Support effects in the hydrogenation of cinnamaldehyde over carbon nanofiber-supported platinum catalysts: Characterization and catalysis

Article abstract of DOI:10.1016/j.jcat.2004.05.026

Carbon nanofiber-supported platinum catalysts were studied to establish the influence of the support surface composition on their performance in the liquid-phase hydrogenation of cinnamaldehyde. By acid-base titration, XPS, and TGA-MS, the number of oxygen-containing surface groups was determined in oxidized carbon nanofiber samples and in platinum catalysts pretreated at 473-973 K. TEM, EXAFS, and H₂-chemisorption studies showed the presence of a narrow and stable platinum particle-size distribution (1-2 nm). The overall catalytic activity increased by a factor of 25 after treatment at 973 K, which is tentatively related to the decreasing amount of oxygen of the carbon fibers. With XPS and H₂-chemisorption experiments carried out at 308 K no clear evidence was found for a change in the electronic structure of the platinum particles induced by the oxygen-containing groups present on/in the surface of the carbon nanofibers. A linear decrease of the hydrogenation activity with an increase of the number of acidic groups on the carbon nanofiber support was found. Therefore, it is suggested that hydrogenation is assisted by adsorption of cinnamaldehyde on the carbon support after removal of the oxygen-containing surface groups.

Full text of DOI:10.1016/j.jcat.2004.05.026

Journal of Catalysis 226 (2004) 215–225
www.elsevier.com/locate/jcat
Support effects in the hydrogenation of cinnamaldehyde
over carbon nanofiber-supported platinum catalysts:
characterization and catalysis
a
a
b
a
a
Marjolein L. Toebes , Yihua Zhang , Jan Hájek , T. Alexander Nijhuis , Johannes H. Bitter ,
a
b
a
a,∗
A. Jos van Dillen , Dmitry Yu. Murzin , Diek C. Koningsberger , Krijn P. de Jong
a
Department of Inorganic Chemistry and Catalysis, Debye Institute, Utrecht University, P.O. Box 80 083, 3508 TB Utrecht, Netherlands
b
Laboratory of Industrial Chemistry, Process Chemistry Group, Åbo Akademi University, Turku/Åbo, Finland
Received 30 January 2004; revised 6 May 2004; accepted 22 May 2004
Abstract
Carbon nanofiber-supported platinum catalysts were studied to establish the influence of the support surface composition on their perfor-
mance in the liquid-phase hydrogenation of cinnamaldehyde. By acid–base titration, XPS, and TGA-MS, the number of oxygen-containing
surface groups was determined in oxidized carbon nanofiber samples and in platinum catalysts pretreated at 473–973 K. TEM, EXAFS,
and H -chemisorption studies showed the presence of a narrow and stable platinum particle-size distribution (1–2 nm). The overall catalytic
2
activity increased by a factor of 25 after treatment at 973 K, which is tentatively related to the decreasing amount of oxygen of the carbon
fibers. With XPS and H -chemisorption experiments carried out at 308 K no clear evidence was found for a change in the electronic structure
2
of the platinum particles induced by the oxygen-containing groups present on/in the surface of the carbon nanofibers. A linear decrease of the
hydrogenation activity with an increase of the number of acidic groups on the carbon nanofiber support was found. Therefore, it is suggested
that hydrogenation is assisted by adsorption of cinnamaldehyde on the carbon support after removal of the oxygen-containing surface groups.

2004 Elsevier Inc. All rights reserved.
Keywords: Carbon nanofibers; CNF; Platinum catalyst; Oxygen-containing surface groups; Cinnamaldehyde hydrogenation; Metal–support interaction
1
. Introduction
nium catalysts gave rise to selectivities of 20–30% [6]. With
activated carbon-supported ruthenium catalysts selectivities
upto 30–40% have been obtained [13]. Often the increased
selectivity toward the unsaturated alcohol is explained in
terms of transfer of π-electrons from the graphitic planes
to the metal particles [7,11,14]. In this way the charge den-
The selective hydrogenation of α, β-unsaturated aldehy-
des to unsaturated alcohols is a key step especially in the
preparation of various fine chemicals [1–4]. In the literature
often noble metals supported on titania and graphitic carbon
are used either with or without a promoter, to obtain a high
alcohol selectivity [1,3,5–9]. Recently several publications
have appeared in which carbon nanofibers (CNF) or multi-
walled carbon nanotubes (MWNT) are examined as catalyst
support [10]. For this type of reactions promising results
were found. For example, in the liquid-phase hydrogena-
tion of cinnamaldehyde with MWNT-supported ruthenium
catalysts [7,11,12] selectivities to cinnamyl alcohol up to
∗
sity on the metal increases, enhancing π backbonding and
CO
thus increasing the probability for the C=O bond hydro-
genation.
Earlier we reported on the influence of oxygen-containing
surface groups in the liquid-phase hydrogenation of cin-
namaldehydeusing CNF-supported ruthenium catalysts [15].
In that study we used well-defined Ru/CNF catalysts with
highly dispersed and uniform ruthenium particles of 1–
9
2% have been observed, whereas alumina-supported ruthe-
2
nm. A clear trend was observed in activity and se-
lectivity with the number of oxygen-containing surface
groups. The rate of cinnamaldehyde conversion was en-
hanced by a factor of 22 after removal of the major-
*
Corresponding author. Fax: +31-30-2511027.
E-mail address: k.p.dejong@chem.uu.nl (K.P. de Jong).
0021-9517/$ – see front matter  2004 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2004.05.026

216
M.L. Toebes et al. / Journal of Catalysis 226 (2004) 215–225
ity of the surface oxygenates which was mainly due to
a strong increase in the hydrogenation rate of the C=C
bond.
Table 1
Sample codes, loadings, and pretreatment conditions of the various CNF-
supported platinum catalysts
Sample
Loading (wt%)
XRF
Gas-phase
prereduction
Heat treatment
A possible explanation for a direct effect of the oxygen-
containing surface groups can be found in the adsorption
of the organic reactant [16–18]. After increasing treatment
temperatures, the CNF surface gradually changes from po-
lar to nonpolar due to the removal of the oxygen-containing
surface groups. This change in polarity might change the
preferential adsorption mode of cinnamaldehyde and en-
able adsorption of the reactant on the CNF support, thereby
influencing the activity and selectivity. An alternative ex-
planation implies the indirect influence of the oxygen-
containing groups on catalysis via the electronic state of
the metal. This model is based on the work of Konings-
berger and co-workers and concerns the interaction between
small (< 2 nm) metal particles and oxidic supports [19–21].
With decreasing electron richness of the support oxygen
atoms, the density of states (DOS) and the Fermi level of
the metal particles shift to a higher binding energy. This ef-
fect has been observed with XPS, FTIR, and XAFS. For
carbon-supported noble metal catalysts this model suggests
that lower amounts of (more electronegative) oxygen in the
carbon support would give rise to a change in the electronic
structure of the metal.
Our previous work on Ru/CNF [15] has not been con-
clusive whether electronic or adsorption effects are predom-
inant in steering catalysis. A fundamental study of the elec-
tronic properties of the ruthenium particles was hampered
by among others the overlap of the carbon and ruthenium
peaks in XPS. For platinum more precise XPS measure-
ments can be carried out while H2 chemisorption can be
related to earlier results. The study of metal–support inter-
action with oxidic supports has largely been performed with
platinum catalysts [19–21].
in N
2
a
PtCNFnp
PtCNF
PtCNF573
PtCNF773
PtCNF973
3.6
3.6
3.6
3.6
3.6
–
–
–
473 K
473 K
473 K
473 K
573 K
773 K
973 K
a
np, not prereduced in the gas-phase.
2. Experimental
2.1. Synthesis carbon nanofiber-supported platinum
catalyst
Carbon nanofibers were grown from H2/CO over a
Ni/SiO2 catalyst as described in [15].
Platinum (intake 5 wt%) was deposited on the fibers ac-
cording to a HDP method. To an acidified suspension (pH 3)
of 5 g CNF in 250 ml demiwater heated up to 363 K
under inert atmosphere, 0.41 g urea (Acros) and 0.52 g
Pt(NH3)4(NO3)2 (Aldrich) were added under vigorous stir-
ring. The pH of the slurry was monitored, to follow the
process and to indicate its completeness. After 18 h the
loaded CNF were filtered and washed thoroughly with demi-
water, dried at 353 K in an N2 flow, and reduced in flowing
H2 at 473 K for 1 h (heating rate = 5 K/min). Next the sam-
ples were exposed to air at RT.
In order to obtain different concentrations of oxygen-
containing surface groups on the CNF, samples of the freshly
reduced catalyst were heat-treated in an N2 flow for 2 h at
573, 773, and 973 K [23]. The catalyst samples together with
their identification codes are listed in Table 1.
2.2. Catalyst characterization
To gain more insight in the nature of the distinct influence
of oxygen-containing surface groups, here we report on the
influence of these groups in CNF-supported platinum cata-
lysts in the liquid-phase hydrogenation of cinnamaldehyde.
This investigation is performed on well-defined Pt/CNF cat-
alysts with metal particles of 1–2 nm with a narrow particle-
size distribution and a tunable amount of support surface
oxygen. A well-defined catalyst with small and uniform
metal particles is needed to enable observation of the elec-
tronic support effect at all [22]. In this paper we discuss
the characteristics of the CNF and the type and number of
oxygen-containing surface groups present after activation
and heat treatment in N2. With XPS and H2-chemisorption
experiments we investigate whether the catalytic perfor-
mance of the Pt/CNF system is related to an electronic effect,
similar to that described for oxide-supported catalysts. The
influence of the oxygen groups on the adsorption of the re-
actant on the CNF support is dealt with in the accompanying
paper.
The numbers of acid sites of the oxidized CNF after the
various heat treatments were determined by standard acid–
base titrations as described in [23]. The total number of ther-
mally removable oxygen-containing groups was determined
by TGA [23].
The CNF-supported platinum catalysts were examined in
a Philips CM-200 FEG TEM and a Tecnai 20 FEG TEM
both operated at 200 kV. After suspending in ethanol under
ultrasonic vibration, samples were brought onto a holey car-
bon film on a copper grid.
Platinum loadings were determined by means of XRF
on a Philips PW 1480. Before analysis the Pt/CNF sam-
ples were pressed into a tablet using a methyl methacrylate
binder.
XPS analyses were performed on a Fisons ESCALAB
2
10I-XL and a Vacuum Generators XPS system as described
earlier [23].
Hydrogen-chemisorption measurements were performed
using a Micromeritics ASAP 2010C. Each sample was dried

M.L. Toebes et al. / Journal of Catalysis 226 (2004) 215–225
217
in He at 393 K for 1 h, and reduced in flowing H2 (flow rate
50 ml/min STP) at 473 K for 2 h, heating rate 5 K/min.
Data for phase shifts and backscattering amplitudes for
Pt–Pt and Pt–C were obtained from FEFF7 and FEFF8 cal-
culations, respectively, and calibrated by the experimental
data of Pt foil and [Rb Pt(CN) ] · 1.5H O. Data analysis
=
−1
After reduction the samples were degassed for 2 h at 10 Pa
at the reduction temperature in order to remove chemisorbed
hydrogen and water. The isotherms were measured at 308 K.
The presented H/Pt ratios are based on the amounts ad-
sorbed at zero pressure found by extrapolation of the linear
part of the isotherm. Calculations have been made with the
total amount of adsorbed hydrogen, because XAFS results of
Oudenhuijzen et al. with Pt/Al2O3 catalysts unambiguously
show that also “weakly” bonded hydrogen is chemisorbed
and must be taken into account [24]. Estimated average par-
ticle sizes and dispersions are based on spherical geometry
and an adsorption stoichiometry of H/Pts = 1. The average
Pt particle size, d, was calculated from
2
4
2
of the CNF-supported platinum catalysts was performed by
multiple shell fitting using the difference file technique in R
1
space (1.6 < R < 3.2 Å) with the XDAP code using both k
and k weighting [27,28].
3
The variances of the magnitude and imaginary part of the
Fourier transforms of fit and data were calculated according
to
ꢀ
n
n
2
[
FT
(R) − FT (R)] dR
exp
model
exp
Variance =
ꢀ
× 100
n
2
[
FT (R)] dR
6Mρsite
1
2
n
n
dD = 10
,
with FT = FT[k χ(k)]. In this study the statistical signif-
ρmetalN
icance of a contribution has been checked by a comparison
where d is the average platinum particle size (nm), D is
the dispersion (H/Pt), M is the atomic weight (Pt =
of the amplitude of (Fit) with the noise level present in the
difference file.
j
1
95.09 g/mol), ρsite is the platinum surface site density
2
3
(
12.5 Pt atoms/nm ), ρmetal the metal density (21.45 g/cm ),
2
3
−1
2
.4. Catalytic experiments
and N the Avogadro constant (6.022 × 10 mol ) giving
d = 1.13/D (nm) [25].
Liquid-phase hydrogenationof cinnamaldehydewas stud-
2
.3. XAFS experiments
ied in a 500-ml stirred semibatch reactor equipped with a
sample port, reagent injection port, gas inlet, and vent.
The catalysts (0.2 g) were (re)-activated in suspension
in i-propanol (220 ml) at 383 K with H2 at a total pres-
sure of 3.0 MPa (hydrogen partial pressure 2.8 MPa) for
XAFS spectra at the Pt L3 edge were taken at the HA-
SYLAB (Hamburg, Germany) synchrotron beamline X1.1,
equipped with a Si(311) double-crystal monochromator. The
monochromator was detuned to 50% of the maximum inten-
sity to avoid higher harmonics present in the X-ray beam.
The measurements were done in transmission mode using
ion chambers filled with an Ar/N2 mixture to have an µx
of 20% in the first and a µx of 80% in the second ioniza-
tion chamber. The energy of the X-ray beam was calibrated
by using a Pt foil between the second and third ionization
chamber.
3
0 min in the stirred autoclave (1500 rpm) prior to the in-
troduction of cinnamaldehyde (3.0 g). All reactions were
executed at 383 K using a total pressure of 5.0 MPa (hy-
drogen partial pressure 4.8 MPa). Samples were withdrawn
periodically and analyzed on a gas chromatograph (Hewlett–
Packard 5890 Series with autosampler) using an HP5 capil-
lary column (30 m × 0.32 mm × 0.25 µm i.d.).
After the hydrogenation experiments the changes in plat-
inum particle size of the spent catalysts were investigated
with TEM. For PtCNF to PtCNF973 no significant modifica-
tion in size and shape of the platinum particles was observed.
The platinum particles in PtCNFnp, however, had sintered to
The powdered catalysts were diluted with boron nitride
and pressed into a self-supporting wafer (calculated to have
an absorbance of 2.5) mounted in a stainless-steel in situ cell
equipped with Be windows. The samples were re-reduced
in situ at 473 K for 30 min (5 K/min) in flowing H2 and
cooled down in an H2 atmosphere. The reduced samples are
considered to be fully covered with hydrogen and are further
denoted by (RED). Next the EXAFS data for the L3 edge
were collected (T = 77 K).
2
–10 nm during the reaction under the rather severe condi-
tions (4.8 MPa H2 and 383 K).
To investigate whether the catalysts became irreversibly
deactivated, three consecutive experiments with one catalyst
batch of PtCNF were executed. Between the runs the cata-
lyst was washed repeatedly with isopropanol and dried. The
activity of the catalysts in the three runs was comparable,
thus proving the absence of irreversible deactivation.
Duplicate experiments were performed with both PtCNF
and PtCNF973 showing rates and selectivities to the various
products to be within an error margin of around 5%, indicat-
ing excellent reproducibility.
Extraction of the EXAFS data from the measured ab-
sorption spectra was performed with the XDAP code [26].
Three scans were averaged and the preedge was subtracted
using a modified Victoreen curve. The background was sub-
tracted employing cubic spline routines with a continuously
adjustable smooth parameter. Normalization was performed
by dividing the data by the edge step at 50 eV after the ad-
sorption edge.

218
M.L. Toebes et al. / Journal of Catalysis 226 (2004) 215–225
Table 2
Number of oxygen atoms/nm on CNF surface as a function of treatment temperature determined with acid–base titration, XPS, and TGA-MS
2
Sample
Titration
XPS
TGA-MS
Number of
O atoms/
O/C atomic
ratio
O atoms/
Weight loss
(%)
O atoms/
2
2
2
2
acid site/nm
nm
nm
nm
CNFox
1.4 ± 0.1
1.0 ± 0.1
2.8 ± 0.2
0.069
0.038
0.032
0.017
9.6 ± 1.4
5.3 ± 0.8
4.4 ± 0.7
2.3 ± 0.3
5.7
3.7
3.3
1.8
9.8 ± 0.8
6.3 ± 0.5
5.6 ± 0.5
3.0 ± 0.3
CNFox, 573 K in N
CNFox, 773 K in N
CNFox, 973 K in N
2.0 ± 0.2
2
2
2
0.2 ± 0.02
0.4 ± 0.04
0.06 ± 0.01
∼ 0.03 ± 0.003
PtCNFnp
PtCNF
PtCNF573
PtCNF773
PtCNF973
0.044
0.038
0.037
0.026
0.016
6.1 ± 0.9
5.3 ± 0.8
5.1 ± 0.8
3.6 ± 0.5
2.2 ± 0.3
5.1
3.8
3.7
2.7
1.5
7.7 ± 0.6
6.5 ± 0.5
6.2 ± 0.5
4.5 ± 0.5
2.6 ± 0.3
2
3
. Results and discussion
.1. Characterization of the CNF support
The CNF are of the fishbone type, i.e., with the graphite
up to 1–3 groups/nm , as earlier proposed by Boehm and
as we found by titration, seem to be more realistic [30].
An explanation for this discrepancy can be found in the as-
sumption that oxygen atoms are not exclusively present at
the surface, but also are built in the subsurface (2–3 nm)
graphene sheets [31,32]. TGA-MS up to 1123 K was used to
establish the number of thermally removable oxygen atoms
from the surface and the bulk of the CNF. Also TGA-MS re-
sults show the presence of oxygen-containing groups in the
oxidized CNF and the partial removal of these groups with
increasing heat treatment. The number of oxygen atoms as
determined using XPS and TGA-MS is comparable when
the error margins are taken into account. From these results
it is concluded that oxygen is not present throughout the car-
bon support, but is mainly located in the outer 2–3 nm of
the fibers with 2–30% present in acidic groups at the exte-
rior surface, i.e., accessible for titration. It is noted that with
increasing treatment temperature the ratio of (acidic oxygen
at exterior)/(total oxygen) drops largely. For a more detailed
investigation on the effect of oxidation and subsequent ther-
mal treatment of CNF we refer to [23].
3
planes oriented at an angle to the central axis and the fibers
interwoven into porous bodies. After removal of the growth
catalyst and the activation treatment with nitric acid for 2 h
the nonmicroporous fibers exhibited a specific surface area
2
of 177 m /g. X-ray diffraction and TEM images demon-
strated that the graphite-like structure of the CNF had not
been affected during activation. From SEM and TEM micro-
graphs an average fiber diameter of 25 nm could be derived
with a narrow diameter distribution.
As stated earlier the activation procedure in nitric acid is
carried out to introduce various types of oxygen functional
groups on the CNF and to remove nonencapsulated nickel
from the growth catalyst. TEM images show that the remain-
ing nickel (ꢀ 0.5 wt%, XRF analysis) is encapsulated by
graphitic envelopes, thus preventing interference in cataly-
sis.
Table 2 lists the number of acid sites (pKa < 7.5) of
oxidized CNF treated at different temperatures in N2 as
determined by titration, as well as the number of oxygen
3.2. Characterization of the CNF-supported platinum
catalysts
2
atoms/nm calculated from XPS and TGA-MS results. Af-
ter activation and drying of the CNF 1.4 acid sites/nm2
The metal loading of the Pt/CNF catalysts as established
with XRF was 3.6 wt% (Table 1). TPR results (not shown)
of the freshly loaded sample demonstrated that all catalysts
were completely reduced at 473 K. Therefore all catalysts
were reduced at this temperature. PtCNFnp was only re-
duced in situ at 383 K preceding the catalytic experiments,
to maintain a support oxygen concentration as large as pos-
sible.
were present. Heat treatment resulted in a gradual removal
of the majority of the acidic oxygen-containing groups to
2
0
.03 acid sites/nm after treatment at 973 K. XPS mea-
surements were done to establish the amount of oxygen in
the outer layer of 2–3 nm of the CNF. After oxidation and
drying an O/C atomic ratio of 0.069 was found. A grad-
ual decrease in the O/C ratio from 0.069 to 0.017 with
increasing heat-treatment temperature was observed. Using
a model the O/C ratios can be converted into a number of
The amount of support oxygen still present on the plat-
inum catalysts after the different treatments is assessed using
XPS and TGA-MS (Table 2). Unfortunately, titration exper-
iments could not be performed, since the presence of plat-
inum interfered with the measurements. This table shows
that due to the application of platinum on the CNF the aver-
2
oxygen atoms/nm [29]. These results (Table 2) show that
2
fibers oxidized in HNO3 for 2 h contain 9.6 O atoms/nm .
Since oxygen-containing surface groups often contain one
or two oxygen atoms, this implies that about 5–10 oxygen-
containing surface groups would be present when it is as-
sumed that all oxygen is located at the outer surface. Values
2
age number of oxygen atoms/nm has decreased from about
9.7 (CNFox) to about 6.9 (PtCNFnp). A more detailed inves-

M.L. Toebes et al. / Journal of Catalysis 226 (2004) 215–225
219
Fig. 1. TEM images of CNF-supported platinum catalysts after reduction (PtCNF).
Fig. 2. TEM images of CNF-supported platinum catalysts after reduction and heat treatment in N (A) PtCNF573, (B) PtCNF773, and (C and D) PtCNF973.
2
tigation on the relation between the oxygen groups and the
metal loading can be found elsewhere [33]. Subsequent re-
duction and heat treatments resulted in a further gradual de-
Table 3
TEM and hydrogen chemisorption results at 308 K on the different CNF-
supported platinum catalysts
Sample
TEM
d (nm)
H chemisorption
2
crease in the amount of support oxygen to ultimately around
2
Irreversible
Total
Hirr/Htot
2
–3 oxygen atoms/nm for PtCNF973, very similar to the
H/Pt
d (nm)
H/Pt
d (nm)
values observed with unloaded CNF.
PtCNF
1–2
1–2
1–2
1–3
0.37
0.63
0.33
0.21
3.1
1.8
3.5
5.4
0.79
1.06
0.74
0.49
1.4
1.1
1.5
2.3
0.45
0.60
0.42
0.43
In Fig. 1 two representative TEM images of PtCNF are
shown. The platinum particles are small and homogeneously
distributed over the CNF surface. The particle-size distri-
bution is very narrow; particles of 1–2 nm are observed.
The average particle size and H/Pt ratio calculated from
hydrogen-chemisorption data are presented in Table 3. Us-
ing the total amount of adsorbed hydrogen a H/Pt of 0.79 an
average platinum particle size of 1.4 nm was calculated for
PtCNF.
PtCNF573
PtCNF773
PtCNF973
After reduction the platinum catalysts were treated in N2
at 573, 773, and 973 K to remove part of the support oxygen-
containing groups. From the TEM images (Figs. 2A–2D) it
is clear that treatment at 573 and 773 K did not affect the
platinum particle size, while treatment at 973 K resulted in a

220
M.L. Toebes et al. / Journal of Catalysis 226 (2004) 215–225
Fig. 3. (A) The EXAFS raw data (k weighted, ꢀk = 3–14 Å) of PtCNF (RED). (B) The corresponding Fourier transforms (k1 weighted, ꢀk = 3–14 Å) of
1
the EXAFS raw data (solid line) and the total fit (dotted line).
small increase in particle size. In Fig. 2D the graphene sheets
of the fishbone CNF covered with some platinum particles
are visible. Hydrogen-chemisorption results (Table 3) show
an unexpected and up to now unexplained increase in H/Pt
from 0.79 for PtCNF to 1.06 for PtCNF573. After treatment
at 773 and 973 K H/Pt ratios of 0.74 and 0.49 were found,
indicating average particle sizes of 1.5 and 2.3 nm, respec-
tively, thereby confirming TEM results.
Table 4
−
1
Fit parameters of EXAFS spectra (ꢀk, 3–14 Å ; ꢀR, 1.6–3.2 Å) and
variances for model spectra of PtCNF (RED) and PtCNF773 (RED)
2
2
1
k
Catalyst
treatment)
Scatter N
R (Å)
ꢀσ (Å ) ꢀE
variance
(%)
10% Im. Abs.
0
(
± 10% ± 0.02 Å ± 5%
(eV)
±
−
−
−
3
3
3
PtCNF
RED)
C(O)s 0.7
C(O) 1.6
2.02
2.65
2.77
2.0×10
1.2
0.39 0.14
(
2.2×10
1.1×10
−0.3
l
Pt
5.2
2.1
EXAFS was used as an additional method to determine
the structural parameters of the platinum particles in PtCNF
and PtCNF773. Fig. 3A shows the experimental EXAFS
−
−
−
3
3
3
PtCNF773 C(O) 0.3
(RED)
s
2.01
2.62
2.76
1.3×10
1.4×10
1.6×10
3.2
1.5
1.3
0.34 0.12
C(O)l 2.0
Pt 5.5
1
−1
data (k weighted, ꢀk = 3–14 Å ) of PtCNF (RED) in
a hydrogen atmosphere. The signal to noise ratio at k =
−
1
4
.4 Å amounts to approximately 150 on both the PtCNF
tively, which is typical for metallic platinum particles in a
hydrogen atmosphere [34]. The Pt–Pt coordination numbers
found are 5.2 (PtCNF) and 5.5 (PtCNF773). This shows
that in both samples very small Pt particles with an av-
erage size of around 1 nm are present [35,36]. The other
parameters of PtCNF and PtCNF773 differ only slightly
or not at all. A full EXAFS study dealing with the struc-
ture of the metal–support interface in the Pt/CNF cata-
lysts, the shortening of the Pt–C distance upon evacua-
tion, and the role of oxygen-containing groups in stabiliz-
ing the platinum particles is published in a separate pa-
per [37].
and the PtCNF773 spectra, showing the data to have an ex-
cellent quality.
The corresponding k1 Fourier transform of the raw
EXAFS data of PtCNF (RED) is displayed in Fig. 3B. This
is a typical example of the Fourier transform of small sup-
ported platinum particles using k1 weighting. It contains
both Pt–Pt and Pt–C(O) contributions. In this study, the
separation of Pt–C and Pt–O contributions is not carried
out, because it is difficult to identify O and C as different
backscatterers since they are neighboring elements in the
Periodic Table. The sum of Pt–C and Pt–O contributions
is denoted as Pt–C(O). The parameters used to fit the ex-
perimental EXAFS data of PtCNF (RED) and PtCNF773
3
.3. Hydrogenation of cinnamaldehyde
(
RED) are given in Table 4. As an example Fig. 3B en-
ables the comparison of the Fourier transforms of the raw
EXAFS data (solid line) and the fit (dotted line) of PtCNF
The hydrogenation of cinnamaldehyde can be presented
by a simplified reaction pathway as shown in Fig. 4. In
this study the main by-products before complete conversion
of cinnamaldehyde were cinnamyl isopropylether and the
di-isopropylacetal of hydrocinnamaldehyde. These products
form via a reaction between cinnamaldehyde or hydrocin-
namaldehyde and the solvent isopropanol.
For PtCNF and PtCNF973 Fig. 5 shows, as a function
of time, the molar composition of the reaction mixture for
cinnamaldehyde (CALD), the primary reaction products hy-
drocinnamaldehyde(HALD) and cinnamyl alcohol (CALC),
(
RED) using the parameters listed in Table 4. This compar-
ison demonstrates the excellent fit quality of all the EXAFS
data. The C(O) and Pt backscatterers were identified using
the difference file technique. The variances in imaginary and
absolute parts were used to determine the fit quality. Fits
with variances below 1% have been considered to represent
good models for the experimental data.
The EXAFS results in Table 4 show a Pt–Pt bond dis-
tance in PtCNF and PtCNF773 of 2.77 and 2.76 Å, respec-

M.L. Toebes et al. / Journal of Catalysis 226 (2004) 215–225
221
pletely converted within 20 min. The difference in activity
will be dealt with in more detail below.
It is clear that the selectivities [selectivity Y(%) = (moles
Y(t)/moles CALD(0)) × 100%] to the main products are
dependent on the catalyst. PtCNF produces cinnamyl alco-
hol and hydrocinnamaldehyde in comparable amounts. On
the other hand, with PtCNF973 hydrocinnamaldehyde is the
main product. In Fig. 6 the selectivities to cinnamyl alcohol
(
CALC) and hydrocinnamaldehyde (HALD) are plotted as
a function of the conversion [conversion (%) = (1 − (moles
CALD(t)/moles CALD(0)))× 100%] and in Table 5 the se-
lectivities at 60% conversion to the various reaction products
and the two main byproducts are represented. For PtCNFnp
SCALC = 58% and SHALD = 22% is found. With increas-
ing treatment temperature the selectivity toward the alcohol
gradually declines and the selectivity to hydrocinnamalde-
hyde rises. For PtCNF973 SCALC = 8%, while SHALD =
74%. Obviously the rate of C=C hydrogenation of cin-
namaldehyde is strongly enhanced with heat treatment of
the catalyst, while the rate of hydrogenation of the C=O
has only slightly increased. Fig. 5 demonstrates, however,
that the secondary hydrogenation of hydrocinnamaldehyde
Fig. 4. Reaction pathway of the hydrogenation of cinnamaldehyde.
and the secondary product hydrocinnamyl alcohol (HALC).
First of all, from these graphs the large difference in hydro-
genation activity between PtCNF and PtCNF973 is strik-
ing. Using PtCNF after 600 min only 80% conversion is
obtained, while using PtCNF973 cinnamaldehyde is com-
Fig. 5. Cinnamaldehyde conversion and product distribution as function of time on stream obtained at 383 K and 4.8 MPa hydrogen over CNF-supported
platinum catalysts (A) PtCNF and (B) PtCNF973.
Fig. 6. Selectivities of different CNF-supported Pt catalysts as a function of conversion to (A) cinnamyl alcohol and (B) hydrocinnamaldehyde.

222
M.L. Toebes et al. / Journal of Catalysis 226 (2004) 215–225
Table 5
Table 6
Selectivity to cinnamyl alcohol, hydrocinnamaldehyde, hydrocinnamyl al-
cohol, and two by-products for cinnamaldehyde hydrogenation with CNF-
supported platinum catalysts determined at 60% conversion
Initial activities, relative activities, and TOF for cinnamaldehyde hydro-
genation with CNF-supported platinum catalysts at 383 K and a hydrogen
pressure of 4.8 MPa
a
Sample
name
S
S
S
HALC
By-products
Sample
name
Initial activity
(mmol/(min gcat))
Relative
TOF
CALC
HALD
(mol%) (mol%) (mol%)
b
−1
Cinnamyl
isopropyl-
ether
Di-isopropylace-
tal of hydrocin-
namaldehyde
activity
(s
)
c
PtCNFnp
PtCNF
PtCNF573
PtCNF773
PtCNF973
0.60
1.9
4.0
11
1.0
3.2
6.7
19
n.d.
0.23
0.39
1.6
PtCNFnp 58
PtCNF 37
PtCNF573 13
PtCNF773 13
22
32
23
61
74
13
15
12
17
18
1
2
11
1
1
6
19
2
15
25
3.3
a
Results are given in mole CALD hydrogenated per mole of platinum
surface atoms in the catalyst per second. The amount of platinum surface
atoms is calculated from the catalyst weight, the metal loading, and the
H/Pt (total H) ratio.
PtCNF973
8
0
0
b
Relative activity is the initial activity divided by the initial activity of
to hydrocinnamyl alcohol is also much faster on PtCNF973;
i.e., the C=O hydrogenation rate seems dependent on the
substrate molecule.
PtCNFnp.
c
n.d., not determined because the H/Pt ratio is not known for this sam-
ple.
Fig. 6 shows that the selectivities to the primary and sec-
ondary products depend on the conversion of cinnamalde-
hyde. Beyond 5% conversion the selectivity of the heat-
treated catalysts remains fairly constant. Another striking
feature is the strong dependence of the amount of formed
by-products with the treatment of the catalyst (Table 5):
PtCNFnp, PtCNF773, and PtCNF973 produce only minor
amounts of the isopropanol ether and acetal, while PtCNF
Table 7
Platinum 4f
and 4f
binding energies of the CNF-supported plat-
7/2
5/2
inum catalysts after reduction at 383 K determined by XPS using Gauss–
Lorenzian peak fits
Sample
name
E (eV)
b
Pt 4f
Pt 4f
5/2
7/2
PtCNFnp
PtCNF
PtCNF573
PtCNF773
PtCNF973
73.4
71.4
71.4
71.4
71.6
76.2
74.6
74.7
74.7
74.6
(
2 and 6%) and especially PtCNF573 (11 and 19%) form
large amounts of these products. It is well known that these
type of reactions are acid-catalyzed [38]. Therefore, it is ex-
pected that with increasing heat treatment of the catalyst
and thus with decreasing support acidity the amount of by-
products drops. However, the low amount of by-products
formed using PtCNFnp is unexpected.
apparent. In contrast, with PtCNF973 the increase in activ-
ity is such that diffusion of the organic compounds in the
catalyst bodies limits the reaction rate [39]. From the latter
calculation the ratio of the intrinsic activity of PtCNF and
PtCNF973 is much larger than the factor of 8 in observed
activitie.
Although the data show many interesting phenomena, in
this paper we mainly focus on the effect of the oxygen-
containing groups on the strongly enhanced activity for
C=C hydrogenation in cinnamaldehyde. From the decrease
in the concentration of cinnamaldehyde with time (Fig. 5)
initial activities were calculated for all samples. These val-
ues are given in Table 6 together with the relative activ-
ity compared to the activity of PtCNFnp and the related
TOF. Both Fig. 5 and Table 6 show the clear differences
in catalytic behavior as a function of treatment tempera-
ture of the catalysts. With increasing treatment temperature
a strong increase in the total activity was found. The ob-
served activity of PtCNF973 was almost 25 times larger
than that of PtCNFnp and almost 8 times larger than that
of PtCNF. The TOFs differ even somewhat more due to
a small increase in mean particle size in PtCNF973. This
increase in total activity is mainly caused by a large in-
crease in hydrocinnamaldehyde formation as can be con-
cluded from Fig. 5.
3.4. Electronic structure of platinum
The hydrogenation results presented here using the Pt/
CNF catalysts and the results reported in an earlier study
on RuCNF catalysts [15] show that the catalytic behav-
ior of CNF-supported noble metals is strongly dependent
on the heat treatment and thus on the amount of oxygen
groups present in/at the support surface. In order to inves-
tigate whether the oxygen-containing groups influence the
electronic properties of the platinum particles, we executed
XPS and H -chemisorption measurements. Using XPS a
shift in platinum binding energy is expected [21,40] and
2
with hydrogen chemisorption the Hstrong/Htotal ratio should
change if a large electronic effect of the support is effec-
tive [41].
In Table 7 the platinum 4f7/2 and 4f5/2 binding energies
are displayed as determined for the CNF-supported platinum
catalysts after reduction using Gauss–Lorenzian peak fits.
For metallic platinum, binding energies of 70.9 (4f7/2) and
Calculations of the Weisz–Prater criterion of PtCNF and
PtCNF973 and calculated concentration profiles in the cata-
lyst particles of these catalysts using a kinetic model, which
includes mass transfer inside the catalyst particles, demon-
strate that with PtCNF internal diffusion limitations are not

M.L. Toebes et al. / Journal of Catalysis 226 (2004) 215–225
223
Fig. 7. Relative total activity for cinnamaldehyde hydrogenation as a function of the amount of acidic groups (A) and as a function of the amount of oxygen
XPS) on the CNF-supported platinum catalysts (B).
(
7
4.3 eV (4f5/2) have been reported [42]. The results show
3.5. Oxygen content of the support
that in PtCNF, PtCNF573, PtCNF773, and PtCNF973 the
platinum binding energy is equal for all samples and around
In Fig. 7A the relative total activity is plotted versus the
amount of acidic oxygen-containing groups/nm of the var-
2
7
1.4 and 74.6 eV, respectively, indicating that there is a shift
ious CNF supports. Because titrimetric determination of the
acidic groups on the Pt/CNF catalysts could not be effec-
tuated due to the interfering action of platinum, we used
the values found with the unloaded CNF support (see Ta-
ble 2). An almost linear relation, especially for the PtCNFnp
to PtCNF773 catalysts, is observed between the total activ-
ities and the concentration of acidic oxygen groups of the
support. When only few acidic groups are present the activ-
ity is high and with increasing surface acidity the activities
drop. The line between PtCNF773 and PtCNF973 is dotted,
since the activity measured using PtCNF973 is influenced by
internal diffusion limitation. Besides the acidic groups also
other oxygen-containing surface groups are present on the
CNF surface. Therefore, in Fig. 7B the relative total activ-
ities are plotted versus the number of oxygen atoms. For
the number of oxygen atoms the values found with XPS
are used. The correlation between activity and oxygen con-
centration is not as clear as between the activity and the
number of acidic support surface groups, since the num-
ber of the acidic oxygen-containing groups is not linearly
proportional to the total amount of oxygen on/in the fibers.
These curves suggest that catalysis is directly influenced by
the acidic oxygen-containing groups at the exterior surface
of the CNF support. The strong increase in activity could
be brought about by the removal of these support surface
groups. A model in which the hydrogenation is assisted by
adsorption of the cinnamaldehyde on the CNF support after
removal of the oxygen surface groups is suggested. A full ki-
netic analysis of a series of cinnamaldehyde hydrogenation
experiments underpins the effects of substrate adsorption on
the support on catalysis [39].
of 0.3–0.5 eV compared to the binding energies of bulk plat-
inum, probably due to the size of the metallic particles. The
binding energy of PtCNFnp is slightly shifted to higher val-
ues, as ionic Pt species are present because of the mild re-
duction conditions.
With oxidic supports change of the electron richness
of the support oxygen atoms was found to influence the
metal binding energies of noble metal particles, i.e., higher
binding energies when applied on supports with a rela-
tively low oxygen electron density. Mojet et al., for in-
stance, found a shift in binding energies of about 1.4 eV
for both Pd 3d3/2 and Pd 3d5/3 in a Pd/LTL with a K/Al
atomic ratio ranging from 0.55 to 1.10 [40]. Our results,
however, indicate that the varying concentrations of oxy-
gen groups on the carbon support do not give rise to these
large shifts in platinum binding energies as determined
with XPS.
In Table 3 the Hirr/Htotal ratios of the various Pt/CNF
catalysts are given. Oudenhuijzen [41] observed that for
small LTL-supported platinum particles the hydrogen cov-
erage and platinum–hydrogen binding strength depend on
the support properties. For the various Pt/CNF catalysts
no clear trend can be observed in the Hstrong/Htotal ra-
tios.
Since we did not observe changes in the platinum bind-
ing energy and H2 chemisorption we think that changes
in electron density of the CNF support by partly remov-
ing the oxygen groups are much smaller than the changes
in the electron density in oxidic-supported catalysts with,
for example, varying K/Al ratios. These small changes in
electronic properties are not revealed with XPS and H2
chemisorption. However, in the accompanying paper differ-
ences are apparent in the order in hydrogen for the hydro-
genation of cinnamaldehyde. Therefore, minor changes in
the electronic properties of the platinum particles that are
still important for catalysis cannot be excluded.
4. Conclusions
Well-defined carbon nanofiber-supported platinum cata-
lysts were prepared with a narrow size distribution, and a

224
M.L. Toebes et al. / Journal of Catalysis 226 (2004) 215–225
mean particle size of 1–2 nm, both features being prerequi-
sites for performing studies on the electronic support effect.
By treatment of carbon nanofibers in nitric acid and subse-
quent loading, reduction, and heat treatment under nitrogen
atmosphere, the concentration of oxygen-containing groups
on the carbon nanofiber surface was varied without a signif-
icant change in platinum particle size.
[6] B. Coq, P.S. Kumbhar, C. Moreau, P. Moreau, M.G. Warawdekar,
J. Mol. Catal. 85 (1993) 215.
[
7] P.S. Kumbhar, B. Coq, C. Moreau, P. Moreau, J.M. Planeix, M.G.
Warawdekar, in: N.M. Gupta, D.K. Chakrabarty (Eds.), Catalysis:
Modern Trends, Narosa, New Delhi, 1995.
[8] A.B. da Silva, E. Jordoa, M.J. Mendes, P. Fouilloux, Appl. Catal.
A 148 (1996) 253.
[9] C.M.G. van de Moesdijk and M.A.R. Bosma, US 4745234 (1988).
The activity and selectivity in the liquid-phase hydro-
genation of cinnamaldehyde was found to be dependent on
the amount of oxygen of the CNF support. The rate of cin-
namaldehyde conversion was enhanced with a factor of 25
after removal of the oxygen groups. The differences in in-
trinsic activities are even much larger, since with the catalyst
treated at high temperatures internal diffusion limits the re-
action rate. This enhanced activity is mainly caused by a
strong increase in the hydrogenation rate of the C=C bond,
while only a slight increase in C=O bond hydrogenation is
observed.
[
10] P. Serp, M. Corrias, P. Kalck, Appl. Catal. A 253 (2003) 337.
[
11] J.M. Planeix, N. Coustel, B. Coq, V. Brotons, P.S. Kumbhar, R. Du-
tartre, P. Geneste, P. Bernier, P.M. Ajayan, J. Am. Chem. Soc. 116
(1994) 7935.
[12] M. Lashdaf, A. Hase, E. Kauppinen, A.O.I. Krause, Catal. Lett. 52
(1998) 199.
[13] S. Galvagno, G. Capanelli, G. Neri, A. Donato, R. Pietropaolo, J. Mol.
Catal. 64 (1991) 237.
[14] T. Vergunst, F. Kapteijn, J.A. Moulijn, Catal. Today 66 (2001) 381.
[15] M.L. Toebes, F.F. Prinsloo, J.H. Bitter, A.J. van Dillen, K.P. de Jong,
J. Catal. 214 (2003) 78.
[16] R. Burch, A.R. Flambard, J. Catal. 78 (1982) 389.
Using XPS and H2 chemisorption up to now no evidence
has been found that the amount of oxygen in the support
influences to a large extent the electronic properties of the
platinum particles. Apparently modification of the electronic
properties of the metal is not the prime cause of the large
changes in catalytic behavior of the catalysts. A linear cor-
relation between the increasing observed hydrogenation ac-
tivity and the decreasing number of acidic groups on the
carbon nanofiber support was found. Therefore, a model in
which the hydrogenation is assisted by adsorption of the
cinnamaldehyde on the CNF support after removal of the
oxygen surface groups is suggested.
[
17] M. Haruta, S. Tsubota, T. Kobayashi, M.J. Genet, B. Delmon, J. Ca-
tal. 144 (1993) 175.
[18] J.H. Bitter, K. Seshan, J.A. Lercher, J. Catal. 171 (1997) 279.
[19] D.E. Ramaker, J. de Graaf, J.A.R. van Veen, D.C. Koningsberger,
J. Catal. 203 (2001) 7.
[20] D.C. Koningsberger, M.K. Oudenhuijzen, J. de Graaf, J.A. van Bok-
hoven, D.E. Ramaker, J. Catal. 216 (2003) 178.
[21] B.L. Mojet, J.T. Miller, D.E. Ramaker, D.C. Koningsberger,
J. Catal. 186 (1999) 373.
[22] A.Yu. Stakheev, L.M. Kustov, Appl. Catal. A 188 (1999) 3.
[23] M.L. Toebes, J.M.P. van Heeswijk, J.H. Bitter, A.J. van Dillen, K.P. de
Jong, Carbon 42 (2004) 307.
[
[
[
[
[
24] M.K. Oudenhuijzen, J.H. Bitter, D.C. Koningsberger, J. Phys. Chem.
B 105 (2001) 4616.
25] J.J.F. Scholten, A.P. Pijpers, A.M.L. Hustings, Catal. Rev.-Sci. Eng. 27
Acknowledgments
(
1985) 151.
26] M. Vaarkamp, J.C. Linders, D.C. Koningsberger, Phys. B 208–209
1995) 159.
The authors are grateful to C. van der Spek and J.W. Geus
for TEM, A. Broersma for the TGA-MS experiments, and
A.J.M. Mens for XPS. We also thank Shell International
B.V. for the XPS measurements. Prof. Dr. M.R. Egmond
and R. Dijkman of the Department Membrane Enzymol-
ogy and Dr. J.M.P. van Heeswijk are acknowledged for their
help with the acid–base titrations. We also thank the HASY-
LAB (Hamburg, Germany) for the opportunity to perform
XAFS measurements at beamline station X1.1. These inves-
tigations have been supported by the Council for Chemical
Sciences of the Netherlands Organization for Scientific Re-
search with financial aid from the Netherlands Technology
Foundation (CW/STW 349-5357).
(
27] D.C. Koningsberger, B.L. Mojet, G.E. van Dorssen, D.E. Ramaker,
Top. Catal. 10 (2000) 143.
28] D.C. Koningsberger, J. de Graaf, B.L. Mojet, D.E. Ramaker, J.T. Mil-
ler, Appl. Catal. A 191 (2000) 205.
[
[
[
29] O.L.J. Gijzeman, unpublished results.
30] H.P. Boehm, Adv. Catal. 16 (1966) 179.
31] T.G. Ros, A.J. van Dillen, J.W. Geus, D.C. Koningsberger, Chem.
Eur. J. 8 (2002) 1151.
[32] H. He, J. Klinowski, M. Fortser, A. Lerf, Chem. Phys. Lett. 287 (1998)
53.
[33] M.L. Toebes, L.M. Tang, M.H. Huis, van’t Veld, J.H. Bitter, A.J. van
Dillen, K.P. de Jong, J. Phys. Chem. B, in press.
[
[
[
[
34] M.K. Oudenhuijzen, J.H. Bitter, D.C. Koningsberger, J. Phys. Chem.
B 105 (2001) 4616.
35] B.J. Kip, F.B.M. Duivenvoorden, D.C. Koningsberger, R. Prins,
J. Catal. 105 (1987) 26.
References
36] J. de Graaf, A.J. van Dillen, K.P. de Jong, D.C. Koningsberger,
J. Catal. 203 (2001) 307.
[
[
[
[
[
1] P. Kluson, L. Cerveny, Appl. Catal. A 128 (1995) 13.
2] V. Ponec, Appl. Catal. A 149 (1997) 27.
3] B. Coq, F. Figueras, Coord. Chem. Rev. 178–180 (1998) 1753.
4] P. Gallezot, D. Richard, Catal. Rev.-Sci. Eng. 40 (1998) 81.
5] A. Giroir-Fendler, D. Richard, P. Gallezot, Stud. Surf. Sci. Catal. 41
37] Y. Zhang, M.L. Toebes, W.E. O’Grady, A. van der Eerden, K.P. de
Jong, D.C. Koningsberger, J. Phys. Chem. B, submitted.
[38] R.T. Morrison, R.N. Boyd, Organic Chemistry, Prentice-Hall, NJ,
1992.
(1988) 171.

M.L. Toebes et al. / Journal of Catalysis 226 (2004) 215–225
225
[39] M.L. Toebes, T.A. Nijhuis, J. Hájek, J.H. Bitter, A.J. van Dillen, D.Y.
[41] M.K. Oudenhuijzen, Ph.D. thesis, Utrecht University, 2002.
Murzin, K.P. de Jong, part II, submitted.
40] B.L. Mojet, M.J. Kappers, J.C. Muijsers, J.W. Niemantsverdriet, J.T.
Miller, F.S. Modica, D.C. Koningsberger, Stud. Surf. Sci. Catal. 84
[
42] C.D. Wagner, W.M. Riggs, L.E. Davis, J.F. Moulder, G.E. Muilen-
berg (Eds.), Handbook of X-Ray Photoelectron Spectroscopy, Perkin–
Elmer Corporation, Eden Prairie, MN, 1978.
[
(1994) 909.