Conversion of levulinic acid derived valeric acid into a liquid transportation fuel of the kerosene type

Article abstract of DOI:10.1016/j.molcata.2013.11.015

In the transformation of lignocellulosic biomass into fuels and chemicals carboncarbon bond formations and rising hydrophobicity are highly desired. The ketonic decarboxylation fits these requirements perfectly as it converts carboxylic acids into ketones forming one carboncarbon bond and eliminates three oxygen atoms as carbon dioxide and water. This reaction is used, in a cascade process, together with a hydrogenation and dehydration catalyst to obtain hydrocarbons in the kerosene range from hexose-derived valeric acid. It is shown that zirconium oxide is a very selective and stable catalyst for this process and when combined with platinum supported on alumina, the oxygen content was reduced to almost zero. Furthermore, it is demonstrated that alumina is superior to active carbon, silica, or zirconium oxide as support for the hydrogenation/dehydration/hydrogenation sequence and a palladium-based catalyst deactivated more rapidly than the platinum catalyst. Hence, under optimized reaction conditions valeric acid is converted into n-nonane with 80% selectivity (together with a 10% of C₁₀-C₁₅ hydrocarbons) in the organic liquid phase upto a 100:1 feed to catalyst ratio [w/w]. The oxygen free hydrocarbon product mixture (85% yield) meets well with the boiling point range of kerosene as evidenced by a simulated distillation. In the gas phase, butane was detected together with mainly carbon dioxide.

Full text of DOI:10.1016/j.molcata.2013.11.015

Journal of Molecular Catalysis A: Chemical 388–389 (2014) 116–122
Contents lists available at ScienceDirect
Journal of Molecular Catalysis A: Chemical
journal homepage: www.elsevier.com/locate/molcata
Conversion of levulinic acid derived valeric acid into a liquid
transportation fuel of the kerosene type
Avelino Cormaa,∗, Borja Oliver-Tomasa, Michael Renza,∗, Irina L. Simakovab
a Instituto de Tecnología Química, Universitat Politècnica de Valencia-Consejo Superior de Investigaciones Científicas (UPV - CSIC), Av. de los Naranjos s/n,
Valencia E-46022, Spain
b Boreskov Institute of Catalysis, Lavrentieva 5, Novosibirsk 630090, Russian Federation
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 26 July 2013
Received in revised form 28 October 2013
Accepted 11 November 2013
Available online 19 November 2013
In the transformation of lignocellulosic biomass into fuels and chemicals carbon carbon bond forma-
tions and rising hydrophobicity are highly desired. The ketonic decarboxylation fits these requirements
perfectly as it converts carboxylic acids into ketones forming one carbon carbon bond and eliminates
three oxygen atoms as carbon dioxide and water. This reaction is used, in a cascade process, together with
a hydrogenation and dehydration catalyst to obtain hydrocarbons in the kerosene range from hexose-
derived valeric acid. It is shown that zirconium oxide is a very selective and stable catalyst for this process
and when combined with platinum supported on alumina, the oxygen content was reduced to almost
zero. Furthermore, it is demonstrated that alumina is superior to active carbon, silica, or zirconium oxide
as support for the hydrogenation/dehydration/hydrogenation sequence and a palladium-based catalyst
deactivated more rapidly than the platinum catalyst. Hence, under optimized reaction conditions valeric
acid is converted into n-nonane with 80% selectivity (together with a 10% of C10–C15 hydrocarbons) in
the organic liquid phase upto a 100:1 feed to catalyst ratio [w/w]. The oxygen free hydrocarbon product
mixture (85% yield) meets well with the boiling point range of kerosene as evidenced by a simulated
distillation. In the gas phase, butane was detected together with mainly carbon dioxide.
Keywords:
Catalyst stability
Ketonic decarboxylation
Ketone hydrodeoxygenation
Pt/alumina
Zirconium oxide
© 2013 Elsevier B.V. All rights reserved.
1. Introduction
building block and manifold conversions into chemicals and fuels
have been proposed.
Carboxylic acids play a predominant role in biomass transforma-
tions into fuels and chemicals when lignocellulose is considered as
a raw material. On one hand they form part of the bio-oils obtained
by pyrolysis of lignocellulosic biomass. In these complex mixtures,
volatile carboxylic acids such as acetic, formic acid, and higher
homologues (acrylic acid, butyric acid, isobutyric acid, valeric acid,
isovaleric acid, etc.) can be found in lower concentration [1–4]. On
the other hand, carboxylic acids such as levulinic acid are available
in a more selective way with yields of 55% (with respect to the hex-
ose content) from ground hardwood [5] or upto 70% from other
raw materials [6] via hydrolysis of hexoses like in the Biofine pro-
cess [5]. In aqueous acid conditions hexoses are first dehydrated to
5-hydroxymethylfurfural (HMF) and this compound is then rehy-
drated and cleaved into levulinic acid and formic acid (Eq. (1)) [7].
Therefore, levulinic acid is generally considered as a biomass-based
(1)
For both, fuels and chemicals, the reduction of the oxygen con-
tent is an important issue. Hereby, the reduction of levulinic acid
into valeric (pentanoic) acid is straightforward via ␥-valerolactone
(GVL) [8,9]. In a certain version of the reaction exploiting the
raw material to a maximum, formic acid co-produced in the re-
hydration of HMF has been used as hydride donor in the first
step [10,11]. For further reducing the oxygen content when treat-
ing carboxylic acids, and increasing the molecular weight at the
same time which is also highly desired in biomass transformations,
the ketonic decarboxylation is a very suitable and environmentally
∗
Corresponding authors at: Instituto de Tecnología Química, Avda. los Naranjos
s/n, Valencia 46022, Spain. Tel.: +34 963870000.
E-mail addresses: acorma@itq.upv.es (A. Corma), mrenz@itq.upv.es (M. Renz).
1381-1169/$ – see front matter © 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.molcata.2013.11.015

A. Corma et al. / Journal of Molecular Catalysis A: Chemical 388–389 (2014) 116–122
117
Scheme 1. Kinetically favored mechanism on zirconium oxide as established by theoretical DFT calculations (see [18]). Two molecules of carboxylic acid are adsorbed in
a different way onto the zirconium oxide surface, one by double deprotonation and the second by dehydroxylation. Then a carbon carbon bond is formed between both
molecules resulting in a ␤-keto acid. The latter is decarboxylated easily and protonation and desorption provides the ketone.
Scheme 2. Reaction sequence involving ketonic decarboxylation, hydrogenation of the resulting ketone to an alcohol, dehydration of this compound, and hydrogenation of
the intermediary olefin. For R = n-octyl see [21]; R = Me, this work.
benign reaction [12]. This reaction is carried out at elevated temper-
atures (350–400 ◦C) in the presence of a catalyst and two molecules
of carboxylic acid are fused to a ketone, expelling one molecule of
carbon dioxide and one of water (Eq. (2)). Further reagents are not
required. Hence, from this sequence, i.e., dehydration, rehydration,
hydrogenation, and ketonic decarboxylation, compounds such as
5-nonanone are available from hexoses (Eq. (3)).
“stabilized” for the reaction in a different way. One molecule is
deprotonated twice to yield an enediolate and the other molecule
is dehydroxylated. Then, the methylene group attacks nucleophli-
cally the carbonyl group of the dehydroxylated acid forming a
␤-keto acid. The latter is decarboxylated to an enolate. Protonation
and desorption of all reaction products allow to close the catalytic
cycle.
This theoretical study has been carried out with zirconium
dioxide as catalyst which has been shown to be the most active
material among zirconium-cerium mixed oxide, cerium oxide, alu-
mina or silica [18]. The catalytic advantages of zirconium oxide
have been confirmed in a further study [20]. Therefore, this is a
material to be tested in multi-step transformations such as pre-
sented for the conversion of fatty acids into diesel and lubricants
[21]. In a cascade reaction the fatty acids have been converted into
fatty ketones over magnesium oxide, the ketones hydrogenated
to alcohols and hydrocarbons have been obtained by subsequent
dehydration and olefin hydrogenation (Scheme 2). The maximum
selectivity towards the corresponding hydrocarbon, i.e., tricosane
from lauric acid, was 58%. However, except for one case and using
approximately the same amount of feed as the weight of catalyst,
deoxygenation was lower than 80% since, generally, the corre-
sponding ketone was observed with more than 20% selectivity. The
activity of the ketonic decarboxylation catalyst decreased with time
on stream.
(2)
(3)
In the present manuscript, we show the possibility of obtain-
ing good yields of kerosene from levulinic acid in a two-bed one
fixed-bed continuous reactor by combining ZrO2 and Pt/Al2O3 in
the presence of hydrogen. The catalyst is stable under reaction
conditions.
Probably the ketonic decarboxylation also plays a decisive role
in the stabilization of the above mentioned bio-oils. When using
them as a co-feed in fluid catalytic cracking units (FCCs) they need
a significant up-grade due to their high oxygen content [13]. Cat-
alysts (e.g., alkali and alkaline earth metal oxides supported on
amorphous silica alumina) and reaction conditions (450 ◦C) [14]
resemble those of the ketonic decarboxylation for which a catalytic
action of these oxides has been established in other cases [15,16].
Furthermore, increased decarboxylation and dehydration has been
observed for the pyrolysis in presence of these catalysts together
with the reduction of the carboxylic acid amount.
Hence, these examples show the increasing interest in the
ketonic decarboxylation for the production of ketones since first
performed more than a century ago [12]. Then, the reaction mech-
anism [17] has been studied in detail by means of theoretical
DFT calculations [18,19]. Thereby, it has been found that a reac-
tion mechanism via a ␤-keto acid requiring hydrogen atoms in
␣-position is kinetically favored (Scheme 1). In a first step, two
molecules of carboxylic acid are adsorbed onto the surface and
2. Experimental part
Valeric acid (98%) and decanoic acid (99%) were purchased
from Sigma–Aldrich and 5-nonanone from Acros. All reagents were
used without any further purification. ZrO2, Al2O3, and SiO2 were
received from ChemPur and activated carbon from Norit. H2PtCl6·6
H2O (99%, Aldrich) and Pd(OAc)2 (Aldrich, 99%) were employed as
noble metal precursors.
The metal catalysts were prepared by the incipient wetness
impregnation method. Solutions with the corresponding concen-
trations of H2PtCl6·6 H2O and of Pd(OAc)2 were prepared in
water and toluene, respectively. The metal oxide supports were
impregnated as pellets (0.4–0.8 mm) and dried in vacuum. In the
case of Pd(OAc)2 several impregnation steps were required to

118
A. Corma et al. / Journal of Molecular Catalysis A: Chemical 388–389 (2014) 116–122
Fig. 2. Conversion and selectivity in the organic liquid phase for the ketonic decar-
boxylation of valeric acid over a fixed bed of zirconium oxide placed into continuous
reactor at shorter contact time and with reactivation steps. ꢀ conversion pentanoic
acid 1st cycle, + selectivity 5-nonanone 1st cycle, ꢁ conversion pentanoic acid 2nd
cycle, − selectivity 5-nonanone 2nd cycle, ꢁ conversion pentanoic acid 3rd cycle, ×
selectivity 5-nonanone 3rd cycle, ꢀ conversion pentanoic acid 4th cycle, * selectivity
5-nonanone 4th cycle.
Fig. 1. Conversion and selectivity in the organic liquid phase for the ketonic
decarboxylation of decanoic acid over a fixed bed of zirconium oxide placed into
continuous reactor. ꢀ conversion pentanoic acid, ꢀ selectivity 10-nonadecanone;
for calculation of the yield see corresponding formula in the experimental part.
Reaction conditions: feed rate for decanoic acid: 0.31 mL/min; nitrogen gas flow:
50 mL/min; 400 ◦C reaction temperature.
Reaction conditions: 1.00 g of catalyst was employed instead of 2.5 g; feed rate for
valeric acid: 0.24 mL/min; nitrogen gas flow: 50 mL/min; 400 ◦C reaction tempera-
ture. After each cycle the catalyst was calcined in-situ. The yield of the water phase
was 47% on average.
achieve the desired metal load. At the end of the impregnation
procedure, all the catalysts were dried in an oven overnight at
100 ◦C.
Reactions were carried out in a reactor described elsewhere
[18] with a back-pressure regulator placed at the exit of the reac-
tor for pressure control. When decanoic acid was employed as
substrate all parts of the reactor that were in contact with the
substrate were heated to 45 ◦C (melting point of decanoic acid
32 ◦C). In general, the feed was passed with a rate of 0.15 mL/min,
and reaction temperature of the catalytic bed was 400 ◦C. At the
exit of the reactor the product was condensed (0 ◦C) and analyzed
off-line by gas chromatography. In the ketonic decarboxylation
with ZrO2 the mass balance for the liquid phase was 97 2% (Cal-
culated with respect to the theoretical yield, resting the weight
of carbon dioxide and water). With Pt/Al2O3, Pt/SiO2, and Pt/C
as a second catalytic bed the weight balance was 85% (Calcu-
lated again with respect to the theoretical yield having in mind
the hydrogen introduced and resting the weight of carbon diox-
ide and two molecules of water). With Pd/Al2O3 and Pt/ZrO2, the
weight balance was 80 and 65%, respectively. With Pt/ZrO2 as single
bed the mass balance was 55–70%. Gaseous effluents were ana-
lyzed off-line on a GC apparatus equipped with three detectors.
Two detectors were thermal conductivity detectors (TCD) for the
hydrogen. Applying a hydrogen pressure of 40 bar and a hydrogen
flow of 470 mL/min the catalyst was treated at the boiling point
of the solvent used for the impregnation step during one hour
and at 400 ◦C for two hours. The heating ramp was 9 ◦C/min in all
cases.
The ketonic decarboxylation of decanoic acid was carried out at
a contact time (W/F) of 0.31 h employing 2.5 g of zirconium oxide.
When the reaction was carried out with valeric acid with a con-
tact time (W/F) of 0.074 h, 1.0 g of zirconium oxide was used. The
study of the bifunctional catalyst for ketonic decarboxylation and
hydrogenation in one bed, i.e., with 3% Pt/ZrO2, was carried out
with 2.5 g of catalyst. In the two-catalytic bed process 2.5 g of zir-
conium oxide was employed for the ketonic decarboxylation in
the first bed and 1.5 g of the hydrogenation catalyst placed down-
stream.
˚
analysis of H2 (separated on a 2 m molecular sieve 5 A column)
and of CO, CO2, and N2 (separated on a 2.5 m molecular sieve
13X column). The third detector was a flame ionization detec-
tor (FID) for the analysis of C1–C5 hydrocarbons separated on a
50 m Plot/Al2O3 column. Main components were CO2, CO, methane,
butane, and pentane depending on the catalyst of the second cat-
alytic bed.
The amount of water obtained for the ketonic decarboxylation
was 47% on average. Its yield might be lowered by incomplete
condensation. The amount of water obtained in the hydrodeoxy-
genation of 5-nonanone at 400 ◦C was 37% on average. Again, yield
might be lowered by incomplete condensation. The yields of water
phases for the combined ketonic decarboxylation and hydrodeoxy-
genation are stated in the captions of the corresponding figures.
The stoichiometrical amount would be two equivalents. A higher
amount might be due to carbon dioxide methanation and a lower
stoichiometric amount due to incomplete condensation as stated
before.
Fig. 3. Conversion and selectivity in the organic liquid phase for the ketonic decar-
boxylation of decanoic acid over a fixed bed of platinum supported onto zirconium
oxide placed into continuous reactor at different hydrogen pressures. ꢀ conversion
decanoic acid, ꢁ selectivity nonane, ꢂ selectivity to 10-nonadecanone, ꢀ selectivity
nonadecane, + selectivity to other hydrocarbons, * selectivity to decane.
Reaction conditions: feed rate for decanoic acid: 0.15 mL/min; hydrogen gas flow:
430 mL/min; 400 ◦C reaction temperature.
The metal-containing catalysts were activated (reduced) in-situ
in the reactor. The reactor was purged with nitrogen and then with

A. Corma et al. / Journal of Molecular Catalysis A: Chemical 388–389 (2014) 116–122
119
Fig. 4. Conversion of 5-nonanone over a Pt/alumina catalytic bed at different temperatures: (a) 100 ◦C, (b) 200 ◦C, (c) 300 ◦C, and (d) 400 ◦C. 5-nonanone can be hydrogenated
to 5-nonanol, the alcohol dehydrated and the resulting olefin again hydrogenated to yield n-nonane. ꢀ conversion 5-nonanone, ꢂ selectivity nonane, ꢁ selectivity to 5-
nonanol, ꢀ selectivity 4-nonene, + selectivity to other hydrocarbons (upto C8), * selectivity to oxygenated compounds.
Reaction conditions: 1.5 g of catalyst was employed; feed rate for 5-nonanone: 0.15 mL/min; hydrogen gas flow 470 mL/min. At 400 ◦C the yield of the water phase was 37%
on average.
The conversion has been calculated by the following formula:
The yield (Y) for the different organic compounds can be calcu-
lated by the following formula:
X/% = 100 − (S × MB/%)/100
Y/% = (S/% × X/% × MB/%)/10,000
being X = Conversion, S = Selectivity of the substrate in the organic
liquid, and MB = mass balance for the organic liquid.
being S = Selectivity of the corresponding compound in the liq-
uid organic phase, X = Conversion and MB = mass balance for the
organic liquid.
Fig. 6. Development of conversion and selectivity in the organic liquid phase
with time on stream when valeric acid was passed over a zirconium oxide and a
Pt/alumina bed. ꢀ conversion pentanoic acid, ꢀ selectivity 5-nonanone, ꢂ selectivity
nonane, ꢁ selectivity to C10–C15 hydrocarbons, + selectivity to other hydrocarbons
(upto C8), * selectivity to oxygenated compounds; for calculation of the yield see
corresponding formula in the experimental part.
Fig. 5. Conversion and selectivities in the organic liquid phase for the transforma-
tion of valeric acid over a zirconium oxide and a Pt/alumina bed at different contact
times W/F. ꢀ conversion pentanoic acid, ꢀ selectivity 5-nonanone, ꢂ selectivity
nonane, ꢁ selectivity to C10–C15 hydrocarbons, + selectivity to other hydrocarbons
(upto C8), * selectivity to oxygenated compounds.
Reaction conditions: hydrogen gas flow: 470 mL/min; 40 bar hydrogen pressure;
400 ◦C reaction temperature. The yield of the water phase was 2.1 equiv. per valeric
acid employed on average.
Reaction conditions: feed rate for valeric acid: 0.15 mL/min; hydrogen gas flow:
470 mL/min; 40 bar hydrogen pressure; 400 ◦C reaction temperature. The yield of
the water phase was 2.6 equiv. per valeric acid employed on average.

120
A. Corma et al. / Journal of Molecular Catalysis A: Chemical 388–389 (2014) 116–122
Fig. 7. Simulated distillation for the product mixture obtained when valeric acid
was passed over a zirconium oxide and a Pt/alumina bed (cf. Fig. 6) at a basis of a
kerosene boiling point range from 145 to 300 ◦C.
Fig. 9. Development of conversion and selectivity in the organic liquid phase
with time on stream when valeric acid was passed over a zirconium oxide and
a Pt/zirconium oxide bed. ꢀ conversion pentanoic acid, ꢀ selectivity 5-nonanone,
ꢂ selectivity nonane, ꢁ selectivity to C10–C15 hydrocarbons, + selectivity to other
hydrocarbons (upto C8), * selectivity to oxygenated compounds.
3. Results and discussion
Reaction conditions: feed rate for valeric acid: 0.15 mL/min; hydrogen gas flow:
470 mL/min; 40 bar hydrogen pressure; 400 ◦C reaction temperature. The yield of
the water phase was 1.2 equiv. per valeric acid employed on average.
The stability of zirconium oxide with time on stream was first
tested under the standard reaction conditions employed for the
selection of this metal oxide as active material for the ketonic
decarboxylation [18]. Hence, decanoic acid was passed as feed with
a contact time, defined as weight of catalyst per weight of feed
passed during one hour (W/F), of 0.31 h. Under these conditions
the conversion was complete and the selectivity towards the
corresponding ketone, i.e., 10-nonadecanone, was >95% in the
organic liquid (Fig. 1). This result does not change during 24 h time
on stream, when already 100 g of substrate was passed per gram of
catalyst and the mass balance for organic and aqueous liquid phase
was 97%. To better study catalyst deactivation, the contact time was
decreased (W/F = 0.074 h), so the conversion was lower than 100%.
The substrate feed was changed to the hexose biomass-derived
valeric acid. The latter was selected as biomass-derived starting
material for the present study. With these changes the initial
conversion was 85% and it decreased approximately 15 percentage
points to 70% when passing more than 90 g of substrate per gram of
catalyst (Fig. 2). At this point the reaction was stopped, the catalyst
reactivated by calcination (500 ◦C, 2 h, air flow of 50 mL/min)
and after this treatment, the initial activity and conversion was
practically recovered (Fig. 2). Hence, the catalytic cycle was carried
out four times with three intermediate activation steps. After the
whole reaction/reactivation sequence the values for the conversion
were approximately 5% lower than the initial ones indicating that
the catalyst can be subjected to cycles of reaction-regeneration.
At this point, we concluded that zirconium oxide was not only
an active and selective catalyst for the ketonic decarboxylation
as reported before [18] but also stable with time on stream.
Therefore, the oxide was considered as a suitable catalyst for the
first step in a combined process including ketonic decarboxylation
and hydrogenation. Furthermore, the change of the carrier gas
from nitrogen to hydrogen should not have a negative effect on
the catalytic performance as demonstrated before [20].
After the fourth in-situ calcination of the catalyst, a thermo-
gravimetric analysis showed that no organic deposits remained on
the surface. In fact, upto a temperature of 200 ◦C only 2% weight
loss was observed that was attributed to water desorption. In the
temperature range from 200 to 800 ◦C the weight reduction was
an additional one percent which might also be due to dehydration
reactions. The BET surface area of the used catalyst was signifi-
cantly decreased from 104 m2/g, measured with the fresh catalyst,
to 78 m2/g after the catalytic reaction.
Since the catalytic behavior of ZrO2 for the ketonic decar-
boxylation was quite good, we decided to prepare a bifunc-
tional metal/ZrO2 catalyst to perform in
a single catalytic
bed where the ketonic decarboxylation and the hydrogena-
tion/dehydration/hdyrogenation of the ketone was formed. It has
been shown that stearic acid [22] or tall oil derived fatty acids [23]
are decarboxylated in the presence of supported platinum or pal-
ladium catalysts with an excellent conversion and high yield of the
corresponding n − 1 hydrocarbons [24]. Interestingly, the reaction
temperature was also elevated, namely, between 300 and 350 ◦C.
An acidic tungsten/zirconium mixed oxide [25] has also been used
as catalyst support, but selectivity towards the corresponding n − 1
hydrocarbon did not exceed 60%. In our work, 3 wt.% of Pt was sup-
ported onto zirconium oxide, placed into a tubular stainless-steel
reactor and reduced in-situ at 400 ◦C during 2 h at 40 bar hydrogen
pressure. For a preliminary study, decanoic acid was preferred over
valeric acid, since the corresponding product, i.e., nonane was eas-
ier to condense and to analyze off-line than butane. From Fig. 3 it
can be seen that, indeed, nonane was the main product in the liq-
uid organic phase with selectivities ranging from 55 to 80% over the
Fig. 8. Development of conversion and selectivity in the organic liquid phase with
time on stream when valeric acid was passed over a zirconium oxide and a Pt/silica
bed. ꢀ conversion pentanoic acid, ꢀ selectivity 5-nonanone, ꢂ selectivity nonane,
ꢁ selectivity 4-nonene, + selectivity to other hydrocarbons (upto C8), * selectivity
5-nonanol.
Reaction conditions: feed rate for valeric acid: 0.15 mL/min; hydrogen gas flow: 470
mL/min; 40 bar hydrogen pressure; 400 ◦C reaction temperature. The yield of the
water phase was 0.75 equiv. per valeric acid employed on average.

A. Corma et al. / Journal of Molecular Catalysis A: Chemical 388–389 (2014) 116–122
121
whole range of tested hydrogen pressures (1–20 bar). The low mass
balance in the liquid phase (55–70%) indicates that further frag-
mentation reactions occur giving hydrocarbons in the gas phase.
Products with nineteen carbon atoms such as nonadecanone or
nonadecane were observed in significant amount (>5%) only at a
one bar hydrogen pressure.
The extensive fragmentation observed with the bifunctional
catalyst prompted us to separate the two processes, i.e., ketonic
decarboxylation and hydrodeoxygenation by using two consecu-
tive catalytic beds. The top bed was formed by ZrO2 which nicely
catalyzes the ketonic decarboxylation, while the second bed was
formed by Pt/Al2O3 as catalyst that should perform well the hydro-
genation reaction. However, in a previous step, the hydrogenation
of a ketone (5-nonanone) was studied on Pt/Al2O3 under differ-
ent reaction temperatures (100–400 ◦C) and pressures (1–40 bar).
The results given in Fig. 4 clearly shows that working at 40 bar
H2, it is possible to perform the hydrogenation of the ketone with
high activity and selectivity at 400 ◦C that matches the tempera-
ture required for the ketonic decarboxylation. Then, a process can
be used that involves one fixed bed reactor with two consecutive
catalytic beds: ZrO2 and Pt/Al2O3, by working at 400 ◦C and 40 bar
of hydrogen.
At this point, valeric acid was passed through a continuous flow
reactor at 400 ◦C with two catalytic beds, i.e., zirconium oxide (2.5 g)
and platinum supported on alumina (1.5 g) while the flow of valeric
acid was 9.0 g h−1. When the contact time was shortened to the half,
i.e., doubling the feed rate from 9.0 to 18.0 g h−1, the selectivity
towards the desired nonane was constant at an 80% level of the
liquid organic phase (yield of the organic phase 85%, Fig. 5) but the
conversion of the intermediary 5-nonanone decreased slightly and
10% of the ketone remained in the effluent of the continuous flow
reactor.
Valeric acid was passed upto a ratio of 100 g per gram of zirco-
nium oxide and 85% was recovered as liquid product. From Fig. 6 it
can be seen that the yield of nonane in the liquid organic phase
did not change with time on stream and no indication of cata-
lyst deactivation was detected. This is a considerable improvement
with respect to the previous system MgO-Pt/Al2O3 employed for
the synthesis of lubricants [21]. Probably, the better stability of the
first catalytic bed with time on stream, avoiding the contact of the
hydrodeoxygenation bed with unreacted carboxylic acids, was also
the reason for the improved stability of the second catalytic bed.
Interestingly, a 10% fraction of hydrocarbon products within a
range of 10–15 carbon atoms was detected in the liquid organic
phase. This revealed that besides the ketonic decarboxylation, fur-
ther carbon carbon bond formation occurred under the present
reaction conditions. The main product (60%) was 4-methylnonane
as identified by GC–MS. We assume that it was formed on the
ketone stage with partly reduced carbon dioxide to methanol
(methylation) or formaldehyde (aldol condensation with subse-
quent hydrogenation). The simulated distillation demonstrated
that the boiling point range of the whole product mixture matched
well with the one of kerosene fuel from 145 to 300 ◦C (Fig. 7).
In summary, it was concluded that the combination of
zirconium oxide as ketonic decarboxylation catalyst together
with platinum supported onto alumina for the hydrogena-
tion/dehydration/hydrogenation sequence was a very suitable
system with respect to deoxygenation and maintaining this cat-
alytic performance.
Fig. 10. Development of conversion and selectivity in the organic liquid phase with
time on stream when valeric acid was passed over a zirconium oxide and a Pt/C
bed. ꢀ conversion pentanoic acid, ꢀ selectivity 5-nonanone, ꢂ selectivity nonane,
ꢁ selectivity to C10–C15 hydrocarbons, + selectivity to other hydrocarbons (upto C8),
* selectivity to oxygenated compounds.
Reaction conditions: feed rate for valeric acid: 0.15 mL/min; hydrogen gas flow:
470 mL/min; 40 bar hydrogen pressure; 400 ◦C reaction temperature. The yield of
the water phase was 1.1 equiv. per valeric acid employed on average.
activity decreased immediately after having passed an amount of
feed equal to twice the weight of the catalyst. In contrast, with
zirconium oxide as support the deoxygenation was almost com-
plete (Fig. 9). However, the selectivity towards the desired nonane
hydrocarbon started at 65% and decreased rapidly to below 50%.
The reason for the low selectivity was the formation of shorter chain
hydrocarbons formed by cracking of nonane or one of its precur-
sors, as we saw before. This fragmentation was also reflected in
the amount of organic liquid recovered that decreased from 85 to
65%. In the gas phase ethane, propane, and butane were detected in
significant amounts, and when they were considered, the mass bal-
ance was 100 2%. With active carbon as support for the platinum
similar results as for alumina were observed. Nonane was obtained
with almost 80% selectivity of the liquid organic phase (mass bal-
ance 85% as for Pt/Al2O3), but a small part of 5-nonanone upto 5%
was still present in the product stream (Fig. 10).
It is assumed that the support helps to dehydrate the interme-
diary alcohol formed by ketone hydrogenation (cf. Scheme 2) and
for this activity a well-balanced Brönsted-acidity is required. In
this respect, it seems that the sites of alumina have the best prop-
erties, closely followed by the active carbon. In the case of silica
the strength of the sites is too low evidenced by a low and easily
deactivated deoxygenation activity. In contrast, zirconium oxide
possesses too strong sites inducing, apart from the desired com-
plete dehydration, additional undesired fragmentation reactions.
These fragmentation reactions lower also the mass balance for the
liquid phase from 85% for the other catalysts to 65% when using
Pt/ZrO2 as hydrodeoxygenation catalyst.
When the platinum (on alumina) was substituted by palladium
the results were very similar or even improved. The selectivity
towards nonane was slightly improved to 85% (versus 80%) and
also a small quantity of C10–C15 hydrocarbons was observed at a
slightly lower mass balance for the liquid phase, i.e., 80% instead of
85% (Fig. 11). However, after passing a 30-fold amount of the weight
of the catalyst, the selectivity for nonane started to decrease and
finally reached half of the initial value. As reported before, with plat-
inum on alumina 100 times the catalyst weight of feed was passed
without any change in the catalytic performance. The reason for the
lower selectivity in the case of palladium was the deactivation of
the hydrogenation activity evidenced by the increasing selectivity
for 5-nonanone (Fig. 11).
We have also considered other Pt supported catalysts. Hence,
as alternative supports silica, active carbon, and zirconium oxide
were used and palladium was also used as an alternative hydro-
genation functionality. With silica as support the hydrogenation
of 5-nonanone was not complete and 20% of this compound
was detected in the reactor effluent at a similar mass balance of
85% for the liquid phase (Fig. 8). Furthermore, the hydrogenation

122
A. Corma et al. / Journal of Molecular Catalysis A: Chemical 388–389 (2014) 116–122
The boiling point range of the product mixture matched well
with the one of kerosene although for an application as jet fuel fur-
ther isomerization might be advantageous. When hexose-derived
valeric acid was employed as substrate, approximately 80% of the
product was n-nonane plus another 10% of C10–C15 hydrocarbons
are present in the liquid organic phase (85% mass balance). Butane
was detected in the gas phase together with carbon monoxide, car-
bon dioxide and methane, and the global mass balance was closed
with 100 2%.
Acknowledgements
The authors thank MINECO (Consolider Ingenio 2010-
MULTICAT, CSD2009-00050 and CTQ2011-27550) and the Spanish
National Research Council (CSIC, Es 2010RU0108). B.O.-T. is
grateful to CSIC for a fellowship (JAE Programme).
Fig. 11. Development of conversion and selectivity in the organic liquid phase
with time on stream when valeric acid was passed over a zirconium oxide and a
Pd/alumina bed. ꢀ conversion pentanoic acid, ꢀ selectivity 5-nonanone, ꢂ selectivity
nonane, ꢁ selectivity to C10–C15 hydrocarbons, + selectivity to other hydrocarbons
(upto C8), * selectivity to oxygenated compounds.
Reaction conditions: feed rate for valeric acid: 0.15 mL/min; hydrogen gas flow:
470 mL/min; 40 bar hydrogen pressure; 400 ◦C reaction temperature. The yield of
the water phase was 1.2 equiv. per valeric acid employed on average.
References
[1] A. Demirbas, Fuel Process. Technol. 88 (2007) 591–597.
[2] K. Sipilä, E. Kuoppala, L.r Fagernäs, A. Oasmaa, Biomass Bioenerg. 14 (1998)
103–113.
[3] C.B. Rasrendra, B. Girisuta, H.H. van de Bovenkamp, J.G.M. Winkelman, E.J.
Leijenhorst, R.H. Venderbosch, M. Windt, D. Meier, H.J. Heeres, Chem. Eng. J.
176–177 (2011) 244–252.
[4] Z. Qi, C. Jie, W. Tiejun, X. Ying, Energ. Convers. Manage. 48 (2007) 87–92.
[5] S.W. Fitzpatrick, US Patent 4897497 (1988).
[6] S. Ritter, Chem. Eng. News 84 (2006) 47.
[7] G. Yang, E.A. Pidko, E.J.M. Hensen, J. Catal. 295 (2012) 122–132.
[8] J.P. Lange, R. Price, P.M. Ayoubm, J. Louis, L. Petrus, L. Clarke, H. Gosselink,
Angew. Chem. Int. Ed. 49 (2010) 4479–4483.
[9] J.C. Serrano-Ruiz, D. Wang, J.A. Dumesic, Green Chem. 12 (2010)
574–577.
[10] H. Heeres, R. Handana, D. Chunai, C.B. Rasrendra, B. Girisuta, H.J. Heeres, Green
Chem. 11 (2009) 1247–1255.
[11] L. Deng, J. Li, D.-M. Lai, Y. Fu, Q.-X. Guo, Angew. Chem. Int. Ed. 48 (2009)
6529–6532.
Therefore, it can be concluded that alumina, in combination with
platinum, was the most efficient material for the selective trans-
formation of the intermediary 5-nonanone into nonane. Palladium
losses rapidly the hydrogenation activity and other supports were
either less selective (zirconium oxide) or less active (active carbon,
silica).
4. Conclusion
[12] M. Renz, Eur. J. Org. Chem. (2005) 979–988.
[13] T.S. Nguyen, M. Zabeti, L. Lefferts, G. Brem, K. Seshan, Biomass Bioenerg. 48
(2013) 100–110.
The combination of zirconium oxide with platinum supported
onto alumina, placed in this order in
a fixed bed continu-
[14] M. Zabeti, T.S. Nguyen, L. Lefferts, H.J. Heeres, K. Seshan, Bioresour. Technol.
118 (2012) 374–381.
ous flow reactor, is a very suitable catalytic system to convert
carboxylic acids into a hydrocarbon mixture. At a reaction tem-
perature of 400 ◦C the ketonic decarboxylation on zirconium oxide
presented high selectivity and excellent stability with time on
stream. It can be reactivated by calcination so that after several
cycles only a 5% decrease in activity was observed. The hydro-
genation/dehydration/hydrogenation catalyst platinum/alumina
provided complete elimination of the oxygen with an excellent sta-
bility with time on stream. Hence, approximately 250 g of valeric
acid were converted into a hydrocarbon mixture with 2.5 g of zirco-
nium oxide and 1.5 g of platinum/alumina without observing any
catalyst deactivation. It was shown that alternative supports such
as silica, zirconium oxide or active carbon or palladium instead of
platinum achieved incomplete deoxygenation or showed a rapid
decrease in the catalytic activity.
[15] M. Renz, A. Corma, Eur. J. Org. Chem. (2004) 2036–2039.
[16] J.A. Sullivan, N. Walsh, L. Sherry, Catal. Lett. 143 (2013) 361–369.
[17] R.W. Snell, B.H. Shanks, Appl. Catal. A: Gen. 451 (2013) 86–93.
[18] A. Pulido, B. Oliver-Tomas, M. Renz, M. Boronat, A. Corma, ChemSusChem 6
(2013) 141–151.
[19] A.V. Ignatchenko, J. Phys. Chem. C 115 (2011) 16012–16018.
[20] Y.A. Zaytseva, V.N. Panchenko, M.N. Simonov, A.A. Shutilov, G.A. Zenkovets, M.
Renz, I.L. Simakova, V.N. Parmon, Top. Catal. 56 (2013) 846–855.
[21] A. Corma, M. Renz, C. Schaverien, ChemSusChem 1 (2008) 739–741.
[22] M. Snare, I. Kubickova, P. Mäki-Arvela, K. Eränen, D.Y. Murzin, Ind. Eng. Chem.
Res. 45 (2006) 5708–5715.
[23] P. Mäki-Arvela, B. Rozmyslowicz, S. Lastari, O. Simakova, K. Eränen, T. Salmi,
D.Y. Murzin, Energ. Fuel. 25 (2011) 2815–2825.
[24] E. Santillan-Jimenez, M. Crocker, J. Chem. Technol. Biotechnol. 87 (2012)
1041–1050.
[25] A.S. Berenblyum, V.Y. Danyushevsky, E.A. Katsman, T.A. Podoplelova, V.R. Flid,
Petrol. Chem. 50 (2010) 305–311.