distribution. A cut of C8–10
As noted above, the mass balance of the reaction to product was collected below 200 °C, accounting for 47 wt% of
produce 6 from ethyl levulinate was mainly a mixture of tri- the mixture and was contaminated with only trace amounts of
meric (C15) products. Rather than esterifying and isolating 6 higher hydrocarbons. Between 200–240 °C mainly C12–13
from this mixture, we submitted the crude condensate to the cycloalkanes were observed, amounting to 48 wt% of the
same decarboxylation–hydrodeoxygenation conditions described product. The pot residue (bp >240 °C) was mainly C13 products
above, which yielded a mixture of branched C8–C13 alkanes. contaminated with trace C12 cycloalkanes and comprised
The overall process involved heating ethyl levulinate with about 5% of the total mass. The density of the distillate
ethanolic NaOEt in the presence of Na2SO4, filtering, and measured at 15 °C was 0.811 g mL−1 and its kinematic vis-
evaporating the solvent to give a crude oil product with a calcu- cosity at −20 °C was 3.38 mm2 s−1. Given that the typical
lated mass recovery of nearly 100%. This material was then density of motor gasoline is 0.74 g mL−1 13
the synthetic fuel
,
treated as shown in Scheme 4 under solvent-free conditions. mixture has a good balance of properties – satsifactory low
The reaction was first carried out with 3.50 g of the crude temperature viscosity and excellent density.
dimer–trimer mixture with careful quantification of the
product (see ESI†), leading to isolated hydrocarbon yields in tering and reusing it three times without washing or treatment
the range of 91–93%. of any kind. The conversion of crude dimer–trimer mixture
Finally, we tested the recyclability of Pd/Al2O3 catalyst by fil-
The product distribution included the same C8–C10 cyclo- was at least 95% in all runs.
pentanes seen in Scheme 3 alongside some acyclic homo-
logues, together with cyclic C12 and C13 products that evidently
derive from of ethyl levulinate trimers that have been triply
and doubly decarboxylated, respectively. The structures of
Conclusions
these products are proposals based on the most likely sites of
addition of a third molecule of ethyl levulinate to 6. In fact,
the mixture was complex, including multiple stereoisomers
and likely also regioisomers. Despite this complexity, the
nearly even distribution of hydrocarbons from C8 to C13
In conclusion, a Stobbe-like condensation of ethyl levulinate
gives a mixture of dimers and trimers in the form of substi-
tuted cyclopentadienes. The use of a dehydrating agent during
the reaction shuts down a competing hydrolysis of the ester
starting material that had previously resulted in the descrip-
Scheme 4 Conversion of the product of the base-induced self-condensation of ethyl levulinate to branched alkanes. Reagents and conditions:
a. NaOEt, Na2SO4, EtOH, 70 °C, 5 h; b. 10% Pd–Al2O3, 4 MPa H2, 320 °C, 5 h.
Green Chem.
This journal is © The Royal Society of Chemistry 2018

View Article Online
Green Chemistry
Paper
tion of very low product yields. Saturation of the cyclo-
pentadiene is facile, and attempts to decarboxylate the result-
ing cyclopentanedicarboxylic acid/ester proceeded best under
2 H. Li, A. Riisager, S. Saravanamurugan, A. Pandey,
R. S. Sangwan, S. Yang and R. Luque, ACS Catal., 2018, 8,
148–187.
high-temperature hydrogenation conditions, leading to
a
3 M. Kaldstrom, M. Lindblad, K. Lamminpaa, S. Wallenius
and S. Toppinen, Ind. Eng. Chem. Res., 2017, 56,
13356–13366.
4 M. Mascal, S. Dutta and I. Gandarias, Angew. Chem., Int.
Ed., 2014, 53, 1854–1857.
mixture of decarboxylation and hydrodeoxygenation products.
The dimer 6 produced a mixture of branched C8–C10 cyclo-
pentanes with high octane ratings. The raw dimer/trimer
mixture likewise delivered C8–C10 plus C12 and C13 products in
high yield with very good fuel properties which, being only
four efficient steps removed from raw biomass, may provide a
competitive basis for cellulosic gasoline production.
5 F. Chang, S. Dutta and M. Mascal, ChemCatChem, 2017, 9,
2622–2626.
6 (a) D. J. Hayes, S. W. Fitzpatrick, M. H. B. Hayes and
J. R. H. Ross, in Biorefineries-Industrial Processes and
Products: Status Quo and Future Directions, ed. B. Kamm,
P. R. Gruber and M. Kamm, Wiley-VCH, Weinheim, 2006,
vol. 1, pp. 144–160; (b) C. Antonetti, D. Licursi, S. Fulignati,
G. Valentini and A. M. R. Galletti, Catalysts, 2016, 6, 196;
(c) M. Mascal and S. Dutta, Top. Curr. Chem., 2014, 353,
41–84.
Conflicts of interest
There are no conflicts to declare.
7 L. Wu, M. Mascal, T. J. Farmer, S. P. Arnaud and
M.-A. Wong Chang, ChemSusChem, 2017, 10, 166–170.
8 S. Pérocheau Arnaud, L. Wu, M.-A. Wong Chang,
J. W. Comerford, T. J. Farmer, M. Schmid, F. Chang, Z. Li
and M. Mascal, Faraday Discuss., 2017, 202, 61–77.
9 P. Duden and R. Freydag, Ber. Dtsch. Chem. Ges., 1903, 36,
944–952.
10 S. Shimizu, Nippon Nogei Kagaku Kaishi, 1950, 23, 288–294.
11 R. Koschinsky, T.-P. Köhli and H. Mayr, Tetrahedron Lett.,
1988, 29, 5641–5644.
Acknowledgements
This work benefited from a grant from the China Scholarship
Council (to Z. L.). Density and viscosity measurements were
performed by Dr Benjamin G. Harvey of the Navy Air Warfare
Center, Weapons Division, China Lake, California, USA. We
also thank Dr Cameron M. Moore at Los Alamos National
Laboratory, Los Alamos, New Mexico, USA for help with RON
calculations.
12 W. L. Kubic Jr., R. W. Jenkins, C. M. Moore,
T. A. Semelsberger and A. D. Sutton, Ind. Eng. Chem. Res.,
2017, 56, 12236–12245.
References
13 Motor Gasolines Technical Review, Chevron Corporation,
1 U. Addepally and C. Thulluri, Fuel, 2015, 159, 935–
2009,
(https://www.chevron.com/-/media/chevron/oper-
942.
ations/documents/motor-gas-tech-review.pdf).
This journal is © The Royal Society of Chemistry 2018
Green Chem.