Phenol hydrogenation over palladium supported on magnesia: Relationship between catalyst structure and performance

Article abstract of DOI:10.1039/b100237f

The gas-phase hydrogenation of phenol was used as a model reaction to probe the dependence of catalytic activity/selectivity on changes in Pd particle size and surface acid-base properties. The catalyst prepared from the acetate precursor exhibited the gr

Full text of DOI:10.1039/b100237f

View Article Online / Journal Homepage / Table of Contents for this issue
Phenol hydrogenation over palladium supported on magnesia:
Relationship between catalyst structure and performance
N. Mahata,a K. V. Raghavan,a V. Vishwanathan,*a C. Parkb and M. A. Keane*c
a Catalysis and Physical Chemistry Division, Indian Institute of Chemical T echnology,
HyderabadÈ500 007, India
b Department of Chemical Engineering, University of L eeds, L eeds, UK L S2 9JT
c Department of Chemical & Materials Engineering, University of Kentucky, L exington
KY 40506 0046, USA. E-mail: makeane=engr.uky.edu
Received 4th January 2001, Accepted 2nd May 2001
First published as an Advance Article on the web 1st June 2001
A series of magnesia-supported palladium catalysts (Pd loading in the range 0.5È7.0 wt.%) has been prepared
by impregnation from aqueous solutions of PdCl , Pd(NH ) Cl and Pd(CH COO) and characterised by
2
3 4
2
3
2
X-ray di†raction (XRD), CO chemisorption and high resolution transmission electron microscopy (HRTEM).
The gas-phase hydrogenation of phenol was employed as a model reaction to probe the dependence of
catalytic activity/selectivity on changes in Pd particle size and surface acidÈbase properties. The catalyst
prepared from the acetate precursor exhibited the greatest Pd dispersion when compared with the chloride and
amine precursors. The surface mobility of the metal chloride resulted in larger Pd particles from the
Cl-containing precursors while the presence of residual Cl in the activated catalysts lowered the hydrogenation
rate and was responsible for a decline in activity with time-on-stream. The e†ects of varying such process
variables as temperature, hydrogen/phenol mol ratio and inlet molar phenol feed rate are presented and
discussed while the question of structure sensitivity is addressed. The reaction exhibited a negative dependence
on phenol partial pressure up to 503 K but a positive dependence was evident at higher temperatures. The
order of the reaction with respect to hydrogen remained positive and was close to unity at 563 K; an apparent
activation energy of 63 kJ mol~1 was recorded. The e†ect of doping the support with calcium and Ñuoride has
shown that modiÐcations to the acidÈbase properties of magnesia can be used to control catalytic
activity/selectivity.
product composition; by modifying the acidÈbase functions of
MgO, catalytic activity/selectivity can be controlled. The
interplay between surface acidity/basicity is accepted as
Introduction
The synthesis of cyclohexanone is of commercial signiÐcance
in the manufacture of caprolactam, a monomer used in the
synthesis of Nylon-6. Cyclohexanone production, on an indus-
trial scale, typically involves either the oxidation of cyclo-
hexane or the hydrogenation of phenol.1 The former route
requires high temperatures and pressures and generates appre-
ciable by-products that lower the product yield and compli-
cate the recovery/separation steps. In the hydrogenation of
phenol, cyclohexanone can be generated in a one- or two-step
process. In the latter case, phenol is Ðrst hydrogenated to
cyclohexanol which is then dehydrogenated, in the second
step, to cyclohexanone. The replacement of a two-step process
by a single-step (phenol ] cyclohexanone) hydrogenation is
economically more advantageous in terms of capital costs and
energy savings.2 Based on the literature to date, Pd catalysts
have shown the most promise in promoting the selective
hydrogenation to cyclohexanone.3h11 Alumina-supported Pd
catalysts, though used extensively in hydrogenation processes,
have a low resistance to deactivation by coke deposition.12
Basic Group I and II oxides can, however, serve as promoters
by modifying acidic supports to increase the coke resistance.
While magnesia as a catalyst support su†ers from the decided
drawback of possessing a low surface area and poor mechani-
cal strength,3 it exhibits some interesting support e†ects in
terms of its ability to stabilize the supported metal.13 The
acidÈbase properties of magnesia can impact on the nature of
the reactant/catalyst interactions and govern the ultimate
having
a
bearing on cyclohexanone selectivity from
phenol3,14,15 but the optimum surface requirements are far
from established.
It is known that calcination at high temperatures and the
addition of alkali/alkaline earth metals can induce signiÐcant
changes in the physico-chemical and catalytic properties of
alumina.16 The possibility of using promoters to modify
Pd/MgO and generate a highly selective phenol hydro-
genation catalyst has yet to be studied in a comprehensive
fashion. The role of metal loading has been
considered3,7,9,17h19 but no conclusive correlation between
Pd dispersion/metal surface area/crystallite size and hydro-
genation activity/selectivity has yet been presented for
Pd/MgO. In this paper, such a correlation is proposed and the
inÑuence of (a) the nature of the Pd catalyst precursor and (b)
the incorporation of calcium and Ñuoride additives on the
overall catalytic performance is reported.
Experimental
Catalyst preparation and activation
Magnesia-supported palladium, where the palladium loading
was in the range 0.5È7.0 wt.%, was prepared by impregnating
a
commercial magnesia (Harshaw MG-0601, surface
area \ 48 m2 g~1 and pore volume \ 1.03 cm3 g~1) with
2712
Phys. Chem. Chem. Phys., 2001, 3, 2712È2719
This journal is ( The Owner Societies 2001
DOI: 10.1039/b100237f

View Article Online
known concentrations of acidiÐed (with HCl, pH 3) aqueous
change, the amount of ammonia adsorbed per unit weight of
catalyst was calculated. After the Ðrst adsorption isotherm, the
catalyst was evacuated for 1 h at the same temperature and
the second isotherm was measured. The parallel isotherms
were extrapolated to zero pressure and the di†erence in
amount adsorbed was taken as the chemisorbed ammonia
content.
solutions of PdCl É xH O. Palladium acetate (Pd(CH COO) )
2
2
3
2
and tetraamine palladium(II) chloride (Pd(NH ) Cl ) were also
3 4
2
used as palladium sources to prepare 1 wt.% Pd/MgO cata-
lysts. The Pd(CH COO) was dissolved in acetonitrile while
3
2
the Pd(NH ) Cl solution was prepared by dissolving PdCl in
3 4
2
conc. NH OH solution (pH 9) at 348 K with appropriate dilu-
tion in deionised water. The magnesia support was also modi-
4
Ðed by impregnation with solutions of calcium nitrate and
ammonium Ñuoride to deliver a 1 wt.% loading prior to the
addition of palladium. The catalysts were dried overnight at
393 K, calcined in air at 723 K for 5 h and reduced at 723 K
for 3 h in Ñowing dry hydrogen (0.26 mol h~1).
Catalytic reactor system
The gas-phase phenol hydrogenation was carried out in a
Ðxed-bed continuous Ñow tubular reactor (2.2 cm id ] 20 cm)
over the temperature range 473 O T /K O 563 and at atmo-
spheric pressure. The catalyst precursor (particle
diameter \ 1400È1800 lm) was calcined/reduced in situ (as
above) and a mixture of phenol in cyclohexane (1 : 2 weight
ratio) was fed to the catalyst bed (ca. 0.5 g) at a Ðxed rate via a
calibrated motorized syringe. Hydrogen served as both
reductant and carrier gas where the hydrogen/phenol inlet
mol ratio was varied from 2 to 11 and the weight hourly space
velocity (WHSV) from 0.02 to 0.054 mol h~1 g~1. The partial
pressures of both reactants were adjusted by dilution with
nitrogen: hydrogen, 0.18È0.88 atm; phenol, 0.06È0.12 atm.
Catalyst characterisation
XRD analysis of the calcined samples was conducted using
a Philips PW 1140 di†ractometer with Ni Ðltered Cu-Ka radi-
ation (j \ 1.54 A). The residual chlorine content in the
calcined/reduced catalysts derived from PdCl
and
2
(Pd(NH ) Cl ) precursors was determined potentiometrically.
3 4
2
A known weight (ca. 0.2 g) of the activated samples was
heated under reÑux in deionised water for 24 h. The water
extract was Ðltered and the chlorine content was measured
Passage of phenol/H through an empty reactor or over the
2
MgO support alone did not result in any detectable conver-
using a Metrohm (Model 278) Autotitrator (AgNO titrant,
3
sion. The liquid products were collected periodically and
analysed by gas chromatography (FID) using a packed
column (Carbowax 20M on Chromosorb W). The overall con-
combined Ag electrode).
Carbon monoxide chemisorption measurements were per-
formed at 298 K using a conventional volumetric apparatus20
where the catalyst was activated in situ, as above. The irrevers-
ible uptake of CO was measured using a dual isotherm tech-
nique, i.e. after determination of the Ðrst isotherm, the samples
were evacuated at 298 K for 30 min and a second isotherm
was obtained. The di†erence between the two isotherms
version of phenol (C ) as a percentage is given by
P
[C H OH] [ [C H OH]
6
5
in
6
6
5
out
] 100
C \
P
[C H OH]
5
in
while process selectivity (as a percentage) in terms of cyclo-
accounts for the amount of CO irreversibly held (CO ) on the
hexanone formation (S ) can be represented by
irr
surface of the catalyst. The ratio CO /Pd, i.e. the ratio of
irr
CJO
[C H O]
surface Pd atoms available for CO adsorption to the total
6
10 out
S
\
] 100
number of Pd atoms, is used in this paper as a measure of
palladium dispersion. Infrared analysis21 revealed that CO
adsorption, under these conditions, was predominantly via a
linear single point attachment. The intensity of the band at ca.
1800 cm~1 attributed to a bridged mode of adsorption was
insigniÐcant when compared with the band at 2070 cm~1 that
CJO
[C H OH] [ [C H OH]
6
5
in
6
5
out
Reactivity is also discussed in this paper in terms of product
yield (Y ) where, in the case of cyclohexanone
[C H O]
6
6
10 out
] 100
Y \
[C H OH]
in
is characteristic of a linear mode of adsorption. A CO /Pd \
5
irr
1 stoichiometry was accordingly taken as a reasonable esti-
mate in determining Pd dispersion. The metal surface area (S,
m2 (g Pd)~1) was calculated from CO uptake using the
relationship22
Results and discussion
The XRD patterns of calcined Pd/MgO with varying Pd
content are shown in Fig. 1. The di†ractograms are character-
ized by strong signals due to the presence of a crystalline
MgO phase. There was no evidence of any brucite (Mg(OH) ),
characteristic peak at 2h \ 38¡, in the calcined samples. With
an increase in Pd content beyond 1 wt.%, the two PdO peaks
S \ CO /1.27 ] 1019 ] metal content
irr
where 1.27 ] 1019 (atoms m~2) corresponds to the number of
surface Pd atoms per unit area of polycrystalline metal
surface. The Pd crystallite size (d, nm) was calculated from the
estimated surface area according to22
2
at inter spacing values of d \ 2.644 and 1.671 A exhibited a
d \ 6 ] 103/oS
where o is the density of palladium metal, i.e. 12.02 g cm~3.
HRTEM analysis was carried out using a Philips CM200
FAGTEM microscope operated at an accelerating voltage of
200 kV. The specimens were prepared by ultrasonic dispersion
in butan-2-ol, evaporating a drop of the resultant suspension
onto a holey carbon support grid. The particle size distribu-
tion proÐles presented in this study are based on a measure-
ment of over 500 individual particles.
Surface acidity was probed by ammonia adsorption at
room temperature. In a typical experiment, ca. 0.2 g of the
activated catalyst was degassed for 3 h at 473 K to a pressure
of 1 ] 10~5 Torr and cooled to ambient temperature. The
ammonia adsorption isotherm was measured by allowing
ammonia from the storage bulb to come in contact with the
catalyst and equilibrate. From the pressure and volume
Fig. 1 XRD patterns for Pd/MgO samples of varying Pd loading
(A \ 1, B \ 2, C \ 3, D \ 5 and E \ 7 wt.%) after calcination in air
at 723 K for 3 h: (…) denotes the MgO phase; PdCl precursor.
2
Phys. Chem. Chem. Phys., 2001, 3, 2712È2719
2713

View Article Online
marked increase in intensity as noted elsewhere23 for higher
Pd loadings. There was no detectable PdO signal where Pd
loading was less than 1 wt.% but this can be attributed to
particle sizes less than the resolution limit (ca. 3 nm) of the
di†ractometer. The inÑuence of the Pd precursor salts on Pd
particle size/dispersion and phenol hydrogenation activity is
shown in Table 1. The degree of Pd dispersion or number of
exposed Pd atoms was determined from the irreversible CO
uptake capacity of each catalyst, a technique that has been
applied elsewhere9,24,25 and shown to yield reliable values.
The Pd catalyst prepared from the acetate precursor realised
the highest metal dispersion when compared with the chloride
and amine precursors. This enhancement of metal dispersion
may be attributed to the larger size of the counter anion26
which serves to space the Pd(II) component apart more e†ec-
tively, thereby limiting agglomeration during calcination to
give smaller PdO crystallites and a subsequent greater disper-
sion of the metallic Pd component after reduction. The
Pd(NH ) Cl precursor is known27,28 to undergo an “auto-
included in Table 1. Each catalyst exhibited a similar size dis-
tribution but there is a greater presence of particles [6 nm in
the case of those activated samples derived from chlorine-
containing precursors. It is instructive to note that the Pd
crystallite diameters determined by HRTEM were consistently
greater than the values extracted from CO chemisorption.
This discrepancy suggests that, despite the IR evidence, there
must be some deviation from an exclusive 1 : 1 CO : Pd stoi-
chiometry on these supported catalysts. Any surface stoichi-
ometry will be dependent on metal loading and the
assignment of one particular value is at best a convenient
approximation. Moreover, CO has been shown to interact
directly with MgO.30 Dorling and Moss,31 in a study of CO
chemisorption on Pt/SiO likewise concluded an exposed
2
Pt/CO adsorption ratio of one from IR analysis but found
that this was not strictly the case and the surface stoichiom-
etry was dependent on the nature of the exposed Pt faces.
Smith et al.32 have also recorded lower (Ni) particle (on SiO )
2
diameters from CO chemisorption when compared with TEM
3 4
2
reductionÏ and the lower dispersion can be linked to the inher-
ent high surface mobility. Moreover, it is well established29
that larger metal particles are typically generated from a chlo-
ride precursor due to crystallite growth involving volatile
metal chloride. The Pd particle size distribution, derived from
HRTEM analysis, is illustrated by the histograms given in
Fig. 2; the surface weighted average particle diameters are
analysis. Indeed, there are many instances in the literature
where the dimensions of supported metal particles that are
measured di†er depending on the nature of the analytical
technique,33h35 an inconsistency that continues to bedevil het-
erogeneous catalysis research. It is, however, usually the case
that each characterisation tool delivers the same trends for a
family of metal catalysts. The latter holds in this study where
Table 1 E†ect of palladium precursor on Pd particle size/dispersion (from CO chemisorption and HRTEM analysis) and phenol hydrogenation
activity over 1 wt.% Pd/MgO: T \ 503 K, WHSV \ 0.027 mol h~1 (g cat)~1; hydrogen/phenol \ 5.4
Crystallite diameter d/nm
Precursor
Dispersion CO /Pd
From CO
From HRTEM
Activity/10~6 mol s~1 g~1
irr
irr
Pd(CH COO)
0.56
0.51
0.37
2.0
2.2
3.0
6.2
7.1
7.6
6.8
3.9
4.6
3
PdCl
2
2
Pd(NH ) Cl
3 4
2
Fig. 2 Palladium particle size distribution in the activated 1 wt.% Pd/MgO catalysts derived from Pd(CH COO) (solid bars), PdCl (open
3
2
2
bars) and Pd(NH ) Cl (cross-hatched bars) precursors. Inset: Pd particle size distribution in the activated 7 wt.% Pd/MgO catalyst derived from
3 4
2
PdCl .
2
2714
Phys. Chem. Chem. Phys., 2001, 3, 2712È2719

View Article Online
Fig. 3 HRTEM of the activated 1 wt.% Pd/MgO catalysts derived from the (a) PdCl , (b) Pd(NH ) Cl and (c) Pd(CH COO) precursors.
3 4
2
2
3
2
the order of increasing particle size was the same for both
techniques. The HRTEM images, shown in Fig. 3, of the 1
wt.% Pd/MgO catalyst illustrates the nature of the Pd metal
dispersion. Selected area electron di†raction analysis revealed
that the Pd metal particles are crystalline and the lattice struc-
ture is in evidence in Fig. 3; the crystallinity of the MgO
support is also apparent.
The palladium catalyst derived from the acetate precursor
delivered the highest hydrogenation activities (Table 1) when
compared with the chloride-derived samples. There was an
appreciable residual chlorine associated with the activated
Pd(NH ) Cl
(3.9 ] 10~5 mol (g cat)~1) and PdCl
3 4
2
2
(3.6 ] 10~5 mol (g cat)~1) samples as determined poten-
tiometrically from the water extracts after heating the samples
under reÑux. A subsequent STEM/energy dispersive X-ray
(EDX) elemental mapping of the reÑuxed solid did not reveal
even trace quantities of chlorine on the surface, i.e. the reÑux
treatment was fully e†ective in extracting the entire chlorine
component from the activated catalysts. Halogens are known
to act as strong poisons of transition metal catalysts35h37 and
the presence of this electron-withdrawing residual chlorine
may induce electronic perturbations in the supported Pd sites.
The latter e†ect can impact on CO interaction(s) and may well
Fig. 4 E†ect of the nature of the palladium precursor on the time
dependent phenol hydrogenation activity over 1 wt.% Pd/MgO. Pal-
ladium source: A \ Pd(CH COO) ; B \ Pd(NH ) Cl ; C \ PdCl ;
3
2
3 4
2
2
T \ 503 K; WHSV \ 0.027 mol h~1 (g cat)~1; hydrogen/
phenol \ 5.4.
indicate that there was no appreciable additional agglomer-
ation of metal crystallites in the Pd-rich samples during acti-
vation. This is supported by XRD analysis where, on the basis
of line broadening, there was no signiÐcant increase in PdO
(calcined samples) or Pd (calcined/reduced samples) size. The
greater particle size for the highest Pd-loaded sample is appar-
ent from the representative HRTEM image shown in Fig. 6
where again metal lattice structure is evident. All the activated
have a bearing on the recorded CO values. We have shown
elsewhere26 that the d-character of supported Pd governs, to a
irr
great extent, phenol hydrogenation activity. Moreover, a
surface contamination by chlorine can disrupt the dissociative
adsorption of hydrogen38 and limit the reactivity of the
surface hydrogen. The chlorine-containing precursors
exhibited a marked loss of activity with time-on-stream as
shown in Fig. 4. The presence of chlorine on catalyst surfaces
is known to promote carbon deposition as demonstrated else-
where.39h41 Any carbon deposits that are formed can occlude
the active sites from incoming phenol with a consequent time-
dependent decline in conversion.
samples derived from the PdCl precursor retained an appre-
2
ciable chlorine content with a Pd/Cl mol ratio of up to 0.96.
The histogram given in the inset to Fig. 2 reveals the broad
(and possibly) bimodal distribution of Pd particle diameters.
The relationship between Pd loading and the ultimate dis-
persion, size and surface area of the supported metal phase is
presented in Fig. 5. We have adopted the Pd dispersion/size
values derived from CO chemisorption that, although they
di†er from the values extracted from the HRTEM analysis,
better represent the nature of the catalytically active surface in
terms of adsorption/reactant activation. The use of TEM
(even at high resolution) to determine metal particle sizes
requires a deal of judgment where insufficient contrast in the
image can hamper an accurate analysis, particularly for diam-
eters of 1 nm and less. The inferred Pd particle size varied in
the range 1.9 to 6.9 nm which represents a decrease in disper-
sion from 0.59 to 0.16. The dispersion decreased markedly as
the Pd loading was increased from 1% to 3 wt.%, above
which it was less dependent on Pd content; metal area (g
cat)~1 increased over this range of Pd loading. The results
Fig. 5 E†ect of Pd loading on the dispersion (…), metal area (])
and crystallite size (L) of activated Pd/MgO: PdCl precursor.
2
Phys. Chem. Chem. Phys., 2001, 3, 2712È2719
2715

View Article Online
al.9 There was, however, a discernible drop in TOF where the
average Pd particle diameter was less than ca. 2 nm and this
can be attributed to an electron deÐciency at the Pd sites. It
has been shown elsewhere43 that Pd particles \3 nm have a
tendency to remain electron deÐcient. There are many
instances in the literature42 where reactivity is strongly inÑu-
enced by the electron density of small supported metal par-
ticles. The presence of residual surface Cl may serve to further
displace electron density from the Pd crystallites and inhibit
reactant(s) activation.
The response of activity/selectivity to alterations in the
hydrogen/phenol molar ratio is illustrated in Fig. 8. The
degree of conversion was very sensitive to the hydrogen
content of the feed and a Ðve-fold increase is recorded in Fig.
8. This increase in activity was again accompanied by a higher
degree of complete hydrogenation to cyclohexanol. While
cyclohexanone was the sole product at hydrogen/phenol \5,
cyclohexanol was increasingly favoured as the hydrogen
content in the feed was increased. An increase in the inlet
phenol molar feed rate at a constant hydrogen/phenol ratio
(5.4) lowered conversion and shifted cyclohexanone selectivity
to higher values (Fig. 9). The inÑuence of reaction temperature
is shown in Fig. 10 where it is evident that phenol conversion
passed through a maximum (T ) at ca. 503 K while cyclo-
Fig. 6 HRTEM of the activated 7 wt.% Pd/MgO catalysts derived
max
hexanone selectivity (S ) was largely temperature invariant
from the PdCl precursor.
CJO
2
over the range 473È563 K. A drop in activity with increasing
The Pd crystallite sizes fall within the so-called mithohedrical
region wherein catalytic reactivity can show a critical depen-
dence on morphology.42 The inÑuence of Pd loading on
phenol conversion and product selectivity is shown in Fig. 7.
Cyclohexanone and cyclohexanol were the only products
detected and there was no evidence of the formation of
cyclohexene-1-ol as a reactive intermediate or of a hydro-
dehydroxylation as observed in the case of supported Ni cata-
lysts.19 The conversion of phenol increased with increasing
metal loading while the cyclohexanone selectivity declined
from almost 100% to \80% at a Pd content [3 wt.%. The
yield of both products increased with an increase in conver-
sion but at Pd loadings in excess of 3 wt.%, the increase in
cyclohexanone yield was marginal (45% to 50%). The yield of
cyclohexanol, on the other hand, was far more sensitive to Pd
content and increased proportionally with an increase in Pd
loading, i.e. as the overall conversion was raised. The latter
e†ect is diagnostic of a stepwise reaction with cyclohexanone
as the partially and cyclohexanol as the fully hydrogenated
product. The turnover frequency (TOF, number of molecules
of phenol converted per second per exposed Pd atom) was
essentially invariant with particle size ([ca. 2 nm) which is
indicative of structure insensitivity as noted by Galvagno et
Fig. 8 E†ect of inlet hydrogen/phenol mole ratio on initial phenol
conversion (C ) and selectivity in terms of cyclohexanone (S ) and
P
CJO
cyclohexanol (S
)
over
1
wt.% Pd/MgO: PdCl precursor;
ChOH
2
T \ 503 K; WHSV \ 0.027 mol h~1 (g cat)~1.
Fig. 7 E†ect of Pd loading on initial phenol conversion (C ) and
Fig. 9 E†ect of phenol molar feed rate on initial conversion (C ,
P
P
selectivity in terms of cyclohexanone (S ) and cyclohexanol (S
)
hatched bars) and selectivity in terms of cyclohexanone (S
, open
CJO
ChOH
CJO
over Pd/MgO: PdCl precursor; T \ 503 K; WHSV \ 0.027 mol
bars) and cyclohexanol (S
, solid bars) over 1 wt.% Pd/MgO:
2
ChOH
h~1 (g cat)~1; hydrogen/phenol \ 5.4.
PdCl precursor; T \ 503 K; hydrogen/phenol \ 5.4.
2
2716
Phys. Chem. Chem. Phys., 2001, 3, 2712È2719

View Article Online
Fig. 11 E†ect of varying the partial pressure of phenol (P ) and
P
Fig. 10 E†ect of reaction temperature on initial phenol conversion
(C ) and selectivity in terms of cyclohexanone (S ) over 1 wt.% Pd/
hydrogen (P ) on initial reaction rate over 1 wt.% Pd/MgO: PdCl
H
2
precursor; T \ 503 K.
P
CJO
MgO: PdCl precursor; WHSV \ 0.027 mol h~1 (g cat)~1;
2
hydrogen/phenol \ 5.4.
concentration which suggests a strong adsorption of phenol
on the catalyst. Indeed, Xu et al.48 have shown by tem-
perature programmed desorption that phenol interacts strong-
ly with magnesia. The switch to positive x values and increase
in magnitude at elevated temperatures reÑect a decrease in
surface coverage and is consistent with the observed drop in
conversion shown in Fig. 10. Chen et al.14 have likewise
reported a negative dependence of rate on phenol concentra-
tion. The order of the reaction (y) with respect to hydrogen
remained positive and was close to unity at 563 K. The tem-
perature dependence of the rate constants yields an apparent
temperature has been noted elsewhere3,8,9,44,45 and attributed
to thermodynamic limitations.3,45 The activity maximum
shown in Fig. 10 could be traversed from either the high or
low temperature side without any loss of activity which, in
e†ect, negates catalyst deactivation as a possible source of the
T
. The contribution of thermodynamic limitations was
max
considered by estimating the pertinent gas-phase equilibrium
reaction constants using the NIST database.46 The calculated
gas-phase equilibrium conversion does show a marked tem-
perature dependence with complete conversion at T \ 443 K
and a continual drop thereafter to trace conversions at 573
K.46 The catalytic and gas-phase equilibrium systems diverge
signiÐcantly in terms of reaction selectivity. Under equilibrium
conditions,47 cyclohexanol is the predominant product at
lower temperatures but cyclohexanone is increasingly pre-
ferred as the reaction temperature is raised. The observed tem-
perature e†ects in our catalytic system can be attributed to
surface phenomena, as concluded by Galvagno et al.9 Given
the observed link between conversion and selectivity for the
range of reaction conditions studied (see Fig. 8È10), it is clear
that hydrogenation occurs stepwise with cyclohexanone as a
reactive intermediate. A non-planar mode of phenol inter-
action with the basic MgO support has been proposed3
through the hydroxy function as opposed to a co-planar
surface arrangement that appears to predominate on acidic
(e.g. Al O ) supports involving a strong interaction with the
activation energy (E ) of 63 kJ mol~1 which is the same as
a
that quoted by Galvagno et al.9 for a closely related Pd/MgO
system. In the only other directly comparable kinetic study,
Gonzalez-Velasco et al.49 have reported the appreciably lower
value of 38 kJ mol~1 for reaction over Pd/CaOÈAl O .
2
3
The strength of phenol adsorption and subsequent hydro-
genation activity/selectivity is known to be dependent on the
acidÈbase properties of the support.3,14,15 The stronger the
interaction of phenol with the surface the greater is the likeli-
hood of complete hydrogenation to cyclohexanol. The addi-
tion of alkali or alkaline earth metals can modify, through
electronic e†ects, catalytic performance.50,51 The e†ect of two
additives (Ca and F) is illustrated in Table 3 where doping
with calcium decreased S
but conversion was unaltered
CJO
while the addition of Ñuoride did not a†ect selectivity but
lowered activity. Addition of a basic oxide promoter has been
shown to result in an electron transfer from the oxide to the
supported metal, thereby increasing the electron density of the
metal site26,52 which can, in turn, impact on conversion
and/or selectivity. In this study, the presence of Ca had no
e†ect on Pd dispersion/crystallite size and any alteration to
the electronic properties of the metal phase, insofar as they
inÑuence hydrogenation activity, were negligible. The decrease
2
3
aromatic nucleus. The former should favour a stepwise addi-
tion of hydrogen and a high cyclohexanone selectivity47 as
observed in this study. The maintenance of a constant S
value at elevated temperatures suggests that the mode of
CJO
adsorption is not temperature sensitive.
The e†ect on rate of altering the partial pressure of phenol
(P ) and hydrogen (P ) was also considered and the results are
in S
over the Pd/CaÈMgO must be due to the availability
P
H
CJO
presented in Fig. 11. Variations in reactant(s) partial pressure
had no e†ect on reaction selectivity. The rate of hydro-
genation (R, mol s~1 (g cat)~1) can be expressed in terms of
partial pressures using the power law
of more basic sites on magnesia which increase the fraction of
strongly bound phenol (as a surface phenolate15), favouring
complete hydrogenation. The incorporation of Ñuoride in the
R \ k(P )x(P )y
Table 2 Kinetic data for phenol hydrogenation over 1 wt.% Pd/
P
H
MgO: PdCl precursor
where x and y are the reaction orders with respect to phenol
and hydrogen, respectively. The values of x and y determined
from the linear relationships (as plotted in Fig. 11) at four
representative temperatures and the corresponding apparent
rate constants (k) are given in Table 2. At the lowest tem-
peratures, the reaction order with respect to phenol was nega-
tive but became positive and increased in magnitude at higher
temperatures. The negative value ([0.5) at 473 K is indicative
of an appreciable inhibition of rate with increasing phenol
2
Reaction order
~1
T /K
phenol (x)
hydrogen (y)
k/mol s~1 (g cat) atm~(x`y)
473
503
533
563
[0.53
[0.33
0.17
0.42
0.50
0.69
0.85
3.7 ] 10~7
7.3 ] 10~7
2.5 ] 10~6
3.8 ] 10~6
0.46
Phys. Chem. Chem. Phys., 2001, 3, 2712È2719
2717

View Article Online
Table 3 E†ect of additives on Pd size/dispersion and the degree of phenol conversion/hydrogenation selectivity: Pd(CH COO) precursor;
3
2
T \ 503 K; WHSV \ 0.027 mol h~1 (g cat)~1; hydrogen/phenol \ 5.4
Selectivity (%)
Catalyst
Dispersion CO /Pd
d/nm from CO
Conversion (%)
S
S
irr
irr
CJO
ChOH
Pd/MgO
Pd/CaÈMgO
Pd/FÈMgO
0.56
0.55
0.36
2.0
2.1
3.1
90
90
54
80
60
80
20
40
20
support served to induce Pd crystallite growth (Table 3) which
is in keeping with our observations regarding the chlorine-
containing Pd precursors. The lower phenol conversion over
Pd/FÈMgO can likewise be attributed to a less e†ective acti-
vation of reactants on the surface. Based on ammonia chemi-
sorption data, the acid site concentration was raised from 7 to
18 mol (g cat)~1 due to the incorporation of F onto the
support. Phenol conversion over this catalyst was character-
ized by a decided decline in activity with time-on-stream, as
shown in Fig. 12, while the parent Pd/MgO and Ca-doped
samples both delivered appreciably more stable conversions.
Loss of activity in the case of Pd/FÈMgO can be ascribed to
coke formation on the acidic sites.
induced a desorption of surface phenol and an accompanying
decrease in reaction rate. The hydrogenation of phenol occurs
stepwise where cyclohexanone and cyclohexanol are the par-
tially and fully hydrogenated products, respectively; alter-
ations to such process variables as Pd loading, H /phenol mol
2
fraction and inlet phenol molar feed rate that serve to increase
the degree of conversion favour complete hydrogenation to
cyclohexanol. The incorporation of CaO into the magnesia
support had no e†ect on conversion but decreased cyclo-
hexanone selectivity by strengthening the phenol/catalyst
interaction(s); the presence of Ñuoride on the support lowered
activity but had no impact on selectivity.
Acknowledgements
The authors wish to thank the University Grants Commission
(UGC), New Delhi, for the award of a Senior Research Fel-
lowship to NM.
Conclusion
The gas-phase hydrogenation of phenol over Pd/MgO can be
classiÐed as a structure insensitive reaction where the inferred
Pd crystallite size falls within the range 3È7 nm. At lower par-
ticle sizes there is a decided drop in TOF which we attribute
to an electron deÐciency of the Pd sites that translates into a
less e†ective surface activation of the reactants. Palladium
dispersion/particle size is dependent on both the nature of the
Pd precursor and Pd loading: chlorine-containing precursors
yield larger Pd particles while dispersion decreases with
increasing Pd content. Palladium particle sizes measured by
CO chemisorption are lower than those obtained from
HRTEM analysis but the trends are the same. The calcined
samples are characterized by crystalline PdO and MgO
References
1
I. Dodgson, K. Griffin, G. Barberis, F. Piganttaro and G.
Tauszik, Chem. Ind. (L ondon), 1989, 830.
J. F. Van Peppen, W. B. Fischer and C. H. Chan, in Catalysis in
Organic Reactions, ed. R. L. Augustine, Marcel Dekker, New
York, 1980, p. 141.
2
3
4
5
6
G. Neri, A. M. Visco, A. Donato, C. Milone, M. Malentacchi and
G. Gubitosa, Appl. Catal. A, 1994, 110, 49.
V. Vishwanathan and S. Narayanan, J. Chem. Soc., Chem.
Commun., 1990, 78.
V. Vishwanathan, M. S. Rajasekhar, G. Sreekanth and S. Naray-
anan, J. Chem. Soc., Faraday T rans., 1991, 87, 3449.
S. Narayanan and R. Unnikrishnan, J. Chem. Soc., Faraday
T rans., 1997, 93, 2009.
S. Narayanan and K. Krishna, Appl. Catal. A, 1996, 147, L253.
P. Claus, H. Berndt, C. Mohr, J. Radnik, E.-J. Shin and M. A.
Keane, J. Catal., 2000, 192, 88.
S. Galvagno, A. Donato, G. Neri and R. Pietropaolo, J. Chem.
T echnol. Biotechnol., 1991, 51, 145.
phases with no evidence of surface Mg(OH) ; crystallinity of
the MgO support is also evident from HRTEM analysis. The
Pd/MgO catalysts derived from PdCl and Pd(NH ) Cl
2
2
3 4 2
retained an appreciable residual surface chlorine content after
activation which served to lower the hydrogenation rate and
induce a loss of activity with time-on-stream; Pd/MgO pre-
pared from Pd(CH COO) delivered a higher and more stable
phenol conversion. The reaction orders with respect to phenol
and hydrogen suggest that the rate-controlling step is a
surface reaction between strongly bound phenol and weakly
adsorbed hydrogen. An increase in reaction temperature
7
8
9
3
2
10 A. K. Talukdar, K. G. Bhattacharyya and S. Sivasanker, Appl.
Catal. A, 1993, 96, 229.
11 N. Itoh and W.-C. Xu, Appl. Catal. A, 1993, 107, 83.
12 N. K. Nag, Catal. L ett., 1994, 24, 37.
13 T. E. Holt, A. D. Logan, S. Chakraborti and A. K. Datye, Appl.
Catal., 1987, 34, 199.
14 Y. Z. Chen, C. W. Liaw and L. I. Lee, Appl. Catal. A, 1999, 177, 1.
15 A. K. Talkudar, K. G. Bhattacharyya and S. Sivasanker, Appl.
Catal. A, 1993, 96, 229.
16 S. Narayanan, B. Prabhu Prasad and V. Vishwanathan, React.
Kinet. Catal. L ett., 1992, 48, 562.
17 S. Narayanan and G. Sreekanth, Ind. J. T echnol., 1993, 31, 507.
18 S. T. Srinivas, L. J. Lakshmi and P. K. Rao, Appl. Catal. A, 1994,
110, 167.
19 E.-J. Shin and M. A. Keane, Ind. Eng. Chem. Res., 2000, 39, 883.
20 N. Mahata and V. Vishwanathan, Ads. Sci. T echnol., 1997, 15,
165.
21 N. Mahata and V. Vishwanathan, unpublished results.
22 J. R. Anderson, Structure of Metallic Catalysts, Academic Press,
London, 1975, p. 296.
23 Y. L. Lam and M. Boudart, J. Catal., 1977, 47, 393.
24 M. A. Vannice, S.-Y. Wang and S. H. Moon, J. Catal., 1981, 71,
152.
25 S. H. Ali and J. G. Goodwin Jr., J. Catal., 1998, 176, 3.
26 N. Mahata and V. Vishwanathan, Ind. J. Chem. A, 1998, 37, 652.
27 W. Zou and R. D. Gonzalez, Catal. L ett., 1992, 12, 73.
Fig. 12 Time-on-stream activity proÐles for the conversion of phenol
over Pd/MgO (…), Pd/CaÈMgO (]) and Pd/FÈMgO (=): T \ 503
K; WHSV \ 0.027 mol h~1 (g cat)~1; hydrogen/phenol \ 5.4.
2718
Phys. Chem. Chem. Phys., 2001, 3, 2712È2719

View Article Online
28 W. Vogel, H. Knozinger, B. T. Carville, W. M. H. Sachtler and
Z. C. Chang, J. Phys. Chem. B, 1998, 102, 1750.
29 C. Hoang-Van, Y. Kachaya, S. J. Teichner, Y. Arnaud and J. A.
Dalmon, Appl. Catal., 1989, 46, 281.
41 R. J. Verderone, C. L. Pieck, M. R. Sad and J. M. Parera, Appl.
Catal., 1986, 21, 329.
42 M. Che and C. O. Bennett, Adv. Catal., 1989, 54, 359.
43 J. C. Vedrine, M. Dufaux, C. Naccache and B. Imelik, J. Chem.
Soc., Faraday T rans., 1978, 74, 440.
44 N. Itoh and W.-C. Xu, Appl. Catal. A, 1993, 107, 83.
45 S. Narayanan and G. Sreekanth, React. Kinet. Catal. L ett., 1993,
51, 449.
46 S. G. Lias, J. F. Liebman, R. D. Levin and S. A. KafaÐ, NIST
Standard Reference Database 25, Structure and Properties,
Version 2.01, January 1994.
30 X. Lu, X. Nin, N. Wang and Q. Zhang, J. Phys. Chem. B, 2000,
104, 10024.
31 T. A. Dorling and R. L. Moss, J. Catal., 1967, 7, 378.
32 J. S. Smith, P. A. Thrower and M. A. Vannice, J. Catal., 1981, 68,
270.
33 A. R. Gonzalez-Elipe, G. Munuera and J. P. Espinos, Surf. Inter-
face Anal., 1990, 16, 375.
34 J. W. E. Coenen, Appl. Catal., 1991, 75, 193.
35 Y. Ohtsuka, J. Mol. Catal., 1989, 54, 225.
36 S. Zhuang, J. Wu, X. Liu, J. Tu, M. Ji and K. Wandelt, Surf. Sci.,
1995, 331, 42.
47 E.-J. Shin and M. A. Keane, J. Catal., 1998, 173, 450.
48 B. Xu, T. Yamaguchi and K. Tanabe, Mater. Chem. Phys., 1980,
19, 291.
49 J. R. Gonzalez-Velasco, J. I. Gutierrez-Oritz, J. A. Gonzalez-
Marcos and A. Romero, React. Kinet. Catal. L ett., 1986, 32, 505.
50 M. E. Dry, T. Shingles, L. J. Bosho† and G. J. Oosthuizen, J.
Catal., 1969, 15, 190.
51 D. T. Wickhaman, B. W. Logdson, S. W. Cowley and C. D.
Butler, J. Catal., 1991, 128, 198.
52 J. H. Lunsford, Stud. Surf. Sci. Catal., 1991, 61, 3.
37 M. Windaw and D. P. Zhang, Catal. T oday, 1996, 30, 99.
38 E.-J. Shin, A. Spiller, G. Tavoularis and M. A. Keane, Phys.
Chem. Chem. Phys., 1999, 1, 3173.
39 A. Chambers and R. T. K. Baker, J. Phys. Chem. B, 1997, 101,
1621.
40 C. F. Cullis, J. E. Manton, G. B. Thomas and H. Wilman, Acta.
Crystallogr., 1959, 12, 382.
Phys. Chem. Chem. Phys., 2001, 3, 2712È2719
2719