Journal of Colloid and Interface Science 266 (2003) 183–194

www.elsevier.com/locate/jcis

Catalyst support effects: gas-phase hydrogenation

of phenol over palladium
Colin Park 1 and Mark A. Keane ∗
Department of Chemical and Materials Engineering, University of Kentucky, Lexington, KY 40506-0046, USA
Received 12 November 2002; accepted 6 February 2003

Abstract
The catalytic action of 10% w/w Pd supported on two forms of graphitic carbon nanofibers (GCN) has been assessed and compared with the
performance of 10% w/w Pd on SiO2 , Ta2 O5 , activated carbon (AC), and graphite. Palladium nitrate served as metal precursor in each case but
the role of the starting metal salt was also considered by examining the action of palladium acetate impregnated SiO2 . The activated catalysts
have been characterized by hydrogen chemisorption, high-resolution transmission electron microscopy, and scanning electron microscopy.
Phenol hydrogenation served as the test reaction, which proceeds in a stepwise fashion involving the partially hydrogenated cyclohexanone
as a reactive intermediate. The occurrence and ramifications of Pd/support interaction(s) are related to hydrogenation activity and selectivity.
The effects of contact time and reaction temperature (398–448 K) are reported and discussed in terms of phenol/catalyst interaction(s).
Hydrogenation kinetics have been adequately represented by a standard pseudo-first-order approximation. The specific activities exhibited
the following sequence of increasing values: Pd/AC < Pd/GCN < Pd/SiO2 ∼ Pd/graphite < Pd/Ta2 O5 . A diversity of product composition
responses to variations in reaction conditions points to the involvement of Pd particle size distribution, Pd particle geometry, and electronic
character in determining overall catalytic behavior.
 2003 Elsevier Inc. All rights reserved.

Keywords: Graphitic carbon nanofibers; Phenol hydrogenation; Palladium; Catalyst support; Selective hydrogenation

1. Introduction drogen treatment represents immediate savings in terms of

fuel and/or chemical recovery. Phenol as a pollutant typi-
An overview of the recent literature (from 1996 on- cally arises in high concentrations in aqueous media [21,22]
ward) on selective catalytic hydrogenation reveals, if not an and, for this reason, we have examined the catalytic trans-
avalanche, certainly a proliferation of papers that describe formation of an aqueous phenol feedstock.
the selective hydrogenation of phenol to cyclohexanone The hydrogenation of phenol in gas [2,3,6,8,13,23–29]
[1–16]. The reason for this upsurge in interest lies principally and liquid [30–32] phases has been reported for a range
in the fact that cyclohexanone is of commercial significance of palladium- [2–4,6,8,10,24,25,27,30–32], platinum- [25,
as a key raw material in the production of both caprolac- 26,29] and nickel- [5,7,10,33] based catalysts. Reaction se-
tam for nylon 6 and adipic acid for nylon 66 [17]. Phenol is, lectivity, in terms of cyclohexanone production, is strongly
on the other hand, an established environmental toxin [18] dependent on reaction conditions [4,5,8,25,27] and, under a
arising from a variety of industrial sources associated with narrow range of conditions, selectivities in excess of 95%
petrochemicals and polymer manufacture [19]. A catalytic have been quoted in the literature [3,9,27,31]. Both activity
hydrogen treatment of such waste to generate reusable raw and selectivity are sensitive to changes in catalyst structural
material is certainly preferable to the standard destructive properties where metal loading [2,4,9,23,24,26,28], surface
route by incineration [20]. A move from incineration to hy- acidity/basicity [5,9,24,25], and the inclusion of promoters
(alkali metals, alkaline earth metals, and lanthanides) [8,
* Corresponding author.
10,24,32] have all been found to impact on catalyst per-
E-mail address: makeane@engr.uky.edu (M.A. Keane).
formance. In order to optimize partial hydrogenation and
1 Current address: Synetix, P.O. Box 1, Belasis Avenue, Billingham, achieve high cyclohexanone selectivity, it has been typically
Cleveland, TS23 1LB, UK. necessary to sacrifice phenol consumption rates as higher ac-
0021-9797/$ – see front matter  2003 Elsevier Inc. All rights reserved.
doi:10.1016/S0021-9797(03)00171-1
184 C. Park, M.A. Keane / Journal of Colloid and Interface Science 266 (2003) 183–194

tivities favor complete hydrogenation to cyclohexanol. The

attainment of a high cyclohexanone selectivity (>95%) at el-
evated conversions (>80%) remains a challenging catalysis
problem.
The use a support to disperse a catalytically active
metal is not only of economic benefit (lower preparation
costs/longer productive lifetime) but the substrate can also
influence catalyst performance through electronic interac-
tions [34–36]. Indeed, the occurrence of “strong metal/sup-
port interactions” (SMSI) and associated catalytic implica-
tions has been the subject of three widely cited reviews
[37–39]. The role of the catalyst support in phenol hydro-
genation has not yet been the subject of a comprehensive
study. In this paper, we focus on supported Pd as the cat-
alytic agent and consider the use of catalytically generated
graphitic carbon nanofibers (GCN) as novel Pd support ma-
terial. The use of structured carbon (nanotubes/nanofibers)
to modify the catalytic behavior of supported metal is now
beginning to attract the interest of the catalysis research
community [40,41]. In this study, we have also examined
the action of SiO2 , Ta2 O5 , activated carbon, and graphite
as support materials. These supports present a range of sur-
face areas and energetics (refractory insulating oxides and
electrically conducting graphitic carbon) that can facilitate
an assessment of the influence of Pd/support interaction(s)
on phenol hydrogenation activity/selectivity. In addition, we Fig. 1. SEM image showing structural features associated with GCN-2.
have considered the possible effect(s) of varying the nature
of the catalyst precursor by examining the action of Pd/SiO2
prepared from palladium nitrate and palladium acetate. necessary to avoid any contribution to phenol hydrogena-
tion from residual Ni and/or Cu; the commercial activated
carbon and graphite samples were also subjected to a dem-
2. Experimental ineralization. The carbonaceous supports were thoroughly
washed with deionized water (until pH approached 7) and
2.1. Catalyst preparation, activation, and characterization oven dried at 383 K for 12 h, and the GCN samples were sub-
jected to a partial oxidation (in a 5% v/v O2 /He mixture) at
The SiO2 (fumed) and Ta2 O5 substrates were supplied 673 K for 2 h to remove the amorphous carbon content. The
by Sigma-Aldrich and used as received. The activated car- carbon nanofibers were characterized by scanning electron
bon (G-60, 100 mesh) was obtained from NORIT (UK) microscopy (SEM), employing a Hitachi S900 field emis-
and the graphite (synthetic 1–2 µm powder) from Sigma- sion SEM, operated at an accelerating voltage of 25 kV;
Aldrich. Two novel graphitic carbon nanofiber supports the sample was deposited on a standard aluminum SEM
were synthesized by the catalytic decomposition of ethyl- holder and coated with gold. The fibrous nature of the carbon
ene, a synthetic route that we have described in some detail growth (with aspect ratios of up to 106 ) is immediately evi-
elsewhere [42–44]. The catalytic GCN growth was pro- dent from the representative SEM image presented in Fig. 1.
moted using unsupported Ni and Ni/Cu catalysts, prepared The Pd loaded (10 ± 0.2% w/w) samples were pre-
by standard precipitation/deposition where the precipitate pared by standard incipient wetness impregnation where a
was thoroughly washed with deionized water, oven dried at 2-butanolic Pd(NO3 )2 (or additionally Pd(C2 H3 O2 )2 in the
383 K overnight, calcined in air at 673 K for 4 h, and then case of SiO2 ) solution was added dropwise at 353 K to
reduced at 723 K in 20% v/v H2 /He for 20 h. The fibers the substrate with constant agitation (500 rpm). Aqueous
generated, using the unsupported Ni catalyst, from a 1/1 v/v solutions were not employed in this study as carbon sup-
C2 H4 /H2 reactant feed at 823 K and a gas hourly space ve- port materials are known to possess hydrophobic properties
locity (GHSV) of 11,300 h−1 are denoted here as GCN-1. leading to difficulties with surface wetting that may ad-
The second set of fibers (GCN-2) were grown from the versely affect the ultimate Pd dispersion. The Pd content
bimetallic Ni/Cu at 848 K employing a 4/1 v/v C2 H4 /H2 feed was determined (to within ±2%) by atomic absorption spec-
at the same GHSV. The catalytically generated GNF prod- trophotometry (VarianSpectra AA-10); samples for analysis
uct was contacted with dilute mineral acid (1 M HNO3 ) for were digested in HF (37% concn.) overnight at ambient tem-
7 days to remove the parent catalyst particles. The latter was perature. The catalyst precursors were subsequently dried
 C. Park, M.A. Keane / Journal of Colloid and Interface Science 266 (2003) 183–194 185

overnight at 383 K and activated by heating (10 K min−1 , factor >0.99). Heat transport effects can also be disregarded
controlled using a Eurotherm 91e temperature programmer) when applying the criteria set down by Mears [47]; the tem-
in 60 cm3 min−1 (monitored using a Humonics Model 529 perature differential between the catalyst particles and bulk
digital flow meter) dry H2 (99.999% v/v) to 523 K, and fluid phase was <1 K. In a series of blank experiments,
this temperature was maintained for 2 h. Nitrogen BET passage of aqueous solutions of phenol in a stream of hydro-
surface area measurements were carried out at 77 K (Mi- gen through the empty reactor or over each of the support
cromeritics TriStar) on freshly reduced samples. Samples materials did not result in any detectable conversion. The
for off-line TEM analysis were cooled (in He) and pas- catalytic data quoted herein represent steady-state values ob-
sivated in a 2% v/v O2 /He mixture at room temperature. tained in the absence of any significant long-term catalyst
The latter step provided a protective oxide layer over the deactivation; repeated catalytic runs generated results that
Pd surface that prevented bulk oxidation upon exposure to were reproducible to within ±7%. The reactor effluent was
the atmosphere. High-resolution transmission electron mi- frozen in a liquid nitrogen trap for subsequent analysis which
croscopy (HRTEM) analysis was carried out using a Philips was made using an AI Cambridge GC94 chromatograph
CM200 FEGTEM microscope equipped with a UTW energy equipped with a flame ionization detector and employing a
dispersive X-ray (EDX) detector (Oxford Instruments) and DB-1 50 m × 0.20 mm i.d., 0.33-µm capillary column (J&W
operated at an accelerating voltage of 200 kV. The specimens Scientific), as described previously [3]. Quantitative analy-
were prepared by ultrasonic dispersion in 2-butanol, evapo- sis was based on relative peak area with acetone as solvent
rating a drop of the resultant suspension onto a holey carbon where analytical repeatability was better than ±0.1% and the
support grid. Selected area electron diffraction (SAED) con- detection limit typically corresponded to a feedstock conver-
firmed that the Pd distributed over each support was present sion less than 0.1 mol%. The overall fractional conversion of
in the metallic form and not as an oxide. The Pd particle phenol (xphenol) is given by
size distribution profiles (and mean Pd sizes) presented in
[phenol]in − [phenol]out
this study are based on a measurement of over 500 individ- xphenol = (2)
ual particles. The specific Pd surface area (SPd , m2 g−1 [phenol]in
Pd ) was
calculated from the relationship [45] while reaction selectivity (as a percentage) in terms of cy-
6 clohexanone formation (SC=O ) can be represented by
SPd = , (1)
ρdPd [C=O]out
SC=O = × 100 (3)
where dPd is the mean particle size and ρ is the Pd specific [phenol]in − [phenol]out
mass (11.97 g cm−3 ). Palladium dispersion was also mea- and percentage cyclohexanone yield (YC=O %) is given by
sured by pulse H2 chemisorption, employing the commercial
[C=O]out
Quantachrome ChemBET 3000 unit; dispersion values were YC=O = × 100, (4)
reproducible to better than ±5%. [phenol]in
where [ ] denotes concentration and in and out refer to the
2.2. Catalytic reactor system reactant/product(s) entering and exiting the reactor. All the
reactants were Analar grade and were used without further
Each catalyst precursor was activated as above in a fixed purification.
bed glass reactor: ratio of catalyst particle to reactor diameter
= 90. The catalytic reactor approximated plug flow behav-
ior and has been fully described previously [5,7] but some 3. Results and discussion
details, pertinent to this study, are given below. A Model
100 (kd Scientific) microprocessor-controlled infusion pump 3.1. Catalyst characterization
was used to deliver aqueous solutions of phenol at a fixed
calibrated flow rate, carried through the catalyst bed in a The Pd-based catalysts considered in this study are listed
stream of dry H2 . The phenol molar feed rate was varied in Table 1, which includes TEM and H2 chemisorption-
from 5.2 × 10−4 to 9.4 × 10−3 h−1 in order to test adherence derived average Pd particle diameters and BET surface
to pseudo-first-order kinetics where H2 was maintained at areas. Representative Pd particle size distributions, based
least 12 times in excess of stoichiometric quantities; the gas- on TEM analysis, are presented in the histograms given
space velocity was kept constant at 3620 h−1 . The catalyst in Fig. 2. The activated Pd/SiO2 catalysts derived from
was supported on a glass frit and a layer of glass beads above the nitrate and acetate precursors, denoted Pd/SiO2 -A and
the catalyst bed served as a preheating zone to ensure that Pd/SiO2 -B, respectively, exhibit similar size distributions,
the reactants were vaporized and reached the reaction tem- albeit the average Pd diameter delivered by the nitrate pre-
perature (398–448 K) before contacting the catalyst; T was cursor is slightly larger. The low-resolution TEM images
constant to within ±1 K. The reactor was operated, using the provided in Fig. 3 serve to illustrate the nature of the
criteria outlined elsewhere [46], with negligible internal or metal dispersion. The Pd phase supported on SiO2 exhibits
external diffusion retardation of reaction rate (effectiveness a spherical morphology (Fig. 3a), suggestive of relatively
186 C. Park, M.A. Keane / Journal of Colloid and Interface Science 266 (2003) 183–194

Table 1
Nitrogen BET surface areas, surface-weighted average Pd particle sizes based on H2 chemisorption (dH ) and TEM (dTEM ) measurements and the associated
specific phenol hydrogenation pseudo-first-order rate constants (k): T = 398 K
Catalyst N2 BET surface area dTEM (nm) dH (nm) 104 k
(m2 g−1 ) (mol h−1 (m2 gPd −1 )−1 )
Pd/SiO2 -Aa 148 8.3 9.7 1.7
Pd/SiO2 -Bb 156 7.8 9.3 1.6
Pd/Ta2 O5 15 11.6 12.9 2.5
Pd/AC 865 16.7 15.9 0.4
Pd/graphite 18 16.0 17.3 1.8
Pd/GCN-1 102 6.9 8.5 0.5
Pd/GCN-2 133 6.4 8.8 1.1
a Pd(NO ) precursor.
3 2
b Pd(C H O ) precursor.
2 3 2 2

and catalyst composition [57] and there have been a num-

ber of reported instances [9,58,59] where the dimensions
of supported metal particles that are measured differ, de-
pending on the analytical technique that is employed. It is,
however, usually the case that each characterization tool
delivers the same trends for a family of metal catalysts.
The latter holds in this study where the order of increasing
particle size was the same for both techniques and fol-
lows the sequence Pd/GCN-2 ∼ Pd/GCN-1 < Pd/SiO2 -B ∼
Pd/SiO2 -A < Pd/Ta2 O5 < Pd/graphite ∼ Pd/AC.
The Pd phase supported on the conventional graphite and
activated carbon substrates exhibited a wider size distrib-
ution when compared with the other supports, diagnostic
of significant Pd mobility/agglomeration during the reduc-
tion step. Graphite has a low (BET) surface area with few
Fig. 2. Palladium particle size distributions associated with freshly activated
edge positions available for depositing any supported metal
Pd/SiO2 -A (a), Pd/AC (b), Pd/GCN-1 (c), and Pd/GCN-2 (d).
with the result that the supported metal phase is typically
present in the form of larger particles [60]. Reduction of
weak metal/support interaction. The size distribution and the Pd-impregnated activated carbon (AC), an essentially
shape(s) of the supported Pd particles are inherent features amorphous material with a high (BET) surface area, yielded
of the interfacial energies associated with each system [48]. a metal phase of comparable dimensions (Table 1) to that
To date, the reported use of the highly refractive Ta2 O5 associated with Pd/graphite but with a broader size range.
oxide as a metal support has been limited and the exist- The Pd particles exhibit an indistinct or globular geometry
ing catalysis-related research has focused on its application (Fig. 3b) that has been shown to be diagnostic of limited
in selective oxidation as applied to pollution remediation metal/support interaction [61]. Carbon fibers are generally
[49,50]. Tantalum oxide is, however, known to possess (to classified as graphitic structures, characterized by a series
some degree) many of the qualities of TiO2 [51,52] with of ordered parallel graphene layers arranged in specific con-
evidence of SMSI [53,54]. The Pd phase supported on formations with an interlayer distance of ca. 0.34 nm [62].
the low surface area Ta2 O5 substrate is characterized by Carbon fiber synthesis is possible by arc discharge and
a symmetrical size distribution and a surface weighted av- plasma decomposition but such methodologies also yield
erage size that is greater than that recorded for Pd/SiO2 . polyhedron carbon nanoparticles (low aspect ratio) and an
In two related studies [55,56], it was shown that the Ni appreciable amorphous carbon component [63,64]. The syn-
phase supported on TiO2 was in the form of significantly thesis of carbon nanofibers via catalytic vapor deposition
larger crystallites than that with SiO2 . It should, however, delivers, on the whole, a more uniform carbon product with
be noted that there is a decided mismatch in the Pd parti- a more feasible scale-up. The carbon product can be tailor-
cle sizes obtained from hydrogen chemisorption (dH ) with made to desired specifications, in terms of lattice orienta-
those derived from TEM analysis (dTEM), as revealed in tion and mechanical/chemical/electrical properties, through
Table 1; dTEM is typically lower than dH . Such discrep- a judicious choice of both catalyst and reaction condi-
ancies suggest some deviation from an exclusive 2:1 H:Pd tions [65–67]. Representative TEM images that illustrate the
adsorption stoichiometry, as is normally applied. The surface structural characteristics of GCN-1 and GCN-2 are shown
stoichiometry is known to be dependent on metal loading in Fig. 4 along with simple schematic representations of the
 C. Park, M.A. Keane / Journal of Colloid and Interface Science 266 (2003) 183–194 187

(a) (b)

(c)

Fig. 3. Representative low-resolution TEM images of (a) Pd/SiO2 -A, (b) Pd/AC, and (c) Pd/GCN-1.

idealized structures as a visual aid. The GCN-1 product was The commonly accepted mechanism [68] involves reactant
predominantly of the “ribbon” form (Fig. 4a) with diame- (C2 H4 as the carbon source in this case) decomposition
ters in the range 50–125 nm (mean diameter = 85 nm) and on the top surface of a metal particle followed by a diffu-
lengths of up to 100 mm. In the ribbon form, the graphene sion of carbon atoms into the metal with precipitation at
layers are oriented parallel to the growing fiber axis [65, other facets of the particle to yield the fiber, which con-
67]. An alternative lattice arrangement, the “herringbone” tinues to grow until the metal particle becomes poisoned
or “fishbone” [65] structure (Fig. 4b) characterizes GCN-2, or completely encapsulated by carbon. The diameter of the
where the parallel carbon platelets are oriented at an angle nanofiber is governed by the dimensions of the seed metal
to the fiber axis. The diameter of GCN-2 extended from particle while the length depends largely on the duration of
50 to 150 nm (mean diameter = 105 nm) with an overall reaction [42,44]. The degree of crystalline order of the car-
length of up to 200 mm. The structure of the catalytically bon product is controlled by various factors including the
generated carbon fiber is largely governed by the crystallo- wetting properties of the metal with graphite and the crys-
graphic orientation of the exposed catalytic metal face(s). tallographic orientation of the metal faces that are in contact
188 C. Park, M.A. Keane / Journal of Colloid and Interface Science 266 (2003) 183–194

(a)

(b)

Fig. 4. Schematic diagrams illustrating an idealized rendering of (a) “ribbon” (GCN-1) and (b) “herringbone” (GCN-2) configurations with representative
TEM images.

with the carbon deposit, features that are ultimately reliant as opposed to conventional graphite where the basal plane
upon the choice of catalyst [66,69,70]. It should be noted is available. Representative high-resolution TEM images are
that the TEM images presented in Fig. 4 correspond to the given in Fig. 5 that illustrate the lattice structure of isolated
GCN product taken directly from the reactor and the pres- Pd particles supported on GCN-1 (lattice fringes also evi-
ence of an amorphous carbon layer on the fiber edges is dent). Routine TEM analysis (Fig. 3c) revealed that the Pd
an artifact of the cooling stage, upon completion of the cat- particles were largely located along the GCN edge sites. Pal-
alytic step. A careful oxidation treatment (see Experimental) ladium supported on both GCN substrates is characterized
was employed to remove this amorphous carbon overlayer; (Fig. 2) by a narrow size distribution (>80% of the parti-
sample weight losses of ca. 5% were recorded during this cles <10 nm) and the smallest average particle diameters
mild oxidative step. The GCN fibers exhibited N2 BET sur- of all the supported Pd catalysts. Pham-Huu et al. [71] also
face areas that are intermediate in magnitude (Table 1) when noted a narrow (centered around 3–5 nm) size distribution
compared with the standard amorphous and graphitic carbon for a 5% w/w Pd loading on comparable GCN material.
supports. The “herringbone” GCN-2 possessed an overall This enhanced dispersion has been attributed to relatively
higher surface area due to the greater number of accessible strong metal/GCN interactions that limit particle growth dur-
edge sites in this more open structure. The GCN supports ing activation [71–73]. The highly crystalline faceted and
provide exposed edge sites for anchoring the Pd particles relatively thin Pd structures shown in Fig. 5 are morpho-
 C. Park, M.A. Keane / Journal of Colloid and Interface Science 266 (2003) 183–194 189

Fig. 6. Series/parallel reaction scheme for the hydrogenation of phenol to

cyclohexanol.

benzene formation (hydrodehydroxylation). Moreover, in

preliminary studies the transformation of methanolic phe-
nol solutions yielded anisole as a major product, formed
through a catalyzed methylation step. The catalytic study
was accordingly limited to an aqueous phenol feed (with va-
porization prior to reaction) at temperatures less than 473 K
in order to asses the intrinsic hydrogenation activity of these
catalysts. Furthermore, there was no detectable condensation
reactions to form dicyclohexylether as has been reported for
Pd on acidic supports [25,75]. A sequential hydrogenation
of phenol has been noted in a number of instances [4,5,7,8,
26] and the reaction can proceed through the series/parallel
scheme shown in Fig. 6 where cyclohexanone serves as a
reactive intermediate and cyclohexanol is the ultimate prod-
uct. Plug flow operation, under steady state, where hydrogen
was maintained far in excess yields the following applicable
reactor/kinetic expression


−1 W
ln(1 − xphenol) = k , (5)
Fphenol
where Fphenol is the inlet molar feed rate (mol h−1 ) and
xphenol is the fractional conversion for a given catalyst
weight (W ); the parameter W/Fphenol (units, g h mol−1 ) has
the physical significance of contact time. From a consid-
eration of gas-phase reaction equilibrium constants, both
(partial and complete) hydrogenation steps are thermody-
namically favored [5]. Over the range of inlet phenol partial
pressures considered in this study, complete conversion un-
der equilibrium conditions is achieved at T
443 K with a
continual drop thereafter to a residual conversion at 573 K.
The catalytic data were far removed from gas-phase equi-
Fig. 5. High-resolution TEM images of isolated Pd particles supported on
Pd/GCN-1. librium conversions and the response of activity to changes
in temperature can be attributed positively to surface reac-
tion phenomena. Application of pseudo-first-order kinetics
logical features that are consistent with the existence of is a practical approach to extract meaningful rate constants
strong interaction between the metal particles and the sup- from conversion data and is employed here as a means of
port medium [74]. The diversity of surface structural features assessing the hydrogenation activity of each supported Pd
associated with these catalysts suggests the possibility of dif- catalyst. The linear relationships between ln(1 − xphenol)−1
ferent (preponderance of) Pd crystallographic faces exposed and W/Fphenol , generated for reactions over representative
to the reactant that can influence overall catalyst perfor- catalysts, shown in Fig. 7, are diagnostic of adherence to
mance. pseudo-first-order kinetics. A least-squares fit, forced to go
through the origin, was used to determine rate constants
3.2. Phenol hydrogenation which were converted to specific values using Eq. (1) and
are recorded in Table 1. We have adopted the Pd dis-
Cyclohexanone and cyclohexanol were the only observed persion/size values derived from H2 chemisorption that,
reaction products in the gas-phase reaction of aqueous phe- although differ from the values extracted from the TEM
nol solutions over each supported Pd catalyst. At tempera- analysis, better represent the nature of the catalytically ac-
tures in excess of ca. 473 K there was evidence of significant tive surface in terms of adsorption/reactant activation. The
190 C. Park, M.A. Keane / Journal of Colloid and Interface Science 266 (2003) 183–194

Table 2
Cyclohexanone selectivities (SC=O ) at comparable phenol fractional con-
versions (xphenol ) over each catalyst: T = 423 K
Catalyst W/Fphenol (g mol−1 h) xphenol SC=O (%)
Pd/SiO2 -Aa 32 0.12 66
Pd/SiO2 -Bb 32 0.10 74
Pd/Ta2 O5 16 0.16 92
Pd/AC 64 0.13 96
Pd/graphite 24 0.07 74
Pd/GCN-1 16 0.07 26
Pd/GCN-2 24 0.11 56

Pd/SiO2 -Aa 96 0.30 58

Pd/SiO2 -Bb 96 0.29 62
Pd/Ta2 O5 64 0.32 91
Pd/AC 192 0.33 93
Pd/graphite 128 0.35 66
Fig. 7. Pseudo-first-order kinetic relationships for the hydrogenation of Pd/GCN-1 64 0.25 24
phenol over Pd/SiO2 -A (Q), Pd/AC ("), and Pd/Ta2 O5 (F): T = 423 K. Pd/GCN-2 64 0.28 44

Pd/SiO2 -Aa 192 0.50 49

use of TEM (even at high resolution) to determine metal Pd/SiO2 -Bb 192 0.47 54
particle sizes requires judgment where insufficient contrast Pd/Ta2 O5 128 0.54 87
Pd/AC 385 0.55 90
in the image can hamper an accurate analysis, particu- Pd/graphite 192 0.45 59
larly for diameters of 1 nm and less. The resultant specific Pd/GCN-1 192 0.57 21
phenol consumption activities exhibit the following trend Pd/GCN-2 128 0.49 33
of increasing values: Pd/AC ∼ Pd/GCN-1 < Pd/GCN-2 < a Pd(NO ) precursor.
3 2
Pd/SiO2 -B ∼ Pd/SiO2 -A ∼ Pd/Graphite < Pd/Ta2 O5 . The b Pd(C H O ) precursor.
2 3 2 2
Pd/SiO2 -A and Pd/SiO2 -B catalysts delivered essentially the
same specific activity, suggesting that the Pd precursor has
little influence on the ultimate metal site efficiency. There
is, however, a decided support effect with a sixfold dif- the lowest specific rate. Bessel et al. have recently reported
ference between the most and least active catalysts. The [74] quite distinct electrocatalytic activities associated with
activity sequence cannot be explained merely on the ba- “ribbon” and “herring bone” GCN forms, suggesting dif-
sis of a particle size effect as earlier work has revealed an ferent electron-donating/withdrawing properties associated
invariant specific rate where the average Pd particle size with the two GCN systems. From a consideration of our re-
(on MgO and SiO2 ) exceeded ca. 2 nm [9,28]. However, sults, the “ribbon” fiber is the less effective support medium
where the Pd particle size was lower than ca. 2 nm there in terms of specific phenol consumption rate.
was an appreciable decrease in reactant turnover [8]. The The product compositions generated from each supported
lower phenol conversions delivered by Pd/AC and Pd/GCN Pd catalyst were essentially independent of time-on-stream
may be attributable (to some degree) to a greater presence (up to 6 h) and the values reported in this paper represent
of smaller Pd particles (0–2.5 nm), as revealed in Fig. 2. It steady-state conversions. The cyclohexanone selectivities
has been established [9,10,32] that the d-character of sup- (SC=O ) at three representative phenol fractional conversions
ported Pd governs, to a great extent, phenol hydrogenation (xphenol = 0.12 ± 0.05, 0.29 ± 0.04, and 0.51 ± 0.06) are
activity where electron-deficient Pd sites have a lower in- compared for each catalyst at the same reaction tempera-
herent turnover rate. Electron transfer between the support ture in Table 2. In every instance an increase in xphenol was
medium and the Pd crystallites can be invoked to explain accompanied by a decrease (to varying degrees) in cyclo-
the observed catalytic phenomena where the electrically hexanone selectivity. The antisympathetic relationship be-
conductive substrates can induce electronic perturbations in tween xphenol and SC=O is to be expected in a sequential hy-
the metal particles. While support-induced electronic effects drogenation scheme where complete hydrogenation should
are considered to be negligible for larger (ca. 10–40 nm) be enhanced at higher xphenol . Similar values of xphenol were
metallic particles [55], the Pd dispersions associated with achieved for each catalyst by varying the W/Fphenol para-
the catalysts considered in this study are such that SMSI meter. The Pd/AC and Pd/Ta2 O5 catalysts delivered very
can contribute directly to the metal site activity. The higher high SC=O values that were largely insensitive to xphenol . At
activities associated with Pd/Ta2 O5 and Pd/graphite may the other extreme, Pd/GCN-1 is characterized by apprecia-
then be ascribed to significant SMSI that enhances the hy- bly lower SC=O that was likewise relatively unaffected by
drogenation efficiency of the supported Pd sites. The high variations in xphenol. The two Pd/SiO2 and Pd/graphite cat-
surface area AC exhibits little or no metal/support interac- alysts exhibited very similar activity/selectivity trends while
tion and it is significant that the Pd/AC sample delivered Pd/GCN-2 was less selective in terms of partial hydrogena-
 C. Park, M.A. Keane / Journal of Colloid and Interface Science 266 (2003) 183–194 191

Table 3
Rate constants for the partial (k1 and k2 ) and complete (k3 ) hydrogenation (see Fig. 6) of phenol over each supported Pd catalyst: T = 423 K
Catalyst k1 (mol g−1 h−1 ) k2 (mol g−1 h−1 ) k3 (mol g−1 h−1 ) k1 /k3
Pd/SiO2 -Aa 2.5 × 10−3 3.1 × 10−3 1.1 × 10−3 2.1
Pd/SiO2 -Bb 2.5 × 10−3 2.8 × 10−3 9.7 × 10−4 2.5
Pd/Ta2 O5 5.3 × 10−3 1.4 × 10−3 6.9 × 10−4 7.7
Pd/AC 2.1 × 10−3 6.0 × 10−5 6.1 × 10−5 36.5
Pd/graphite 2.4 × 10−3 2.0 × 10−3 6.9 × 10−4 3.5
Pd/GCN-1 1.1 × 10−3 2.3 × 10−3 3.3 × 10−3 0.3
Pd/GCN-2 2.7 × 10−3 7.1 × 10−3 2.4 × 10−3 1.1
a Pd(NO ) precursor.
3 2
b Pd(C H O ) precursor.
2 3 2 2

tion but the associated SC=O values were greater than those
recorded for Pd/GCN-1.
Applying pseudo-first-order conditions at quasi-steady
state for irreversible hydrogenation steps, three differential
equations describe the system


d[phenol]
Fphenol = −(k1 + k3 )[phenol], (6)
dW


d[C=O]
Fphenol = k1 [phenol] − k2 [C=O], (7)
dW


d[C–OH]
Fphenol = k3 [phenol] + k2 [C=O]. (8)
dW
Integration of these rate equations provides the following
dependence of product mol fraction (α) on space time
  Fig. 8. Experimental (symbols) and calculated (lines: based on a se-
W
αphenol = exp −(k1 + k3 ) , (9) ries/parallel scheme) product molar compositions (α) resulting from the
Fphenol hydrogenation of phenol over Pd/SiO2 -A (Q), Pd/Ta2 O5 (F), Pd/AC ("),


  Pd/GCN-1 (a), and Pd/GCN-2 (e): T = 423 K.
k1 W
αC=O = exp −(k1 + k3 )
k2 − (k1 + k3 ) Fphenol
  substituent [24]; the former is more likely to result in com-
W
− exp −k2 , (10) plete hydrogenation [7]. A one site dissociative adsorption of
Fphenol
both phenol and hydrogen on palladium has been proposed
αC–OH = 1 − αphenol − αC=O . (11)
by Chen et al. [4] to favor high cyclohexanone selectivi-
Taking the experimentally determined α values at various ties. In the case of interaction through the aromatic nucleus,
contact times (W/Fphenol ), the resultant nonlinear equations the ring may be coplanar with [76] or perpendicular to
were solved to yield values for the stepwise rate constants. [77] the surface. Cyclohexanone production can follow three
The applicability of this scheme can be assessed in Fig. 8 routes: (a) direct hydrogenation or addition of two hydro-
wherein representative experimental and calculated prod- gen molecules/four hydrogen atoms; (b) hydrogenation to
uct compositions are presented: the agreement is uniformly cyclohexanol followed by a dehydrogenation step; (c) for-
good. The rate constants for the partial (k1 and k2 ) as mation of cyclohexene-1-ol followed by a tautomerism. The
opposed to complete (k3 ) hydrogenation of phenol are pre- stronger the interaction of phenol with the surface the greater
sented in Table 3. The value of the k1 /k3 ratio signifies the likelihood of complete hydrogenation to cyclohexanol.
the relative importance of the two possible hydrogenation It is not possible, from the combination of catalyst char-
pathways, stepwise and concerted, respectively. The high acterization and catalysis data presented in this paper, to
value of this rate constant ratio recorded for the less ac- conclusively identify the predominant surface arrangement
tive Pd/AC is to be expected as are the significantly lower that leads to the observed cyclohexanone selectivity and how
values associated with both Pd/SiO2 catalysts. The surface this is dependent on reaction conditions. Certain features
requirement(s) for partial as opposed to complete phenol hy- should, however, be flagged that are suggestive of a sup-
drogenation has not been established in the literature and port effect in modifying phenol hydrogenation selectivity.
the reaction mechanism is not fully understood. Depend- While Pd/Ta2 O5 and Pd/graphite delivered the highest spe-
ing on the nature of the catalyst, phenol can interact with cific phenol consumption rate constants (k1 +k3 , see Table 1)
the surface through the aromatic ring and/or the hydroxyl the associated k1 /k3 ratios indicate a greater relative impor-
192 C. Park, M.A. Keane / Journal of Colloid and Interface Science 266 (2003) 183–194

Fig. 10. Cyclohexanone yield (YC=O ) as a function of temperature in the

Fig. 9. Relationship between cyclohexanone selectivity (SC=O ) and frac-
conversion of phenol over Pd/SiO2 -A (Q), Pd/SiO2 -B (P), Pd/Ta2 O5 (F),
tional phenol conversion (xphenol ) over Pd/SiO2 -A (Q), Pd/SiO2 -B (P),
Pd/AC ("), Pd/graphite (2), Pd/GCN-1 (a), and Pd/GCN-2 (e):
Pd/Ta2 O5 (F), Pd/AC ("), Pd/graphite (2), Pd/GCN-1 (a), and Pd/
W/Fphenol = 128 g mol−1 h.
GCN-2 (e) where temperature was varied: W/Fphenol = 128 g mol−1 h.

by reactant [7,28]. In contrast to the trends shown in Fig. 8,

tance of the partial hydrogenation step when compared with the reduction in temperature that resulted in increased phe-
the less active Pd/SiO2 . Moreover, both Pd/GCN catalysts, nol conversion served to enhance the selectivity with which
while less active than Pd/Ta2 O5 , Pd/graphite, and Pd/SiO2 , the partially hydrogenated cyclohexanone was produced.
exhibit a significantly greater tendency to promote complete The exceptions to this response are Pd/AC and Pd/GCN-1
hydrogenation to cyclohexanol (lower k1 /k3 ), behavior that where the dependence was slight. The temperature-induced
is quite distinct from Pd/AC. It has been shown elsewhere enhancement of SC=O can be considered to result from a
[78–80] that transition metals supported on structured car- desorption of reactant and reactive intermediate from the
bon deliver quite distinct activities and/or selectivities when surface, the net effect of which is to enrich the product
compared with conventional amorphous supports. The use of stream with the partially hydrogenated cyclohexanone. This
fibrous supports also has decided benefits in terms of greater effect is particularly marked in the case of Pd/GCN-2 which
mechanical strength and overall greater applicability in a suggests the involvement of quite different surface adsorp-
range of reactor configurations [81]. Given the diversity of tion energetics. A plot of cyclohexanone yield (YC=O ) vs
observed activity/selectivity values associated with the range temperature (Fig. 10) reveals the diversity in catalyst per-
of structural features particular to this family of Pd catalysts, formance associated with Pd dispersed on the different
it seems likely that the Pd particle morphology, exposed supports. The observed drop in yield is due to lower frac-
crystallographic orientations, and electronic character can tional phenol conversions. The exception is Pd/AC where
influence the ultimate product composition. The Pd structure conversion was enhanced at higher temperatures and des-
(geometric/electronic) is ultimately dependent on the choice orption phenomena do not appear to play a role over this
of support [82], which can, in turn, influence the nature of temperature interval. The highest YC=O is associated with
the chemisorbed intermediate(s) and the intrinsic surface re- Pd/Ta2 O5 at the lowest reaction temperature, a result that we
activity [83,84]. It is worth noting that Molchanov et al. ascribe to a high phenol conversion due to electron donation
[85] have provided some persuasive evidence to suggest that from the substrate combined with a phenol/surface interac-
unsaturated hydrocarbon hydrogenation selectivity over Ni tion that favors stepwise hydrogenation. The Pd/AC catalyst
supported on structured carbon is dependent on the crystal- represents the other extreme where weaker metal/support in-
lographic planes accessible to the reactants as well as the Ni teraction appears to result in lower intrinsic activity that is
electronic state that they link with surface morphology. elevated somewhat at higher temperatures without appar-
The relationship between SC=O and xphenol where the ently impacting on reactant/surface interactions to deliver
contact time was kept constant and temperature varied is il- the highest YC=O at 448 K.
lustrated in Fig. 9: a variety of responses is in evidence. With
the exception of the less active Pd/AC and Pd/GCN-1, the
steady-state phenol conversion decreased over the tempera- 4. Conclusions
ture interval 398
T
448 K. A drop in phenol conversion
with increasing temperature has been reported previously Impregnation and activation of (10% w/w) Pd on various
[7,24,25,27,28] and, in the absence of thermodynamic lim- oxide and carbon- (including novel GCN) based substrates
itations, has been attributed to a temperature-induced de- have led to significant variations in the intrinsic catalytic
crease of the fraction of the catalyst surface that is covered phenol hydrogenation activity/selectivity. The Pd phase ex-
 C. Park, M.A. Keane / Journal of Colloid and Interface Science 266 (2003) 183–194 193

hibited a morphological diversity where the average Pd [19] S.K. Ong, A.R. Bowers, J. Environ. Eng. 16 (1990) 1013.
particle sizes spanned the range 6–17 nm. A pseudo-first- [20] D.R. van der Vaart, E.G. Marchand, A. Bagely-Pride, Crit. Rev.
order kinetic analysis has been used to deliver specific Environ. Sci. Tech. 24 (1994) 303.
[21] L.J. Alemany, M.A. Bañares, E. Pardo, F. Martin, M. Galán-Fereres,
rate constants that are indicative of a degree of structure J.M. Blasco, Appl. Catal. B: Environ. 13 (1997) 289.
sensitivity in that the intrinsic rate was higher for those [22] G.B. Shulpin, M.M. Bochkova, G.V. Nizova, N.B. Kozlova, Appl.
systems exhibiting SMSI and increased in the following Catal. B: Environ. 12 (1997) 1.
sequence: Pd/AC ∼ Pd/GCN-1 < Pd/GCN-2 < Pd/SiO2 ∼ [23] S. Narayanan, G. Sreekanth, Ind. J. Technol. 31 (1993) 507.
Pd/graphite < Pd/Ta2 O5 . In the case of Pd/SiO2 , the use of [24] G. Neri, A.M. Visco, A. Donato, C. Milone, M. Malentacchi, G.
Gubitosa, Appl. Catal. A: Gen. 110 (1994) 49.
palladium acetate or nitrate as the metal precursor has lit- [25] A.K. Talukdar, K.G. Bhattacharyya, S. Sivasanker, Appl. Catal. A:
tle effect on catalyst productivity. Variations in selectivity Gen. 96 (1993) 229.
can be ascribed to differences in Pd/support interaction(s) [26] S.T. Srinivas, L.J. Lakshmi, P.K. Rao, Appl. Catal. A: Gen. 110 (1994)
that impact on the supported Pd microstructure where, in 167.
the case of Pd dispersed on AC and Ta2 O5 , the metal can [27] N. Itoh, W.-C. Xu, Appl. Catal. A: Gen. 107 (1993) 83.
[28] S. Galvagno, A. Donato, G. Neri, R. Pietropaolo, J. Chem. Technol.
be considered to adopt a crystallographic orientation that fa- Biotechnol. 51 (1991) 145.
vors partial hydrogenation to cyclohexanone whereas Pd on [29] S.T. Srinivas, P.K. Rao, J. Chem. Soc. Chem. Commun. (1993) 33.
GCN promotes complete hydrogenation to the alcohol. A de- [30] U.M. Tang, M.Y. Huang, Y.Y. Jiang, Macromol. Rapid Commun. 15
crease in the effective contact time (at constant T ) lowered (1994) 527.
fractional conversion with a preferential formation of cyclo- [31] M. Higashijima, S. Nishimura, Bull. Chem. Soc. Jpn. 64 (1992) 2955.
[32] N. Mahata, V. Vishwanathan, Ind. J. Chem. 37A (1998) 652.
hexanone. Where the contact time was maintained constant [33] S. Narayanan, G. Sreekanth, React. Catal. Lett. 51 (1993) 449.
an increase in reaction temperature lowered both conversion [34] G.M. Schwab, Adv. Catal. 27 (1982) 1.
and cyclohexanone selectivity (with the exception of Pd/AC [35] R. van Hardeveld, F. Hartog, Adv. Catal. 18 (1972) 75.
and Pd-GCN-1), a result that is ascribed to desorption phe- [36] V. Ponec, Stud. Surf. Sci. Catal. 11 (1982) 63.
nomena. [37] A.Y. Stakheev, L.M. Kustov, Appl. Catal. A: Gen. 188 (1999) 3.
[38] B.C. Gates, Chem. Rev. 95 (1995) 511.
[39] G.C. Bond, Chem. Soc. Rev. 20 (1991) 441.
[40] T.G. Ros, D.E. Keller, A.J. van Dillen, J.W. Geus, D.C. Koningsberger,
Acknowledgments J. Catal. 211 (2002) 85.
[41] C. Liang, Z. Li, J. Qiu, C. Li, J. Catal. 21 (2002) 278.
We are grateful to Gonzalo Pina and Alan Dozier for [42] C. Park, M.A. Keane, Chem. Phys. Chem. 2 (2001) 101.
[43] C. Park, M.A. Keane, J. Colloid Interface Sci. 250 (2002) 37.
assistance with the catalysis and catalyst characterization
[44] C. Park, M.A. Keane, Langmuir 17 (2001) 8386.
measurements. [45] M.A. Armendia, V. Borau, I.M. Garcia, C. Jimenez, F. Lafont,
A. Marinas, J.M. Marinas, F.J. Urbano, J. Catal. 187 (1999) 392.
[46] G. Tavoularis, M.A. Keane, J. Chem. Technol. Biotechnol. 74 (1999)
References 60.
[47] D.E. Mears, Ind. Eng. Chem. Proc. Des. Dev. 10 (1971) 541.
[1] S. Narayanan, R. Unnikrishnan, J. Chem. Soc. Faraday Trans. 93 [48] R.T.K. Baker, J. Catal. 63 (1980) 523.
(1997) 2009. [49] T. Ushikubo, Catal. Today 57 (2000) 331.
[2] S. Narayanan, K. Krishna, Appl. Catal. A: Gen. 147 (1996) L253. [50] J. Oi-Uchisawa, A. Obuchi, R. Enomoto, J. Xu, T. Nanba, S. Liu,
[3] P. Claus, H. Berndt, C. Mohr, J. Radnik, E.-J. Shin, M.A. Keane, S. Kushiyama, Appl. Catal. B: Environ. 32 (2001) 257.
J. Catal. 192 (2000) 88. [51] S.J. Tauster, S.C. Fung, J. Catal. 55 (1978) 29.
[4] Y.Z. Chen, C.W. Liaw, L.I. Lee, Appl. Catal. A: Gen. 177 (1999) 1. [52] S.J. Tauster, S.C. Fung, R.T.K. Baker, J.A. Horsley, Science 211
[5] E.-J. Shin, M.A. Keane, Ind. Eng. Chem. Res. 39 (2000) 883. (1981) 1121.
[6] N. Mahata, V. Vishwanathan, Catal. Today 49 (1999) 65. [53] J.H.A. Martens, R. Prins, D.C. Koningsberger, Catal. Lett. 2 (1989)
[7] E.-J. Shin, M.A. Keane, J. Catal. 173 (1998) 450. 211.
[8] S.G. Shore, E. Ding, C. Park, M.A. Keane, Catal. Commun. 3 (2002) [54] E. Ruckenstein, H.Y. Wang, J. Catal. 190 (2000) 32.
77. [55] M.C.J. Bradford, M.A. Vannice, Appl. Catal. A: Gen. 142 (1996) 73.
[9] N. Mahata, K.V. Raghavan, V. Vishwanathan, C. Park, M.A. Keane, [56] M.A. Ermakova, D.Yu. Ermakov, G.G. Kuvshinov, L.M. Plyasova,
Phys. Chem. Chem. Phys. 3 (2001) 2712. J. Catal. 187 (1999) 77.
[10] V. Vishwanathan, N. Mahata, M.A. Keane, React. Kinet. Catal. [57] E. Boellaard, A.M. van der Kraan, Geus, J.W., Appl. Catal. A:
Lett. 72 (2001) 297. Gen. 224 (2002) 1.
[11] N. Mahata, V. Vishwanathan, J. Catal. 196 (2000) 262. [58] J.S. Smith, P.A. Thrower, M.A. Vannice, J. Catal. 68 (1981) 270.
[12] S. Narayanan, K. Krishna, Appl. Catal. A: Gen. 198 (2000) 13. [59] J.W.E. Coenen, Appl. Catal. 75 (1991) 193.
[13] N. Mahata, K.V. Raghavan, V. Vishwanathan, Appl. Catal. A: Gen. 182 [60] R.T.K. Baker, E.B. Prestridge, G.B. McVicker, J. Catal. 89 (1984) 422.
(1999) 183. [61] T. Baird, J.R. Fryer, B. Grant, Nature 233 (1971) 329.
[14] S. Scire, C. Crisafulli, R. Maggiore, S. Minicio, S. Galvagno, Appl. [62] R.T.K. Baker, E.B. Prestridge, R.L. Garten, J. Catal. 59 (1979) 293.
Surf. Sci. 136 (1998) 311. [63] N. Hatta, K. Murata, Chem. Phys. Lett. 217 (1994) 398.
[15] S.T. Srinivas, P.K. Rao, J. Catal. 179 (1998) 1. [64] T.W. Ebbesen, P.M. Ajayan, Nature 358 (1992) 220.
[16] K. Amouzegar, O. Savadogo, J. Appl. Electrochem. 27 (1997) 539. [65] V. Ivanov, A. Fonseca, J.B. Nagy, A. Lucas, P. Lambin, D. Bernaerts,
[17] I. Dodgson, K. Griffen, G. Barberis, F. Pignatoro, G. Tauszik, Chem. X.B. Yang, Carbon 33 (1995) 1727.
Ind. (1989) 830. [66] N.M. Rodriguez, J. Mater. Res. 8 (1993) 3233.
[18] E.Y. Bezuglaya, A.B. Shchutskaya, I.V. Smirnova, Atmos. Environ. 27 [67] R.T.K. Baker, J. Adhesion 52 (1995) 13.
(1993) 773. [68] K.P. de Jong, J.W. Geus, Catal. Rev.-Sci. Eng. 42 (2000) 481.
194 C. Park, M.A. Keane / Journal of Colloid and Interface Science 266 (2003) 183–194

[69] R.T.K. Baker, M.A. Barber, P.S. Harris, F.S. Feates, R.J. Waite, [77] R. Taylor, K.H. Lundlum, J. Phys. Chem. 76 (1972) 2882.
J. Catal. 26 (1972) 51. [78] K. Tanabe, T. Nishizaki, in: G.C. Bond, P.B. Wells, F.C. Tomkins
[70] C. Park, N.M. Rodriguez, R.T.K. Baker, J. Catal. 169 (1997) 212. (Eds.), Proceedings of the 6th International Congress on Catalysis,
[71] C. Pham-Huu, N. Keller, G. Ehret, L.J. Carbonniere, R. Ziessel, M.J. Catalysis Society, Letchworth, 1977, p. 863.
Ledoux, J. Mol. Catal. A: Chem. 170 (2001) 155. [79] C. Park, R.T.K. Baker, J. Phys. Chem. B 102 (1998) 5168.
[72] J.M. Planeix, N. Coustel, B. Coq, V. Brotons, P.S. Kumbhar, R. [80] F. Salman, C. Park, R.T.K. Baker, Catal. Today 53 (1999) 385.
Dutartre, P. Geneste, P. Bernier, P.M. Ajayan, J. Am. Chem. Soc. 116 [81] V.V. Molchanov, V.V. Chesnokov, R.A. Buyanov, N.A. Zaitseva,
(1994) 7935. Kinet. Catal. 39 (1998) 378.
[73] M.L. Toebes, J.A. van Dillen, K.P. de Jong, J. Mol. Catal. A: [82] Yu. Matatov-Meytal, M. Sheintuch, Appl. Catal. A: Gen. 231 (2002)
Chem. 173 (2001) 75. 1.
[74] C.A. Bessel, K. Laubernds, N.M. Rodriguez, R.T.K. Baker, J. Phys. [83] M. Che, C.O. Bennett, Adv. Catal. 36 (1989) 55.
Chem. 105 (2001) 1115. [84] I. Efremenko, J. Mol. Catal. A: Chem. 173 (2001) 19.
[75] C. Park, R.T.K. Baker, J. Phys. Chem. B 103 (1999) 2453. [85] V.V. Molchanov, V.V. Chesnokov, R.A. Buyanov, N.A. Zaitseva,
[76] G. Neri, A.M. Visco, A. Donato, C. Milone, M. Malentacchi, G. V.I. Zaikovskii, L.M. Plyasova, V.I. Bukhtiyarov, I.P. Prosvirin, B.N.
Gubitosa, Appl. Catal. A: Gen. 110 (1994) 49. Novgorodov, Kinet. Catal. 39 (1998) 386.