Sie sind auf Seite 1von 11

Journal of Catalysis 318 (2014) 151161

Contents lists available at ScienceDirect

Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat

Kinetic and FTIR studies of 2-methyltetrahydrofuran


hydrodeoxygenation on Ni2P/SiO2
Ara Cho a,b,1, Hosoo Kim b, Ayako Iino a, Atsushi Takagaki a, S. Ted Oyama a,c,
a
Department of Chemical Systems Engineering, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan
b
Corporate R&D, LG Chem Research Park, 188, Munji-ro, Yuseong-gu, Daejeon 305-738, Republic of Korea
c
Department of Chemical Engineering, Virginia Tech, Blacksburg, VA 24061, USA

a r t i c l e i n f o a b s t r a c t

Article history: The hydrodeoxygenation of 2-methyltetrahydrofuran (2-MTHF) at a medium pressure of 0.5 MPa is
Received 10 June 2014 studied over a Ni2P/SiO2 catalyst. Contact time studies show the formation of n-pentanal, 1-pentanol,
Revised 24 July 2014 2-pentanone, 2-pentanol, and 2-pentene, as reaction intermediates and the production of pentane and
Accepted 26 July 2014
butane as major products. The results are consistent with adsorption of 2-MTHF followed by rate-
Available online 27 August 2014
determining ring-opening to form either 1-pentoxide or 2-pentoxide alkoxide intermediates. Subsequent
hydrogen-transfer steps produce the various intermediates, a decarbonylation step of the pentanal forms
Keywords:
n-butane and CO, and further hydrodeoxygenation steps result in n-pentane. Fitting of the results using a
Hydrodeoxygenation
2-Methyltetrahydrofuran
rake mechanism that considers adsorbed intermediates gives excellent agreement with experimental
Ni2P/SiO2 data, and agrees with a simulation with a simpler rst-order model. The more detailed rake analysis indi-
Rake mechanism cates that the surface species from the 1-pentoxide intermediate are ten-fold more plentiful than those
In situ FTIR produced from the 2-pentoxide intermediate. In situ infrared measurements support this reaction
mechanism.
2014 Elsevier Inc. All rights reserved.

1. Introduction Phosphides are particularly promising and the subject has been
reviewed [25,26].
The utilization of biomass to replace fossil fuels is a topic of Previous work in our laboratory on the reaction of guaiacol [27]
current interest due to increasing concerns about global climate and ethanol [28] showed that Ni2P is an effective catalyst for
change induced by the release of carbon dioxide and the desire deoxygenation and this has been conrmed in a number of HDO
to develop renewable sources of energy [14]. Many methods for studies such as with guaiacol [29], anisole [23,22], dibenzofuran
the conversion of biomass to fuels have been considered [5,6], [30], and 4-methoxyphenol [31]. Recent work in our laboratory
including the production of bioethanol [7], biodiesel [8], Fischer [32,33] employed the model compound 2-methyltetrahydrofuran
Tropsch liquids [9,10], and pyrolysis oil [11]. The route involving (2-MTHF) as a substrate for deoxygenation because there are
pyrolysis oil is competitive for converting biomass to naphtha substantial amounts of furans contained in pyrolysis oil. Also, the
and diesel [12,13]. presence of the methyl group gives the compound two routes for
A problem with pyrolysis oil is that it contains considerable ring-opening and this can give information about the nature of
amounts of oxygen (40 wt%) [11], which reduces its heating value the active site that is involved in the reaction. An initial study at
and makes the oil unstable. For this reason there has been 0.1 MPa [32] showed that the order of activity was Ni2P > WP >
signicant work in the hydrodeoxygenation (HDO) of the oil, either MoP > CoP > FeP, and a contact time analysis for Ni2P showed that
as it is produced [14] or in separate units [15]. Many catalysts small amounts of intermediate alkenes were formed rst, then
have been tried, including suldes [16], zeolites [17], carbides/ small amounts of 2-pentanone, and n-pentanal and nally large
nitrides [18], phosphides [1922], and phosphide alloys [23,24]. amounts of butane and pentane products. A subsequent study at
0.5 MPa on Ni2P [33] showed slight differences in intermediates
Corresponding author at: Department of Chemical Systems Engineering, The such that small amounts of alcohols were formed rst instead of
University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan. alkenes, then 2-pentanone, and nally butane and pentane. That
E-mail addresses: ted_oyama@chemsys.t.u-tokyo.ac.jp, oyama@vt.edu study used a simple rst-order kinetic scheme to simulate the net-
(S. Ted Oyama). work, and it was concluded that the initial steps of ring-opening
1
Current address: Corporate R&D, LG Chem Research Park, 188, Munji-ro, Yuseong-
were the rate-determining steps in the sequence.
gu, Daejeon 305-738, Republic of Korea.

http://dx.doi.org/10.1016/j.jcat.2014.07.021
0021-9517/ 2014 Elsevier Inc. All rights reserved.
152 A. Cho et al. / Journal of Catalysis 318 (2014) 151161

Nomenclature

Symbol description units NV number of measured variables ()


ni total number of measurements over all experiments for Wj weight of experiment j ()
measured variable i () z predicted model response ()
NE number of experiments () ~z measured response ()
NM number of measurements () ki heteroscedasticity parameter for measured variable i ()

The objective of the present study is to provide a more in-depth a Ni foil. The storage ring was operated at 2.5 GeV with a beam
kinetic analysis of the reaction network at the intermediate current of 300 mA. For the EXAFS measurements, the samples were
pressure of 0.5 MPa that takes into consideration adsorbed species. prepared in the form of pellets (/ = 1 cm) with 60 mg of calcined
In the previous work, [33] measurements were carried out to a Ni2P/SiO2 precursor and, for the reference, a mixture of bulk Ni2P
contact time of just under 0.1 s and conversions of 1%. In the (6 mg) and BN (36 mg) was used. The pellet of the Ni2P/SiO2
present paper, the contact time was reduced to about 0.01 s and sample was reduced in a quartz reactor in the same manner as
conversions of 0.1% were studied. A further goal is to monitor the the catalyst sample at 723 K (450 C). After reduction, the reactor
adsorbed species using in situ infrared spectroscopy also at containing sample was carefully moved into a glove box without
0.5 MPa to provide independent evidence for the reaction exposing it to air and the sample was taken out and taped on both
sequence. Selectivity measurements are carried out with attention sides of the pellet with polyamide tape. The measurements were
placed on the low contact time region to check on previous results, done at room temperature in transmission mode. The EXAFS spec-
and to determine the formation of n-pentanal, a key reaction tra were analyzed using WinXAS3.1. Theoretical scattering ampli-
intermediate. tudes and phase shifts for the Ni2P were calculated by FEFF8
code [35]. Curve tting was conducted using the three dominant
2. Experimental shells (2 NiP, 4 NiP and 4-NiNi) [36], and the reducing factor
(So2) was xed as 0.8.
2.1. Catalyst synthesis and characterization
2.2. Contact time analysis
The nickel phosphide catalyst was synthesized by incipient wet-
ness impregnation followed by temperature programmed reduc- The effect of contact time on the conversion and selectivity for
tion (TPR) as described elsewhere [34]. Briey, a silica Cabosil the HDO of 2-MTHF was studied in a xed-bed ow reactor in the
support was impregnated with aqueous solutions of Ni(NO3)26H2O gas phase at 573 K and 0.5 MPa over the Ni2P/SiO2 catalyst. Prior to
and (NH4)2HPO4, which was followed by drying at 383 K for 12 h the reaction, passivated catalysts were re-reduced with hydrogen
and calcination at 773 K for 6 h in static air. The initial Ni/P ratio at 723 K for 2 h at atmospheric pressure. The reactant, 2-MTHF,
in the impregnating solution was kept at 1/2. The calcined precur- containing 5 vol% of n-heptane as an internal standard, was fed
sors were pelletized and sieved to particles of 6501180 lm diam- using a liquid pump to a vaporizer before delivery to the reactor.
eter, and the resulting particles were reduced to phosphide by a TPR During the reaction, products were sampled every 1 h and were
method in 1 l/min of hydrogen. The temperature was increased at a analyzed on-line using a Shimadzu 10A gas chromatograph (GC)
rate of 3 K/min from room temperature to the nal temperature of equipped with a silicone capillary column and ame ionization
863 K, which was determined by TPR of samples and the tempera- detector (FID). The analytical method was improved compared to
ture was maintained at the nal state for 2 h. After cooling to room past work by increasing the integrator sensitivity. Contact time
temperature in He ow, reduced samples were passivated in 0.5% was changed in the range of 0.010.31 s while keeping the inlet
O2/He mixture at room temperature for 3 h before being taken molar ratio of H2 to 2-MTHF constant at 19. Contact time (s) was
out from the TPR reactor. The levels of Ni loading in the Ni2P/SiO2 dened as s (s) = V/v, where V is the volume of the catalyst bed
catalyst were 1.16 mmol (g support)1. and v is the ow rate of hydrogen. Total conversion was kept below
Carbon monoxide (CO) uptakes on the samples were measured 11% for every measurement except one point.
by a pulse injection method using a catalyst analyzer (CHEMBET-
3000, Quantachrome Instruments) equipped with a thermal con- 2.3. Fourier transform infrared spectroscopy
ductivity detector to estimate the dispersion of active sites on
the catalysts. Prior to adsorption, 0.3 g of passivated samples were Temperature programmed desorption (TPD) of 2-MTHF
placed in a quartz reactor and were re-reduced with hydrogen at adsorbed on the Ni2P/SiO2 catalyst was studied using infrared
723 K for 1 h and then cooled to room temperature in helium ow. spectroscopy under nitrogen or hydrogen ow. For the in situ
Pulses of volume 25 ll of 10% CO/He were injected using a helium transmission infrared measurement of the Ni2P/SiO2 catalyst sur-
carrier owing at 100 ml min1 to a reactor containing the catalyst face, the sample was prepared in the form of self-supporting disk
at room temperature until saturation of the sample surface with of diameter 1 cm (15 mg) and was placed in the center of an IR
CO was achieved. cell equipped with KBr windows (Fig. 1). To reduce the dead vol-
Transmission electron microscopy (TEM) images of Ni2P/SiO2 ume of the IR cell, stainless-steel rods with a hole (/ = 0.8 cm) were
were taken with a JEOL JEM 2000EXII operated at 200 kV. The inserted in both sides of the IR cell and KBr rods were inserted into
Ni2P/SiO2 catalysts were ground to ne particles and dispersed into those holes to minimize absorption by gas-phase species in the IR
ethanol and put on a carbon-coated copper grid. spectra. Before dosing the reactants, the sample was reduced at
In situ extended X-ray absorption ne-structure (EXAFS) mea- 623 K for 3 h in a ow of 100 ml (NTP)/min of hydrogen. After cool-
surements were made at the BL9C, beam line of the Photon Factory ing to 423 K in a N2 or H2 ow, background spectra were obtained
in the Institute of Materials Structure Science, High-Energy Accel- at 423, 448, 473, 498, 523, 548, 573, and 598 K, respectively. After
erator Research Organization (KEK-IMSS-PF). The EXAFS spectra cooling again to 423 K, 2-MTHF was injected to the sample with N2
were obtained in transmittance mode, and were calibrated using or H2 as the carrier and then the sample was purged with the same
A. Cho et al. / Journal of Catalysis 318 (2014) 151161 153

Fig. 1. Schematic diagram of in situ IR cell.

gas for 0.5 h to remove gas-phase species. IR spectra were collected simplex algorithm that is computationally cheap since it only
in situ at increasing temperature from 423 to 598 K under N2 or H2 involves function evaluations and no gradient or Hessian informa-
ow with a JASCO FT/IR-6100 equipped with a MCT detector with a tion [37]. It can require many more iterations to converge than gra-
resolution of 1 cm1 in the region of 40001000 cm1. dient based methods; however, it may be able to solve complex
Transient spectra for desorption of 2-MTHF adsorbed on the problems which fail to converge using FEASOPT.
Ni2P/SiO2 catalyst were obtained at constant temperature, 523 K
under nitrogen or hydrogen ow using the same reactor as used
3. Results and discussion
for the other IR measurements (Fig. 1). After pretreatment at
623 K the samples were cooled to 523 K in a N2 or H2 ow and
A transmission electron microscopy (TEM) image of a freshly
background spectra were obtained in the absence of 2-MTHF. Then
reduced and passivated Ni2P/SiO2 catalyst (Fig. 1a) shows highly
2-MTHF was injected to the sample with N2 or H2 as the carrier
dispersed particles with narrow size distribution. The particle size
until saturation was achieved. After saturation, IR spectra were col-
distribution for 100 particles (Fig. 1b) shows that the average par-
lected under N2 or H2 ow without 2-MTHF injection keeping the
ticle diameter was 3.7 nm.
temperature constant at 523 K. Measurement was made every
Structural characterization of a freshly reduced sample placed
60 s with a resolution of 4 cm1 in the region of 4000500 cm1
in a cell with Kapton windows was carried out with extended
using 100 scans spectrum1. (For the TPD experiments, measure-
X-ray absorption ne-structure (EXAFS) at room temperature.
ments were made with a resolution of 4 cm1 in the region of
The EXAFS spectrum shows two peaks (Fig. 2) due to NiP and
40001000 cm1 using 200 scans spectrum1).
NiNi bonds (Table 2) at 0.225 nm and 0.26 nm, respectively,
which agree well with previously reported values of 0.23 and
2.4. Parameter estimation 0.26 nm [34]. The low coordination numbers are consistent with
particle sizes in the vicinity of 4 nm, in agreement with the TEM
The proposed mechanism was modeled with Aspen Custom results [38] (see Fig. 3).
Modeler (ACM), a commercially available software package devel- Earlier, we had estimated particle sizes from the CO uptake and
oped by AspenTech, and then the kinetic parameters were esti- this gave a particle size of 22 nm [33]. We suspect that the CO
mated by a mathematical routine that optimizes the t of model uptake was low because of site blockage by phosphorus, and this
results to measured values from the experiment data by varying gave the large value. Low values of chemisorptions were found ear-
the kinetic parameters. The following maximum likelihood log lier [39]. The present TEM results are more reliable because they
function was used for the objective function, which denes an opti- are a direct determination.
mal t. Table 2 summarizes the selected crystallographic Ni2P standard
8 2 !2 3 data which was used in the curve tting and the tting parameters
<1 XNV XNE NM
Xij 2 ~zijk  zijk (bond distance, coordination number, and DebyeWaller factor) of
min n ln2p 1 ni ln 4ni Wj 5
:2 i1 i j1 k1
~zkijki the EXAFS spectra for the bulk Ni2P reference and the Ni2P/SiO2
) catalyst. The rst and the second shells correspond to two different
NE NM
X Xij   NiP bonds and the third shell corresponds to a NiNi bond. For the
ki ln W j ~zijk
Ni2P/SiO2 catalyst in the rst shell, the coordination number was
j1 k1
almost same as that of bulk Ni2P, whereas coordination numbers
For solving this optimization problem, FEASOPT and Nelder in the second and the third shells appeared larger and smaller,
Mead solvers were used in sequence. The experiment data did respectively, than those of bulk Ni2P. It was reported that smaller
not contain sufcient information to provide good initial guesses Ni2P crystallites have stronger NiP interaction and are decorated
to accurately determine the 16 kinetic parameters. Therefore, ini- with phosphorus on the surface resulting in larger coordination
tial guesses for the kinetic parameters of the mechanism were rst number in the second shell (NiP) [34,40]. It was also suggested
obtained using FEASOPT solver, a feasible path successive qua- that the decrease in the coordination number in the third shell
dratic programming which can rapidly move the model parame- (NiNi) may due to extra phosphorus inducing reduction of the
ters in a correct direction based on gradient information. Next, a NiNi coordination [37]. The EXAFS analysis results conrm that
NelderMead solver was used to get a better solution from the the highly dispersed Ni2P was formed on the SiO2 support, which
initial guesses by minimizing the objective function based on the is consistent with the TEM results.
154 A. Cho et al. / Journal of Catalysis 318 (2014) 151161

(a)

k -weighted FT / a.u.
Bulk Ni2P

3
Ni2P/SiO2

50 nm
0 2 4 6
R (10 nm)

Fig. 3. Fourier transform Ni K-edge EXAFS oscillations of Ni2P/SiO2 reduced in H2 at


(b) 40
723 K (450 C) for 8 h and bulk Ni2P.
Average 3.7 nm
35

30 of the 20-fold difference in scales (70% vs. 3.5%). One set of mea-
surements was made for the entire data, but the points were taken
25
Counts

at increasing and decreasing contact times to establish that the


20
catalyst was stable during the experiment. The selectivity to
n-pentane increases monotonically as expected for a nal product
15 in a sequence. The selectivity to n-butane also increases but much
more gradually, because it is a product of an intermediate,
10
n-pentanal, that can react to form both n-butane by decarbonyla-
5 tion and n-pentane by sequential hydrogenation.
The data in Fig. 4 are largely in agreement with previous results
0
1 2 3 4 5 6 7 8 9 10 [33], except that an improved analytical method was used here
Particle size / nm that enabled the observation of n-pentanal, an intermediate that
appears at the lowest concentration (Fig. 4b). Based on the present
Fig. 2. (a) TEM image of Ni2P/SiO2 and (b) particle distributions of Ni2P/SiO2 100 contact time results, the overall reaction pathway for the 2-MTHF
particles.
HDO over Ni2P/SiO2 is modied so that n-pentanal is shown explic-
itly (Fig. 5). The pathway has two branches, with ring-opening by
CO bond breaking occurring at either the more hindered (k1) or
Reactivity results in the form of selectivity vs. contact time were the more open (k5) position. Previous work carried out also at
used to determine the reaction network (Fig. 4). The top panel 0.5 MPa showed that these rate constants were small compared
(Fig. 4a) shows the selectivity to the main products, and the bot- to the rest of the rate constants and, thus, that the ring-opening
tom panel to the minor products (Fig. 4b). Note should be made step was rate-determining [31].

Table 1
Textural properties of Ni2P catalyst.

Catalyst BET area CO uptake Crystallite size from CO uptake Crystallite size from TEM
Ni2P/SiO2 214 m2/g 55 lmol/g 22 nm 3.7 nm

Table 2
EXAFS parameters for the bulk Ni2P and Ni2P/SiO2.

NiP (I) NiP (II) NiNi (I and II)


Reference R-factor (%)
CN 2 2 1 4 2 2 4
R (nm) 0.2209 0.2266 0.2369 0.2457 0.2605 0.2613 0.2678
Bulk Ni2P 0.74
CN 1.8 1.2 3.9
R (nm) 0.2240 0.2348 0.2611
r2 (105 nm2) 1.89 0.26 7.52
Ni2P/SiO2 0.78
CN 1.7 2.2 2.9
R (nm) 0.2219 0.2349 0.2626
r2 (105 nm2) 0.17 1.24 6.71

CN: coordination number, R: bond distance, r2: DebyeWaller factor.


R-factor: residual-factor.
A. Cho et al. / Journal of Catalysis 318 (2014) 151161 155

70 sequence that considers adsorption steps is developed here based


(a) on a rake mechanism. A rake mechanism considers a sequence of
60 steps on a surface with each intermediate being capable of desorb-
ing [41]. The term rake is used because the written out mechanism
50 n-Pentane with the vertical adsorption/reaction steps and the horizontal sur-
Selectivity / %

n-Butane face steps resembles the prongs of the device used to gather leaves
40 (Fig. 6). In the present case, the simplest possible description of a
2-Pentanone
realistic reaction network was targeted, with each major observed
30 species assigned a corresponding surface intermediate. The two
species produced with the lowest selectivity (<0.5%), n-pentanal
20 and 2-pentene, were omitted in order to minimize the number of
tting variables and keep the simulation within proper statistical
10 bounds. Actually, a simulation with those two species was
attempted (Supplementary information), and although good ts
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35
to the data were obtained, a unique set of rate constants could
Contact time / s not be produced. The errors produced by omitting the two species
3.5 are likely to be small because of their very low levels.
(b) 2-Pentanol The treated sequence (Fig. 6) follows that considered before
3.0 1-Pentanol (Fig. 5) and begins with the adsorption of the reactant 2-MTHF.
Adsorption occurs via the oxygen atom, which activates the
2-Pentene
2.5 adjacent bonds, and makes them susceptible to reaction. This
Selectivity / %

Pentanal
occurs by ring-opening, and as before, there are two branches (this
2.0 time left and right), which, respectively, produce 2-pentoxide or
1-pentoxide intermediates. Further reactions are hydrogen-
1.5
transfer steps that produce alcohol precursors or bond-shift steps
that produce alkyl groups.
1.0
The experimental data used in the parameter estimation were
obtained under excess of hydrogen. So, hydrogen adsorption on
0.5
the surface was not considered.
0.0 In the diagram above, [A], [B], [D], [E], [G], and [I] (lmol cm3)
0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 are concentrations of 2-methyltetrahydrofuran, 2-pentanone, 2-
Contact time / s pentanol, n-pentane, 1-pentanol, and n-butane, respectively, in
the gaseous phase. [A*], [B*], [D*], [E*], [G*], and [H*] (lmol cm2)
Fig. 4. Effect of contact time on the product selectivity over the Ni2P/SiO2 catalyst:
are concentrations of the surface intermediates. The symbol [*] rep-
Contact time vs. selectivity at 573 K (300 C) and 0.5 MPa. (a) Majority products and
(b) minority products. resents the concentration of the empty adsorption sites on the cat-
alyst. The rate constants for the adsorption of each component are
denoted as kA, kB, kD, and kG (cm3 lmol1 s1). The rate constants
The 2-pentene selectivity shows a uniform rise with decreasing for desorption are denoted as k-A, k-B, k-D, k-E, k-G, k-H, and k-I
contact time, which would normally indicate that it is a primary (s1). The constants k1, k2, k4, k5, and k7 (s1) are the rate constants
intermediate. However, its very low level suggests that it is a prod- of the surface reactions. The kinetic rate law equations for the
uct of 2-pentanol dehydration, and that the lack of a maximum in major species are given below (Table 3).
selectivity is probably due to lack of data at very low contact time. The rates of reaction for the formation of the surface intermedi-
A limitation of the prior analysis is that the reaction network ates are given below (Table 4). These equations are solved simulta-
was modeled by a series of rst-order steps, without consideration neously with those of the gaseous species.
of adsorbed intermediates or the chemistry of each step. Although The rate constants obtained from the tting are summarized
such treatments are common, they can miss essential steps in the below (Table 5), and the calculated concentration curves are plot-
reaction pathway by not considering surface species [25]. To ted (Fig. 7). The points are the actual data presented earlier (Fig. 4).
improve the kinetic analysis, a more realistic description of the The kinetic ts for the observable gas-phase species show excellent

Fig. 5. Proposed reaction pathway for the 2-MTHF HDO at 0.5 MPa over Ni2P/SiO2 (rate constants are reported in previous work [33]).
156 A. Cho et al. / Journal of Catalysis 318 (2014) 151161

Fig. 6. Rake mechanism for the 2-MTHF HDO over the Ni2P/SiO2 and simplied nomenclature.

Table 3
Rate expressions for observable gaseous species.

dA dB
V
S ds
kA A kA A  V
S ds
kB B kB B 
dD  dE
V
S ds
kD D kD D  V
S ds
kE E  kH H 
dG dI
V
S ds
kG G kG G  V
S ds
kI G 
S
V 1:0  104 cm, where S is the total surface area of Ni2P and V is the volume of the catalyst bed

Table 4
Rate expressions for the surface intermediates.

dA  dB 
ds
kA A  kA k1 k2 A  ds
kB B k1 A   kB k4 B 
dD    dE 
ds
kB D k4 B   kD k5 D  ds
k5 D   kE E 
dG  dH 
ds
kG G k2 A   kG kI k7 G  ds
k7 G   kH H 
The total concentration of the adsorption sites, L is given as follows:
 
mol g1 104 cm2
CO uptake
L Effectiv e surface area of Ni2 P
5542lm 2 g1 m2
1:3  104 lmol cm2
Effectiv e surface area of Ni2 P q6Dc q = 7.09 g cm3 Dc = average crystallite size
L = [*] + [A*] + [B*] + [D*] + [E*] + [G*] + [H*]

agreement with the experimental data. The conversion of 2-MTHF However, it should be noted that the tting method used here
can be calculated from the decrease in 2-MTHF concentration. It is involving the optimization and solution of differential algebraic
linear with contact time and is 3.1% at s = 0.32 s. This also indicates equations performed in a two tier approach cannot guarantee that
that product inhibition is not occurring, as if this were the case the the obtained solution is a global minimum, though it should be
conversion would be proportionally lower at higher contact time. close.
The standard error estimates for each species (Table 6) indicate The reaction scheme together with the rate constants is sum-
low values. These are consistent with the good ts of the data. marized below (Fig. 8) in order to ease visualization of the results.
A. Cho et al. / Journal of Catalysis 318 (2014) 151161 157

Table 5
Estimated kinetic parameters from experimental data.

Rate constant (s1) Rate constant (s1) Rate constant (cm3 lmol1 s1)
k-A 0.04 k1 46 kA 1.4
k-B 11,491 k2 82 kB 10
k-D 2810 k4 11,564 kD 15
k-E 12,111 k5 11,067 kG 16
k-G 88 k7 1157
k-H 814
k-I 452

10 metallic sites. The desorption rate constants are shown to the right
2-MTHF of the vertical arrows and show an interesting trend. The constants
Concentration (mol/cm3)

n-Pentane are substantially larger for the three intermediates to the right of
n-Butane
adsorbed 2-MTHF than for the two intermediates to the left. The
1 intermediates on the right are formed from secondary carbons
and those on the left from terminal primary carbons, and differ-
2-Pentanone ences in reactivity are expected, perhaps by stabilization of transi-
2-Pentanol tion states with partial carbonium ion character on the carbons.
0.1
The analysis also allows calculation of concentrations of the
suggested surface intermediates. Plots of these surface concentra-
1-Pentanol tions show that they do not depend greatly on contact time
0.01 (Fig. 9). The most abundant intermediate is adsorbed 2-MTHF,
the reactant. This is reasonable, as its further reaction by ring open-
0.0 0.1 0.2 0.3 ing is rate-determining for both branches.
Contact time, (s) Quantitative comparisons with experimental quantities are
possible. From Fig. 9, the average concentration of adsorbed
Fig. 7. Comparison of experimental and estimated concentrations of 2-MTHF, n- 2-MTHF (A*), which is the major surface species, can be calculated
pentane, n-butane, 2-pentanone, 2-pentanol, and 1-pentanol at 573 K (300 C) and
0.5 MPa.
to be 1.3  105 lmol/cm2 for s = 0.21 s. To put this value in per-
spective, recall that the total surface area of the sample was
214 m2/g and the CO uptake of the supported Ni2P phase was
55 lmol/g (Table 1). This gives a site density of n = (55 lmol/g)/
Table 6
Estimation result: standard error estimates. (214 m2/g)(104 cm2/m2) = 2.6  105 lmol/cm2, which, as should
be, is higher than the concentration of A* calculated above. In terms
Measured variable Standard error of estimates
of coverages this is the equivalent of h = 0.50. In a previous study
[A] 6.85E02 [31] of the kinetics of the reaction at the same temperature
[B] 9.62E03
(573 K) and pressure (0.5 MPa), a LangmuirHinshelwood equation
[D] 2.28E04
[E] 2.19E02 was obtained which allowed calculation of the adsorbed 2-MTHF
[G] 9.13E04 coverage. The calculated value at s = 0.21 s was h = 0.40, which
[I] 5.27E03 agrees well with the h = 0.50 determination here by a completely
independent method. Since the sites are based on the chemisorp-
tion of CO, these calculated coverages are on the Ni2P surface
and do not count the support.
The vertical reactions represent adsorption and desorption steps, The second most abundant species are the two species with pri-
while the horizontal reactions represent surface transformation mary carbons, G* and H* (protonated 1-alkoxide and 1-alkyl), in the
steps. The sequence begins with the adsorption of the reactant left branch of the reaction scheme, whose concentrations are about
2-MTHF (indicated). Once adsorbed the 2-MTHF can undergo ring a tenth of that of the adsorbed 2-MTHF. The least abundant species
opening in two manners, on the unsubstituted side (left sequence) are the three species with secondary carbons, D*, B*, and E* (proton-
or on the methyl side (right sequence). These are indicated in the ated 2-alkoxide, 2-alkoxide, and 2-alkyl), in the right branch of the
horizontal direction (Fig. 8). The ratio of rate constants for ring reaction scheme, whose concentration are about one hundredth of
opening on the unsubstituted side to the methyl substituted side that of the adsorbed 2-MTHF. These results are completely consis-
is 82 s1/46 s1 = 1.8. This agrees with a previous determination tent with the magnitude of the rate constants discussed earlier
using a simple rst-order sequence of steps [33] of 2.2. It should (Fig. 8), with the species with the smallest desorption constants
be noted that the higher value of the rate constant for the forma- on the left branch having the larger coverages. The species on the
tion of adsorbed 1-pentoxide (82 s1) compared to the lower value right branch are more reactive as they desorb more rapidly, and
of the rate constant for the formation of adsorbed 2-pentoxide their coverages are smaller. This lower surface concentration is in
(46 s1) is not inconsistent with the observation of 2-pentanone accord with expectations for highly reactive species.
as a major gas-phase intermediate, since the rate constant for its An important consideration is whether the adsorbed species
desorption (11,491 s1) is extremely high compared to those of can be observed and distinguished by a technique such as infrared
the products from the left branch. spectroscopy. At rst glance, all the suggested adsorbed intermedi-
The adsorption rate constants are shown in the vertical direc- ates appear to be similar. They all are alkyl or alkoxide species with
tion in red to the left of the arrows. These constants follow the similar structural units, so look impossible to differentiate. How-
order: alcohols > ketone > ether, which is in the order of electron ever, close examination of the structures reveals a distinguishing
donating ability of the functional groups, and probably indicates feature, which is the number of methylene groups (CH2). The
that there are interactions of these groups with electron-accepting adsorbed reactant A* has three CH2 groups, the two species to the
158 A. Cho et al. / Journal of Catalysis 318 (2014) 151161

Fig. 8. Summary of the rake mechanism and kinetic parameter estimations. The units of the rate constants kA, kB, kD, and kG are cm3 lmol1 s1 and those for the surface
reactions k-A, k-B, k-D, k-E, k-G, k-H, k-I, k1, k2, k4, k5, and k7 are s1.

high wavenumbers, 3737 and 3661 cm1, which are due to SiOH
and POH, respectively [34,42]. They appear negative because the
10
-5
[A*] hydroxyl groups are consumed by interaction with the 2-MTHF.
1.3 x 10 -5 Both sets of spectra also show aliphatic CH stretching bands,
Concentration (mol/cm2)

29702870 cm1 [43], and CH bending and scissoring bands,


-6 14601380 cm1 [44], that are due to adsorbed 2-MTHF and/or
10 [H*] CH3
O
some reaction intermediates. A peak at around 1700 cm1 can be
[G*] +
* assigned to adsorbed 2-MTHF and/or 2-methylfuran. A very small
feature at 2483 cm1 is due to PH vibrations. A doublet band at
-7
10
[D*] 2365/2330 cm1 is due to atmospheric CO2 outside the in situ cell.
[B*] A band at 2035 cm1 is due to linearly adsorbed CO on Ni2P, which
[E*] appears at slightly lower wavenumber than in other reports
-8
10 [45,46], possibly because of the high dispersion of the present sam-
ple. In the IR spectra in N2 of 2-MTHF adsorbed on the Ni2P/SiO2
0.0 0.1 0.2 0.3
(Fig. 10a), all of signals decrease and negative peaks are recovered
Contact time, (s) as temperature increases. This is noteworthy for the case for the
Fig. 9. Estimated concentrations of the reaction intermediates. OH stretches at high wavenumber, which go to the baseline level
already at 453 K, and indicate that desorption from the support
is complete at that temperature. Thus, the overall decrease in peak
left have four CH2 groups, while the three species to the right each intensity is due to desorption of 2-MTHF from rst the support and
have two CH2 groups (Table 7). Thus, it should be possible to dis- then the Ni2P.
tinguish between the species by focusing on the CH2 vibration in The IR bands in H2 of 2-MTHF adsorbed on the Ni2P/SiO2
the infrared spectrum. This was duly found to be the case. (Fig. 10b) show a similar tendency to decrease with increasing
Infrared spectra of the adsorbed species were obtained in an in temperature as those taken in N2. However, it can be readily seen
situ cell (Fig. 1) at two sets of conditions, in a N2 ow stream and in that although the intensities begin at a comparable level, in H2 the
a H2 ow stream. First, results for adsorption of 2-MTHF as temper- adsorbate signals do not decrease as rapidly with rising tempera-
ature was progressively raised are reported (Fig. 10). ture as in N2, but rather tend to persist. The retention of the band
The spectra at the adsorbed 2-MTHF in the inert N2 and reactive intensities in the presence of H2 can be explained by the transfor-
H2 atmospheres display similar features, but also important differ- mation of adsorbed 2-MTHF to adsorbed ring-opened products
ences which will be discussed momentarily. First, covering similar- (Fig. 8) by hydrogen-transfer steps. Thus, instead of desorbing,
ities, the spectra at the two conditions show two negative peaks at the adsorbed 2-MTHF reacts with hydrogen to form other adsorbed

Table 7
Differences between adsorbed species.

Left branch products Reactant Right branch products


H* G* A* D* B* E*
4 CH2 4 CH2 3 CH2 2 CH2 2 CH2 2 CH2
A. Cho et al. / Journal of Catalysis 318 (2014) 151161 159

2970
(a)Under N2 flow
2934 0.05
2876
1385
3737 3661 1420
2365/2330 1461
Absorbance

448 K 2483 2035


473 K
498 K
523 K
548 K
573 K
598 K

4000 3500 3000 2500 2000 1500


Wavenumbers (cm-1)

2964
(b) Under H 2 flow 2936 0.05
2878
1384
3738 2365/2327 1419
1460
Absorbance

3666
2483 2041 1570 Fig. 11. In situ FTIR spectra of 2-MTHF adsorbed on the Ni2P/SiO2 catalyst at various
448 K 1695 temperatures: 33002750 cm1 (a) under N2 ow and (b) under H2 ow.
473 K
498 K
523 K There is another signicant change observed when the carrier
548 K gas is changed from inert N2 to reactive H2. A sharp strong feature
573 K 2060 scarcely present under N2 appears at 2035 cm1, which is due to
598 K linearly adsorbed CO on Ni2P. There are also smaller intensity
bands in the range 18201920 cm1 due to bridging CO species.
4000 3500 3000 2500 2000 1500 The formation of CO is due to a decarbonylation reaction of a ter-
Wavenumbers (cm-1) minal aldehyde. At steady state this results in equal amounts of
CO and n-butane, which is observed experimentally [31]. The rea-
Fig. 10. In situ FTIR spectra of 2-MTHF adsorbed on the Ni2P/SiO2 catalyst at various
son the decarbonylation occurs in the presence of H2 is that the
temperatures. (a) Under N2 ow and (b) under H2 ow.

(a) Under N2 flow


species which populate the surface before themselves desorbing.
The question is which species are on the surface? 2970-2876

Absorbance
0.1
The answer can be obtained by examining an expanded region of
the CH stretching region (Fig. 11). There are three bands which
under H2 appear at 2964, 2936, and 2878 cm1 (Fig. 11b). Similar
2035
bands appear under N2 (Fig. 11a). These can be assigned predomi-
nantly to the stretching frequencies of the CH bonds in CH3, CH2,
and CH groups [47,48]. In actuality some overlap occurs, e.g. due
to symmetric and asymmetric band contributions. The notable 2
4
in
aspect of the spectra under H2 is that the central band at
/m

6
2936 cm1 due mainly to CH2 vibrations is visibly stronger than 8
e
m

10
Ti

under N2. Its intensity is almost equal to that of the rst band at 4000 3500 3000 2500 2000 1500
-1
2964 cm1 due to CH3 vibrations. This is a clear indication that Wavenumbers / cm
the species formed by hydrogenation of the adsorbed 2-MTHF have
more CH2 bonds than in N2 and points to the majority surface spe- (b)Under H2 flow
cies after 2-MTHF being those on the left branch (Table 6). This is 2041
Absorbance

exactly as predicted in the tting calculations (Fig. 9) which indi- 2964-2878 0.1
cated that the terminal alkoxy and alkyl species on the left branch
are more numerous than the secondary species on the right branch.
The spectra under N2 can be integrated to give an estimate of
the heat of adsorption of 2-MTHF. A linear vant Hoff plot was
obtained (Supplementary information), which gave a heat of
adsorption of 12.0 kJ/mol. 2
The main ndings obtained by raising the temperature are 4
in

6
/m

duplicated in transient spectra at constant temperature (Fig. 12). 8


e
m

The peaks close to the high 2900 cm1 range are due to the CH 10
Ti

4000 3500 3000 2500 2000 1500


stretches of adsorbed species, and it can be seen that in N2 the high -1
Wavenumbers / cm
wavenumber band due to CH3 groups is dominant, while in H2, a
double peak caused by CH3 and CH2 contributions is the main Fig. 12. Transient spectra of 2-MTHF on Ni2P at 523 K (250 C). (a) Desorption in N2
feature. and (b) reaction in H2.
160 A. Cho et al. / Journal of Catalysis 318 (2014) 151161

Fig. 13. Possible steps in the transformation of 2-MTHF. (a) Ring-opening and (b) decarbonylation.

precursor aldehyde needs to be formed by a ring-opening reaction ten-fold more abundant than those with secondary carbon
(Fig. 13a). A possible mechanism of reaction for this portion of the substitution.
network indicates that a hydride is also necessary for the decarb- 6. Measurements of adsorbed species at reaction conditions
onylation (Fig. 13b). with Fourier transform infrared spectroscopy conrm that
To summarize, this work investigated the hydrodeoxygenation the species with primary carbon substitution are more
(HDO) pathway of the biomass model compound 2-meth- plentiful on the surface.
yltetrahydrofuran (2-MTHF) over a Ni2P/SiO2 catalyst at 0.5 MPa
and 573 K (300 C). The results were analyzed using a rake mech-
anism where a sequence of reactions occur on the surface of the Acknowledgments
catalyst and each intermediate species can desorb. The 2-MTHF
reactant is interesting because in the course of reaction it can This work was supported by the US Department of Energy, Ofce
undergo ring-opening in two manners, by cleavage of the ring of Basic Energy Sciences, through Grant DE-FG02-963414669,
CO bond on the more hindered side to give intermediates with the National Science Foundation through Grant CBET-144316,
oxygen on a primary carbon or cleavage of the ring CO bond on and the Japan Ministry of Agriculture, Forestry, and Fisheries
the more open side to give intermediates with oxygen on a second- (Norinsuisansho).
ary carbon. The TEM pictures in this work were taken in Research Hub for
The studies here conrm previous work that showed that both Advanced Nano Characterization, The University of Tokyo,
pathways are possible. However, the present studies go consider- supported by the Ministry of Education, Culture, Sports, Science
ably further, by considering surface intermediates and providing and Technology (MEXT), Japan.
evidence for the rst time that the ring-opening pathway that
results in bonding through a primary carbon results in intermedi- Appendix A. Supplementary material
ates at high levels of concentration. This is conrmed by in situ FTIR
measurements that show that the intermediates bonded through Supplementary data associated with this article can be found, in
primary carbons are more plentiful than those bonded through sec- the online version, at http://dx.doi.org/10.1016/j.jcat.2014.07.021.
ondary carbons. This gives a complete and consistent picture of the
reaction pathways in the HDO of 2-methyltetrahydrofuran. References

4. Conclusions [1] M. Ni, J. Sen, E. Smeets, E. Stehfest, D.P. Van Vuuren, Glob. Change Biol.
Bioenergy 4 (2012) 130147.
[2] J. Zakzeski, P.C.A. Bruijnincx, A.L. Jongerius, B.M. Weckhuysen, Chem. Rev. 110
The main ndings of this study are the following: (2010) 35523599.
[3] J.Y. He, C. Zhao, J.A. Lercher, J. Am. Chem. Soc. 134 (2012) 2076820775.
1. The hydrodeoxygenation of 2-methyl tetrahydrofuran, a [4] M.A. Gonzalez-Borja, D.E. Resasco, Energy Fuels 25 (2011) 41554162.
[5] W.D. Huang, Y.-H.P. Zhang, PLoS One 6 (2011) e22113, http://dx.doi.org/
cyclic saturated ether compound, over a Ni2P/SiO2 catalyst 10.1371/journal.pone.0022113.
at 0.5 MPa proceeds through a reaction network in which [6] J.C. Serrano-Ruiz, J.A. Dumesic, Energy Environ. Sci. 4 (2011) 83.
ring-opening can occur on either the side with the methyl [7] P. Brjesson, Appl. Energy 86 (2009) 589.
[8] E.J. Steen, Y. Kang, G. Bokinsky, Z. Hu, A. Schirmer, et al., Nature 463 (2010)
group or on the unfunctionalized side. 559.
2. Ring-opening on the side with the methyl group is slightly [9] C.N. Hamelinck, A.P.C. Faaij, Energy Policy 34 (2006) 3268.
less favored and produces intermediates with secondary [10] M.J.A. Tijmensen, A.P.C. Faaij, C.N. Hamelinck, M.R.M. van Hardeveld, Biomass
Bioenergy 23 (2002) 129.
carbon substitution such as 2-pentanone and 2-pentanol [11] A.V. Bridgwater, Biomass Bioenergy 38 (2012) 68.
before formation of the nal product n-pentane. [12] A. Dermibas, Energy Convers. Manage. 42 (2001) 1357.
3. Ring-opening on the unsubstituted side is more favored [13] M.M. Wright, D.E. Daugaard, J.A. Satrio, R.C. Brown, Fuel 89 (2010) S2S10.
[14] O.D. Mante, F. Agblevor, S.T. Oyama, R. McClung, Appl. Catal. A: Gen. 445446
and produces intermediates with primary carbon substitu- (2012) 312320.
tion such as n-pentanal and 1-pentanol before formation of [15] Ofei D. Mante, Foster A. Agblevor, S.T. Oyama, Fuel 117 (2014) 649659.
the same nal product n-pentane. [16] B. Yoosuk, D. Tumnantong, P. Prasassarakich, Chem. Eng. Sci. 79 (2012) 17.
[17] O.I. Senol, E.-M. Ryymin, T.-R. Viljavam, A.O.I. Krause, J. Mol. Catal. A 277
4. The n-pentanal can undergo a decarbonylation step to pro-
(2007) 107.
duce CO which adsorbs on the surface. [18] Sara Boullosa-Eiras, R. Ldeng, H. Bergem, M. Stcker, L. Hannevold, E.A.
5. Kinetic modeling of the reaction network with a rake mech- Blekkan, Catal. Today (2013).
anism that considers plausible adsorbed intermediates [19] X. Duan, Y. Teng, A. Wang, V.M. Kogan, X. Li, Y. Wang, J. Catal. 261 (2009) 232.
[20] V.M.L. Whiffen, K.J. Smith, Energy Fuels 24 (2010) 4728.
gives excellent ts to contact time data, and predicts that [21] A.-M. Alexander, J.S.J. Hargreaves, Chem. Soc. Rev. 39 (2010) 4388.
the species formed with primary carbon substitution are [22] K.L. Li, R.J. Wang, J.X. Chen, Energy Fuels 25 (2011) 854.
A. Cho et al. / Journal of Catalysis 318 (2014) 151161 161

[23] D.J. Rensel, S. Rouvimov, M.E. Gin, J.C. Hicks, J. Catal. 305 (2013) 256. [37] W.H. Press, S.A. Teukolsky, W.T. Vetterling, B.P. Flannery, Numerical Recipes in
[24] A. Cho, J. Shin, A. Takagaki, R. Kikuchi, S.T. Oyama, Top. Catal. 55 (2012) 969. FORTRAN. The Art of Scientic Computing, second ed., Cambridge University
[25] R. Prins, M.E. Bussell, Catal. Lett. 142 (2012) 1413. Press, 1992.
[26] S.T. Oyama, T. Gott, H. Zhao, Y.-K. Lee, Catal. Today 143 (2009) 94. [38] D. Bazin, D. Sayers, J.J. Rehr, J. Phys. Chem. B 101 (1997) 11040.
[27] H.Y. Zhao, D. Li, P. Bui, S.T. Oyama, Appl. Catal. A: Gen. 391 (2011) 305. [39] S.T. Oyama, X. Wang, Y.-K. Lee, W.-J. Chun, J. Catal. 221 (2004) 263.
[28] D. Li, P. Bui, H. Zhao, S.T. Oyama, T. Dou, Z.H. Shen, J. Catal. 290 (2012) 1. [40] Y.-K. Lee, Y. Shu, S.T. Oyama, Appl. Catal. A: Gen. 322 (2007) 191.
[29] J.-S. Moon, E.-G. Kim, Y.-K. Lee, J. Catal. 311 (2014) 144. [41] F.X. Cormerais, G. Perot, F. Chevalier, M. Guisnet, J. Chem. Res. S (1980) 362363.
[30] J.A. Cecilia, A. Infantes-Molina, E. Rodrguez-Castelln, A. Jimnez-Lpez, S.T. [42] S.T. Oyama, Y.-K. Lee, J. Phys. Chem. B 109 (2005) 2109.
Oyama, Appl. Catal. B: Environ. 136137 (2013) 140. [43] K.A. Layman, M.E. Bussell, J. Phys. Chem. B 108 (2004) 15791.
[31] V. Whiffen, K. Smith, Top. Catal. 55 (2012) 981. [44] T.M. El-Gogary, Spectrochim. Acta, Part A 57 (2001) 1405.
[32] P. Bui, J.A. Cecilia, S.T. Oyama, A. Takagaki, A. Infantes-Molina, H. Zhao, D. Li, E. [45] K.A. Layman, M.E. Bussell, J. Phys. Chem. B 108 (2004) 10930.
Rodrguez-Castelln, A. Jimnez Lpez, J. Catal. 294 (2012) 184. [46] T.I. Kornyi, E. Pfeifer, J. Mihly, K. Fttinger, J. Phys. Chem. A 112 (2008) 5126.
[33] A. Iino, A. Cho, A. Takagaki, R. Kikuchi, S.T. Oyama, J. Catal. 311 (2014) 17. [47] G. Socrates, Infrared Characteristic Group Frequencies, Wiley, New York, 1994.
[34] S.T. Oyama, Y.-K. Lee, J. Catal. 258 (2008) 393. [48] D. Lin-Vien, N.B. Colthup, W.G. Fateley, J.G. Grasselli, The Handbook of Infrared
[35] A.L. Ankudinov, B. Ravel, J.J. Rehr, S.D. Conradson, Phys. Rev. B 58 (1998) 7565. and Raman Characteristic Frequencies of Organic Molecules, Academic Press,
[36] Y.-K. Lee, S.T. Oyama, J. Catal. 239 (2006) 376. Boston, 1991. pp. 155178.

Das könnte Ihnen auch gefallen