Hydrodeoxygenation (HDO) of methyl palmitate over bifunctional Rh/ZrO2 catalyst: Insights into reaction mechanism via kinetic modeling

Article abstract of DOI:10.1016/j.apcata.2016.08.030

Hydrodeoxygenation (HDO) of triglycerides into hydrocarbons is a novel catalytic process for the production of green biofuels. In this work, the HDO reaction mechanism over Rh/ZrO₂ catalyst was studied by selecting methyl palmitate as a model compound. HDO of methyl palmitate proceeded initially via the hydrogenolysis into palmitic acid intermediate, followed by sequential hydrogenation-decarbonylation reaction into pentadecane via aldehyde intermediate. Bifunctional mechanism of the Rh/ZrO₂ catalyst is advocated for the HDO process, in which both Rh sites and oxygen vacancy sites on ZrO₂ synergistically contribute to the catalysis. The interface between Rh nanoparticle and support was proposed to host the most active sites. Based on our earlier work, a surface reaction mechanism was proposed and slightly modified to develop a set of mechanistic kinetic models. The mechanistic model consisting of two distinct types of adsorption sites for oxygenated components and H₂, gave a good fitting to the kinetic data over a broad range of reaction conditions and conversion levels.

Full text of DOI:10.1016/j.apcata.2016.08.030

Accepted Manuscript
Title: Hydrodeoxygenation (HDO) of methyl palmitate over
bifunctional Rh/ZrO2 catalyst: insights into reaction
mechanism via kinetic modeling
Author: Yuwei Bie Juha Lehtonen Jaana Kanervo
PII:
S0926-860X(16)30448-3
DOI:
Reference:
http://dx.doi.org/doi:10.1016/j.apcata.2016.08.030
APCATA 15993
To appear in:
Applied Catalysis A: General
Received date:
Revised date:
Accepted date:
24-6-2016
29-8-2016
30-8-2016
Please cite this article as: Yuwei Bie, Juha Lehtonen, Jaana Kanervo,
Hydrodeoxygenation (HDO) of methyl palmitate over bifunctional Rh/ZrO2 catalyst:
insights into reaction mechanism via kinetic modeling, Applied Catalysis A, General
http://dx.doi.org/10.1016/j.apcata.2016.08.030
This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.

Hydrodeoxygenation (HDO) of methyl palmitate over bifunctional Rh/ZrO2
catalyst: insights into reaction mechanism via kinetic modeling
Yuwei Bie*, Juha Lehtonen, Jaana Kanervo
Industrial Chemistry Research Group, Department of Biotechnology and Chemical Technology,
School of Chemical Technology, Aalto University, PO BOX 16100, FI-00076, Espoo, Finland
Corresponding author: yuweibie.chem@gmail.com
Graphical abstract
1

10
8
C15H32
Hydrocarbon
H2
C15H31COOCH3
Ester
HDO Kinetics
6
Mechanism
H
4
spillover
Rh
c H
2
c
c c c
c
ZrO2
0
0
20
40
60

Highlights:
 New understanding into the bifunctional Rh/ZrO2 catalyst were gained for the HDO reaction
of methyl palmitate into biofuels: the most active sites are located at perimeter of Rh metal.
 The HDO activity of Rh/ZrO2 was found to be strongly correlated with the Rh metal area,
while bare ZrO2 possessed a weak activity for certain HDO reaction pathways.
 Mechanistic kinetic model was derived based on a surface reaction mechanism and can
successfully describe the HDO reaction kinetics.
 Mechanism insights and kinetic modeling results can be used in HDO catalyst innovation
and process design.
2

Abstract
Hydrodeoxygenation (HDO) of triglycerides into hydrocarbons is a novel catalytic process for the
production of green biofuels. In this work, the HDO reaction mechanism over Rh/ZrO2 catalyst was
studied by selecting methyl palmitate as a model compound. HDO of methyl palmitate proceeded
initially via the hydrogenolysis into palmitic acid intermediate, followed by sequential
hydrogenation-decarbonylation reaction into pentadecane via aldehyde intermediate. Bifunctional
mechanism of the Rh/ZrO2 catalyst is advocated for the HDO process, in which both Rh sites and
oxygen vacancy sites on ZrO2 synergistically contribute to the catalysis. The interface between Rh
nanoparticle and support was proposed to host the most active sites. Based on our earlier work, a
surface reaction mechanism was proposed and slightly modified to develop a set of mechanistic
kinetic models. The mechanistic model consisting of two distinct types of adsorption sites for
oxygenated components and H2, gave a good fitting to the kinetic data over a broad range of
reaction conditions and conversion levels.
Keywords: hydrodeoxygenation (HDO), mechanism, Rh/ZrO2, kinetic modeling, methyl palmitate,
biofuel
3

1. Introduction
Biofuel production from renewable resources is receiving considerable attention in order to tackle
the increasing energy demand and environmental concerns caused by the use of fossil fuels.
Triglyceride-based materials (vegetable oils, waste oils/fats, and microalgae oil) are considered to
be promising renewable feedstocks to produce biodiesel (fatty acid methyl esters) via
transesterification with methanol [1-2]. However, the widespread use of biodiesel is limited due to
its relatively high oxygen content causing some bad fuel properties [3]. Therefore, attention has
lately shifted to the deoxygenation of triglyceride-based biofeeds through hydrodeoxygenation
(HDO) process, in which oxygen is eliminated either as H2O or CO/CO2 [4]. HDO technology has
the advantage of yielding hydrocarbon-based biofuels which have the same or even better properties
than conventional petroleum fuels [3, 4].
Fundamental issues of the HDO process, i.e. the catalyst development and reaction pathways, have
been widely studied by using model compounds such as fatty acids, fatty acid esters, and
triglycerides because of their structural similarity to vegetable oils [5-12]. HDO processes typically
involve a complex reaction network, involving hydrogenolysis, hydrogenation, dehydration,
decarbonylation, decarboxylation, etc. [13]. The nature of catalyst considerably influences the
reaction pathways. The HDO catalysts are mainly divided into two categories: conventional
hydrotreating catalysts that need to be presulfided, such as CoMo/ γ-Al2O3 and NiMo/γ-Al2O3 [5-7],
and non-conventional catalysts, such as supported noble metal/transition metal catalysts (e.g. Pt, Pd,
Rh and Ni) [8-12, 14-20] as well as metal carbides and nitrides [21]. Nonconventional catalysts
have recently gained the most interest for research because of the need for avoiding catalyst
presulfidation.
For the development of novel HDO catalyst, Yakovlev et al. [22] proposed that a bifunctional
catalyst comprising a reducible metal oxide (e.g. ZrO2 and CeO2) and a noble/transition metal in its
reduced state is the most attractive and active for HDO process. Stimulated by the advantage of
bifunctionality, our group investigated ZrO2 supported Rh and Pt catalysts for HDO of methyl
heptanoate as model compound in liquid phase [8]. Monometallic Rh catalyst was found to show
higher HDO activity and hydrocarbon selectivity than Pt and bimetallic RhPt catalysts. Lercher and
coworkers [14-16] tested a series of supported Ni and Pd catalysts for the liquid phase HDO
reaction of palmitic acid into hydrocarbons at 260 °C. ZrO2 supported catalysts were found to
4

exhibit superior activity compared to other supports, such as C, SiO2 and Al2O3. The mechanistic
explanation was that HDO of palmitic acid can be catalyzed not only by metal sites but also
synergistically by the oxygen vacancy sites on ZrO2. It was proposed that oxygen vacancy sites can
serve as active sites for the conversion of palmitic acid into aldehyde via activating a carboxylate
group and forming a ketene species [16].
Despite a number of studies on catalyst development, detailed reaction mechanisms and kinetic
studies for HDO process are seldom reported because of the typical complex reaction network.
Mechanistic understanding of major reaction pathways is very useful in guiding experimental
design and catalyst synthesis. In the existing kinetic study, a detail kinetic model based on reaction
mechanism is largely lacking, although there have been several attempts using power law model
[17-19]. A suitable mechanistic model could also help to reveal the reaction mechanism. In our
recent study [17], we proposed a comprehensive surface reaction mechanism for the HDO reaction
of methyl heptanoate over Rh/ZrO2 catalyst by gathering the literature information. Furthermore,
we derived a set of mechanistic kinetic models consisting of two types of active sites, one being
responsible for hydrogen dissociation, and the other responsible for the adsorption of oxygenated
species. The derived mechanistic model has been successfully used to describe the HDO kinetics of
methyl heptanoate over broad experimental conditions.
The objective of this work is to test whether the previously proposed mechanistic model is
applicable for the HDO of methyl palmitate, a more realistic and heavier model compound. It has
been found that the fatty acid esters with different chain length exhibited different kinetic behaviors.
For this purpose, the HDO kinetics of methyl palmitate were experimentally investigated over
Rh/ZrO2 catalyst. A series of different Rh/ZrO2 catalysts were prepared by using different
calcination temperature of ZrO2 support, leading to different catalyst surface area and metal
dispersion. By combining the experimental and modeling study, we are aiming to reveal some new
mechanism insights in order to design and optimize HDO catalyst and process.
2. Experimental
2.1 Catalyst preparation and characterization
A series of Rh/ZrO2 catalysts were prepared by incipient wetness impregnation technique with an
aqueous solution of Rh (NO3)3 (Aldrich). ZrO2 (MEL, ECO100) as support was calcined at different
temperatures (500, 700 and 900 °C) for 10 h in synthetic air flow before the Rh impregnation. In
this manner, the dispersion of Rh nanoparticles on ZrO2 was varied. Different metal loadings (0.4 –
5

0.7 wt%) were achieved by varying the amount of Rh precursor. After impregnation, catalyst was
settled overnight and then dried under vacuum at 40 °C for 1.5 h and then at 60 °C for 30 min. The
dried catalyst was finally calcined at 450 °C for 2 h in a synthetic air flow (4 L/h). All the prepared
Rh/ZrO2 catalysts were characterized with nitrogen physisorption for BET-surface area, and with
hydrogen chemisorption for the metal surface area and metal dispersion. The actual Rh loading was
measured by using X-ray fluorescence (XRF) and inductively coupled plasma atomic emission
spectroscopy (ICP-AES) method. More details about the preparation procedures and
characterization methods can be found elsewhere [8, 17].
2.2 Reaction kinetic
2.2.1. Catalytic activity of Rh/ZrO2
Liquid phase HDO reaction of methyl palmitate over Rh/ZrO2 catalyst was investigated in a batch
reactor (50 mL) under isothermal condition and static H2 pressure. All the above-mentioned
Rh/ZrO2 catalysts prepared with different supports were tested at HDO reaction conditions. In a
typical run, Rh/ZrO2 or bareZrO2 (0.44 g, crushed size: 0.20 – 0.25 mm) was placed inside a
spinning catalyst basket equipped as an reactor impeller, and reduced in situ with 20 bar of static H2
at 350 °C for 1 h. After the reduction, 30 mL of reaction mixture, i.e. 10 wt.% methyl palmitate
dissolved in dodecane, was introduced into the reactor. Reaction mixture was then heated up to the
target temperature (270 °C) with a minor stirring and low hydrogen pressure (30 bar). Subsequently,
the reactor was pressurized up to the target H2 pressure (80 bar H2) and the stirring was started,
marking the start of HDO reaction. In the course of experiment, 4 to 6 liquid samples (about 0.6
mL for each) were withdrawn from a sampling line for quantitative GC analysis using an Innowax
capillary column and a FID detector. At the end of reaction, gas phase sample was qualitatively
analyzed with off-line GC. Sampling from liquid phase led to a minor pressure drop, which was
compensated by re-charging H2 to restore the original target pressure.
2.2.2 Kinetic experiments for modeling
A standard batch of Rh/ZrO2 (ZrO2 calcined at 700 °C) was selected for further kinetic studies.
Kinetic experiments for methyl palmitate HDO were conducted at four temperatures (240, 255, 270
and 300 °C) with an initial H2 pressure of 80 bar. The effect of H2 pressures (40 bar and 80 bar) for
HDO kinetics was studied at 300 °C. In addition, separate HDO experiments using palmitic acid
and hexadecanol as reactants were performed to confirm their nature as reaction intermediates and
to validate the kinetic model. Preliminary tests for external and internal mass transfer resistance
6

were performed at 300 °C with varying stirring speeds (500 700 rpm) and size ranges of catalyst
particle mm). It was confirmed that HDO reaction was conducted in kinetic controlled
regime in this work by applying a vigorous stirring (700 rpm) and small catalyst particles (0.20
0.25 mm).
3. Experimental results and discussion
3.1 Reaction pathways on Rh/ZrO2
In a typical run, HDO of methyl palmitate over Rh/ZrO2 (ZrO2 calcined at 700 °C) was studied at
270 °C and 80 bar of H2. Fig. 1 shows the concentration profile of the major components in the
liquid phase as a function of time. In the course of experiment, oxygenated intermediates, such as
palmitic acid, hexadecanol, palmityl palmitate in small amount, and hexadecanal in traces, were
observed. In the end, pentadecane and small amount of hexadecane were yielded as the final
deoxygenation products. Pentadecane (C15) was significantly more dominant than hexadecane
(C16). Unsaturated hydrocarbons were not detected at all. The concentration of palmitic acid and
hexadecanol initially increased and then gradually decreased, indicating their nature as reaction
intermediates. Moreover, palmitic acid appeared to be the primary intermediate, which is far more
abundant than hexadecanol. Light gases (CH4, CO and CO2) were detected in the gas phase. Solvent
cracking reaction was not observed until reaching full conversion of the oxygenated compounds (i.e.
methyl palmitate and palmitic acid) only at 300 °C.
Fig. 1. Typical concentration profiles of reaction components in HDO of methyl palmitate. Reaction
conditions: 270 °C, 80 bar H2, Rh/ZrO2 catalyst (0.44 g, 0.4 wt.% Rh, ZrO2 calcined at 700 °C).
7

Based on the product results, we can deduce the HDO reaction pathways of methyl palmitate over
Rh/ZrO2 catalyst, as shown in Fig. 2. The initial step is the hydrogenolysis of methyl palmitate,
producing palmitic acid and gaseous CH4. Palmitic acid can be further hydrogenated into
hexadecanal via the removal of water, followed by decarbonylation reaction into pentadecane and
CO. Meanwhile, hexadecanol was yielded via the hydrogenation of hexadecanal. Meanwhile,
hexadecanol can be reversibly transformed to hexadecanal via its dehydrogenation. These routes
can be further confirmed in separate experiments using palmitic acid or hexadecanol as reactant.
Moreover, the water-gas-shift (WGS) and methanation reactions may have occurred, affecting the
concentration of CO and CO2 in the gas phase.
In the HDO reaction network over Rh/ZrO2, palmitic acid is a key intermediate, which is finally
transformed mainly into pentadecane. Even in the HDO of hexadecanol, the yield of hexadecane
formed was less than 2.0% at almost complete conversion. It demonstrated that the dehydration
reaction of hexadecanol was highly suppressed over Rh/ZrO2 catalyst, whereas this route is
remarkable over traditional sulfided NiMo/Al2O3 catalyst [6]. The direct decarboxylation or
decarbonylation of palmitic acid, has been widely reported over carbon supported metal catalysts
especially under hydrogen-poor conditions [14, 20]. However, the direct deoxygenation of methyl
palmitate and palmitic acid was considered to be negligible in this work. We performed additional
experiments for both methyl palmitate and palmitic acid under inert gas atmosphere (80 bar N2) at
270 °C. Comparing with the experiments with H2, both the reaction rate and the pentadecane yield
were more than 10 times smaller. Heavy palmityl palmitate was formed via the esterification of
palmitic acid with hexadecanol or the transesterification of methyl palmitate with hexadecanol.
However, its yield was quite low (< 2.0 mol-%) over Rh/ZrO2. Thus, these side reactions were not
shown in Fig. 2.
Fig. 2. Reaction pathways for HDO of methyl palmitate over Rh/ZrO2 catalyst in the presence of H2
3.2 The catalytic role of ZrO2
8

The catalytic role of ZrO2 for HDO of methyl palmitate was studied at 270 °C and 80 bar H2. Table
1 compares the conversion and product distribution results obtained with bare ZrO2 and Rh/ZrO2.
ZrO2 in the absence of metal possessed a minor but measurable HDO activity. The methyl palmitate
conversion was just 6% over ZrO2 after 30 min; however, reached 18% over Rh/ZrO2 after 3.5 min.
The initial reaction rate was over 30 times higher over Rh/ZrO2 than ZrO2.
Table 1. Comparison of catalyst performance for ZrO2 and Rh/ZrO2 in HDO of methyl palmitate
Catalysta Initial
Product selectivity at 18% conversion
Rateb
palmitic
acid
hexadecanol hexadecanal heptadecane palmityl
palmitate
palmitone
(mmol/h)
0.018
Rh/ZrO2 0.60
ZrO2
12
63
6.0
4.2
24
4.6
32
28
24
0.0
0.4
0.0
b
c
a: ZrO2 calcined at 700 °C. : Initial reaction rate calculated within 10 % conversion. : Product
selectivity is calculated as “the amount of reactant consumed for each individual product divided by
the amount of totally consumed reactant”
Pentadecane and oxygenated intermediates, e.g. palmitic acid, hexadecanol and hexadecanal were
all observed over bare ZrO2, indicating that most of the reaction routes in Fig. 2 have occurred to
some extent on ZrO2. This is in agreement with the bifunctionality of ZrO2 supported metal catalyst.
In addition, two heavy products, palmitone and palmityl palmitate, were formed in the presence of
pure ZrO2. From Table 1, the palmitic acid selectivity was much lower on bare ZrO2 than Rh/ZrO2,
whereas the hexadecanal selectivity and heavy products were more evident on ZrO2. The
accumulation of hexadecanal indicates that decarbonylation of aldehyde into pentadecane was
suppressed in the absence of metal. The formation of palmitone on ZrO2 can be attributed to
ketonization of palmitic acid [9, 14]. Our results demonstrate that the presence or absence of metal
on ZrO2 can lead to different preferences of reaction pathways. The deposition of metal can
facilitate some key reactions for hydrocarbon formation such as decarbonylation, and eliminate
some undesired reactions such as ketonization.
3.3 The effect of support calcination on catalyst performance
3.3.1 Characterization
The effect of ZrO2 calcination temperature (500, 700, and 900 °C) on the catalytic properties of
Rh/ZrO2 was investigated in order to increase the understanding of the nature of active sites for
HDO. Catalyst characterization by BET-surface area and H2-chemisorption allows us to correlate
9

the changes in catalyst physicochemical properties with the calcination temperature. Increasing the
calcination temperature for ZrO2 support before Rh impregnation significantly decreased the
specific surface area of ZrO2 and Rh/ZrO2 catalysts. For example, the Rh/ZrO2 catalysts prepared
after calcining ZrO2 at either 700 °C or 500 °C had surface area in the range of 70-80 m2/g.
However, the surface area decreased down to just 23 m2/g when calcining ZrO2 at 900 °C. There
was no appreciable difference in the surface areas of ZrO2 and Rh/ZrO2 upon impregnation of Rh.
The decrease in the catalyst surface areas directly led to a decrease in the metal dispersion and the
active metal surface area. The dispersion of Rh nanoparticles dropped by half when the support
calcination temperature increased from 700 °C to 900 °C. In line with this, the specific surface area
of Rh (based on H2-chemisorption) correspondingly declined.
Increasing the metal loading from 0.4 wt.% to 0.73 wt.% on the same type of ZrO2 did not
obviously change the catalyst surface area and the Rh dispersion. This implies that the amount of
metal particles of relatively similar size increased with higher metal loading, whereas larger metal
particles were not formed. It appears that the ZrO2 can provide surface sites that are capable of
anchoring Rh nanoparticles of a stable nanoparticle size in the studied range of metal loading.
Based on the literature, the anchoring sites could be tentatively t-OH or oxygen vacancies on ZrO2
[15, 23, 24].
Table 2. Characterization results
Catalyst
Rh % BET surface area
m2/g
Metal surface area, m2/(g
cat.)
Dispersion,
%
-
ZrO2(900°C)
ZrO2(700°C)
ZrO2(500°C)
-
-
-
24.9
74.1
81.2
-
-
Rh/ZrO2(900 °C)a 0.40 23.8
Rh/ZrO2(700°C)a 0.43 73.8
Rh/ZrO2(500°C)a 0.40 80.1
0.71
1.41
1.47
2.23
41%
75%
79%
73%
Rh/ZrO2(700°C)b 0.73 70.7
b
a: 0.4 wt% Rh loading; : 0.73 wt% Rh loading, Stoichiometry ratio of Rh atom and H2 is 2
(dissociation). 500, 700, and 900 °C denote the calcination temperature of ZrO2 support.
10

3.3.2 Comparison of catalyst activity
All the Rh/ZrO2 catalysts listed in Table 2 were tested for HDO of methyl palmitate. The reaction
pathways described in Fig. 2 were all observed over these catalysts, regardless of their metal
dispersion and metal loading. However, their catalytic activities are different. Fig. 3 presents the
correlation between the initial hydrogenolysis rate and the metal surface area of Rh/ZrO2 catalysts.
It was observed that the intrinsic hydrogenolysis activity of Rh/ZrO2 increased proportionally to the
amount of metal surface area. The turnover frequency (TOF) is approximately of the same
magnitude (0.3-0.5 s-1) for all the tested catalysts, calculated by the initial reaction rate divided by
the number of metal sites (estimated based on H2 chemisorption).
Obviously, the difference in the catalyst activity is attributed to the difference in the number of
active sites on Rh/ZrO2. Our results confirm the significance of metal sites in catalyst activity but
do not provide evidence that the active sites would be metallic Rh. Considering the bifunctionality
of Rh/ZrO2 for HDO process, the physicochemical properties of ZrO2 (i.e. surface property) may
play a significant role in catalyst activity.
Fig. 3. Correlation between initial reaction rate and metal surface area. Reaction conditions: 270 °C;
80 bar H2, 10 wt.% methyl palmitate in dodecane, Rh/ZrO2 catalyst (ZrO2 calcined at different
temperatures: 500, 700 and 900 °C).
4. Kinetic modeling
We recently developed a set of mechanistic models for the HDO of methyl heptanoate over
Rh/ZrO2 catalyst [17]. It was assumed that HDO of methyl ester proceeds via the initial dissociative
adsorption of ester into carboxylate species, followed by consecutive surface hydrogenation-
11

decarbonylation steps into hydrocarbons. In this section, we test whether the previous surface
mechanism is valid for methyl palmitate over different type of Rh/ZrO2 catalyst by fitting the
derived kinetic model (presented below) against broad kinetic data.
4.1 Surface reaction steps
Fig. 4a is a simplified HDO reaction network of methyl palmitate based on a lumping method. The
principle for the simplification can be found in Ref. [17]. Four major reaction routes are considered
for kinetic modeling. The rest of minor reaction steps, resulting in only 4.0% of products in yield,
were not considered. The kinetically relevant components are mainly methyl palmitate, palmitic
acid, hexadecanol and pentadecane. Fig. 4b illustrates the surface reaction mechanism in more
detail. Table 3 shows the proposed elementary reaction steps.
R1 represents the methyl palmitate hydrogenolysis into palmitic acid (see Eqs. (1) – (3) in Table 3).
Methyl palmitate initially dissociates on the oxygen vacancy sites on ZrO2 (denoted as S1), forming
a palmitate and methyl surface species. Hydrogen is assumed to be dissociated by Rh sites (denoted
as S2). Methyl species easily react with the dissociated hydrogen to form a gaseous methane.
Palmitate species can react with surface hydrogen to form palmitic acid, which is detected in liquid
phase.
R2 represents the sequential hydrogenation of palmitic acid into hexadecanol and R3 represents the
sequential hydrogenation-decarbonylation of palmitic acid into pentadecane. The palmitate formed
in R1 can be dehydrated into a ketene and then hydrogenated into an aldehyde. Trace amounts of
hexadecanal were detected in the bulk phase, implying a high reactivity and low surface coverage of
aldehyde. Hexadecanal is further hydrogenated into hexadecanol via forming a hexadecoxy species
(R2) or rapidly decarbonylated into pentadecane (R3). R2 and R3 are regarded as two independent
reactions with distinct rate constants. They are expressed by Eqs. (4) – (9) in Table 3. However,
based on the mechanism hypothesis, they share a common rate determining step (RDS), which is
the final addition of a hydrogen atom into the carbonyl species. Thus, the rate expressions of R2 and
R3 are expected to be the same.
R4 represents the reversible dehydrogenation-decarbonylation of hexadecanol into pentadecance.
The surface reaction mechanism is described by Eqs (7) – (9) in Table 3. C16 aldehyde is formed
via dissociation of hexadecanol, which is in fast equilibrium. The decarbonylation of aldehyde is
assumed to the RDS.
12

Fig. 4. Proposed surface reaction mechanism for HDO of methyl palmitate over Rh/ZrO2: (a)
simplified HDO reaction pathways (R1 – R4); (b) proposed surface reaction steps (ads.: adsorbed
surface species).
Table 3. Proposed elementary reaction steps.
Surface reaction steps
Routes
R1
Note
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
RDS for
C C C C C C
Quasi-equilibrium
R1
Quasi-equilibrium
Quasi-equilibrium
Quasi-equilibrium
R1
C C C C
R2&R3
R2&R3
C C C C C
C C C C C
R2&R3 RDS for &
C C C C
Quasi-equilibrium
Quasi-equilibrium
R2&R4
R2&R4
C C C
C C
R3&R4 RDS for
C C C C
4.2 Kinetic model
The surface elementary steps (Table 3) allow us to derive the kinetic rate equations for R1 – R4,
which are expressed in Eqs (10) – (13). The derivation method is similar to that used in Ref. [17].
13

k1 cA  KH2 cH 2
R1 
KBcB
KC cC
(10)
(11)
(12)
(13)
(1
(1

)(1 KH2 cH2 )
KH2 cH2
KH2 cH2
k2  KB  KH2 cB cH2
R2 
KBcB
KH2 cH2
KC cC

)(1 KH2 cH2 )
KH2 cH2
k3  KB  KH2 cB cH2
R3 
R4
KBcB
KC cC
(1
(1

)(1 KH2 cH2 )
KH2 cH2
KH2 cH2
k4  KC cC
KBcB
KCcC

) KH2 cH2
KH2 cH2
KH2 cH2
Ci is the liquid phase concentration of components (i = A, B and C, which means methyl palmitate,
palmitic acid and hexadecanol, respectively). kj denotes the rate constant of reaction route Rj (j = 1,
2..). Kk (k = B, C and H2) is the adsorption constant of surface species (e.g. palmitate, hexadecoxy
and hydrogen atom). Other surface species and light components, such as aldehyde, ketene, hexane,
H2O, CO, CO2 and CH4, were all neglected in the kinetic model.
It is worth mentioning that the surface reaction step in Eq. (1) is slightly modified from that in
earlier work, where the reaction of methyl species with hydrogen into methane was defined as an
independent elementary step [17]. Thus, in this work, Eq. (10) for the rate equation of R1 is
accordingly adjusted, where hydrogen concentration appears. The modified kinetic equation is able
to better describe the effect of hydrogen pressure on kinetics.
4.3. General considerations for parameter estimation
HDO reaction of methyl palmitate was performed in a well-mixed three-phase batch reactor in the
absence of external and internal mass transfer resistances. Thus, a pseudo-homogeneous reactor
model can be applied to describe the mass balances of components in the gas and liquid phases,
respectively. Rate expressions Eqs and the stoichiometry in the reaction network are
taken into account in the mass balance equations. In the reactor model, the temperature is kept
constant, and the total pressure remained nearly unchanged. This can be evidenced by the
stoichiometric relationship of overall HDO reaction equation: C15H31COOCH3 + 2H2 C15H32 +
CO + CH4 + H2O. The amount of produced gaseous products was almost equivalent to that of
consumed hydrogen. Parameter estimation was performed in KINFIT program [25], which is
14

integrated with a process simulator FLOWBAT [26] for thermodynamic related calculations,
because the reactor system in question was thermodynamically non-ideal. The concentration of
hydrogen in the organic liquid phase was estimated using PSRK method by vapor-liquid-
equilibrium (VLE) subroutine in FLOWBAT.
Kinetic parameters were estimated by minimizing the objective function: the weighted sum of the
squared differences between the experimental and the modeled concentrations. The liquid phase
concentrations of methyl palmitate, palmitic acid, hexadecanol, and pentadecane were considered in
the objective function. Concentration data at different time intervals collected under four
temperatures and three H2 pressures and extra palmitic acid experiments (two temperatures) are
simultaneously used for parameter estimation (148 data points in total). Mass balance equations
(ordinary differential equation) were solved by using LSODE solver based on backward difference
method [27]. The significance of the overall regression is tested by means of a T-test at the 95%
confidence level. More detailed information about the batch reactor model and the numerical
strategy can refer to Ref. [17].
The temperature dependency of rate constant kj and adsorption equilibrium constant Kk are
described by Arrhenius and Van’t off equations, respectively. Both equations were re-
parameterized using average temperature to decrease the correlation between parameters:
 E


1
1
a, j
(14)
(15)
k j  kref, j. exp
( 
T
)




R
Tref.
 H


1
1
ads,k
Kk  Kref,k .exp
( 
)




.
R
T
Tref.
Thus, the kinetic parameters to be estimated are kj and Kk at the reference temperature (270 °C),
activation energy ( ) and adsorption enthalpy ( ). In order to decrease the number of
parameters, the temperature dependency of the adsorption constants for palmitic acid (KB) and
hexadecanol (KC) were considered to be equal. Moreover, the activation energies of k2 and k3 were
fixed to be equal values as well, because R2 and R3 share a common RDS.
4.4 Modeling results
Fig. 5 shows the experimental concentrations and the calculated results for the HDO of methyl
palmitate. The mechanistic model gave a good fitting to the experimental results obtained at the
temperature range from 240 to 300 °C. It was also found that increasing reaction temperature
15

clearly accelerated the overall deoxygenation reaction of oxygenates into the final hydrocarbon
product.
Fig. 5. Comparison of the experimental (dotted line) and predicted concentration (solid line) in the
liquid-phase HDO of methyl palmitate in a batch reactor. Reaction conditions: 240 – 300 °C; 80 bar
H2, Rh/ZrO2 catalyst (ZrO2 calcined at 700 °C)
Fig. 6 depicts the model fit to the kinetic data collected at 300 °C with two hydrogen pressures, i.e.
40 bar and 80 bar respectively. It can be seen that the mechanistic model is able to describe the
effect of hydrogen pressure on the HDO kinetics of methyl palmitate reasonably well. The kinetic
data also indicate that increasing hydrogen pressure from 40 bar to 80 bar only slightly accelerated
the overall HDO reaction. It might be attributed to the fact that a high surface coverage of hydrogen
was already achieved at the catalyst surface (up to 70%, calculated by the estimated hydrogen
adsorption constant).
16

Fig. 6. Effect of hydrogen pressure on the kinetics of methyl palmitate HDO reaction in a batch
reactor. Dotted line: experimental; solid line: predicted. Red color: 40 bar; black color: 80 bar.
Reaction conditions: 40 and 80 bar H2, 300 °C, Rh/ZrO2 catalyst (ZrO2 calcined at 700 °C).
Additional HDO experiments of palmitic acid were performed at 250 and 280 °C. The experimental
data points were simultaneously used for parameter fitting. Fig. 7 shows a good agreement between
the experimental results and the predicted values in both experiments.
Fig. 7. Comparison of the experimental (dotted line) and predicted concentration (solid line) in the
liquid-phase HDO of palmitic acid in a batch reactor. Reaction conditions: 250 °C (left) and 280 °C
(right); 80 bar of H2
Based on the discussion above, the mechanistic model led to a satisfactory fit to the HDO reaction
kinetics of methyl palmitate over a wide range of reaction conditions and conversion levels.
Moreover, similar type of reaction mechanism and kinetic model also performed well for the HDO
reaction kinetics of methyl heptanoate over a different type of Rh/ZrO2 catalyst (ZrO2 calcination
17

temperature 900 °C). Thus, the validity of the mechanistic model can be somehow supported. Table
4 presents the estimated kinetic parameters along with their standard errors within 95% confidence
interval based on the T-test. The model parameters were physically meaningful and statistically
significant based on the calculated T-value. The calculated R square value for the overall fitting is
over 0.96, which is in agreement with the observed good model fit.
Table 4. Estimated kinetic parameters with 95% confidence interval.
Parameter
Unit
kg-1·s-1
mol ·m3 ·kg-1·s-1
mol ·m3 ·kg-1·s-1
mol ·m3 ·kg-1·s-1
kJ/mol
Value
2.76 ± 0.18
174 ± 33
T-value
29
5.3
9
344 ± 71
3322 ± 901
92 ± 8.3
19
21
kJ/mol
121 ± 21.5
10
kJ/mol
m3/mol
m3/mol
m3/mol
98 ± 14.0
5.1
3.9
3.8
8.7
4.3
4.6
5.7 ± 2.5 10-3
4.5 ± 3.0 10-3
4.7 ± 1.1 10-3
-68 ± -11
kJ/mol
kJ/mol
-31 ± -6.0
Rate constants are straightforward indicators of the reactivity of various oxygenated compounds in
the HDO process. The rate constant for R2 (k2) is two times smaller than that for R3 (k3), implying
that Rh/ZrO2 catalyst favored sequential hydrogenation-decarbonylation of fatty acid into C15
hydrocarbon more than sequential hydrogenation into C16 alcohol. The rate constant for the
hexadecanol dehydrogenation-decarbonylation reaction is much larger than the others, indicating
that hexadecanol is more reactive than palmitic acid and methyl palmitate.
5. Insights into reaction mechanism and active sites
In this study, the starting model compound is methyl palmitate and palmitic acid is the major
intermediate in the course of HDO reaction. Thus, the conversion of palmitic acid seems to be the
key step in the whole HDO process of methyl ester. Thus, a promising HDO catalyst should be able
to effectively activate palmitic acid to catalyze the following hydrogenation reaction into aldehyde.
18

The experimental results over Rh/ZrO2 in this work are consistent with the bifunctional mechanism
proposed on ZrO2 supported metal catalysts in HDO literature [14, 16, 22]. ZrO2 manifests an
ability to adsorb and activate oxygenates (i.e. palmitate species), whereas the presence of metal
accounts for a higher activity for hydrocarbon formation. According to the mechanistic hypothesis,
palmitate species adsorbed on the oxygen vacancies could be hydrogenated into reactive aldehyde,
followed by decarbonylation into heptadecane in the presence of metal [14, 16]. Metal (Rh) is
essential to provide active hydrogen for the following hydrogenation steps. It would be useful to
apply advanced surface characterization means under real reaction conditions to capture some direct
evidences on the oxygen vacancy theory; which is however beyond the scope of this study.
Similar reaction pathways were observed over both Rh/ZrO2 and Ni/ZrO2 [14, 15], even though
they may have different physicochemical properties. It can be thus speculated that Rh functions
similar to transition metal when ZrO2 is used as catalyst support. The nature of support can
influence the reaction pathways, e.g. direct deoxygenation (i.e. decarboxylation and
decarbonylation reaction) has been widely reported to take place over carbon-supported metal
catalysts (e.g. Rh/C and Pd/C) in both hydrogen poor and rich conditions [9, 10, 14]. In this work,
the direct deoxygenation route was negligible based on the additional experiments conducted under
inert gas.
In the absence of metal, the catalyst activity for deoxygenation reaction was very low, whereas
ketonization of carboxylic acids occurred remarkably on bare ZrO2. According to literature, a
ketone can be formed via the coupling reaction of the carboxylate with the ketene or enolate
intermediate, which are formed through the abstraction of α-hydrogen of the carboxylate species at
the oxygen vacancy sites on reducible oxides such as ZrO2, TiO2, and CeO2 [14, 28-30]. However,
palmityl palmitate and palmitone, which are dominant products on bare zirconia, were found to be
negligible in the presence of Rh. This indicates that metal may have partly displaced the active sites
for ketonization reaction on zirconia while new active phase, i.e. synergistic active sites, may be
created for competitive reaction pathways.
From Fig. 3, the conversion rate of methyl palmitate was strongly correlated with the metal surface
area in the studied range of Rh loadings. One possible explanation could be that the synergy effect
between ZrO2 and Rh increased as a function of Rh loading in this work. In other words, the
number of synergy active sites have increased along with the metal surface area in the case of low
metal loading and small metal particles. Thus, we propose that the metal-support interface could be
the most relevant active phase, as illustrated in Fig. 8 (in grey color). These active sites at the
19

perimeter of Rh nanoparticles may possess unique catalytic properties for HDO reaction because
the redox properties and/or hydroxyl chemistry can be promoted by the metal [31]. They may have
the greatest affinity to adsorb carboxylates and can be restored as a result of facile hydrogen
dissociation by Rh.
The dissociated hydrogen may spillover onto the surrounding ZrO2 to accelerate the HDO reaction
[16]. Nonetheless, the rate of hydrogen spillover seems not to be the rate limiting step for HDO
reaction kinetics because the hydrogen effect was observed insignificant. It is worth noting that the
acidity of support may influence the rate of H2 spillover [32]. However, ZrO2 in this work is
regarded as weakly acidic, which can be supported by the suppressed dehydration reaction of
hexadecanol on Rh/ZrO2. In spite of multiple active zones for HDO reaction, it is expected that the
HDO activity of active sites that are further from metal is lower than that at the immediate
perimeter.
Fig. 8. Concept of active site on Rh/ZrO2 catalyst for HDO of methyl palmitate. : Oxygen
vacancies on ZrO2; H: adsorbed hydrogen atom
Moreover,
6. Conclusions
Liquid-phase hydrodeoxygenation (HDO) of methyl palmitate, a model compound for vegetable oil,
was performed over Rh/ZrO2 catalyst in a batch reactor in the absence of external and internal mass
transfer resistances. It was found that HDO of methyl palmitate proceeded initially via
hydrogenolysis to palmitic acid, followed by sequential hydrogenation-decarbonylation reaction
steps into pentadecane via an aldehyde intermediate.
A bifunctional mechanism of the Rh/ZrO2 catalyst is advocated for the HDO reaction: both Rh
metal and ZrO2 contribute to the catalysis due to a synergy effect. By gathering literature
information, a surface reaction mechanism for the key reaction pathways was proposed, allowing
for the derivation of a kinetic model. The derived model consisting of two distinct types of active
sites, for both H2 and oxygenates, performed well in describing the experimental observations under
20

versatile reaction conditions. Moreover, a mechanistic model has the advantage of predicting
kinetics over a wider range of conversion, which is unlike the empirical power-law model with
fixed reaction orders.
The active phase on Rh/ZrO2 is further clarified for HDO process of fatty acid esters into biofuel
products. Different activity zones may exist simultaneously: 1) at a low-activity zone far away from
the metal particles, HDO reaction slowly proceeds on the oxygen vacancies in a similar mechanism
as occurred on bare ZrO2; 2) at a medium-activity zone nearby the metal particles, the dissociated
hydrogen by metal can spillover and accelerate the HDO on ZrO2; 3) the most-active zone is located
at the perimeter of Rh metal nanoparticle. At this zone, the support is electronically modified so that
the oxygen vacancies have better redox properties and therefore higher activity for the
deoxygenation pathway.
Acknowledgement
Graduate School in Chemical Engineering (GSCE) in Finland, Academy of Finland, and School of
Chemical Technology (Aalto University, Finland) are acknowledged for financial support. Prof.
Leon Lefferts from University of Twente, The Netherlands (FiDiPro Professor at Aalto University)
and Prof. Outi Krause from Aalto University are especially thanked for helpful discussion. Mr. José
Luis González Escobedo and Liam Gillan are thanked for language proofing.
21

Reference
[1] G.W. Huber, S. Iborra, A. Corma, Chem. Rev. 106 (2006) 4044–4098.
[2] D.C. Elliott, Energy & Fuels. 21 (2007) 1792–1815
[3] A. Demirbas, Energy Convers. Manage. 50 (2009) 14–34
[4] E. Furimsky, Appl. Catal., A. 199 (2000) 147–190.
[5] D Kubička, L Kaluža, Appl Catal., A. 372 (2010) 199–208.
[6] İ Şenol, E.M. Ryymin, T.R. Viljava, A.O.I. Krause, J. Mol. Catal. A 268 (2007) 1–8.
[7] İ Şenol, T R Viljava, A I Krause, Appl Catal A 7 –244.
[8] Y. Bie, A. Gutierrez, T.R. Viljava, J.M. Kanervo, J. Lehtonen, Ind. Eng. Chem. Res. 52 (2013)
11544–11551.
[9] L. Boda, G. Onyestyák, H. Solt, F. Lónyi, J. Valyon, A. Thernesz, Appl. Catal. A 374 (2010)
158–169
[10] M nåre, I Kubičková, P Mäki-Arvela, K. Eränen, D.Y. Murzin, Ind. Eng. Chem. Res. 45
(2006) 5708–5715.
[11] M nåre, I Kubičková, P Mäki-Arvela, D. Chichova, K. Eränen, D.Y. Murzin, Fuel. 87 (2008)
933–945.
[12] P.T.D. Do, M. Chiappero, L.L. Lobban, D.E., Resasco, Catal. Lett. 130 (2009) 9-18.
[13] R.W. Gosselink, S.A.W. Hollak, S.-W. Chang, J. van Haveren, K.P. de Jong, J.H. Bitter, D.S.
van Es, ChemSusChem. 6 (2013) 1576–1594.
[14] B.X. Peng, C. Zhao, S. Kasakov, S. Foraita, J.A. Lercher, Chem. Eur. J. 19 (2013) 4732–4741.
[15] B. Peng, X. Yuan, C. Zhao, J.A. Lercher, J. Am. Chem. Soc. 134 (2012) 9400–9405.
[16] S. Foraita, J.L. Fulton, Z.A. Chase, A. Vjunov, P. Xu, E. Baráth, D.M. Camaioni, C. Zhao, J.A.
Lercher, Chem. Eur. J. 21 (2015) 2423–2434.
[17] Y. Bie; J. Kanervo; J. Lehtonen, Ind. Eng. Chem. Res. 54 (2015) 11986–11996
[18] P. Kumar, S.R. Yenumala, S.K. Maity, D. Shee, Appl. Catal. A 471 (2014) 28–38.
[19] H. Zhang, H. Lin, W. Wang, Y. Zheng, P. Hu, Appl. Catal. B 150–151 (2014) 238–248.
[20] B Rozmysłowicz, P Mäki-Arvela, A. Tokarev, A.R. Leino, K. Eränen, D.Y. Murzin, Ind. Eng.
Chem. Res. 51 (2012) 8922-8927.
[21] J. Han, J. Duan, P. Chen, H. Lou, X. Zheng, H. Hong, ChemSusChem 5 (2012) 727–733
[22] V.A. Yakovlev, S.A. Khromova, O.V. Sherstyuk, V.O. Dundich, D.Y. Ermakov, V.M.
Novopashina, M.Y. Lebedev, O. Bulavchenko, V.N. Parmon, Catal. Today, 144 (2009), 362
[23] Y. Sakata, C.A. van Tol-Koutstaal, V. Ponec, J. Catal. 169 (1997) 13
22

[24] R. Pestman, R.M. Koster, J.A.Z. Pieters, V. Ponec, J. Catal. 168 (1997) 255–264.
[25] J.R. Aittamaa, K.I. Keskinen, Flowbat-User’s Instruction Manual, Laboratory of Chemical
Engineering Helsinki University of Technology, Finland, 2000.
[26] J.R. Aittamaa, K.I. Keskinen, Kinfit User’s Instructions Manual, Laboratory of Chemical
Engineering, Aalto University, Finland, 2013.
[27] A.C. Hindmarsh, in: R.S. Stepleman (Eds.), Scientific Computing, Amsterdam, North-Holland,
1983, pp. 55–64.
[28] T.N. Pham, D. Shi, D.E. Resasco, Top. Catal. 57 (2014) 706–714
[29] T.N. Pham, T. Sooknoi, S.P. Crossley, and D.E. Resasco, ACS Catal. 3 (2013) − 7
[30] T. Yokoyama, N. Yamagata, Appl. Catal. A 221 (2001) 227–239
[31] C. Diagne, H. Idriss, A. Kiennemann, Catal. Commun. 3 (2002) 565–571
[32] W. C. Conner, J. L. Falconer, Chem. Rev. 95 (1995) 759–708
23