Comparison of HMF hydrodeoxygenation over different metal catalysts in a continuous flow reactor

Article abstract of DOI:10.1016/j.apcata.2015.10.009

The three-phase hydrodeoxygenation (HDO) of 5-hydroxymethylfurfural (HMF) and hydrogenation of 2,5-dimethylfuran (DMF) were studied over six carbon-supported metal catalysts (Pt, Pd, Ir, Ru, Ni, and Co) using a tubular flow reactor with 1-propanol solvent, at 180 °C and 33 bar. By varying the space time in the reactor, the reaction of HMF is shown to be sequential, with HMF reacting first to furfuryl ethers and other partially hydrogenated products, which then form 2,5-dimethylfuran (DMF). Ring-opened products and 2,5-dimethyltetrahydrofuran (DMTHF) were produced only from reaction of DMF. Rate constants for the pseudo-first-order sequential reactions were obtained for each of the metals. The selectivities for the reaction of DMF varied with the metal catalyst, with Pd forming primarily DMTHF, Ir forming a mixture of DMTHF and open-ring products, and the other metals forming primarily open-ring products. Catalyst stabilities followed the order Pt ～ Ir > Pd > Ni > Co > Ru. Since the stability order correlated with carbon balances in the product (>93% for Pt; <75% for Ru), deactivation appears to be caused by deposition of humins on the catalyst.

Full text of DOI:10.1016/j.apcata.2015.10.009

Applied Catalysis A: General 508 (2015) 86–93
Contents lists available at ScienceDirect
Applied Catalysis A: General
journal homepage: www.elsevier.com/locate/apcata
Comparison of HMF hydrodeoxygenation over different metal
catalysts in a continuous flow reactor
Jing Luoa, Lisandra Arroyo-Ramíreza, Jifeng Weia, Hongseok Yunb,
Christopher B. Murrayb, Raymond J. Gortea,∗
a Department of Chemical & Biomolecular Engineering, University of Pennsylvania, Philadelphia, PA 19104, United States
b Department of Chemistry, University of Pennsylvania, Philadelphia, PA 19104, United States
a r t i c l e i n f o
a b s t r a c t
Article history:
The three-phase hydrodeoxygenation (HDO) of 5-hydroxymethylfurfural (HMF) and hydrogenation of
2,5-dimethylfuran (DMF) were studied over six carbon-supported metal catalysts (Pt, Pd, Ir, Ru, Ni, and
Co) using a tubular flow reactor with 1-propanol solvent, at 180 ◦C and 33 bar. By varying the space time
in the reactor, the reaction of HMF is shown to be sequential, with HMF reacting first to furfuryl ethers and
other partially hydrogenated products, which then form 2,5-dimethylfuran (DMF). Ring-opened products
and 2,5-dimethyltetrahydrofuran (DMTHF) were produced only from reaction of DMF. Rate constants for
the pseudo-first-order sequential reactions were obtained for each of the metals. The selectivities for the
reaction of DMF varied with the metal catalyst, with Pd forming primarily DMTHF, Ir forming a mixture
of DMTHF and open-ring products, and the other metals forming primarily open-ring products. Catalyst
stabilities followed the order Pt ∼ Ir > Pd > Ni > Co > Ru. Since the stability order correlated with carbon
balances in the product (>93% for Pt; <75% for Ru), deactivation appears to be caused by deposition of
humins on the catalyst.
Received 20 August 2015
Received in revised form 5 October 2015
Accepted 6 October 2015
Available online 22 October 2015
Keywords:
5-Hydroxymethylfurfural
Hydrodeoxygenation
Dimethyl furan
Continuous flow reactor
Metal catalyst
© 2015 Elsevier B.V. All rights reserved.
1. Introduction
H2 using a Ru/C catalyst in tetrahydrofuran (THF); but Saha et al.
[8] achieved only a 3% DMF yield using the similar catalyst and
Hydrodeoxygenation (HDO) of 5-hydroxymethylfurfural (HMF)
to 2,5-dimethylfuran (DMF) has received a great deal of attention in
the past few years because of its importance for converting biomass
into petrochemical replacements. HMF can be produced by dehy-
dration of glucose in reasonably high yields [1–3] but remains too
reactive and unstable for use as a fuel. The amount of expensive
H2 required to convert HMF to DMF is relatively small compared
to what would be required for conversion of HMF to alkanes. DMF
appears to have very good properties as a blending component for
gasoline [4] and can also be converted to p-xylene by reaction with
ethylene [5].
While a number of groups have reported high yields for the reac-
tion of HMF to DMF [6–7], there are very large variations in the
reported selectivities, sometimes for what appear to be nearly the
same catalyst and reaction conditions. For example, Hu, et al. [6]
reported yields to DMF as high as 95% for the reaction of HMF with
solvent. In another example, it has been reported that Pt catalysts
are not selective for the production of DMF due formation of over-
hydrogenated products, with some studies showing selectivities as
low as 10% [7]. However, in recent work from our laboratory [9],
yields as high as 60% were obtained on a Pt/C catalyst.
Determining the intrinsic selectivity for this reaction over a par-
ticular catalyst is complicated by the fact that it is a three-phase
reaction (gas-phase H2; liquid-phase HMF, DMF, and solvent; solid
catalyst) which is usually carried out in a semi-batch reactor. Dif-
fusion of H2 to the catalyst surface is likely to be the rate-limiting
step for most conditions. The catalyst volume is also typically a
small fraction of the liquid volume, necessitating residence times
on the order of hours in order to achieve high conversions. This
means that initially formed products can undergo secondary reac-
tions that in turn complicate analysis of the product selectivities
[9]. Catalyst stability is also difficult to study in these systems.
In recent work from our laboratory, the reaction of HMF to DMF
was studied over a Pt/C catalyst in both semi-batch and tubular
flow reactors, under similar pressure and temperature conditions
[9]. In agreement with previous reports that suggested Pt is a
poor catalyst for this reaction, the selectivity to DMF was very
∗
Corresponding author at: Department of Chemical and Biomolecular Engi-
neering, University of Pennsylvania, 311ATowne Building, 220 S. 33rd Street,
Philadelphia, PA 19104, United States. Fax: +1 215 573 2093.
E-mail addresses: jingluo@seas.upenn.edu, gorte@seas.upenn.edu (R.J. Gorte).
http://dx.doi.org/10.1016/j.apcata.2015.10.009
0926-860X/© 2015 Elsevier B.V. All rights reserved.

J. Luo et al. / Applied Catalysis A: General 508 (2015) 86–93
87
Scheme 1. Reaction network for HMF hydrodeoxygenation using alcohols as solvent.
low under semi-batch conditions. However, reasonable selectivi-
ties and yields could be obtained in the tubular flow reactor. By
varying the residence time in the flow reactor, it was possible to
show that the initial hydrogenation occurred on the carbonyl and
alcohol functionalities, ultimately forming DMF. DMF itself could be
considered an intermediate product that was further converted to
open-ring and hydrogenated-ring products at longer space times.
Therefore, the reaction network over Pt/C should be considered
sequential, with all products going through DMF. The high selectiv-
ities in the flow reactor were possible because the reactant contact
time was on the order of minutes. Poor selectivity was observed in
the semi-batch reactor because DMF was further converted.
The flow reactor also provided a convenient method for char-
acterizing catalytic performance. By changing the flow rate and
the catalyst loading, it was possible to vary the contact time
by nearly two orders of magnitude, allowing measurement of
the pseudo-first-order rate constants for the reaction network
under steady-state conditions, with various solvents and at dif-
ferent temperatures. Obviously, the fact that this is a three-phase
reaction makes a detailed understanding of these rate constants
complicated; however, the rate constants do allow quantitative
comparisons of different reactor conditions and different catalysts.
Rates of catalyst deactivation could be easily monitored by follow-
ing the conversion and selectivity as a function of time.
In the present work, we set out to compare the reaction of HMF
to DMF over a series of carbon-supported catalysts, including Pt, Ir,
Pd, Ni, Co, and Ru. These metals were chosen because they are all
good hydrogenation catalysts, which have been shown to exhibit a
wide range of selectivities for this reaction. All the reactions were
performed using 1-propanol as the solvent, at a 180 ◦C and 33 bar
pressure. Interestingly, the overall reaction network was found to
be similar with all six metals in that DMF is an intermediate product.
The six metals do show variations in the rates for each of the reac-
tion steps, different selectivities for over-hydrogenated products
formed from DMF, and different rates of deactivation.
2. Experimental
The catalysts were all prepared in our lab using carbon black
(Vulcan XC-72R) as the support and all had 10-wt% metal loadings.
Fig. 1. Conversion and product distribution for the HDO reaction of HMF over a 10-wt% Pt/C catalyst as a function of reactor space time. The overall product distribution
is given in (a) while a more detailed description of the over-hydrogenated products (product group D) is shown in (b). Reaction conditions: 33 bar and 180 ◦C. (᭹) HMF
conversion, (ꢀ) product group B, (ꢁ) DMF, (᭿) product group D, (ᮀ) DMTHF, (ꢂ) 2-hexanone, (ꢀ) 2,5-hexandione, (ꢃ) 2-propoxyhexane, (ꢁ) 2,5-dipropoxyhexane.

88
J. Luo et al. / Applied Catalysis A: General 508 (2015) 86–93
Fig. 2. Conversion and product distribution for the HDO reaction of HMF over a 10-wt% Ir/C catalyst as a function of reactor space time. The overall product distribution
is given in (a) while a more detailed description of the over-hydrogenated products (product group D) is shown in (b). Reaction conditions: 33 bar and 180 ◦C. (᭹) HMF
conversion, (ꢀ) product group B, (ꢁ) DMF, (᭿) product group D, (ᮀ) DMTHF, (ꢂ) 2-hexanone, (ꢀ) 2,5-hexandione, (ꢃ) 2-propoxyhexane, (ꢁ) 2,5-dipropoxyhexane.
The metals were added to the support by impregnation, using a
water/ethanol (5:1) solution of tetraammineplatinum (II) nitrate
(Pt(NH3)4(NO3)2, 99.99%, Alfa Aesar), tetraamminepalladium(II)
nitrate solution, (Pd(NH3)4(NO3)2, 99.9%, Alfa Aesar), cobalt(II)
nitrate hexahydrate (Co(NO3)2·6H2O, 99%, Aldrich), ruthenium(III)
nitrosylnitrate solution, (N4O10Ru, Alfa Aesar), or iridium(III) chlo-
ride hydrate (Cl3Ir·xH2O, Aldrich). The dried powders were reduced
by flowing a 5% H2/He mixture over the catalysts while ramping
the temperature at 2 ◦C min−1 to 400 ◦C, followed by heating to
500 ◦C with heating ramp of 1 ◦C min−1. The catalysts were then
held at this temperature for 2 h. The catalysts were then placed in
the flow reactor. We attempted to measure metal dispersions by
selective chemisorption but found the results to be unreliable, in
agreement with another recent report for carbon-supported metals
[10]; therefore, TEM measurements were performed on each of the
catalysts used in this study, with results shown in the Supplemental
materials section, Figs. S1–S6.
The reaction of HMF with H2 was carried out in the high-
pressure, flow reactor that has been described in detail elsewhere
[9]. The tubular reactor was a 20-cm long, stainless-steel tube with
a 4.6-mm ID and ¼-inch OD, passed through a tube furnace. The
liquid feed, a mixture of 1 g HMF (99%, Sigma–Aldrich) and 100 mL
1-propanol (99.9%, Fisher Scientific), was introduced into the reac-
tor by an HPLC pump (Series I+, Scientific System Inc.), which could
also monitor the total pressure in the reactor. The pressure within
the reactor was controlled by a back pressure regulator (KPB series,
Swagelok) that was connected at the outlet of reactor. The reactor
pressure was fixed at 33 bar for all the experiments performed in
this study. Pure H2 (Airgas, UHP grade) was supplied from a high
delivery pressure regulator (Airgas) to the reactor through 8 feet
of capillary tubing (0.002-inch ID, Valco Instrument, Inc.). The H2
flow rate was calibrated in separate experiments as a function of the
cylinder outlet pressure and pressure drop across the capillary tube.
For a typical experiment, the liquid flow rate was set as 0.2 mL/min,
while the H2 flow rate was 20 mL/min (STP). The ratio of liquid and
gas flow rates was kept constant. A bubble meter at the reactor exit
was used to check that the H2 flow rates were maintained.
measurements, each catalyst was pretreated by heating to 250 ◦C
in 1 bar of flowing H2 for 30 min. Fresh samples were used for each
experiment at a specified reaction condition.
The reaction products were collected at room temperature
and immediately injected into a GC–MS (QP-5000, Shimadzu) for
analysis using a syringe. The GC–MS was equipped with an HP-
Innowax capillary column (Agilent Technologies). Both liquid-and
gas-phase products were examined, but the gas-phase prod-
ucts were found to consist of only H2 and solvent vapor under
the conditions of this study. Product selectivities were quanti-
fied using solutions with known concentrations of HMF, DMF,
dimethyl-tetrahydrofuran (DMTHF), 2-hexanone, 2-hexanol, and
2,5-hexandione (all purchased from Sigma–Aldrich). For quan-
tification of other furan-based, intermediate products, the GC
sensitivity was assumed to be the same as that for HMF. For open-
ring, ether products, the GC sensitivity was assumed to be the same
as 2-hexanone or 2,5-hexandione. The typical run time for each
experiment was 3 h, and each GC sample analysis time was 30 min.
Because all of the catalysts underwent deactivation to some extent,
the data which we refer to as “initial performance” were typically
chosen from the second or third measurement (40–60 min after
starting the reaction).
In order to further characterize the DMF reaction kinetics and
the distribution of products formed from DMF, experiments were
also performed with a mixture of 0.76 g DMF and 100 mL of 1-
propanol as the feed. The molar concentration of DMF in this case
is the same as that used in the HMF experiments. No reaction was
observed in the absence of a catalyst under the conditions of this
study.
3. Results
In the previous HDO study of HMF over Pt/C in 1-propanol,
the reaction network was shown to be sequential [9], as indi-
cated in Scheme 1. In this scheme, the HMF (A) first reacts
to a group of partially hydrogenated compounds and propyl
ether products formed from those compounds (B). Specific prod-
ucts that were formed include 5-propoxymethyl-2-furanmethanol,
5-propoxymethyl-2-methylfuran (ether-methyl furan), 5-methyl
furfural (MF), 5-methyl-2-furanmethanol (methyl furfuryl alcohol),
and 2,5-bis(hydroxymethyl) furan (BHMF). All of these react further
to form DMF (C), which was then converted to over-hydrogenated
The catalyst samples (between 0.025 and 0.24 g, depending on
the desired range of space velocities to be tested) were packed into
the middle portion of the reactor and held in place by glass wool.
An inert glass tube was placed downstream from the catalyst in
order to prevent the catalyst bed from moving due to the reactant
flow and to minimize the open volume of the reactor. Prior to rate

J. Luo et al. / Applied Catalysis A: General 508 (2015) 86–93
89
Fig. 3. Conversion and product distribution for the HDO reaction of HMF over a 10-wt% Pd/C catalyst as a function of reactor space time. The overall product distribution
is given in (a) while a more detailed description of the over-hydrogenated products (product group D) is shown in (b). Reaction conditions: 33 bar and 180 ◦C. (᭹) HMF
conversion, (ꢀ) product group B, (ꢁ) DMF, (᭿) product group D, (ᮀ) DMTHF, (ꢂ) 2-hexanone, (ꢀ) 2,5-hexandione, (ꢃ) 2-propoxyhexane, (ꢁ) 2,5-dipropoxyhexane.
products (D), mainly DMTHF, 2-hexanone, 2-hexanol, and 2,5-
hexandione.
The results for Ir/C in Fig. 2 are the most similar to that of Pt/C.
Because the results for Ir are shown to higher space times, the
DMF conversion, shown in Fig. 2(a), is higher and the production of
over-hydrogenated products more extensive. The product carbon
balance was again greater than 90% at all space times. The most sig-
nificant difference between the data for Pt and Ir is in the products
formed from DMF, with the major products shown in Fig. 2(b). Ring-
opening products, 2,5-dipropoxyhexane and 2,5-hexanedione, are
still majority products but significant amounts of DMTHF are also
formed. Results for Pd/C follow the trend. As shown in Fig. 3(a), the
activity of the 10-wt% Pd/C catalyst is slightly lower than that of
the Pt and Ir catalysts, so that higher space times were needed for
high conversion of the HMF. However, ring opening is negligible
in the secondary products, Fig. 3(b), with DMTHF forming almost
exclusively. The product carbon balance on Pd/C was only 85% for
typical reaction conditions and a significant fraction (as much as
10%) of the over-hydrogenated products could not be identified.
As shown by the data for Ni/C, Fig. 4, even higher space times are
required for high conversions on Ni. However, the yield of DMF at
the optimal space time was reasonably high, greater than 50%; and
the major secondary product was 2,5-hexandione. As with Pd, the
product carbon balance was only 85%.
Similar to the case for Ni catalyst, data for Co/C, Fig. 5, and Ru/C,
Fig. 6, indicate that high space times are needed to achieve high
conversions. Also, 2,5-hexandione is the major secondary product.
The primary difference between these two metals and the others
is the carbon balance was very poor. Only 65–75% of the carbon
entering the reactor as HMF could be accounted for in the products.
Catalyst stability for the various metals loosely correlated with
the carbon balances observed in reaction. Fig. 7 is a plot of the
relative HMF conversion, normalized to the initial conversion, as
a function of time on stream for the six catalysts, for reaction at
180 ◦C, 33 bar, using a constant space time, W/F, of 0.25 g min/L.
For these flow conditions, the initial conversions were as follows:
83.4% for Pt/C, 64.5% for Ir/C, 36.4% for Pd/C, 47.0% for Ni/C, 36.7%
for Co/C, and 40.6% for Ru/C. The figure shows that deactivation was
modest on Pt, Ir, and Pd and much more severe over Ni, Co, and Ru.
Although some of the differences in deactivation between the cat-
alysts are due to the variations in initial activity, it is still apparent
that deactivation was most severe on Ru and Co, the two catalysts
for which the carbon balances were also worst. Therefore, it is likely
deactivation results from humin formation on the catalyst and that
Data which demonstrates this for a Pt/C catalyst are shown in
Fig. 1. In Fig. 1(a), we have plotted the conversion as a function
of space time, along with the yields of DMF and the sum of com-
pounds listed as B and D. The space times are given as the weight of
the metal catalyst divided by the volumetric flow rates of the liq-
uid. The data presented here were obtained at 180 ◦C and 33 bar but
qualitatively similar results were obtained at lower temperatures
[9]. The product carbon balance in this example was better than 93%
at all space times. Fig. 1(a) indicates that the conversion was greater
than 50% for even the shortest times. With increasing space time,
the partially hydrogenated compounds, B, decreased steadily while
the DMF initially increased, then decreased. A more detailed analy-
sis of the partially hydrogenated compounds is given in Table 1. The
over-hydrogenated products, D, are not formed initially but only
begin to form at higher space times. These results are a strong indi-
cation that this is a series reaction, since there is no hydrogenation
of the furan ring or formation of open-ring products before DMF
production levels off and begins to decline. Fig. 1(b) gives a more
complete analysis of the major over-hydrogenated products. For
Pt/C, these are 1-propoxy-1-methyl-pentane (2-propoxyhexane)
and 1,4-dipropoxy-1,4-dimethyl-butane (2,5-dipropoxyhexane),
which are reductive-etherification products formed by reaction
of 1-propanol with 2-hexanone and 2,5-hexandione, respectively.
Only a small amount of DMTHF was formed, demonstrating that
secondary reactions on Pt give primarily open-ring products. Small
amounts (<5%) of unidentified products were formed at the highest
space times.
Similar experiments were performed on 10-wt% Ir/C, Pd/C, Ni/C,
Co/C, and Ru/C, with results shown in Fig. 2 through 6 and Table 1.
The reaction experiments were again performed at 180 ◦C and
33 bar but we varied the amounts of catalyst that were loaded into
the reactor in order to achieve high conversions for all the metals.
The results for each metal were qualitatively the same in the fol-
lowing ways: (1) The HMF conversions increased with space time,
as expected. (2) The partially hydrogenated compounds, B, appear
to be the first products formed, generally decreasing with space
time except for low HMF conversions. (3) The DMF yields initially
increase with space time, then decline. (4) Over-hydrogenated
products only begin to form at higher space times and their for-
mation coincides with declining DMF yields.

90
J. Luo et al. / Applied Catalysis A: General 508 (2015) 86–93
Fig. 4. Conversion and product distribution for the HDO reaction of HMF over a 10-wt% Ni/C catalyst as a function of reactor space time. The overall product distribution
is given in (a) while a more detailed description of the over-hydrogenated products (product group D) is shown in (b). Reaction conditions: 33 bar and 180 ◦C. (᭹) HMF
conversion, (ꢀ) product group B, (ꢁ) DMF, (᭿) product group D, (ᮀ) DMTHF, (ꢂ) 2-hexanone, (ꢀ) 2,5-hexandione, (ꢃ) 2-propoxyhexane, (ꢁ) 2,5-dipropoxyhexane.
Fig. 5. Conversion and product distribution for the HDO reaction of HMF over a 10-wt% Co/C catalyst as a function of reactor space time. The overall product distribution
is given in (a) while a more detailed description of the over-hydrogenated products (product group D) is shown in (b). Reaction conditions: 33 bar and 180 ◦C. (᭹) HMF
conversion, (ꢀ) product group B, (ꢁ) DMF, (᭿) product group D, (ᮀ) DMTHF, (ꢂ) 2-hexanone, (ꢀ) 2,5-hexandione, (ꢃ) 2-propoxyhexane, (ꢁ) 2,5-dipropoxyhexane.
Fig. 6. Conversion and product distribution for the HDO reaction of HMF over a 10-wt% Ru/C catalyst as a function of reactor space time. The overall product distribution
is given in (a) while a more detailed description of the over-hydrogenated products (product group D) is shown in (b). Reaction conditions: 33 bar and 180 ◦C. (᭹) HMF
conversion, (ꢀ) product group B, (ꢁ) DMF, (᭿) product group D, (ᮀ) DMTHF, (ꢂ) 2-hexanone, (ꢀ) 2,5-hexandione, (ꢃ) 2-propoxyhexane, (ꢁ) 2,5-dipropoxyhexane.

J. Luo et al. / Applied Catalysis A: General 508 (2015) 86–93
91
Table 1
Yields of partially hydrogenated compounds over different metal catalysts at a space time of 0.25 g min/mL.
Catalyst
Pt
Ir
Pd
Ni
Co
Ru
18.3
7.3
11.4
8.3
2.5
1.5
2.6
2.2
1.4
<1
<1
1.1
<1
<1
<1
2.0
9.3
14.5
1.8
–
–
–
–
1.2
1.3
1.5
1.8
–
–
<1
Table 2
the missing carbon from the mass balance is depositing onto the
catalyst.
Pseudo first-order rate constants for HMF hydrodeoxygenation over the various
metal catalysts, for reaction in 1-propanol at 180 ◦C and 33 bar.
To obtain a rough estimate of the amount of carbon that may
have deposited on the Ru and Co catalysts, we assumed an average
carbon balance 70% throughout the reaction and a typical liquid
flow rate of 0.2 mL/min. Based on these assumptions, approxi-
mately 0.10 g of HMF-derived compounds could have deposited
onto 0.05 g catalyst after 3 h. Although this amount is large, it is
possible that this much of the reactant was retained on the cata-
lyst, given the high surface area and low density of the activated
carbon support. Because it is very difficult to weigh the wet cata-
lyst after taking it out of the reactor, it was not possible for us to
determine with certainty whether this is the reason for the poor
carbon balance but the calculation does explain the deactivation
behavior.
ꢂ
Catalyst
k1 (min−1
)
k2 (min−1
)
k3 (min−1
)
k3 (min−1
)
Pt
Ir
Pd
Ni
Co
Ru
5.8
3.5
1.5
1.4
1.1
1.0
9.2
8.8
5.7
3.1
3.7
4.3
1.1
1.2
0.56
0.81
0.14
0.10
0.16
0.48
0.77
0.12
0.09
0.14
ꢂ
Note: k3 was calculated from rate measurements using DMF as the feed.
during the reaction of HMF is high at a space time of 4 g min/mL,
a result which is also found when DMF is fed to the reactor at
this space time. More importantly, the major products from the
reaction of DMF agreed with that obtained as secondary prod-
ucts in the reaction of HMF. For example, significant amounts of
2-propoxyhexane were produced over Pt/C from DMF reaction at
high space time. High yields of 2,5-dipropoxyhexane were observed
with Ir/C. DMTHF was the primary product formed over Pd/C. For
Ni/C, Co/C and Ru/C, ring-opening or ring-hydrogenation products
were also observed with increasing space time. The major differ-
ence between results for the reaction of DMF and HMF is that
the yields of 2,5-hexandione (and its further reaction product, 2,5-
dipropoxyhexane) are lower in the reaction of DMF. This is almost
certainly due to the much lower water content in the reactant
stream when DMF is fed to the reactor, since the reaction of HMF to
DMF produces water and 2,5-hexandione is formed by hydrolysis.
(Note: Some water was present even in the DMF experiments due
to the reaction of 1-propanol to form dipropyl ether, a product that
was observed in the GC/MS.)
3.1. Reaction of DMF
To help support the conclusion that the over-hydrogenated
compounds in Fig. 1 through 6 result from the reaction of DMF, we
investigated the products formed by feeding DMF to the reactor.
The experiments were carried out with a solution of DMF in 1-
propanol (at the same molar concentration as HMF in 1-propanol),
varying W/F over the six catalysts at 180 ◦C and a total pressure
of 33 bar. The data are summarized in Fig. 8. The reaction of DMF
in this case differs from the conditions of the sequential reactions
for HMF in that water is formed in the HDO reaction to form DMF,
while no water was present when DMF was fed to the reactor.
First, there is qualitative agreement between the conversion of
DMF in the plots of Fig. 8 and the data in Fig. 1 through 6. For
example, with Ir/C, the conversion of DMF to secondary products
3.2. Kinetic models
In an attempt to quantify the differences between the metals, we
modeled the reaction of HMF as a series of first-order, sequential
reactions. Using the reaction network shown in Scheme 1, k1 is
defined as the rate constant for the reaction A → B, k2 for B → C,
and k3 for C → D. The data in Fig. 1 through 6 were then fit to give
the rate constants shown in Table 2. Table 2 also reports values for
k3ꢂ, which were obtained for the reaction C → D from the reaction
data for DMF in Fig. 8. All of the rate constants should obviously
be used for comparison purposes only, because the reaction occurs
in a complex, three-phase environment. Also, given that all of the
catalysts were prepared with the same weight, not mole, percent
metal, comparison of rate constants for the various metals is also
only qualitative. It was not possible to include humin formation in
the reaction network. First, it should be noticed that there is good
agreement between k3 and k3ꢂ, providing additional evidence that
the reaction is sequential on each of the metal catalysts. The fact
that k3 values were slightly higher in some cases can be explained
by the presence of water formed from the reaction of HMF and
the enhanced rates for ring-opening reactions that the presence of
Fig. 7. Normalized conversions for HMF hydrodeoxygenation over each of the
six metal catalysts as
a function of measurement time. Reaction condition:
W/F = 0.25 g min/mL, 33 bar and 180 ◦C. (ᮀ) Pt/C, (ꢀ) Ir/C, (ꢁ) Pd/C, (ꢂ) Ni/C, (♦)
Co/C, (ꢃ) Ru/C.

92
J. Luo et al. / Applied Catalysis A: General 508 (2015) 86–93
Fig. 8. Conversion and product distributions for the reaction of DMF in 1-propanol over the various metal catalysts. Reaction condition: 33 bar and 180 ◦C. (a) 10-wt% Pt/C;
(b) 10-wt% Ir/C; (c) 10-wt% Pd/C; (d) 10-wt% Ni/C; (e) 10-wt% Co/C; (f) 10-wt% Ru/C. (᭹) DMF conversion, (ᮀ) DMTHF, (ꢂ) 2-hexanone, (♦) 2-hexanol, (ꢀ) 2,5-hexandione,
(ꢃ) 2-propoxyhexane, (ꢁ) 2,5-dipropoxyhexane.
water causes. Second, the k1 values in the table quantify our earlier
qualitative conclusion that Pt and Ir catalysts were more active for
HMF conversion, while the Pd, Ni, Co, and Ru catalysts exhibited
much lower rates. The Pt and Ir catalysts also have much higher
values for k3, which is not desirable for achieving high yields of
DMF. The relatively high yields obtained for DMF over Ni/C, even
though the rates for DMF formation are relatively low, result from
the fact that k3 is also relatively low. It seems apparent that the
optimal catalyst would have a high k1:k3 ratio.

J. Luo et al. / Applied Catalysis A: General 508 (2015) 86–93
93
4. Discussion
from other laboratories has shown that water tends to promote
ring-opening reactions [12]. For these reasons, the present work
focused on using alcohols as the solvent.
The most important lesson from the results of the present study
is that the reaction network for hydrodeoxygenation of HMF is very
similar for all six of the metal catalysts that were studied. Indeed,
the shape of the curves for products as a function of residence
time are distinct in showing that the reaction is sequential, with
DMF formed as an intermediate product. Over-hydrogenated prod-
ucts were not observed at low space times and appear only as the
DMF selectivities begin to decline. This indicates that hydrogena-
tion of the furan ring and formation of ring-opened products occur
through DMF and do not occur until alcohol and carbonyl func-
tionalities are removed from the HMF. The fact that this reaction
is sequential has very important implications since selectivity, not
rate, is the crucial issue in the choice of catalyst. All of the metals we
studied were effective in forming DMF. Whether the metal catalyst
was capable of achieving high yields was determined more by its
reactivity towards DMF. To achieve higher selectivities, one should
target catalysts that are inactive for reactions of DMF.
It seems apparent that understanding the reaction network for
HDO of HMF is critical for catalyst characterization and for the
discovery of new materials. Because the most selective materials
reported in the literature for the HDO reaction of HMF are bimetallic
catalysts [7,13–18], it will be interesting to see whether the reac-
tion on these materials is again sequential or whether it would be
better to describe these materials in terms of parallel reaction net-
work. Future work in our laboratory will attempt to address this
question with these other types of catalysts.
5. Conclusions
The hydrodeoxygenation of HMF to DMF occurs via a sequential
reaction network on Pt, Ir, Pd, Ni, Co, and Ru, with the selectivity
determined by the ratio of rates for the reaction of DMF to over-
hydrogenated products and the formation of DMF. While all of the
metals initially formed DMF from the HMF, DMF was hydrogenated
to different products over the different metals. Catalyst stabilities
were also metal dependent. The results indicate that the selectivi-
ties to DMF are a strong function of the type of reactor used to make
the measurements.
Although the etherification intermediate compounds are not
desired products, our previous study has shown that the forma-
tion of these ether products can improve the HDO rate over Pt/C
[9]. Since those ether compounds can be further converted to DMF,
the yield of the ethers can be simply decreased by increasing space
time or reaction temperature.
The sequential nature of the reaction also has important con-
sequences for how catalysts should be characterized. For example,
if catalysts are simply compared by analyzing the products after a
fixed time in a semi-batch reactor, incorrect conclusions may be
reached if the reaction time is near the optimum value for one cat-
alyst and past the optimum for another. For example, the present
work suggests that reports of poor selectivity with Pt are likely to
be caused, at least in part, by the high activity of Pt. Previous batch
reactor measurements implied that Pt is not a selective catalyst,
whereas it was the most selective (for DMF) of the metals examined
in the present flow-reactor study.
Catalyst stability may also play an important role in the yields
observed on different metals. For example, among the pure metals,
Ru has been reported to be one of the more selective metals [6],
much better than Pt [7]. As discussed above, it seems likely that
poor selectivity with Pt is at least partially due to the high activ-
ity and good stability, which leads to high rates for the reaction of
DMF to form other products. On the other hand, Ru was observed to
undergo severe deactivation in our study, preventing further reac-
tion of DMF. Therefore, poor stability may be responsible for some
previous reports from semi-batch studies that indicated Ru was the
most selective of the pure metals [6].
Acknowledgement
We acknowledge support from the Catalysis Center for Energy
Innovation, an Energy Frontier Research Center funded by the U.S.
Department of Energy, Office of Science, Office of Basic Energy Sci-
ences under Award no. DE-SC0001004.
Appendix A. Supplementary data
Supplementary data associated with this article can be found,
in the online version, at http://dx.doi.org/10.1016/j.apcata.2015.10.
009.
References
[1] T. Wang, M. Nolte, B. Shanks, Green Chem. 16 (2014) 548.
[2] M. Dashtban, A. Gilbert, P. Fatehi, R. Soc, Chem. Adv. 4 (2014) 2037.
[3] A. Torres, P. Daoutidis, M. Tsapatsis, Energy Environ. Sci. 10 (2010) 1560.
[4] M. Papachristos, J. Swithenbank, G. Priestman, S. Stournas, P. Polysis, E. Lois, J.
Inst. Energy 64 (1991) 113.
[5] C. Williams, C. Chang, P. Do, N. Nikbin, S. Caratzoulas, D. Vlachos, R. Lobo, W.
Fan, P. Dauenhauer, ACS Catal. 2 (2012) 935.
[6] L. Hu, X. Tang, J. Xu, Z. Wu, L. Lin, S. Liu, Ind. Eng. Chem. Res. 53 (2014) 3056.
[7] G. Wang, J. Hilgert, F. Richter, F. Wang, H. Bongard, B. Spliethoff, C.
Weidenthaler, F. Schüth, Nat. Mater. 13 (2014) 294.
The various metals did show differences for whether DMF reacts
via furan ring-hydrogenation or ring-opening. This can probably be
explained by the relative adsorption energies of furan rings on the
various metal surfaces. For example, it has been argued that [11]
the furan ring is strongly absorbed on Ni, which in turn promotes
the weakening of the C O bond, leading to ring-opening. On Pd,
the interaction is weaker and ring-hydrogenation is preferred.
While we did not examine solvent effects in the present study,
the choice of solvent is clearly an important variable in these reac-
tions. In a previous publication from our laboratory [9], HDO rates
and selectivities over Pt/C were found to be better with alcohols
than with the non-polar solvent, toluene. On the other hand, work
[8] B. Saha, C. Bohn, M. Abu-Omar, ChemSusChem 7 (2014) 3095.
[9] J. Luo, L. Arroyo-Ramírez, R. Gorte, D. Tzoulaki, D. Vlachos, AIChE J. 61 (2015)
590.
[10] J. Tengco, Y. Lugo-José, J. Monnier, J. Regalbuto, Catal. Today 246 (2015) 9.
[11] S. Sitthisa, D. Resasco, Catal. Lett. 141 (2011) 784.
[12] Y. Nakagawa, M. Tamura, K. Tomishige, ACS Catal. 3 (2013) 2655.
[13] S. Nishimura, N. Ikeda, K. Ebitani, Catal. Today 232 (2014) 89.
[14] S. Sitthisa, W. An, D. Resasco, J. Catal. 284 (2011) 90.
[15] M. Myint, Y. Yan, J. Chen, J. Phys. Chem. C 118 (2014) 11340.
[16] J. Lee, Y. Kim, G. Huber, Green Chem. 16 (2014) 708.
[17] Y. Huang, M. Chen, L. Yan, Q. Guo, Y. Fu, ChemSusChem 7 (2014) 1068.
[18] Y. Zu, P. Yang, J. Wang, X. Liu, J. Ren, G. Lu, Y. Wang, Appl. Catal. B 146 (2014)
244.