Preparation and activity of small rhodium metal particles on fishbone carbon nanofibres

Article abstract of DOI:10.1016/S0021-9517(02)93695-X

Very small Rh metal particles were deposited on HNO₃/H₂SO₄-oxidized fishbone carbon nanofibers (CNF) using different impregnation and ion-exchange techniques. Carbon nanofibers of the fishbone type were prepared by catalytic decomposition of CH₄ on a Ni/Al₂O₃ catalyst. The surface oxidation with HNO₃/H₂SO₄ was a prerequisite for a good interaction between aqueous catalyst precursor solutions and the CNF. Particle sizes of 1.1-2.1 nm were obtained with 1 wt % Rh metal loadings. The smallest Rh particles were the result of impregnation, subsequent calcination at 150°C, and reduction at 140°C or ion exchange at elevated temperature with an aqueous solution of RhCl₃·2H₂O and reduction at 200°C. All catalysts were quite active in the hydrogenation of cyclohexene. The metal particle size did not have a major impact on the catalytic activity, indicating that other variables, such as the clustering of support bodies in the liquid phase, were more critical.

Full text of DOI:10.1016/S0021-9517(02)93695-X

Journal of Catalysis 211, 85–102 (2002)
doi:10.1006/jcat.2002.3695
Preparation and Activity of Small Rhodium Metal Particles
on Fishbone Carbon Nanofibres
T. G. Ros, D. E. Keller, A. J. van Dillen, J. W. Geus, and D. C. Koningsberger1
Department of Inorganic Chemistry and Catalysis, Debye Institute, Utrecht University, P.O. Box 80083, 3508 TB Utrecht, The Netherlands
Received February 7, 2002; revised May 23, 2002; accepted May 23, 2002
dure, however, is too expensive to envisage thus-produced
carbon nanotubes as a support on a commercial scale. Car-
bon nanofibres, on the other hand, are a promising catalyst
support material. The perspective is that when grown in a
fluidised bed reactor, carbon nanofibres can be obtained
at low cost, making commercial application possible (1).
The properties of CNFs make them especially suitable for
liquid-phase reactions. The fibres are mechanically strong
and can withstand the forces involved in vigorous agitation
of the reaction medium. Furthermore, the skeins of fibres
possess a mesoporous macrostructure, which decreases
the possibility of encountering diffusion limitations during
catalytic reactions in the liquid phase with respect to micro-
porous materials, such as activated carbon. The structure of
the CNFs can be tuned by changing the growth conditions
and their hydrophobicity can be controlled by surface ox-
idation. The fibres are chemically inert and can be used in
strongly acidic or basic liquids. Finally, after deactivation
of the catalyst, CNFs can be readiliy combusted to recover
the precious metal component of spent catalysts without
leaving ash residues.
The use of carbon nanofibres and nanotubes as catalyst
support has been reviewed recently (1). A strong incentive
for their use is the replacement of activated carbon (AC) for
liquid-phase reactions. The properties of ACs are difficult
to control. Unsatisfactory reproducibility as well as the mi-
croporosity of activated carbon has often impeded catalyst
development. The sometimes inadequate mechanical prop-
erties of AC may lead to filtration problems in liquid-phase
applications.
Hoogenraad and co-workers (2–6) extensively studied
the use of Pd/CNF for the hydrogenation of nitrobenzene
to aniline. The authors prepared fibres with the graphite
layers parallel to the fibre axis (parallel fibres) and fibres
with the graphite layers at an angle with the fibre axis (fish-
bone fibres). The smallest Pd particles were obtained when
fibres that were oxidized with HNO3 were ion exchanged
with a positively charged Pd–ammonia complex under the
exclusion or air. With parallel fibres a maximal loading of
3wt%wasobtained, whilewithfishbonefibresonly1.5wt%
Pd could be exchanged (3). On parallel fibres Pd particles
of about 1.5 nm were obtained (6). The Pd/CNF catalyst
A number of different impregnation and ion-exchange proce-
dures have been employed to synthesize very small rhodium metal
particlesonHNO3/H2SO4-oxidizedfishbonecarbonnanofibres. The
surface-oxidation of the nanofibres with HNO3/H2SO4 is a prereq-
uisite for a good interaction between aqueous catalyst precursor
solutions and the fibres. Depending upon the preparation tech-
nique applied and using 1 wt% rhodium metal loadings average
particle sizes ranging from 1.1 to 2.1 nm were detected with XAFS
spectroscopy. The rhodium metal particles are so small that metal–
support interactions on carbon nanofibres can be investigated with
XAFS spectroscopy. All catalysts are highly active in the liquid-
phasehydrogenationofcyclohexene. Nosignificanteffectofparticle
size on the catalytic activity is observed, suggesting that other fac-
tors, such as clustering of the support particles in the liquid phase,
c
are much more important.
ꢀ 2002 Elsevier Science (USA)
Key Words: carbon nanofibres; EXAFS spectroscopy; heteroge-
neous catalysis; hydrogenation; rhodium metal particles.
1. INTRODUCTION
Currently, there is much interest in the physical and
chemical properties of carbon nanotubes and carbon
nanofibres (CNFs). Whereas carbon nanotubes are usually
synthesized by arc discharge techniques, CNFs are obtained
by catalytic decomposition of carbon-containing gases on
small supported metal particles. The arc discharge tech-
nique can produce carbon nanotubes only in small quan-
tities. Furthermore, thus-produced tubes are polluted with
other forms of carbon, such as carbon onions, fullerenes,
or amorphous carbon. Carbon nanofibres, on the other
hand, do not contain other forms of carbon. Additionally,
no heteroatoms, such as sulphur, are incorporated during
synthesis. The metal and support of the catalyst originally
used to grow the CNFs, however, have to be removed af-
ter synthesis. Investigation of use of carbon nanotubes and
carbon nanofibres as catalyst support has only started re-
cently (1–15). Carbon nanotubes are mostly used to study
the effects of this novel carbon material on the properties
of the final supported catalyst. The arc discharge proce-
1 To whom correspondence should be addressed. Fax: 030-2511027.
E-mail: D.C.Koningsberger@chem.uu.nl.
85
0021-9517/02 $35.00
c
ꢀ 2002 Elsevier Science (USA)
All rights reserved.

86
ROS ET AL.
turned out to be more active than a Pd/AC catalyst of a
This work describes the deposition of small rhodium
higher metal loading (2). Though the authors did not offer metal particles on fishbone carbon nanofibres by several
an explanation for the higher activity of the fibre-supported impregnation and ion-exchange techniques. The objective
catalyst, an effect of the size or the clustering of the carbon is to assess which fibre pretreatment procedures are needed
support bodies is very likely.
to achieve an optimal interaction of the nanofibres with the
Baker and co-workers studied the activity of several met- catalyst precursor. Very small (1–2 nm) metal particles can
als on CNFs. Fe or FeCu precursors were impregnated on be deposited on surface-oxidized CNFs. Furthermore, the
HCl-treated CNFs, ultimately resulting in particles of an influence of calcination and reduction procedures on the
average size of 12 nm at 5 wt% loading (7). In spite of the final metal particle size is studied. The size of the parti-
relatively large size of the metal crystallites, the activity of cles provides the opportunity to study metal–support in-
these catalysts was considerably higher in the hydrogena- teractions with XAFS spectroscopy, which requires very
tion of ethylene than Fe(Cu) was on activated carbon or small particles and a narrow particle size distribution. These
alumina. The authors pointed out the possible relationship studies could shed more light on the unusual activities and
between this activity enhancement and a unique metal– selectivities described above. The Rh/CNF catalysts were
support interaction between the metal particles and the tested in the liquid-phase hydrogenation of cyclohexene.
basal planes of the fibres. In later publications (8–10), Ni Veryhighactivitieswerefound. Moreover, theresultsmight
was impregnated on CNFs, resulting in 6- to 8-nm-size indicate that other factors, such as clustering of the catalyst
particles. Again, activity improvement compared to that support particles in the liquid phase, are more important in
of nickel metal on classical supports was obtained in the determining the activity of the catalyst than is metal particle
hydrogenation of 1-butene and 1,3-butadiene (8). Further- size.
more, differently structured carbon nanofibres displayed
different activities and selectivities (9). Preferential block-
ing of either the armchair or the zigzag graphitic edge
2. EXPERIMENTAL
faces with P or B, respectively, indicated that the effects
could be ascribed to the catalytic action of Ni residing on
2.1. Growth of Fishbone Carbon Nanofibres
the zigzag faces (10). Similar effects were found when the
Carbon nanofibres of the fishbone type were produced
hydrogenation of crotonaldehyde on Ni/CNF was studied by catalytic decomposition of CH4 on a Ni/Al2O3 catalyst
(11). The nanofibre-supported catalysts were more active (2–4, 16).
than Ni/Al2O3. Moreover, when alumina or “ribbon” CNFs
The Ni/Al2O3 catalyst with 30 wt% Ni metal loading was
were used as the supports, selective hydrogenation of the synthesized by the deposition–precipitation technique (17).
double bond took place, whereas with “platelet” CNFs the Alumina (Alon-C, Degussa) was suspended in an acidified
aldehyde functional group was converted. The results were aqueous solution of nickel nitrate (Acros, 99%) and di-
related to the adsorption of crotonaldehyde on nickel par- luted ammonia was injected within 2 h at room temperature
ticles on the zigzag faces, which supposedly have a special under vigorous stirring until the pH had reached a value of
atomic arrangement.
8.5. After overnight stirring, the suspension was filtered,
Other researchers found surprising selectivities in se- washed, and dried at 120◦C. Finally, the catalyst was cal-
lective hydrogenation reactions as well. Ledoux and co- cined at 600◦C in stagnant air for 3 h.
workers (12) impregnated Pd on acid-treated CNFs and
For the production of fishbone CNFs, 0.5 g of the 30 wt%
obtained 3- to 5-nm-size particles at 5 wt% loading. A very Ni/Al2O3 catalyst was reduced at 600◦C in 14% H2 in N2
high activity and a 98% selectivity towards the aldehyde (flow rate, 350 ml/min) in a vertical tubular reactor (diam-
was found with the hydrogenation of cinnamaldehyde at eter, 3 cm) for 2 h. After the temperature was decreased to
atmospheric pressure and 80◦C. Planeix et al. (13), on the 570◦C, methane (50% in N2, flow rate of 450 ml/min) was
other hand, observed a selectivity towards cinnamyl alco- passed through the catalyst bed for 6.5 h. The yield of fibres
hol of 92% with 3- to 7-nm-size Ru particles on nonpurified amounted to approximately 12 g.
nanotubes (0.2 wt% loading).
The synthesis of the fishbone CNFs was confirmed with
Finally, it is worth mentioning the work of Ang et al. (14) high-resolution TEM (18).
and of Keyser and Prinsloo (15). Ang deposited 2-nm-size
Pd particles on HNO3/H2SO4-oxidized carbon nanotubes
by using an electroless plating technique. He reported that
2.2. Catalyst Synthesis
the tubes were densely coated with particles, indicating a
high metal loading. Prinsloo used deposition–precipitation
from a homogeneous solution to obtain HNO3-oxidized
carbon nanofibres densely coated with cobalt particles of
7-nm diameter.
2.2.1. Impregnation. Untreated (as synthesized) fish-
bone carbon nanofibres as well as CNFs that were surface
oxidized with concentrated nitric acid or a mixture of con-
centrated nitric and sulphuric acid were used for the im-
pregnation of an aqueous solution of RhCl3 · 2H2O.

SMALL RHODIUM PARTICLES ON CARBON NANOFIBRES
87
Ten grams of CNFs were boiled in 40 ml of HNO3 or
2.2.2. Ion exchange. Only fibres that were surface ox-
80 ml of HNO3/H2SO4 (1 : 1) for 60 min (HNO3, Lamers & idized with HNO3/H2SO4 (see above) were used for ion
Pleuger, 65%, pure; H2SO4, Merck, 95–97%, p.a.). Upon exchange. As ion exchange with RhCl3 · 2H2O in water
cooling and dilution with demineralized water, the sus- was not successful at room temperature, 1 g of fibres
pensions were filtered over a Teflon membrane filter with was refluxed in 30 ml of a 3.3 × 10−3 mol/L solution of
a pore diameter of 0.2 µm, washed with water until the RhCl3 · 2H2O in demineralized water for 48 h (maximum
washings showed no significant acidity, and dried at 120◦C loading 1 wt%). Next, the ion-exchanged fibres were fil-
for 16 h.
tered and dried at 120◦C for 17 h. Subsequently, the cata-
The total pore volume of the differently treated fibres lysts were reduced either with H2 (10% H2/Ar, 5.6◦C/min,
was determined by adding demineralized water to degassed 200◦C, 20 min) or with NaBH4. Reduction with NaBH4
CNFs with a syringe until the fibres had a sticky appearance. was carried out by dispersing 1 g of ion-exchanged cata-
The values obtained were about 0.4 ml/g. Next, degassed lyst in 10 ml of ethanol (Merck, p.a.) and by adding 55 mg
CNFs were impregnated in static vacuum with an aque- of NaBH4 (Alfa, 99%) in 10 ml of ethanol. The evolution
ous solution of RhCl3 · 2H2O with a concentration of about of gas indicated an immediate reaction. After being stirred
0.3 mol/L in order to get a rhodium metal loading of 1 wt%. for 30 min at room temperature, the reduced catalyst was
Degassing of the fibres was performed by evacuation at filtered from the colourless solution, washed with ethanol,
room temperature for several hours. The Rh-impregnated and dried under vacuum at room temperature for 17 h.
fibres were dried at room temperature for 36 h and, subse-
The ion-exchange procedure was also carried out in a 4%
ammonia solution at room temperature and at refluxing
quently, dried at 120◦C for 17 h.
The Rh-impregnated fibres were subjected to several conditions for 48 h. Again, the solution contained enough
calcination–reduction procedures. The catalysts were either rhodium to obtain a 1 wt% catalyst. These catalysts were
calcined and then reduced or directly reduced. Calcination reduced at 200◦C in H2 (10% H2/Ar, 5.6◦C/min, 20 min).
took place at 150 or 250◦C (stagnant air, 5◦C/min, 1 h), while
Table 1 summarizes the several ion-exchange procedures
the subsequent reduction temperature was 140 or 250◦C together with the accompanying catalyst codes.
(10% H2/Ar, 5.6◦C/min, 20 min). The reduction tempera-
ture for the direct reduction was either 160 or 250◦C. The re-
2.3. Characterization
ductions at lower temperatures were monitored with a hot
wire detector (HWD). The several calcination–reduction
treatments along with the catalyst codes are shown in
Table1fortheHNO3/H2SO4-oxidizedfibres. Theuntreated
Characterization of the catalysts was carried out by scan-
ning electron microscopy (SEM), transmission electron
microscopy (TEM), and X-ray absorption fine structure
(XAFS) spectroscopy.
and the HNO3-oxidized fibres were not coded.
2.3.1. Electronmicroscopy. SEMimageswere obtained
on a Philips XL30FEG microscope at an accelerating volt-
age of 20 kV. TEM samples were prepared by dispersing
ground catalyst in ethanol using ultrasound and drying a
TABLE 1
Catalyst Codes for HNO3/H2SO4-Oxidized Carbon Nanofibres
drop on a copper grid covered by a holey carbon film. They
Code
Description
were imaged in a Philips CM-200 FEG electron microscope
operating at 200 kV and in a Philips EM-420 electron mi-
croscope operated at an accelerating voltage of 120 kV. The
CM-200 microscope was equipped with a field emission gun
and an EDX facility for elemental analysis.
Rh/CNF(imp, c150, r140)
Impregnated, calcined at 150◦C,
reduced at 140◦C
Rh/CNF(imp, c250, r140)
Impregnated, calcined at 250◦C,
reduced at 140◦C
Rh/CNF(imp, r160)
Rh/CNF(imp, c150, r250)
Impregnated, reduced at 160◦C
Impregnated, calcined at 150◦C,
reduced at 250◦C
2.3.2. XAFS data collection. XAFS spectra at the Rh
K-edge (23,220 eV) were obtained at the HASYLAB
(Hamburg, Germany) synchrotron beamline X1.1, which
was equipped with a Si(311) double-crystal monochroma-
tor. The monochromator was detuned to 50% of the maxi-
mum intensity to minimize the presence of higher harmon-
ics. All measurements were performed in the transmission
mode using ionization chambers filled with an Ar/N2 mix-
ture to have an absorbance of 20 and 80% in the first and
the second ionization chambers, respectively.
Rh/CNF(imp, c250, r250)
Impregnated, calcined at 250◦C,
reduced at 250◦C
Rh/CNF(imp, r250)
Rh/CNF(IET)
Rh/CNF(IET, r200)
Impregnated, reduced at 250◦C
Ion exchanged at 100◦C, not reduced
Ion exchanged at 100◦C, reduced
at 200◦C
Rh/CNF(IET, rNaBH4)
Rh/CNF(IEN, r200)
Rh/CNF(IENT, r200)
Ion exchanged at 100◦C, reduced
with NaBH4
Ion exchanged with NH3, reduced
at 200◦C
Ion exchanged with NH3 at 100◦C,
reduced at 200◦C
The powdered CNF samples (300 mg) were pressed
into self-supporting wafers and mounted into an in situ
EXAFScellequippedwithBewindows(19). Duetothelow
Note. All catalysts contain 1 wt% Rh.

88
ROS ET AL.
concentration of Rh (1 wt%) and the low absorption of car-
2.3.5. R-space fitting, difference file technique, and weight
bon, the calculated edge step was about 0.2. The Rh2O3 ref- factor kn. The EXAFS data-analysis program XDAP al-
erence compound (Aldrich, 99.8%) was diluted with boron lows multiple-shell fitting in R-space by minimizing the
nitride (Alfa, 99.8%) and pressed into a self-supporting residuals between both the magnitude and the imaginary
wafer (calculated edge step 1). The rhodium reference part of the Fourier transforms of the data and the fit.
foil (Aldrich, 99.9%, 25-µm thickness) was mounted on R-space fitting has important advantages compared to the
the sample holder with Kapton tape. All samples (except usuallyappliedfittingink-spaceandisextensivelydiscussed
Rh/CNF(IET) and Rh/CNF(IET, rNaBH4)) were in situ re- in a recent paper by Koningsberger et al. (23).
duced (10% H2/He, 5◦C/min, 1 h) and cooled to liquid ni-
trogen temperature prior to measurement. All gasses were phase-corrected Fourier transforms to resolve the differ-
purified before use. ent contributions in the EXAFS data (23). The difference
The difference file technique was applied together with
It should be noted that with all catalysts, edge steps in the file technique allows one to optimize each individual con-
X-ray absorption spectra were obtained that were equal to tribution with respect to the other contributions present
the theoretical edge step for 1 wt% Rh on carbon. In other in the EXAFS spectrum. If the experimental spectrum is
words, all catalysts indeed contain 1 wt% Rh.
composed of different contributions, then
2.3.3. XAFS data processing. Extraction of the EXAFS
data from the measured absorption spectra was performed
with the XDAP code developed by Vaarkamp et al. (20)
Two scans were averaged and the pre-edge was subtracted
using a modified Victoreen curve (21). The background was
subtracted employing cubic spline routines with a continu-
ously adjustable smooth parameter (22, 23). Normalization
was performed by dividing the data by the height of the
absorption edge at 50 eV, leading to normalised EXAFS
data.
N
ꢀ
Exp.Data =
(Fit)i ,
[1]
i=1
where (Fit)i represents the fitted contribution of coordina-
tion shell i. For each individual contribution the following
equation should then be valid:
N
ꢀ
(Fit)j = Exp.Data −
(Fit)i .
[2]
i=1 and i= j
2.3.4. EXAFS phase shifts and backscattering amplitudes.
Data for phase shifts and backscattering amplitudes were
obtained from the reference compounds: Rh foil for Rh–Rh
and Rh2O3 for Rh–O contributions (23). It is well-known
that differences in backscattering amplitude and phase shift
of neighbours and next nearest neighbours in the periodic
system are too small to make a distinction between the two
scatterers (24). Therefore, the contribution of Rh–C was
fitted with the Rh–O reference. Table 2 gives the crystallo-
graphic data and the forward and inverse Fourier transform
ranges used to create the EXAFS references. Both the ref-
erence spectra and the samples were measured at liquid
nitrogen temperature. This means that no temperature ef-
fect has to be included in the difference in Debye–Waller
factor (ꢀσ2) between sample and reference as obtained in
the EXAFS data analysis.
The right-hand side of this equation is further denoted as
the difference file of shell j. A good fit is only obtained if the
total fit and each individual contributing coordination shell
describe correctly the experimental EXAFS and the differ-
ence file, respectively. In this way not only the total EXAFS
fit but also the individual fits of all separate contributions
can be determined reliably.
Both high-Z (e.g., metal–metal) and low-Z (e.g., metal–
oxygen) contributions are present in the EXAFS data col-
lected on small metal particles dispersed on high-surface-
area supports. The low-Z contributions may arise from the
support. A k3 weighting emphasizes the high-Z contribu-
tions to the spectrum, since high-Z elements have more
scattering power at high values of k than do low-Z elements.
However, the use of a k3-weighted EXAFS spectrum or
Fourier transform makes the analysis much less sensitive
to the presence of low-Z contributions in the EXAFS data.
In this study, the EXAFS fits have been checked by apply-
ing k1, k2, and k3 weightings in order to be certain that the
results are the same for all weightings.
TABLE 2
Crystallographic Data and the Forward and Inverse Fourier
Transform Ranges Used to Produce the EXAFS References
2.3.6. Number of independent data points, variance, and
statistical significance. The number of independent data
points (Nindp) was determined as outlined in (27, 28):
Atom
pair
Ref.
Forward −F1T Inverse FT
compd. Ref. Nref Rref (A) kn (ꢀK) (A
)
(ꢀR) (A)
˚
˚
˚
Rh–Rh Rh foil
Rh–O Rh2O3
25
26
12
6
2.69
2.05
k3
k1
2.7–22.7
3.7–16.3
1.4–3.1
0.0–2.3
Nindp = (2ꢀkꢀR/π) + 2.
[3]
a
a It was established with XRD that the Rh2O3 was in the hexagonal
form.
The variances of the magnitude and imaginary part of the
Fourier transforms of fit and data were calculated according

SMALL RHODIUM PARTICLES ON CARBON NANOFIBRES
89
to the equation
ꢁ ꢂ
where dp is the particle diameter. DeA is given by
DeA = (εp DA)/τp,
ꢃ
FTnmodel(R) − FTn (R) 2 d R
[8]
ꢁ ꢂ
ꢃexp
Variance =
× 100, [4]
FTnexp(R) 2 d R
where DA is the diffusion coefficient of reactant A, εp is
the void fraction of the particle, and τp is the tortuosity of
the particle. The reaction rate constant kv,p can be obtained
from
with FTn = FT[knχ(k)]. In this study the statistical signifi-
cance of a contribution has been checked by a comparison
of the amplitude of (Fit)j with the noise level present in the
difference file (the noise in the difference file is essentially
the same as the noise in the experimental data).
−k1t
CA = C
e
,
[9]
A0
assuming a first-order reaction. CA is the concentration of
reactant at time t, CA0 is the initial concentration, and k1 is
the reaction rate constant per unit reactor volume. k1 can
easily be converted to kv,p by multiplying it with the ratio
of reactor volume and catalyst volume.
Because the kv,p thus obtained is an observed reaction
rate constant, which can be influenced by diffusion, the cal-
culation should be made cyclic: η should be used to calculate
an intrinsic reaction rate constant by dividing kv,p with η and
the calculation should be repeated until no more changes
are observed. In this way, a good approximation of the in-
trinsic reaction rate constant can be obtained.
2.3.7. Estimation of particle size. The average particle
size of the various catalysts was estimated from the Rh–
Rh coordination numbers obtained with EXAFS by using
a computer program (29). The program “cuts” spherical or
semispherical particles with different diameters from a Rh
fcc crystal and calculates the average coordination number
and the number of atoms in the particle.
2.4. Catalytic Test Experiments
The catalytic activity of the Rh/CNF catalysts in the hy-
drogenation of cyclohexene was tested in a semibatch slurry
reactor, operated at a constant pressure of 1200 mbar of
H2. The thermostated, double-walled reaction vessel was
equipped with vertical baffles and a gas-tight stirrer with
a hollow shaft and blades for gas recirculation. The stirrer
was operated at 2000 rpm. During reaction, the hydrogen
consumption was automatically monitored by a mass flow
meter. It was assured that under the conditions used, the
rate of dissolution of H2 in the solvent was higher than the
maximum measurable rate of H2 uptake.
In a typical experiment, the reaction vessel was filled with
50 mg of powdered catalyst and 100 ml of ethanol (Merck
p.a.). Next, the reactor was thermostated at 25◦C, evacu-
ated, filled with H2, and pressurized. After being stirred
for 1 h under H2 in order to reduce the catalyst, the reac-
tion was started by injection of 1 ml of cyclohexene (Acros,
99%) with a syringe.
3. RESULTS
3.1. Scanning Electron Microscopy
Figure 1 displays two SEM images of powdered fishbone
carbon nanofibres treated with HNO3/H2SO4. Figure 1A
shows that the bodies of entangled fibres in this sample ex-
hibit a broad particle size distribution. However, no nanofi-
bre particles larger than about 50 µm are observed. Most
CNF particles have a dimension between 5 and 20 µm. At
larger magnification (Fig. 1B) the individual fibres become
visible. The fibres are entangled and form an open, macro-
porous structure. The pore structure of the fibres consists
only of mesopores, which are located between individual fi-
bres in a cluster (31). The fibres themselves are not porous
and, consequently, metal particles will only be located at
the outer surface of the fibres.
The presence or absence of internal diffusion limitations
was checked by calculation of the internal effectiveness fac-
tor η, using the expression (30)
3.2. Transmission Electron Microscopy
The size and distribution of the rhodium particles on un-
treated, HNO3-oxidized and HNO3/H2SO4-oxidized fibres
were examined with TEM (see Fig. 2). Within the untreated
CNFs rhodium particles are found only very locally. No
rhodium particles are observed on the carbon nanofibres
(see Fig. 2A). However, many rhodium particles have been
deposited on the holey carbon film (see Fig. 2B). Most of
the deposited rhodium particles are smaller than 10 nm,
but larger particles up to 80 nm have been found as well.
Detected rhodium is always associated with the presence of
Al2O3 inthese fibres(originatingfromtheNi/Al2O3 growth
catalyst).
η = tanh(ϕ)/ϕ,
[5]
with
√
ϕ = 1/avꢂ (kv,p/DeA),
[6]
where ϕ is the Thiele modulus, kv,p is the reaction rate con-
stant per unit catalyst volume, and DeA is the effective dif-
fusion constant for reactant A. avꢂ is the ratio between the
external surface area and the volume of the catalyst parti-
cle. Values of η close to 1 indicate that internal diffusion
limitations are absent.
For spherical particles avꢂ is given by
With the HNO3-oxidized carbon nanofibres most fibres
are also bare (see Fig. 2C). Only occasionally are large Rh
avꢂ = 6/dp,
[7]

90
ROS ET AL.
FIG. 1. SEM images of HNO3/H2SO4-oxidized fishbone carbon nanofibres at different magnifications. (A) The skeins of fibres; (B) the individual
CNFs.
particles detected on the CNFs. EDX analysis revealed that 20 min (right-hand part of the figure). The figure shows
Rh was only present within clusters of fibres. An example that the reduction of Rh/CNF(imp, c250, r140) had just
is shown in Fig. 2D. These rhodium particles are probably completed at the termination of the program. The same
large as well.
results were obtained for Rh/CNF(imp, c150, r150) and
In spite of an extended study, no rhodium particles could Rh/CNF(imp, r160). The reductions at higher temperatures
be imaged at all on the fibres that had been treated with were not monitored.
HNO3/H2SO4. The diffraction contrast of the fibres them-
selves hinders the imaging of very small rhodium particles.
However, when an EDX analysis was performed on a single
3.4. XAFS Spectroscopy
fibre (see Figs. 2E and 2F) some rhodium could be detected.
The Cu observed originates from the copper grid, the Al
comes from some Al2O3 that remained in the sample after
the oxidation treatment, and the Si is probably introduced
during the oxidation treatment (in a glass round-bottomed
flask).
3.4.1. XAFS analysis. Figure 4 shows the experimental
EXAFS data of Rh/CNF(imp, c150, r140) (Fig. 4A) and
Rh/CNF(imp, c150, r250) (Fig. 4C). The data q−u1ality is ex-
˚
cellent. The signal-to-noise ratio at k = 4 A amounts
to approximately 120 for Rh/CNF(imp, c150, r140) and 90
for Rh/CNF(imp, c150, r250). It can clearly be seen that
the EXAFS intensity of Rh/CNF(imp, c150, r250) at larger
values of k (k > 10 A−1) is higher than for Rh/CNF(imp,
˚
3.3. Reduction of the Catalysts
c150, r140), indicating the presence of larger rhodium parti-
The impregnated catalysts were first reduced at the low- cles. The corresponding k1–Fourier transforms of both cata-
est reduction temperature possible (140◦C for the reduction lysts (Figs. 4B and 4D) confirm this observation. The peak
of Rh2O3 and 160◦C for the reduction of RhCl3 · 2H2O). at 2.5 A originating from the Rh–Rh contribution has a
˚
These temperature levels and the temperatures for higher intensity with Rh/CNF(imp, c150, r250) than with
the calcination of RhCl3 · 2H2O were determined using Rh/CNF(imp, c150, r140).
in situ temperature-programmed XRD on pure Rh2O3 and
RhCl3 · 2H2O.
The Fourier transform (FT) of the experimental EXAFS
dataofRh/CNF(imp, c150, r140)togetherwiththeFTofthe
Figure 3 shows the reduction profile of Rh/CNF(imp, total fit is shown in Fig. 5A. The fit quality is excellent for the
˚
c250, r140). The left part of the figure exhibits the reduc- used fit range (ꢀR = 1–2.8 A). In Rh/CNF(imp, c150, r140)
tion profile during the temperature ramp up to 140◦C. Be- Rh–Rh and Rh–C contributions could be established. The
cause the plotted temperature is the temperature within the phase-corrected Fourier transforms of the fits of the two
reactor and the temperature controller uses the tempera- individual contributions together with their corresponding
ture of the furnace, the temperature of the catalyst bed is difference files are shown in Fig. 5B and 5C for the Rh–
lower than 140◦C after the ramp. After a period of stabiliza- Rh and Rh–C contributions, respectively. The fit quality
tion of the temperature (during which the H2 consumption of each individual contribution is very good. The statisti-
is not monitored), the catalyst bed reaches a temperature cal significance of the fits is demonstrated in Fig. 5D. The
of 140◦C and the reduction is continued isothermally for calculated k0-weighted Rh–Rh and Rh–C EXAFS spectra

SMALL RHODIUM PARTICLES ON CARBON NANOFIBRES
91
FIG. 2. (A, B) TEM images of rhodium impregnated untreated CNFs (calc. 150◦C, red. 140◦C). (C, D) TEM images of rhodium impregnated
HNO3-oxidized CNFs (calc. 150◦C, red. 140◦C). (E) TEM image of Rh/CNF(imp, c150, r140). (F) EDX spot analysis of area indicated with a white
arrow in (E).
that are shown in the figure can be compared with the dot-
Table 3 presents the fit parameters of the EXAFS spectra
ted lines approximately indicating the maximum (peak to resulting from the multiple-shell fits in R-space and the
peak) noise levels. It is shown that the Rh–C contribution variances of the imaginary and absolute part of the fits for
has a higher intensity than the noise level up to high k val- all catalysts. All variances obtained are well below 1%,
ues, demonstrating that reliable fit results for the weaker indicating that excellent fit results for all catalysts have
Rh–C contribution have been obtained as well.
been acquired (23). Furthermore, the calculated number of

92
ROS ET AL.
independent data points (Nindp, see Eq. [3]) is 14, show-
ing that a fit with three shells is possible (Nindp = 12). With
Rh/CNF(imp, c150, r140) the Rh–Rh coordination num-
ber amounts to 5.8. Furthermore, 2.2 carbon atoms are
˚
distinguished at a distance of 2.7 A. Upon reduction at
250◦C the Rh–Rh coordination number increases to 7.7
[Rh/CNF(imp, c150, r250)]. An increase can also be ob-
served with the catalysts that were calcined at 250◦C and
those that were directly reduced. Furthermore, calcination
at either 150 or 250◦C does not influence the coordination
number of the rhodium atoms obtained. Direct reduction
results in a larger coordination number at low reduc-
tion temperatures than does calcination–reduction. At a re-
duction temperature of 250◦C this effect almost disappears.
FIG. 3. Hydrogen consumption profile during reduction of Rh/CNF
(imp, c250, r140).
FIG. 4. (A) Experimental k1-weighted EXAFS spectrum of Rh/CNF(imp, c150, r140). (B) Corresponding uncorrected Fourier transform (k1
weighted, ꢀk = 3.9–14 A−1). (C) Experimental k1-weighted EXAFS spectrum of Rh/CNF(imp, c150, r250). (D) Corresponding uncorrected Fourier
˚
transform (k1 weighted, ꢀk = 3.9–14 A−1).
˚

SMALL RHODIUM PARTICLES ON CARBON NANOFIBRES
93
−1
FIG. 5. Fit results of Rh/CNF(imp, c150, r140) (k1 weighted, ꢀk = 3.9–14 A , ꢀR = 1–2.8 A). (A) Fourier transform of experimental
EXAFS data and total fit (uncorrected). (B) Rhodium–rhodium fit and corresponding difference file (Rh–Rh phase corrected). (C) Rhodium–carbon
fit and corresponding difference file (Rh–C phase corrected). (D) Calculated k0-weighted EXAFS spectrum of Rh–Rh and Rh–C fit. The two dotted
lines indicate the noise level of the spectra.
˚
˚
With all impregnated catalysts a Rh–C contribution can be Fig. 6B, 6D, and 6F. Before reduction (Fig. 6B) mainly con-
˚
distinguished.
tributions at 1.6 A arising from low-Z scatterers are present.
˚
Figures 6A, 6C, and 6E display the experimental Upon reduction with H2 (Fig. 6D) an intense peak at 2.4 A
EXAFS data of Rh/CNF(IET), Rh/CNF(IET, r200), and points to the formation of rhodium particles. After reduc-
Rh/CNF(IET, rNaBH4), respectively. The signal-to-noise tion with NaBH4 (Fig. 6F) only a small Rh–Rh contribution
ratio at k = 4 A−1 of these spectra ranges from 50 to 70. The at the same distance can be distinguished.
˚
EXAFS intensity of Rh/CNF(IET) (Fig. 6A) is low at large
Figure 7 shows the fit results of Rh/CNF(IET, rNaBH4).
k values (k > 10 A−1), indicating that high-Z scatterers are Again, good results are obtained for the total fit and sep-
absent in this spectrum. Upon reduction with H2 (Fig. 6C) arate Rh–Rh, Rh–O, and Rh–C contributions. At a dis-
˚
˚
the spectrum totally changes. The high EXAFS intensity at tance of 2.68 A an average number of 3.8 Rh neighbours
large k values (k > 10 A−1) shows that a significant Rh–Rh is present (see Table 3). The coordination number of oxy-
˚
˚
˚
contribution is present. The spectrum of the catalyst that gen neighbours amounts to 4.7 at 2.05 A. Finally, 1.5 carbon
was reduced with NaBH4 (Fig. 6E) displays characteristics atoms are visible at a distance of 2.29 A. Before reduction
intermediate to that of the other catalysts. The k2–Fourier [Rh/CNF(IET)] no Rh–Rh can be fitted in the FT of the
transforms of the experimental EXAFS data are shown in experimental EXAFS data. Four oxygen atoms are visible

94
ROS ET AL.
TABLE 3
Fit Parameters of EXAFS Spectra and Variances of Fits (see Eq. [4])a
Variancec (%)
N
R
ꢀσ2
ꢀE0
(eV) ( 10%)
2
˚
˚
Catalyst
Atom
(
5%)b
(A) ( 1%)
(A ) ( 5%)
Im
Abs
0.21
Rh/CNF(imp, c150, r140)
Rh/CNF(imp, c250, r140)
Rh/CNF(imp, r160)
Rh
C
Rh
C
Rh
C
5.8
2.2
5.9
2.0
6.7
1.8
7.7
2.3
7.5
2.5
7.8
2.5
4.0
2.0
2.7
5.9
2.2
3.8
4.7
1.5
8.9
8.7
2.68
2.70
2.68
2.70
2.68
2.70
2.68
2.70
2.67
2.70
2.68
2.70
2.05
2.27
2.86
2.68
2.70
2.68
2.05
2.29
2.69
2.69
0.002
0.000
0.002
0.000
0.002
0.000
0.002
0.000
0.003
0.000
0.002
0.000
0.001
0.004
0.001
0.004
0.000
0.010
0.001
0.008
0.001
0.001
5.5
−3.6
7.0
0.37
0.62
0.23
0.18
0.35
0.26
0.19
0.32
0.13
0.10
0.21
0.15
0.05
−3.2
7.2
−3.4
7.2
Rh/CNF(imp, c150, r250)
Rh/CNF(imp, c250, r250)
Rh/CNF(imp, r250)
Rh
C
Rh
C
Rh
C
O
−7.3
7.2
−6.9
7.5
−7.1
−6.4
8.0
Rh/CNF(IET)d
Cl
C
−4.9
Rh/CNF(IET, r200)
Rh
C
Rh
O
9.4
0.19
0.49
0.11
0.20
−1.8
7.8
−2.1
−5.6
0.3
Rh/CNF(IET, rNaBH4)d
C
Rh/CNF(IEN, r200)
Rh/CNF(IENT, r200)
Rh
Rh
0.19
0.23
0.15
0.17
−0.8
a
−1
1
˚
˚
ꢀk = 3.9–14 A ; ꢀR = 1–2.8 A; k weighted.
b For Rh–Rh. For low-Z scatterers the accuracy is estimated at 10–20%.
c Variance of total fit.
d k2 fitted.
˚
˚
points were fitted with logarithmic functions. Figure 9B dis-
plays the relationship between the particle size and the
number of atoms in the particle. The results obtained corre-
spond well with data obtained by Kip et al. (32). The figures
clearly show that the outcome depends considerably upon
the assumed particle shape. However, the geometry of the
Rh/CNF particles is not known. The smallest particles could
well be a semispherical shape, whereas the larger particles
could adapt a more spherical shape. Therefore, the average
of the particle size obtained for spherical and semispheri-
cal particles was taken. Table 4 displays the results for all
at 2.05 A, two chlorine atoms reside at 2.27 A, and 2.7
˚
carbon atoms are distinguishable at 2.86 A. After reduc-
tion with H2 [Rh/CNF(IET, r200)] the coordination num-
ber of Rh amounts to 5.9, and 2.2 carbon atoms are present
˚
at 2.70 A.
Figure 8 shows the Rh–C fits and the corresponding dif-
ference files of Rh/CNF(imp, c150, r140) and Rh/CNF(IET,
rNaBH4). After reduction with H2 a rhodium–carbon con-
tribution is present at 2.70 A, while after the use of NaBH4
as reductant a Rh–C interaction is present at a much shorter
˚
˚
distance, 2.29 A.
The ion-exchanged catalysts that were prepared in the
presence of NH3 and reduced at 200◦C show the largest
Rh–Rh coordination number (see Table 3). Synthesis at
room temperature or reflux temperature does not affect
the coordination number significantly. Values of 8.9 and 8.7
were obtained, respectively. No Rh–C contributions could
be fitted.
TABLE 4
Estimated Average Particle Sizes for All Catalysts
Catalyst
N
D (nm)
Rh–Rh
Rh/CNF(imp, c150, r140)
Rh/CNF(imp, c250, r140)
Rh/CNF(imp, r160)
Rh/CNF(imp, c150, r250)
Rh/CNF(imp, c250, r250)
Rh/CNF(imp, r250)
Rh/CNF(IET, r200)
Rh/CNF(IEN, r200)
Rh/CNF(IENT, r200)
5.8
5.9
6.7
7.7
7.5
7.8
5.9
8.9
8.7
1.1
1.1
1.4
1.7
1.6
1.7
1.1
2.1
2.0
3.4.2. Estimation of particle size. Figure 9A shows the
relationship between the particle size and the coordination
number for spherical and semispherical particles. Because
in some cases no additional rhodium atoms are included in
the modelled particle upon increase of its size with 0.1 nm,
some points have the same coordination number. The data

SMALL RHODIUM PARTICLES ON CARBON NANOFIBRES
95
FIG. 6. Experimental k1-weighted EXAFS spectra of Rh/CNF(IET) (A), Rh/CNF(IET, r200) (C), and Rh/CNF(IET, rNaBH4) (E) and the
corresponding uncorrected Fourier transforms (B, D, and F, respectively) (k2 weighted, ꢀk = 3.9–14 A−1).
˚
catalysts. The size of the rhodium particles in the catalysts This curve was integrated to get an H2 uptake pattern,
ranges from 1.1 to 2.1 nm.
which, in turn, was used to calculate the initial activity (i.e.,
the activity in mol of H2/gcat h−1 at t = 0). The results are
summarized in Table 5 together with the estimated particle
sizes. Figure 11 shows a plot of the initial activities against
the particle size. All activities could be reproduced well
within 10%.
3.5. Catalytic Test Experiments
The catalytic measurements resulted in a plot of the H2
flow against time [see Fig. 10 for Rh/CNF(imp, c150, r140)].

96
ROS ET AL.
−1
˚
˚
FIG. 7. Fit results of Rh/CNF(IET, rNaBH4) (ꢀk = 3.9–14 A , ꢀR = 1–2.8 A). (A) Fourier transform of experimental EXAFS data
and total fit (k2 weighted, uncorrected). (B) Rhodium–rhodium fit and corresponding difference file (k1 weighted, Rh–Rh phase corrected).
(C) Rhodium–oxygen fit and corresponding difference file (k1 weighted, Rh–O phase corrected). (D) Rhodium–carbon fit and corresponding dif-
ference file (k1 weighted, Rh–C phase corrected).
It is important to determine whether the measured activ-
TABLE 5
Initial Activities and Estimated Particle Sizes for All Catalysts
Initial activity
ity is influenced by diffusion of the reactants and products
to and from the active sites. In general, three mass transfer
phenomena can be distinguished for the reactants in hy-
drogenation reactions (30): (i) transfer of hydrogen from
the gas phase to the liquid phase, (ii) transport of the reac-
tants to the external surface of the catalyst particle (exter-
nal mass transfer), (iii) transport of the reactants from the
external surface of the particle through the pores to the ac-
tive sites on the interior surface (internal mass transfer). As
was pointed out under Experimental, under the conditions
used, the rate of dissolution of H2 in the solvent is higher
than the maximum measurable rate of H2 uptake. In other
words, the rate is independent of stirrer frequency. Stirring
at a higher speed decreases the stagnant layer around a
catalyst particle. In the case of external diffusion limita-
tion, an effect of stirrer frequency on the activity should
Catalysta
(mol H2/gcat h−1
)
D (nm)
1. Rh/CNF(imp, c150, r140)
2. Rh/CNF(imp, c250, r140)
3. Rh/CNF(imp, r160)
4. Rh/CNF(imp, c150, r250)
5. Rh/CNF(imp, c250, r250)
6. Rh/CNF(imp, r250)
7. Rh/CNF(IET, r200)
- Rh/CNF(IET, rNaBH4)
8. Rh/CNF(IEN, r200)
0.78
0.71b
0.74
1.13
0.86
0.83
1.12
1.04
0.86
0.83
1.1
1.1
1.4
1.7
1.6
1.7
1.1
—
2.1
2.0
9. Rh/CNF(IENT, r200)
a Catalyst quantity, used 50 mg.
b With 100 mg the same value was obtained.

SMALL RHODIUM PARTICLES ON CARBON NANOFIBRES
97
FIG. 8. Rhodium–carbon fits and corresponding difference files (k1 weighted, Rh–C phase corrected) of (A) Rh/CNF(imp, c150, r140) and
(B) Rh/CNF(IET, rNaBH4).
be expected. Therefore, it is highly unlikely that the reac- in order to decrease the particle size even further. Both
tion rate is affected by transfer of H2 to the liquid phase catalysts showed about a threefold increase in activity. Al-
or by diffusion of reactants to the external surface of the though these results must be interpreted with care, because
catalyst particle. What remains is the question of whether other factors, such as the clustering of support bodies, can
there is internal diffusion limitation or not. An easy check also play a role when a catalyst is subjected to such me-
is to decrease the catalyst particle size. If the reaction rate chanical forces, the outcome of this experiment points to
is not affected by this procedure, the reaction is not limited the presence of internal diffusion limitation. To shed more
by internal diffusion. However, the catalysts used in this light on this diffusion phenomenon the internal effective-
study are already powdered and it is difficult to control par- ness factor η was calculated. Estimations of the values of
ticle size on such a small scale. Therefore, as a rough check, the parameters needed to calculate η are shown in Table 6.
two catalysts [Rh/CNF(imp, c250, r140) and Rh/CNF(imp,
It was ascertained that the reaction studied was indeed
c250, r250)] were powdered with a mortar very thoroughly first order with respect to cyclohexene [plotting ln(C0/C)
FIG. 9. (A) First shell coordination number against particle diameter for spherical and semispherical particles. (B) Number of atoms against
particle diameter for spherical and semispherical particles.

98
ROS ET AL.
FIG. 10. H2 flow and H2 uptake curves of Rh/CNF(imp, c150, r140).
against t resulted in a straight line]. Cyclohexene concen- the bulk solution is much lower than that of cyclohexene
trations were calculated from the initial concentration and (3.5 × 10−3 (33) vs. 0.1 mol/L, respectively), the rate of the
the H2 uptake pattern. Calculated k1 values at all concen- reaction is probably limited by the availability of H2. Conse-
trations of cyclohexene during reaction with Rh/CNF(imp, quently, the initial activities obtained for the Rh/CNF cata-
c250, r250) were averaged. For particles of 20 µm, which is lysts should only be considered qualitatively. As is shown in
a reasonable value for these catalysts (see Fig. 1A), a Thiele Eq. [6]theThielemodulusϕ andtherewiththeeffectiveness
modulus ϕ of 0.46 and an effectiveness factor η of 0.94 is ob- factor η is still always dependent upon the reaction rate kv,p
.
tained. We would like to point out that the cyclic calculation If we assume that the limitation with respect to H2 is about
referred to under Experimental did not have any significant the same for all catalysts, different intrinsic activities will
effect on the value of η. In other words, the observed still lead to different observed reaction rates, although the
first-rate order in cyclohexene is not influenced by diffu- differenceswouldbesmallerthanwithoutanydiffusionlim-
sion. This calculation also shows that it is unlikely that itations. It should be noted that after having powdered both
the reaction is limited by internal diffusion of cyclohex- catalysts severely, the catalyst Rh/CNF(imp, c250, r250) was
ene. However, considering that the concentration of H2 in still more active than the catalyst Rh/CNF(imp, c250, r140).
TABLE 6
Values of the Parameters Needed to Calculate the Internal
Effectiveness Factor η for Cyclohexene (see Eqs. [5]–[9])
Parameter
Value
dp
DA
2 × 10−5
m
10−9 m2 s−1
0.4 ml g−1
2.25 × 10−3 kg m−3
0.47
a
Total pore volumeb
ρfibres
c
εp
τ
2
DeA
2.4 × 10−10 m2 s−1
1.7 × 10−3 s−1
4.1 s−1
d
k1
kv,p
ϕ
0.46
η
0.94
a Typical for liquid phase.
b See Experimental section.
c
∼
ρgraphite
ρ
(34).
=
fibres
FIG. 11. Initial activity against particle diameter for all catalysts.
d Averaged over all cyclohexene concentrations during reaction.

SMALL RHODIUM PARTICLES ON CARBON NANOFIBRES
99
This supports the assumption that the observed activities nanofibres from a growth catalyst with an easily removable
are related to intrinsic activities. support, such as silica. Growth from a Ni/C support would
Table 5 and Figure 11 show that not much effect of be another option, since the carbon would not pollute the
rhodium metal particle size on the catalytic activity is ob- fibres. It would then become possible to obtain pure un-
served. With the exception of Rh/CNF(imp, c150, r250) treated fibres, which could be impregnated with an organic
and Rh/CNF(IET, r200) all catalysts exhibit about the same solution. This could be another interesting method of syn-
activity.
thesis of metal particles on CNFs.
4.2. Preparation of Small Rhodium Particles
on Surface-Oxidized Carbon Nanofibres
4. DISCUSSION
4.1. Pretreatment of Carbon Nanofibres
The EXAFS results have shown that with all catalysts
The TEM results have shown that after impregnation very small rhodium particles have been formed on the fish-
with an aqueous solution of RhCl3 · 2H2O, small rhodium bone carbon nanofibres. With Rh/CNF(IET, rNaBH4) no
metalparticlesaredepositedonlyoncarbonnanofibresthat particle size could be estimated based on the Rh–Rh coor-
are surface oxidized with HNO3/H2SO4. These findings can dination number, because the particles are partly oxidized.
be explained by taking into account the wetting proper- The reduction procedure with NaBH4 was executed ex situ
ties of the differently treated fibres. It has been shown be- and the catalyst was exposed to air afterwards. Most proba-
fore (31) that untreated fibres do not show any affinity for bly, theparticlesconsistofarhodiummetalcoresurrounded
water. Fibres treated with HNO3 have some affin- by an oxidic phase. We roughly estimate that the particle
ity for water, whereas only HNO3/H2SO4-oxidized fi- size ranges between 1.5 and 2 nm. A larger diameter is not
bres are well wetted by this solvent. Consequently, dis- likely, because a Rh–C contribution is still observed.
solved RhCl3 · 2H2O can only reach the surface of the
With the impregnated catalysts the particle size ranges
HNO3/H2SO4-oxidized CNFs. With the untreated fibres from 1.1 to 1.7 nm. The results obtained with the low-
the rhodium particles are mainly deposited on or near the temperature reductions, i.e., 140–160◦C, have shown that
Al2O3 particles that are still present and are deposited as calcination at 150 or 250◦C makes no difference for the
some large metal lumps. The affinity of the particles for particle size obtained. However, direct reduction results in
Al2O3 is also not very large, because ultrasonic treatment larger particles. This difference can be explained by higher
during the TEM sample preparation dispatches the parti- mobility of rhodium chloride under reducing than under
cles from the Al2O3 support (see Fig. 2B). With the HNO3- oxidising conditions. Furthermore, the reduction temper-
oxidized fibres about 50% of the original amount of Al2O3 ature of the direct reduction is 20◦C higher than that of
is removed during the oxidation treatment (31), decreas- the reduction after calcination. The EXAFS results show
ing the chance that the Rh will be deposited on or near that the particles are completely reduced after reduction at
this oxidic carrier. Furthermore, the fibres have developed 140–160◦C. No Rh–O contributions could be fitted in the
some affinity for water. This results in large rhodium par- spectrum. This result agrees well with the profiles obtained
ticles between the fibres. Prinsloo obtained similar results during temperature-programmed reduction. Furthermore,
with the impregnation of an iron precursor onto HNO3- no chlorine is present in the coordination sphere of the
oxidized CNFs (35). The Al2O3 within the skeins of fibre rhodium, although it is still possible that chlorine is present
that were treated with HNO3/H2SO4 is removed by about near the rhodium particle. When the catalysts are reduced
80% (31). Furthermore, the remaining Al2O3 is most prob- at 250◦C, no differences in particle size are observed any-
ably α-Al2O3 of a low surface area, present in the high- more. It is, therefore, likely that at this temperature the
surface-area γ -Al2O3 used for the synthesis of the carbon mobility of the rhodium metal atoms and not the rhodium
nanofibres. Therefore, not much influence of the remain- chloride is decisive for the final size of the particles.
ing alumina on the formation of Rh particles on the carbon
With the ion-exchanged catalysts, stirring at room tem-
nanofibres is to be expected. The HNO3/H2SO4-oxidized perature with an aqueous solution of RhCl3 · 2H2O was not
fibres are nicely wetted with water and small rhodium par- successful. This result can be explained by the relatively
ticles are formed after reduction. In view of the above- highacidityofthissolution. Thesolutionof3.7×10−3 mol/L
described results, the research was continued with only the used had a pH of 2.9. Hoogenraad and co-workers (4) found
HNO3/H2SO4-oxidized impregnated catalysts.
that the isoelectric point of HNO3-oxidized fibres is about
The TEM results demonstrate that the wetting prop- 2.3. At pH values above this point the fibres are negatively
erties of the carbon nanofibres are crucial for success of charged. The pH of the solution used for ion exchange was,
the impregnation. Furthermore, quite a large influence of therefore, probably very close to the isoelectric point of the
the remaining Al2O3 has been observed. Only after the se- fibres and there would have been barely any driving force
vere oxidation treatment does alumina largely dissolve in for ion exchange. Furthermore, the structure and, there-
the strong acid. These results call for the growth of carbon fore, the charge of RhCl3 · 2H2O in water was uncertain.

100
ROS ET AL.
Complexes such as [Rh(H2O)6]3+, [Rh(OH)3Cl3]3−, or erage diameter of 6–8 nm. The results of this work demon-
[RhCl6]3− may have been present in solution (36).
strate that it is possible to synthesize very small metal par-
When an aqueous solution of RhCl3 · 2H2O is refluxed ticles on oxidized CNFs. With particles of 1–2 nm the pre-
with the oxidized fibres, on the other hand, a rhodium cious metal component of the catalyst will be optimally
complex is formed on the CNFs [see EXAFS data of used. In other words, almost all metal atoms are surface
Rh/CNF(IET)]. The synthesis procedure resembles that atoms. Furthermore, with particles of this size, metal–
of procedures used with rhodium complex chemistry (37, support interactions can be studied. Several researchers
38). It is, therefore, likely that rhodium has formed a com- found increased activities or selectivities when carbon
plex with the carboxyl groups on the surface of the fibres. nanofibres or nanotubes were used as the support (7–13).
The negatively charged COO− group and the two chloride More information about the interaction between the metal
ions compensate for the charge of the Rh(III) ion. Three and the CNF support could shed some light on these
water molecules probably complete the coordination phenomena.
sphere of the central rhodium ion. Subsequent reduction
with H2 at 200◦C gives extremely small rhodium particles,
4.3. Metal–Support Interaction in Rh/CNFs
1.1 nm in size. By optimization of the reduction procedure
it is probably possible to synthesize even smaller particles
using this approach. Reduction with NaBH4 gives larger
The size of the particles obtained is so small that the
metal–support contact can be observed. With the exception
of Rh/CNF(IEN, r200) and Rh/CNF(IENT, r200) all cata-
particles, 1.5–2 nm in size. Probably this is due to the rel-
lysts show a Rh–C contribution. The size of the particles in
atively rough method of reduction [all NaBH4 was added
Rh/CNF(IEN, r200) and Rh/CNF(IENT, r200) is too large
to observe this contribution. An interesting observation can
be made by comparison of a catalyst that is reduced with H2
[Rh/CNF(imp, c150, r140)] and the NaBH4-reduced cata-
at once to the fibres that were (badly) stirred in a round-
bottomed flask].
Another successful ion exchange procedure involved ad-
dition of NH3 to the aqueous solution of RhCl3 · 2H2O.
lyst [Rh/CNF(IET, rNaBH4)]. Upon reduction with H2 the
Two factors play a role with this procedure: (i) the pH
of the solution is raised, increasing the driving force of
ion exchange between the negatively charged CNF surface
and a positively charged complex, and (ii) most probably
the positively charged [Rh(NH3)5Cl]2+ complex is formed
(39, 40). No effect of an increase of temperature during
ion exchange on the final size of the rhodium particles was
found using this synthesis procedure. However, the metal
particles formed using this ion-exchange method are much
larger than with other catalyst preparation techniques. This
is probably due to the high mobility of metal ammine com-
plexes under reducing atmospheres (41, 42). Careful drying
and calcination procedures could decrease the particle size
obtained considerably.
˚
Rh–C distance is observed at 2.70 A, whereas after reduc-
˚
tion with NaBH4 this distance is evident at 2.29 A. This
difference can be explained by assuming the presence of hy-
drogen at the interface of the rhodium metal particle and
the carbon nanofibre support with the H2-reduced cata-
lyst. A schematic depiction of rhodium particles on carbon
nanofibres in Rh/CNF(imp, c150, r140) and Rh/CNF(IET,
rNaBH4) is shown in Fig. 12. As a rhodium–carbon con-
tribution is observed it is likely that the particle has a
large interface area with the support. A semispherical shape
could then be assumed. With Rh/CNF(imp, c150, r140) the
rhodium particle is fully metallic. In between the particle
andthecarbonnanofibresupportinterfacehydrogenissitu-
ated, creating a large Rh–C distance. With Rh/CNF(IET,
rNaBH4)therhodiumparticleis partlyoxidized.Arhodium
core is surrounded by an oxidic phase. In this case, no in-
terface hydrogen is present and the Rh–C distance is short.
The phenomenon of interface hydrogen has repeat-
edly been observed with several metals, such as Rh, Ir,
Theparticlesizesobtainedwiththesynthesisofthe1wt%
rhodium-on-fishbone carbon nanofibre catalysts are very
small compared to results reported in the literature. De-
position of rhodium on carbon nanofibres or nanotubes
has not been reported yet, but for other metals several
reports exist. Hoogenraad and co-workers (6) synthesized
Pd on HNO3-oxidized parallel CNFs using ion exchange
of [Pd(NH3)4](NO3)2 and subsequent reduction at 200◦C.
They obtained an EXAFS coordination number of 7.9 and
estimated the size of the particles to be 1.5 nm. Ang et al.
(14) obtained Pd particles 2 nm in size on HNO3/H2SO4-
oxidized carbon nanotubes by electroless plating. Ledoux
and co-workers (12) impregnated palladium nitrate on
CNFs and observed 3- to 5-nm particles, while Planeix et al.
(13) prepared 3- to 7-nm Ru particles on carbon nanotubes.
Finally, Park and Baker (9) impregnated a solution of nickel
FIG. 12. Schematic depiction of rhodium particles on carbon nanofi-
nitrate in ethanol on CNFs and detected particles of an av- bres in Rh/CNF(imp, c150, r140) (left) and Rh/CNF(IET, rNaBH4)(right).

SMALL RHODIUM PARTICLES ON CARBON NANOFIBRES
101
and Pt, on oxidic carriers (γ -Al2O3, TiO2, zeolite-Y) by zirconia. An optimum size of about 3 nm was found. This
Koningsberger and co-workers (43–46). Reduction of the behavior they related to the strength and mode of adsorp-
catalysts at low temperature, i.e., below 300◦C, results in a tion of cyclohexene, although the adsorption characteris-
˚
long metal–oxygen distance, about 2.6–2.7 A. Upon subse- tics of H2 could play a role as well. The same conclusion
quent evacuation at 200◦C the M–O distance contracts to was drawn by Busser et al. (53), who, surprisingly, found a
˚
about 2.1–2.2 A. Ensuing reduction restores the long M–O decrease in TOF with increasing particle size for polymer-
interaction. These results clearly show that the occurrence stabilized rhodium colloids. Augustine et al. (54) reported
of the long M–O interaction is associated with hydrogen that with Pt catalysts several alkene hydrogenation sites ex-
and that the presence of interface hydrogen is a likely ex- ist that are related to the mode of adsorption of hydrogen.
planation for the above-described observations. With the Mu¨ller et al. (55) showed that the adsorption and desorption
Rh/CNF catalysts the difference in Rh–C distance can ex- as well as the Faradaic hydrogenation of cyclohexene are
plained by the different reducing agents used. Whereas structure sensitive. The rate of hydrogenation was higher
reduction with H2 results in a long Rh–C distance, in ac- on Pt(110) or roughened surfaces.
cordance with earlier observations, reduction with NaBH4
The results reported in this work, however, indicate that
gives a short Rh–C distance, because the catalyst does not the effect of metal particle size on the activity in the liquid-
come into contact with molecular hydrogen. Thus, reduc- phase hydrogenation of cyclohexene is low. In our opinion,
tion with NaBH4 can be compared to evacuation of the other effects, such as the clustering of the support bodies in
catalyst at elevated temperature.
the liquid phase, the effect of water present in the system,
and the influence of oxygen-containing groups of the sur-
face, are much more important. It should be stressed that
all activities obtained in this study could be reproduced well
within 10%. The high activities of Rh/CNF(imp, c150, r250)
and Rh/CNF(IET, r200) are striking and reflect the fact that
many more factors than just metal particle size play a role
in liquid-phase reactions. As yet we do not have enough
knowledge about the system under investigation to give a
conclusive explanation.
4.4. Catalytic Test Experiments
The presence of internal diffusion limitations with re-
spect to H2 during the hydrogenation of cyclohexene with
the Rh/CNF catalysts is remarkable. With liquid-phase re-
actions external diffusion limitations would rather be ex-
pected to take place, because the stagnant layer around the
catalyst particle is large compared to the size of the parti-
cle. Furthermore, the carbon nanofibres have a mesoporous
macrostructure, the loading of active metal component is
low (1 wt%), and the concentration of cyclohexene is only
1% v/v. However, the reaction is executed at a H2 pressure
of only 1200 mbar and the catalysts turn out to be extremely
active in this test reaction. A simple solution would be to
increase the hydrogen pressure in order to increase the sol-
ubility of this gas in the solvent. Unfortunately, this remedy
was not possible with our experimental setup.
As was pointed out above, the activities of the catalysts
still qualitatively reflect the intrinsic activities. Figure 11
shows that, with the exception of Rh/CNF(imp, c150, r250)
andRh/CNF(IET, r200), noeffectofrhodiummetalparticle
size on catalytic activity is observed. As the rhodium metal
surface area decreases with increasing particle size, this
means that the turnover frequency (TOF) would increase
significantly with the rhodium particle size. The lack of ef-
fect of the particle size on the catalytic activity is quite sur-
prising, as cyclohexene hydrogenation is generally reported
to be a structure-insensitive reaction. With a structure-
insensitive reaction a decrease of activity should be found
with increasing particle size. Boudart and co-workers, for
instance, found that the liquid-phase hydrogenation of cy-
clohexene was independent of particle size for Rh/Al2O3
(49), Pd/SiO2 andPd/Al2O3 (50), andNi/SiO2 (51)catalysts.
However, Jackson and co-workers (52) did find a depen-
dence of metal crystallite size on the activity in cyclohex-
5. CONCLUSIONS
This study describes the synthesis of very small rhodium
metal particles on HNO3/H2SO4-oxidized fishbone carbon
nanofibres using several methods. The surface oxidation
with HNO3/H2SO4 is a prerequisite for a good interac-
tion between aqueous catalyst precursor solutions and the
CNFs. Particle sizes ranging from 1.1 to 2.1 nm are ob-
tained with 1 wt% Rh metal loadings. The smallest rhodium
particles result from impregnation, subsequent calcina-
tion at 150◦C, and reduction at 140◦C or from ion ex-
change at elevated temperature with an aqueous solution
of RhCl3 · 2H2O and reduction at 200◦C. The size of the par-
ticles makes investigation of the metal–support interaction
on carbon nanofibres possible. All catalysts are very active
in the hydrogenation of cyclohexene. Not much effect of
the metal particle size on the catalytic acitivity is observed,
suggesting that other parameters, such as the clustering of
support bodies in the liquid phase, are more important.
ACKNOWLEDGMENTS
We thank the HASYLAB (Hamburg, Germany) for the opportunity to
perform EXAFS measurements at beamline station X1.1. We are grateful
for the collection of the EXAFS data by the EXAFS measurement team.
This work was supported by the Netherlands’ Organisation for Scientific
ene hydrogenation for Pd on alumina, carbon, silica, and Research (NWO).

102
ROS ET AL.
30. Kapteijn, F., Marin, G. B., and Moulijn, J. A., in “Catalysis—An
REFERENCES
Integrated Approach to Homogeneous, Heterogeneous, and Indus-
trial Catalysis” (J. A. Moulijn, P. W. N. M. van Leeuwen, and R. A.
van Santen, Eds.), p. 251. Elsevier, Amsterdam, 1993.
31. Ros, T. G., van Dillen, A. J., Geus, J. W., and Koningsberger, D. C.,
Chem. Eur. J. 8, 1151 (2002).
32. Kip, B. J., Duivenvoorden, F. B. M., Koningsberger, D. C., and Prins, R.,
J. Catal. 105, 26 (1987).
33. Young, C. L., “Hydrogen and Deuterium,” Solubility Data Series,
Vol. 5/6. Pergamon, Oxford, 1981.
34. Fenelonov, V. B., Avdeeva, L. B., Zheivot, V. I., Okkel’, L. G.,
Goncharova, O. V., and Pimneva, L. G., Kinet. Catal. 34, 483
(1993).
35. Prinsloo, F. F., M. Sc. thesis. University of Cape Town, South Afrika,
2000.
1. de Jong, K. P., and Geus, J. W., Catal. Rev.–Sci. Eng. 42, 481 (2000).
2. Hoogenraad, M. S., Ph.D. thesis. Utrecht University, Utrecht, The
Netherlands, 1995.
3. Hoogenraad, M. S., Onwezen, M. F., van Dillen, A. J., and Geus, J. W.,
Stud. Surf. Sci. Catal. 101, 1331 (1995).
4. Geus, J. W., Hoogenraad, M. S., and van Dillen, A. J., in “Synthesis and
Properties of Advanced Catalytic Materials” (E. Iglesia, P. W. Lednor,
D. A. Nagaki, and L. T. Thompson, Eds.), p. 87. Materials Res. Soc.,
Pittsburgh, 1995.
5. Hoogenraad, M. S., van Leeuwarden, R. A. G. M. M., van Breda
Vriesman, G. J. B., Broersma, A., van Dillen, A. J., and Geus, J. W.,
in “Preparation of Catalysts VI” (G. Poncelet, et al., Eds.), p. 263.
Elsevier, Amsterdam, 1995.
6. Mojet, B. L., Hoogenraad, M. S., van Dillen, A. J., Geus, J. W., and
Koningsberger, D. C., J. Chem. Soc. Faraday Trans. 93, 4371 (1997).
7. Rodriguez, N. M., Kim, M.-S., and Baker, R. T. K., J. Phys. Chem. 98,
13108 (1994).
8. Chambers, A., Nemes, T., Rodriguez, N. M., and Baker, R. T. K.,
J. Phys. Chem. B 102, 2251 (1998).
9. Park, C., and Baker, R. T. K., J. Phys. Chem. B 102, 5168 (1998).
10. Park, C., and Baker, R. T. K., J. Phys. Chem. B 103, 2453 (1999).
11. Salman, F., Park, C., and Baker, R. T. K., Catal. Today 53, 385 (1999).
12. Pham-Huu, C., Keller, N., Charbonniere, L. J., Ziessel, R., and Ledoux,
M. J., Chem. Commun. 1871 (2000).
13. Planeix, J. M., Coustel, N., Coq, B., Brotons, V., Kumbhar, P. S.,
Dutartre, R., Geneste, P., Bernier, P., and Ajayan, P. M., J. Am. Chem.
Soc. 116, 7935 (1994).
36. Meyer, R. J., and Pietsch, E., “Gmelins Handbuch der Anorganischen
Chemie: Ruthenium/Rhodium,” 8th ed. Verlag Chemie G. M. B. H.,
Berlin, 1938.
37. Bukanova, A. E., Prokof’eva, I. V., and Shubochkin, L. K., Russ.
J. Inorg. Chem. 25, 875 (1980).
38. Frye, H., Luschak, C., and Chinn, D., Z. Naturforsch. B 22, 268 (1967).
39. See Ref. 36.
40. Bramley, R., Figgis, B. N., and Nyholm, R. S., J. Chem. Soc. A 861
(1967).
41. Dalla Batta, R. A., and Boudart, M., in “Proceedings, 5th International
Congress on Catalysis, Palm Beach, 1972” (J. W. Hightower, Ed.),
p. 1329. North Holland, Amsterdam, 1973.
42. Zou, W., and Gonzales, R. D., J. Catal. 133, 202 (1992).
43. van Zon, J. B. A. D., Koningsberger, D. C., van‘t Blik, H. F. J., and
Sayers, D. E., J. Chem. Phys. 82, 5742 (1986).
44. Koningsberger, D. C., Martens, J. H. A., Prins, R., Short, D. R., and
Sayers, D. E., J. Phys. Chem. 90, 3047 (1986).
45. Vaarkamp, M., Modica, F. S., Miller, J. T., and Koningsberger, D. C.,
J. Catal. 163, 294 (1996).
14. Ang, L.-M., Andy Hor, T. S., Xu, G.-Q., Tung, C.-H., Zhao, S., and
Wang, J. L. S., Chem. Mater. 11, 2115 (1999).
15. Keyser, M., and Prinsloo, F. F., submitted for publication.
16. Teunissen, W., Ph.D. thesis. Utrecht University, Utrecht, The
Netherlands, 2000.
17. Geus, J. W., Stud. Surf. Sci. Catal. 16, 1 (1983).
18. Ros, T. G., van Dillen, A. J., Geus, J. W., and Koningsberger, D. C.,
Chem. Phys. Phys. Chem. 3, 209 (2002 ).
19. Vaarkamp, M., Mojet, B. L., Modica, F. S., Miller, J. T., and
Koningsberger, D. C., J. Phys. Chem. 99, 16067 (1995).
20. Vaarkamp, M., et al., http://www.xsi.nl.
21. Vaarkamp, M., Dring, I., Oldman, R. J., Stern, E. A., and
Koningberger, D. C., Phys. Rev. B 50, 7872 (1994).
22. Cook, J. W., Jr., and Sayers, D. E., J. Appl. Phys. 52, 5024 (1981).
23. Koningsberger, D. C., Mojet, B. L., van Dorssen, G. E., and Ramaker,
D. E., Top. Catal. 10, 143 (2000).
24. Teo, B. K., and Lee, P. A., J. Am. Chem. Soc. 101, 2815 (1979).
25. Wyckhoff, R. W. G., “Crystal Structures,” Vol. 1, 2nd ed. Wiley,
New York, 1963.
46. Koningsberger, D. C., van Zon, F. B. M., Vaarkamp, M., and
Mun˜oz-Paez, A., in “X-Ray Absorption Fine Structure for Cata-
lysts and Surfaces” (Y. Iwasawa, Ed.), Series on Synchrotron Radi-
ation Techniques and Applications, Vol. 2, p. 257. World Scientific,
Singapore, 1996.
47. van Dorssen, G. E., Ph.D. thesis. Utrecht University, Utrecht, The
Netherlands, 1999.
48. Koningsberger, D. C., de Graaf, J., Mojet, B. L., Ramaker, D. E., and
Miller, J. T., Appl. Catal. A 191, 205 (2000).
49. Boudart, M., and Sajkowski, D. J., Faraday Discuss. 92, 57 (1991).
50. Gonzo, E. E., and Boudart, M., J. Catal. 52, 462 (1978).
51. Boudart, M., and Cheng, W.-C., J. Catal. 106, 134 (1987).
52. Jackson, S. D., Kelly, G. J., Watson, S. R., and Gulickx, R., Appl. Catal.
A 187, 161 (1999).
26. Coey, J. M. D., Acta Crystallogr. Sect. B 26, 1876 (1970).
27. Koningsberger, D. C., Jpn. J. Appl. Phys. 32, 877 (1993).
28. Stern, E. A., Phys. Rev. B 48, 9825 (1993).
29. Voogt, E. H., “THEBE, BOLLIE and NUBIE Computer Program.”
Surface Science Div., Debye Inst., Utrecht Univ., Utrecht, The Nether-
lands, 1996.
53. Busser, G. W., van Ommen, J. G., and Lercher, J. A., Adv. Catal. Nanos-
truct. Mater. 213 (1996).
54. Augustine, R. L., Thompson, M. M., and Doran, M. A., J. Chem. Soc.
Chem. Commun. 1173 (1987).
55. Mu¨ller, U., Schmiemann, U., Du¨lberg, A., and Baltruschat, H., Surf.
Sci. 335, 333 (1995).