Structure sensitivity and kinetics of D-glucose oxidation to D-gluconic acid over carbon-supported gold catalysts

Article abstract of DOI:10.1016/j.jcat.2004.01.010

The heterogeneously catalyzed oxidation of D-glucose to D-gluconic acid over Au/C catalysts has been studied. A series of Au/C catalysts were prepared by the gold sol method with different reducing agents and different kinds of carbon support providing Au mean particle diameters in the range 3-6 nm. The activities of these catalysts with respect to D-glucose oxidation were compared, and several aspects influencing activity, especially Au particle size, were discussed. The influence of reaction conditions (T=30-90°C, pH 7.0-9.5) on the kinetics of the D-glucose oxidation has been examined using the most active Au/C catalyst. By a detailed analysis of all reaction products under different reaction conditions, a reaction network of the D-glucose oxidation is presented, and a reaction mechanism for D-glucose oxidation that explains the influence of pH on reaction rate is proposed. Ensuring that D-glucose oxidation takes place in the kinetic regime (sufficient stirring rate and airflow rate), a semiempirical model based on a Langmuir-Hinshelwood-type reaction pathway is assumed. At 50°C and pH 9.5 kinetic parameters were calculated by an optimization routine. The resulting concentration courses of D-glucose and D-gluconic acid were in good agreement with the experimental data. All experiments were carried out in a semibatch reactor under pH control at atmospheric pressure.

Full text of DOI:10.1016/j.jcat.2004.01.010

Journal of Catalysis 223 (2004) 122–133
www.elsevier.com/locate/jcat
Structure sensitivity and kinetics of D-glucose oxidation to D-gluconic
acid over carbon-supported gold catalysts
Y. Önal, S. Schimpf, and P. Claus ∗
Department of Chemistry, Chemical Technology II, Darmstadt University of Technology, Petersenstrasse 20, D-64287 Darmstadt, Germany
Received 10 October 2003; revised 6 January 2004; accepted 6 January 2004
Abstract
The heterogeneously catalyzed oxidation of D-glucose to D-gluconic acid over Au/C catalysts has been studied. A series of Au/C catalysts
were prepared by the gold sol method with different reducing agents and different kinds of carbon support providing Au mean particle
diameters in the range 3–6 nm. The activities of these catalysts with respect to D-glucose oxidation were compared, and several aspects
◦
influencing activity, especially Au particle size, were discussed. The influence of reaction conditions (T = 30–90 C, pH 7.0–9.5) on the
kinetics of the D-glucose oxidation has been examined using the most active Au/C catalyst. By a detailed analysis of all reaction products
under different reaction conditions, a reaction network of the D-glucose oxidation is presented, and a reaction mechanism for D-glucose
oxidation that explains the influence of pH on reaction rate is proposed. Ensuring that D-glucose oxidation takes place in the kinetic regime
(sufficient stirring rate and airflow rate), a semiempirical model based on a Langmuir–Hinshelwood-type reaction pathway is assumed. At
◦
50 C and pH 9.5 kinetic parameters were calculated by an optimization routine. The resulting concentration courses of D-glucose and
D-gluconic acid were in good agreement with the experimental data. All experiments were carried out in a semibatch reactor under pH
control at atmospheric pressure.
2004 Elsevier Inc. All rights reserved.
Keywords: D-Glucose oxidation; Au/C catalysts; Structure sensitivity; Specific gold surface area; Gold nanoparticle; Gold sol method; Mass transfer;
Langmuir–Hinshelwood kinetics
1. Introduction
or on oxides [6] is the preferred catalyst for the oxidation
of functional groups (–OH, C=O). Even when their mean
gold particle sizes were similar, carbon proved to be more
advantageous as a support than oxidic materials in the liquid-
phase oxidation of alcohols and aldehydes. The gold sol
method is an easy and effective way of preparing highly ac-
tive carbon-supported Au catalysts with small gold particle
diameters (4–9 nm) [7–9]. The particle diameter can thereby
be easily influenced by variation of certain preparation pa-
rameters (reduction agent, concentrations of reagents [8],
temperature [10]). By use of this preparation method, Au/C-
catalyzed oxidation of D-glucose to D-gluconic acid in the
aqueous phase has already been examined by Prati et al. [5].
In particular, the influence of pH value on reaction rate was
studied in comparison with other catalysts based on Pt group
metals. Au/C was found to be the most active catalyst inde-
pendent of pH value, and it could be shown that the activity
of the Au/C catalyst was less sensitive to lower pH values.
With respect to our research program on the catalytic
chemistry of renewable resources including the oxida-
D-Gluconic acid is an important intermediate in the field
of food industry and pharmaceutical applications and is usu-
ally produced by enzymatic oxidation of D-glucose [1] with
an estimated market of 60,000 tons per year [2]. Besides
this route, recent developments have shown the potential of
heterogeneous catalysts for oxidizing glucose with oxygen
or air. Most of the catalysts are based on supported plat-
inum group metals, and various factors controlling activity
and selectivity were studied. The main disadvantage is de-
activation of the catalysts with increasing reaction time, i.e.,
with conversion. Improvementof activity/selectivity and sta-
bility can be achieved in the presence of a second metal in
ex situ-prepared bimetallic catalysts (especially bismuth as
promotor) [3] and by the nature of the individual metal [4].
With respect to the latter, gold on activated charcoal [5]
*
Corresponding author.
E-mail address: claus@ct.chemie.tu-darmstadt.de (P. Claus).
0021-9517/$ – see front matter
doi:10.1016/j.jcat.2004.01.010
2004 Elsevier Inc. All rights reserved.

Y. Önal et al. / Journal of Catalysis 223 (2004) 122–133
123
tion [11] and hydrogenation of sugars [12] and stimulated
by the promising results of liquid-phase oxidation using
gold catalysts [5], we report first in this article the char-
acterization of carbon-supported gold catalysts by transmis-
sion electron microscopy (TEM) and X-ray photoelectron
spectroscopy (XPS) with the aim of studying the influence
of gold particle size and surface composition on activity
and selectivity (structure sensitivity) of glucose oxidation
over gold catalysts. Then, we focus on the kinetics under
optimized reaction conditions (pH 9.5, T = 50 ◦C) which
has never been investigated before. For this purpose, Au/C
catalysts with different mean Au particle diameters were
prepared by the gold sol method using different reducing
agents (NaBH4 and tetrahydroxymethylphosphonium chlo-
ride, THPC) and carbon supports. The kinetics of glucose
oxidation was estimated by appropriate experiments with
the most active Au/C catalyst at different glucose and cat-
alyst concentrations in the kinetic regime at 50 ◦C and pH
9.5. Parameters of the semiempirical reaction rate, which
was proposed assuming a Langmuir–Hinshelwood model,
were calculated by an optimization routine.
Scheme 1. Overview of the preparation/activation procedure.
is illustrated in Scheme 1. The amounts of the reagents given
in the text were calculated in order to prepare 5 g Au/C cat-
alyst with an Au loading of 1 wt%.
2.1.2.1. THPC/NaOH method H2O (232 mL) was first
mixed with (0.2 M) NaOH (7.5 mL), and afterward 5 mL
THPC (1.2 mL of the 78 wt% THPC solution was diluted
with water to 100 mL) was added into the alkaline solution
while the mixture was stirred. After 2 more min of stirring,
6 mL HAuCl4 (0.043 M) was introduced into the solution.
Formation of the gold sol (brown color of solution) occurred
immediately. Then, 5 g carbon was suspended in 500 mL
H2O and agitated for 15 min in an ultrasonic bath. The gold
sol was then added to the support suspension.
2. Experimental
2.1.2.2. PVA/NaBH4 method HAuCl4 (0.0877 g, 258
mmol) was dissolved in 1.725 L H2O and then, while the
mixture was stirred, mixed with 4 mL PVA solution (2 wt%).
Afterward, 10 mL NaBH4 (0.1 M) was dropped into the
mixture. Again, the formation of the gold sol occurred im-
mediately. Then 5 g carbon was added to the gold sol.
Both with the THPC/NaOH method and the NaBH4/PVA
method, immobilization of the gold sol on the support
was completed after 1–3 days. During this time the gold
sol became colorless. The adsorbed support was then fil-
tered, washed with water to free the filtrate of chloride
(AgNO3 test), dried, and mortared. The organic scaffold
was removed from the support by inert gas treatment
(3 h/350 ◦C/He). The catalysts were activated by a reduc-
ing process (3 h/350◦C/H2).
2.1. Catalyst preparation
2.1.1. Reagents
THPC (78 wt%), polyvinyl alcohol (PVA) (> 98%), and
NaOH (0.2 M) were purchased from Merck and HAuCl4·
3H2O (99.99% ACS, Au 49.5% min.) from Alfa Aesar. Two
types of carbon supports were used, namely, Black Pearls
2000 (GP3755) and Vulcan XC72R (GP3759), which were
kindly listed by CABOT GmbH. Characteristic support data
are listed in Table 1.
2.1.2. Preparation procedure
Au/C catalysts were prepared by the gold sol method with
two different types of reducing agents, NaBH4 and THPC,
which lead to small Au particle sizes as reported in [7–10].
Depending on the reduction agent used, the preparation
methods are slightly different as described in the literature.
The principle of the preparation procedures described below
2.2. Catalyst characterization
The metal content of the Au/C catalysts was determined
by optical emission spectroscopy with inductive coupled
plasma (ICP-OES). For this purpose the samples were solu-
bilized in aqua regia (HCl/HNO3).
Table 1
a
Characteristic data of the carbon supports used
Mean particle diameters of the gold on the carbon sup-
port were determined by TEM analysis. The samples were
dispersed in isopropanol, agitated in an ultrasonic bath, and
deposited on a commercially available Cu support net (grid
300, Plano). Particle size of the reduced samples was exam-
ined in the light- and darkfield modes in association with
selected area electron diffraction (SAED) with a JEM 100C
(100 kV). Mean particle diameters were calculated on the as-
sumption of a logarithmic normal distribution of the particle
sizes.
Property
Black Pearls 2000
GP3755
Vulcan XC72R
GP3759
2
Specific surface area (m /g)
Micropore volume (mL/g)
Iodine number (mg/g)
Total sulfur (%)
1635
0.59
1400
1.0
9.5
200
150
245
0.06
253
0.5
6.0
25
pH
+325 mesh residue [max ppm]
Density (g/L)
270
a
Data from manufacturer [23].

124
Y. Önal et al. / Journal of Catalysis 223 (2004) 122–133
X-Ray photoelectron spectra were obtained with a VG
Table 2
Au content of the prepared Au/C catalysts resulting from ICP-OES analysis
ESCALAB 220 iXL equipped with a monochromated Al-Kα
radiation source (1486.6 eV) and pass energies of 150 eV
(overview spectra) and 25 eV (high-resolution mode). De-
tails of both the procedure applied to the gold catalysts and
the data evaluation were published recently [13]. Peak posi-
tions could be determined with a precision of 0.1 eV. Gold
catalysts were analyzed as prepared and after two different
pretreatment procedures, i.e., heating in flowing argon and
hydrogen at 350 ◦C. This pretreatment was carried out in
a gas cell integrated into the lock of the system at the de-
sired temperature and in the gas flux. Thus, it was possible
to transfer the sample from this gas cell to the UHV analy-
sis chamber without air contact. The pressure in the analysis
chamber was below 1 × 10−6 Pa.
Catalyst
Au content (wt%)
Au/C GP3755 PVA
Au/C GP3759 PVA
Au/C GP3755 THPC
Au/C GP3759 THPC
0.60
0.42
0.70
0.48
discrepancy between the calculated and analyzed Au con-
tent can be explained by a nonquantitative immobilization of
the gold sol on the support, probably depending on the mi-
crostructure and chemical composition of the carbon varying
both the distribution of the gold particles and their interac-
tion with the support.
TEM images of the catalyst surface are shown in Fig. 1
together with the average gold particle diameter and its stan-
dard deviation. The catalysts contain nanosized gold parti-
cles in the range 3 to 6 nm. By variation of the preparation
method (PVA/NaBH4 vs THPC/NaOH, see Section 2.1.2)
and use of the same support (GP3755 or GP3759), gold par-
ticle diameter decreases as shown in Fig. 1 by comparing
the upper and lower TEM images. A more striking decrease
in gold particle size was observed when only the carbon
support was varied for the same preparation method ap-
plied (compare left- and right-hand side images in Fig. 1).
Generally, the results of TEM analysis clearly show that
the mean Au particle size is considerably smaller on the
carbon support GP3759 (Vulcan) than on GP3755 (Black
Pearls), independent of the catalyst preparation method ap-
plied.
Whereas the characterization of oxide-supported gold
catalysts by X-ray photoelectron spectroscopy exhibited pro-
nounced shifts in Au 4f7/2 binding energy to lower val-
ues [13] compared with the binding energy of metallic gold
(84.2 eV [14]), XPS measurements on the carbon-supported
gold catalysts in this study gave an average value of 84.15
0.05 eV for binding energy, independent of the pretreatment
method applied. This corresponds to the value of metallic
gold.
2.3. Liquid-phase oxidation of glucose
Catalytic runs were carried out in a 250-mL five-necked
glass reactor. The reactor was equipped with several inlets
and detectors to monitor the reaction conditions. During the
experiment, the temperature of the three-phase reaction mix-
ture and the pH value (pH electrode from HANNA Instru-
ments), which also depends on temperature, were indicated.
By a drop funnel filled with NaOH solution (4 wt%), the pH
value could be maintained at a certain value. Air was bub-
bled through the reaction mixture at a constant flow rate at
atmospheric pressure. The reactor was additionally equipped
with a cooler, so that any evaporation of the liquid phase
could be neglected. A hotplate equipped with a stirrer pro-
vided the heat needed to achieve the desired reaction tem-
perature and an intensive mixing of the reaction mixture.
A typical catalytic run was carried out with 100 mL
D-glucose solution (4 wt%) and 432 mg Au/C catalyst (mo-
lar ratio of D-glucose to Au: 1000). Both the catalyst and the
D-glucose solution were first mixed together, and at a low
stirring rate the mixture was heated to the desired temper-
ature. The reaction was started by accelerating the stirring
rate, switching on the air supply, and adjusting to a certain
pH value by dropping an adequate amount of NaOH solu-
tion. During the whole reaction, pH was kept at the desired
value by continuous NaOH drop. The products of the reac-
tion mixture at different reaction periods were quantitatively
analyzed by HPLC (Bio-Rad, Aminex HPX-87H). The reac-
tion was complete when there was no significant change of
pH value.
3.2. Optimization of the reaction conditions
3.2.1. Reaction temperature
The influence of reaction temperature (other reaction con-
ditions remained constant) on the reaction rate of D-glucose
oxidation was estimated at an alkaline (pH 9.5) as well as
neutral (pH 7.0) pH value. The catalyst used was Au/C
GP 3759 THPC, which had proved to be particularly ac-
tive in previous experiments. The initial reaction rates of
D-gluconic acid formation at different reaction temperatures
and pH values are depicted in Fig. 2. An optimal tempera-
ture range was found at which the reaction rate increased as
high as possible. This range is around 50 ◦C at pH 9.5 and
60 ◦C at pH 7.0. The decrease in reaction rate of D-gluconic
acid formation at higher temperatures can be explained by
the side reactions of D-glucose in the aqueous phase which
3. Results
3.1. Catalyst characterization
Gold content of the prepared Au/C catalysts is summa-
rized in Table 2. According to the ICP-OES analysis, the
amount of gold on each support, independent of the prepa-
ration method, was less than 1 wt% (0.42–0.70 wt%). The

Y. Önal et al. / Journal of Catalysis 223 (2004) 122–133
125
Fig. 1. TEM images and mean gold particle diameters.
Fig. 2. Initial reaction rates of D-gluconic acid formation at different reac-
tion temperatures and pH values (catalyst: Au/C GP3759 THPC).
Fig. 3. Formation of by-products at pH 7.0 (t = 300 min) and pH 9.5
(t = 60 min) as a function of reaction temperature (catalyst: Au/C GP3759
THPC).
are favored at higher temperatures and pH values, as shown
in Fig. 3. Therefore, significant by-products of D-glucose ox-
idation are fructose, mannose, glycolaldehyde, sorbitol, and
maltose as described in the reaction network of glucose ox-
idation given in Section 4.3 (cf. Fig. 11). The decrease in
selectivity to D-gluconic acid with increasing temperature is
shown in Fig. 4. At alkaline pH values, the degree of depen-
dency of the reaction rate and selectivity to D-gluconic acid
on temperature is higher than at lower pH values.

126
Y. Önal et al. / Journal of Catalysis 223 (2004) 122–133
Table 3
Calculated Au specific surface areas of different catalysts based on ICP-
OES and TEM results
2
Catalyst
A
spec
(m /g)
Au/C GP 3755 PVA
Au/C GP 3759 PVA
Au/C GP 3755 THPC
Au/C GP 3759 THPC
51.0
84.0
62.2
94.2
Fig. 5. Initial reaction rates of D-gluconic acid formation and selectivities
◦
to D-gluconic acid (pH 9.5 and T = 50 C), as a function of specific Au
surface area: F, GP3755; 2, GP3759.
D-glucose oxidation at 50 ◦C and pH 9.5, as these values
represent the optimized reaction conditions with respect to
reaction rate and selectivity as presented above. The goal
was to find a correlation between catalytic activity and
Au particle diameter which has not been investigated so
far.
Fig. 4. Selectivity to D-gluconic acid (top) and conversion of glucose
(bottom) at different pH and temperature values (catalyst: Au/C GP3759
THPC).
3.2.2. pH value
Because the Au contents of the catalysts differ widely
(Table 2), catalytic activity could not be correlated only
with Au mean particle diameter. Therefore, specific sur-
face area (Aspec) of the active component of each cata-
lyst was calculated based on the results of ICP-OES and
TEM analysis by assuming closed-shell gold particles of
nearly spherical shape [15] (Table 3). As the distribution
of the particle diameter of each catalyst was nearly sym-
metric, mean particle diameters were used to calculate the
specific surface areas. Initial reaction rates of D-gluconic
acid formation and selectivities to D-gluconic acid as a
function of specific surface area are illustrated in Fig. 5.
In the field examined, catalytic activity correlates expo-
nentially with specific surface area, whereas selectivity to
D-gluconic acid does not change with increasing specific
surface area. According to the results presented, there must
be a particle size effect of Au/C catalysts in the aqueous-
phase oxidation of D-glucose to D-gluconic acid and, thus,
the structure sensitivity of the reaction is satisfactorily
shown.
Au catalysts are less sensitive to lower pH values with
respect to their activity in the oxidation of D-glucose than
catalysts based on Pd group metals according to [5]. As
shown in Fig. 2, the reaction rate of D-gluconic acid for-
mation could be strongly increased at alkaline pH values.
By increasing the pH value of the reaction mixture from 7.0
to 9.5 (T = 50 ◦C), the initial reaction rate of D-gluconic
acid formation could be accelerated by a factor of 3.2. An
important drawback of this phenomenon is the decrease in
selectivity to D-gluconic acid at higher pH values (Fig. 4),
mainly by isomerization of D-glucose to fructose. However,
in contrast to the results presented in [5], the activity of Au/C
catalysts in the present study exhibited significant depen-
dence on the pH value.
3.3. Influence of the Au particle diameter on the catalytic
activity
Catalytic activity of catalysts with different mean Au
particle diameters (Fig. 1) was examined with respect to

Y. Önal et al. / Journal of Catalysis 223 (2004) 122–133
127
Fig. 7. Initial reaction rates of D-gluconic acid formation and selectivities
to D-gluconic acid (t = 30 min) at different initial D-glucose concentrations
◦
(T = 50 C, pH 9.5, catalyst: Au/C GP3759 THPC).
Fig. 6. Initial reaction rates of D-gluconic acid formation at different catalyst
concentrations (T = 50 C, pH 9.5, catalyst: Au/C GP3759 THPC).
◦
3.4. Kinetics
3.4.1. Influence of catalyst concentration and D-glucose
concentration on reaction rate
To ensure that the reaction rate is not limited by mass
transfer at the phase boundaries, the effects of stirring rate as
well as airflow rate on reaction rate were examined at 50 ◦C
and pH 9.5. It was found that at stirring rates higher than
900 rpm and airflow rates higher than 1600 mL/min, the
overall reaction rate is not limited by external mass transfer
(diffusion). Kinetic measurements with varied catalyst and
D-glucose concentrations were therefore carried out at a stir-
ring rate of 900 rpm and an airflow rate of 2000 mL/min.
The influence of catalyst concentration on the reaction rate
of D-gluconic acid formation at 50 ◦C and pH 9.5 with the
Au/C GP3759 THPC catalyst is presented in Fig. 6, which
demonstrates the asymptotic dependence of reaction rate on
catalyst concentration.
Initial reaction rates of D-gluconic acid formation and se-
lectivities to D-gluconic acid at different initial D-glucose
concentrations are shown in Fig. 7. Obviously the depen-
dence of reaction rate on initial D-glucose concentration is
negligible. If the overall reaction rate of D-glucose oxida-
tion was limited by adsorption of D-glucose on the catalyst
surface, the reaction rate should show a strong increase with
increasing initial D-glucose concentration. According to the
results presented here, the overall reaction rate is limited by
the surface reaction or the desorption of D-gluconic acid.
As the isomerization of D-glucose to fructose increases with
increasing initial D-glucose concentration, the selectivity to
D-gluconic acid slightly decreases.
Fig. 8. D-Glucose oxidation over Au according to the Langmuir–
Hinshelwood model.
tive dehydrogenation mechanism commonly passed in the
liquid-phase oxidation of alcohols and aldehydes is well es-
tablished in the literature [16] and was also adopted in the
oxidation of D-glucose. Sorption of the D-glucose hydrate on
the catalyst surface is represented by the sorption constant
KG, whereas k1 is the rate constant for the dehydrogena-
tion reaction on the catalyst surface and KGS is the sorption
constant of D-gluconic acid. Hydrides bound to the catalyst
surface as a consequence of dehydrogenationof the substrate
subsequently react with dissociatively adsorbed oxygen to
form water, which is desorbed. The oxidation reaction of
the hydrides should be fast compared with the dehydrogena-
tion reaction. Thus, the whole reaction mechanism can be
reduced to three elementary reactions, namely, adsorption of
D-glucose, surface reaction of D-glucose, and desorption of
gluconic acid.
According to the results presented in Section 3.4.1, it
was ruled out that the reaction is limited by adsorption of
D-glucose on the catalyst surface. As all kinetic experiments
were performed in an alkaline medium, the desorption of
gluconic acid should also be faster than the surface reac-
tion [10,17]. Thus, it can be concluded that the surface reac-
tion is the rate-determining step of the overall reaction and
is expressed in the equation
3.4.2. Kinetic model
Oxidation of D-glucose in the aqueous phase over Au/C
catalysts with oxygen can be described by a Langmuir–
Hinshelwood-type model which is schematically repre-
sented by the catalytic cycle shown in Fig. 8. The oxida-
rsurface = rOR = k1θG − k−1θGS
.
(1)

128
Y. Önal et al. / Journal of Catalysis 223 (2004) 122–133
Consequently, the other elementary steps are in equilibrium,
rSor,G = rSor,GS = 0, (2)
4. Discussion
4.1. Influence of reaction conditions on kinetics
The effect of temperature on reaction rate and selectiv-
ity to D-gluconic acid was examined with the catalyst Au/C
GP3759 THPC at pH 7.0 as well as at pH 9.5. In both
cases, the highest reaction rates were obtained in the tem-
perature regime 50 to 60 ◦C, while the temperature has a
more important effect on the reaction rate at higher pH val-
ues. According to ICP-OES analysis of the product solu-
tions obtained after glucose oxidation, Au leaching could be
neglected within the detection limit of the method applied
(< 2 mg/L) at any reaction temperature. Thus, a decrease
in activity caused by catalyst deactivation, i.e., by loss of
metal, with increasing reaction temperature can be excluded.
In the literature [5], analysis of fresh and used catalysts re-
vealed leaching of gold when the oxidation was carried out
at pH 7, whereas no leaching was observed at pH 9.5. In
the present study, however, no correlation of the dependence
of the reaction rate on pH value with Au leaching could be
found based on ICP-OES analysis. The temperature effect
can be explained by the following phenomena. At both pH
7.0 and pH 9.5, selectivity to D-gluconic acid decreases with
increasing temperature due to side reactions, as discussed
in Section 3.2.1. Moreover, another aspect that has to be
taken into account is the solubility of oxygen in water, which
decreases with increasing temperature [19]. Thus, the con-
centration of oxygen in the aqueous phase should limit the
overall reaction rate of D-glucose oxidation at higher reac-
tion temperatures. On the other hand, at lower temperatures,
the reaction is limited by kinetics as shown in Fig. 2.
and the overall reaction rate of D-gluconic acid formation
can be expressed as
d[GS]
dt
k1KG[G]
rOR
=
=
· MC.
(3)
1 + KG[G] + KGS[GS]
Because the rate of conversion of D-glucose is faster than
the rate of formation of D-gluconic acid due to additional
side reactions, the latter were considered by a further kinetic
term in the overall reaction rate of D-glucose consumption:
d[G]
dt
KGk1[G]
−
=
· MC + kP[G].
(4)
1 + KG[G] + KGS[GS]
Taking the selectivity of the reaction at 50 ◦C and pH 9.5
and already calculated values of kinetic parameters for D-
glucose oxidation over Pd/Al2O3 catalysts [18] (Table 4)
into account, the empirical constant kP was suggested to be
6 × 10−5
s
−1. The product of KG and k1 was expressed as
kOx in the program code for the optimization routine. Initial
values for the parameter optimization are from the litera-
ture [18]. Fig. 9 presents the kinetic parameters determined
and the concentration courses of D-glucose and D-gluconic
acid calculated, which are in good agreement with the ex-
perimental data obtained at 50 ◦C and pH 9.5 with the Au/C
GP3759 THPC catalyst.
Table 4
The pH value of the reaction mixture has a crucial ef-
fect on the reaction rate of D-gluconic acid formation. The
reaction rate clearly accelerates with increasing pH value.
A possible explanation for this phenomenon is the neglect
of catalyst deactivation by self-poisoning in alkaline solu-
tion, as D-gluconic acid is then deprotonated and, thus, no
longer capable of blocking active centers on the catalyst sur-
face [3,4,17]. Fig. 10 illustrates the concentration courses of
D-gluconic acid and temporal reaction rates of D-gluconic
acid formation at 50 ◦C and pH 7.0 or 9.5, respectively. After
certain reaction periods which correspond to the same con-
version degrees at different reaction conditions, the decrease
in reaction rate at pH 7.0 was greater than at pH 9.5, namely,
by a factor of nearly 2. Thus, significant catalyst deactivation
by self-poisoning should be taken into account when the re-
action is not performed in an alkaline reaction medium. On
the other hand, selectivity to D-gluconic acid decreases at
higher pH values (ꢀ 9.5) as a result of isomerization and
irreversible by-product formation catalyzed in alkaline solu-
tions (Fig. 11).
◦
Kinetic parameters of D-glucose oxidation over Pd/Al O catalysts at 50 C
and pH 9 [18]
2
3
−4
−1 −1
s
k
= 7.9 × 10 L g
Ox
−4
−1
K
= 138.0 × 10 L mmol
G
−4
−1
K
= 279.0 × 10 L mmol
GS
4
−1 −1
6
−1
,
k
= 7.1 × 10 L g
s
−1
, K = 4.9 × 10 L mmol
Ox
G
5
−3
−1 −1
s
K
= 7.2 × 10 L mmol , k = 4.6 × 10 mmol g
GS
1
If, however, only self-poisoning was responsible for the
different catalytic activities at different pH values, reaction
rates at the beginning of each experiment should be at least
similar without consideration of the pH value, because at the
Fig. 9. Comparison of calculated (—) and experimental (∗) concentrations
◦
versus time of D-glucose and D-gluconic acid (T = 50 C, pH 9.5, catalyst:
Au/C GP3759 THPC) and kinetic parameters.

Y. Önal et al. / Journal of Catalysis 223 (2004) 122–133
129
ied by calculating specific Au surfaces and correlating them
with the initial reaction rates. A significant dependence of
catalyst activity in the oxidation of D-glucose in the aqueous
phase on specific Au surface area and, hence, Au particle
diameter could be shown (see Fig. 5). Such a dependence
can usually be related to differences in the adsorption be-
havior of the reactant on different crystal sites. As the metal
particle size decreases, the relative proportions of surface
atoms on corners and edges increase, implying the adsorp-
tion/hydrogenation of D-glucose, probably via the carbonyl
group, on these highly exposed sites. Indeed, the latter were
identified as active sites for the gold-catalyzed hydrogena-
tion of the C=O bond of unsaturated aldehydes [20].
Although the observed rate enhancement of glucose oxi-
dation with increasing specific Au surface area (i.e., reduc-
tion of particle size) is the expected sympathetic behavior, it
has to be noted that structure sensitivity can be caused by a
cooperative effect of particle size, degree of rounding, and
portion of multiple-twinned particles as shown by our group
in a recent study with oxide-supportedgold catalysts [21]. To
elucidate structure sensitivity in terms of gold particle mor-
phology, high-resolution transmission electron microscopy
should be involved in further studies.
Furthermore, it is important to mention that in this study,
focusing only on the particle size effect, different kinds of
support materials and reducing agents (THPC and NaBH4)
were used to prepare gold catalysts with various Au parti-
cle diameters; the trends were reported in Section 3.1. In
principle, the generation of particle size itself can be in-
fluenced by the type of support and the reducing agent.
However, from the rates of D-gluconic acid formation, using
the catalysts Au/C GP3759 PVA and Au/C GP3759 THPC,
we observed that gold particles with diameters of 3.7 nm
(Aspec = 84.0 m2/g) and 3.3 nm (Aspec = 94.2 m2/g) (see
Fig. 10. Temporal concentration and reaction rate courses at pH 7.0 and 9.5
and T = 50 C. Catalyst: Au/C GP3759 THPC.
◦
early stages of the reaction deactivation of catalyst should
not yet occur. According to Fig. 10, the initial reaction
rates differ significantly (3.15 mmolL−1 min−1 at pH 9.5 vs
1.00 mmol L−1 min−1 at pH 7.0). These data indicate that
higher pH values obviously also favor the adsorption of D-
glucose on the catalyst surface or the surface reaction.
4.2. Effect of Au particle size on catalyst activity
Although the intent was to prepare Au/C catalysts with
the same metal loading (1 wt%), the values of Au content
analyzed by ICP-OES differed widely (0.42–0.70 wt%). As
a result the structure sensitivity of glucose oxidation over Au
could not be reliably correlated with the mean particle size of
the metal without considering Au loading. Thus, the effect of
Au particle diameter on the activity of the catalyst was stud-
Fig. 11. Reaction network of the heterogeneously catalyzed D-glucose oxidation reaction in alkaline solution and at higher temperatures.

130
Y. Önal et al. / Journal of Catalysis 223 (2004) 122–133
Fig. 1 and Table 3) gave rise to different activities (see Fig. 5)
depending on the reducing agent (NaBH4 vs THPC) but not
on the support (Vulcan, which was identical in both cases).
Nevertheless, for a more reliable examination of particle size
effect, one should use the same support material and the
same reducing agent and prepare the catalysts under differ-
ent preparation conditions. This could be achieved, e.g., by
variation of the concentration of the reducing agent or, as
shown recently by us [21], by variation of pretreatment time
and temperatures, leading to variation in Au particle size;
however, this can be accompanied by a change in particle
morphology.
Note also that each carbon support itself seems to affect
differently the reactivity of the same gold particle size, in-
dicating a specific metal–support interaction as shown for
the liquid-phase oxidation of ethylene glycol by a combi-
nation of XPS and activity data of catalysts differing only
in carbon support [22]. From XPS investigation of the gold
catalysts in our study, the atomic Au/C ratios, based on cor-
rected intensities of the Au 4f7/2 and C 1s signals, were
calculated. However, regardless of the pretreatment proce-
dure (as prepared or heated in argon or hydrogen at 350 ◦C),
the average ratio was considerably higher for gold supported
on Vulcan-type carbon, (6.2 1.0) × 10−4, than for gold
on Black Pearls, (3.5 0.3) × 10−4, indicating primarily a
higher gold dispersion of the former according to the TEM
results (dAu = 3.3 and 3.7 nm for Au on Vulcan vs dAu = 5.0
and 6.1 nm for Au on Black Pearls).
at higher reaction temperatures and pH values and reacts fur-
ther, as shown by our HPLC analysis, mainly to glycolalde-
hyde, glyceraldehyde, and dihydroxyacetone.These reaction
pathways are probably a kind of retro-aldol reaction. To our
knowledge, in this study the formation of sorbitol, a well-
known reaction product of the hydrogenationof glucose over
precious metal catalysts [12], has been reported as another
side reaction in D-glucose oxidation at higher temperatures
and pH values, for the first time. This phenomenon is impor-
tant evidence of the oxidative dehydrogenation mechanism
already mentioned in Section 3.4.2 which is passed in the
formation of gluconic acid. According to the reaction mech-
anism of aldehyde oxidation to acids already proposed [16],
the following reaction steps should be considered, and are
presented in Fig. 12. In the aqueous phase, D-glucose ex-
ists in the pyranose form with 99.9% and in the open chain
structure with 0.1%. The open-chain D-glucose converts in
aqueous solution into the hydrate which should be stable
because D-glucose is a very instable aldehyde. After adsorp-
tion on the catalyst surface, the hydrate is dehydrogenated
and eventually desorbed. The desorption of gluconic acid is
thereby faster if the substrate is deprotonated (pH > 7.0);
otherwise catalyst deactivation by self-poisoning cannot be
neglected. Another reason for the pH dependence of reaction
rate is formation of the hydrate, which is faster in weakly al-
kaline or acidic media. Because in acidic media, the overall
reaction rate should be limited by desorption of D-gluconic
acid, a weakly alkaline solution should be preferred. Hy-
drides bound to the surface react with adsorbed OHad to form
water, which is also desorbed.
4.3. Reaction network of D-glucose oxidation
The reaction rate of sorbitol formation correlates with the
catalytic activity concerning D-glucose oxidation which is
an important hint for the above-postulated reaction mech-
anism. The formation of sorbitol can be explained by a
partially hydride-covered catalyst surface as an immediate
consequence of the oxidative dehydrogenation mechanism.
Because the degree of hydride coverage should correlate
with the catalytic activity concerning D-glucose oxidation,
it is not surprising that the reaction rate of sorbitol formation
correlates with the reaction rate of D-gluconic acid forma-
tion. The formation of sorbitol can certainly not be explained
The selectivity of the liquid-phase oxidation of D-glucose
as a polyfunctional molecule to D-gluconic acid depends
greatly on the correct choice of reaction conditions. If the
reaction temperature and pH are too high (T > 70 ◦C, pH
> 9), a number of side reactions of D-glucose, presented in
Fig. 11, have to be considered. In experiments carried out
at higher pH values (> 7.0), formation of fructose occurred
at any temperature. Formation of mannose occurred only at
higher temperatures (> 70 ◦C) and higher pH values (> 9.0).
D-Glucose as well as its isomerization products are not stable
Fig. 12. Simplified reaction mechanism of the heterogeneously catalyzed D-glucose oxidation.

Y. Önal et al. / Journal of Catalysis 223 (2004) 122–133
131
by a Cannizzaro-like side reaction of D-glucose in solution,
because only aldehydes without hydrogen in α-position are
subject to the Cannizzaro reaction in strongly alkaline solu-
tion.
4.4. Kinetic model
First, it was proved that the kinetic experiments were per-
formed under reaction conditions (sufficient stirring rate,
airflow rate) at which by external mass transfer at the
phase boundaries does not limit the overall reaction rate.
Then the rate-determining step according to the Langmuir–
Hinshelwood model was indicated by studying the influence
of different catalyst concentrations and initial D-glucose
concentrations on the reaction rate of D-glucose oxidation.
Although the reaction rate correlates with catalyst concen-
tration, the rate-determining step of the heterogeneously
catalyzed reaction cannot be stated, as all elementary steps
according to the Langmuir–Hinshelwood model (adsorption,
surface reaction, and desorption) can be influenced by cata-
lyst concentration. In another series of experiments, it could
be shown that the reaction rate does not significantly change
at different initial D-glucose concentrations. Therefore, it
could be proved that adsorption of the substrate on the cat-
alyst surface is certainly not the rate-determining step. A
detailed differentiation, by experiments, of whether the sur-
face reaction or the desorption of D-gluconic acid is the
rate-determining step has not been carried out yet. Because
at lower pH values (< 7.0) the catalyst is deactivated by
self-poisoning [17], desorption of D-gluconic acid should
certainly not be a rate-limiting factor, if the reaction is run
in an alkaline medium as it was in our kinetic experiments.
The overall reaction rate should therefore be limited by the
surface reaction and can be described by the semiempirical
model according to Eqs. (3) and (4). Thus, it was assumed
that the equilibrium of the surface reaction is widely on the
product side.
(a)
(b)
(c)
To check the reliability of the kinetic parameters deter-
mined, the latter were varied and the resulting temporal con-
centration profiles of D-glucose were compared (Fig. 13).
The degree of variation of the kinetic parameters was al-
ways the same, so that the effect of the parameters on the
concentration profiles of D-glucose could be estimated more
reliably. If the value of kOx is changed, the degree of de-
viation of the calculated concentration profiles will be very
high. In particular, the slope of the concentration profile at
the beginning of the reaction depends on kOx. Provided that
this slope is accurately determined, the calculated value for
kOx (k1) should be credible.
Fig. 13. Concentration courses of D-glucose at varied kinetic parame-
3
3
−3
−3
ters: (a) k = 63.3 × 10 –77.9 × 10 (k = 13 × 10 –15 × 10 ).
Ox
1
5
6
4
4
(b) K = 4 × 10 –5 × 10 . (c) K = 67 × 10 –76 × 10 .
G
GS
If the sorption equilibrium constant KG is varied, the
change in the concentration profiles of D-glucose is less dis-
tinctive. If one wants to determine a reliable value for KG,
above all the transition areas of the concentration profiles
of D-glucose and D-gluconic acid according to Fig. 14 have
to be analyzed precisely by a large number of HPLC sam-
ples. In this transition area the D-glucose concentration has
Fig. 14. Transition area (marked) of the concentration courses of D-glucose
and D-gluconic acid.

132
Y. Önal et al. / Journal of Catalysis 223 (2004) 122–133
already decreased to a small value, and the catalyst surface
should not be covered completely with D-glucose. Thus, the
adsorption of D-glucose is the rate-determining step of the
reaction and a modeling that represents this transition area
correctly should provide a reliable value for KG.
The concentration profiles of D-glucose hardly change,
when the value of KGS is varied. That means that the des-
orption of D-gluconic acid is always very fast compared with
the rates of the other elementary steps, so that the overall re-
action rate is never affected by this step.
Acknowledgments
The authors thank Dr. C. Mohr for the TEM analysis
of the catalysts at the Max-Planck-Institute for Microstruc-
ture Physics Halle, and Dr. J. Radnik (Institute for Applied
Chemistry Berlin) for the XPS measurements.
Appendix A. Nomenclature
[G] /[GS] D-Glucose/D-gluconic acid concentration;
[G]LS/[GS]SL D-Glucose/D-gluconic acid concentration on
the catalyst surface;
To check the optimization routine, initial values for the
kinetic parameters were varied widely and their influence on
the calculated parameters was analyzed. The resulting con-
centration courses were all in good agreement with the ex-
perimental data. It is notable that the adsorption constant of
D-gluconic acid was always smaller than that of D-glucose.
Furthermore, the values calculated for k1 were always simi-
lar independent of the model applied.
kads,G/kdes,G Adsorption/desorption constant of D-glucose;
k1/k
Reaction rate constants of the surface reaction;
−1
kdes,GS/kads,GS Desorption/adsorption constant of D-gluco-
nic acid;
kP
Empirical reaction constant of side reactions of D-
glucose;
θG
θGS
D-Glucose coverage grade of catalyst surface;
D-Gluconic acid coverage grade of catalyst surface;
5. Conclusions
rsurface Reaction rate of surface reaction;
rOR Overall reaction rate;
This investigation highlights the following points:
rSor,G Rate of D-glucose sorption;
rSor,GS Rate of D-gluconic acid sorption;
(i) Nanosized gold particles in the range of 3–6 nm pre-
pared on Black Pearls and Vulcan-type carbons are ac-
tive in liquid-phase oxidation of D-glucose to gluconic
acid. The most active gold catalyst had a specific gold
surface area of 94.2 m2/g.
MC
X
Catalyst concentration;
Conversion of D-glucose;
Selectivity to D-gluconic acid.
S
(ii) Reaction conditions could be optimized with respect
to the reaction rate of D-gluconic acid formation and
the selectivity of the reaction. The best results were
obtained at 50 ◦C and pH 9.5. Reasons for the effect
of temperature and pH on reaction rate of D-gluconic
acid formation were proposed, and on the basis of these
arguments together with the carefully estimated analy-
sis of reaction products, a reaction network was pre-
sented. The reaction could be successfully described by
an oxidative dehydrogenation mechanism in the aque-
ous phase which could be confirmed by experimental
observations.
(iii) It could also be shown that D-glucose oxidation is a
structure-sensitive reaction as a significant increase in
reaction rate could be observed with increasing Au spe-
cific surface area under the same reaction conditions.
(iv) Kinetic runs were carried out neglecting mass transfer
at the phase boundaries by intensive stirring and high
volumetric flow rate of air. By analyzing the effect of
catalyst concentration, initial D-glucose concentration,
and pH value of the liquid phase on reaction rate, it
could be shown that the surface reaction is the rate-
limiting step according to a semiempirical model based
on a Langmuir–Hinshelwood-type pathway. On the ba-
sis of these results, kinetic parameters were calculated
for the first time by an optimization routine and the re-
liability of the determined parameters was checked by
different methods.
References
[1] Ullmanns Encyclopedia of Industrial Chemistry, Electronic Release,
Wiley–VCH, 2003; http://www.mrw.interscience.wiley.com/ueic/
articles/a12_449/frame.html.
[2] F.W. Lichtenthaler, Acc. Chem. Res. 35 (2002) 728–737.
[3] M. Besson, P. Gallezot, in: R.A. Sheldon, H. van Bekkum (Eds.),
Fine Chemicals through Heterogeneous Catalysis, Wiley–VCH, 2001,
pp. 509–510, Chap. 9.3.
[4] M. Besson, P. Gallezot, Catal. Today 81 (2003) 547–559.
[5] S. Biella, L. Prati, M. Rossi, J. Catal. 206 (2002) 242–247.
[6] A. Martin, H. Berndt, I. Pitsch, A. Dittmar, in: Proceedings of 7th
European Workshop on Selective Oxidation, EuropaCat VI, Inns-
bruck, Austria, September 1–2, 2003; http://www.icp.csic.es/iso2003/
Programme%20ISO03.doc.
[7] S. Biella, G.L. Castiglioni, C. Fumagalli, L. Prati, M. Rossi, Catal.
Today 72 (2002) 43–49.
[8] F. Porta, L. Prati, M. Rossi, S. Coluccia, G. Martra, Catal. Today 61
(2000) 165–172.
[9] C. Bianchi, F. Porta, L. Prati, M. Rossi, Top. Catal. 13 (2000) 231–233.
[10] J.D. Grunwaldt, C. Kiener, C. Wögerbauer, A. Baiker, J. Catal. 181
(1999) 223–232.
[11] P. Claus, S. Schimpf, Y. Önal, in: Proceedings of 18th North American
Catalysis Society Meeting, Cancun, Mexico, June 1–6, 2003, p. 365.
[12] B. Kusserow, S. Schimpf, P. Claus, Adv. Synth. Catal. 345 (2003) 1–
11.
[13] J. Radnik, C. Mohr, P. Claus, Phys. Chem. Chem. Phys. 5 (2003) 172–
177.
[14] NIST database for XP spectroscopy; http://srdata.nist.gov/xps.
[15] J.M. Montejano-Carrizales, F. Aguilera-Granja, J.L. Moran-Lopez,
Nanostruct. Mater. 8 (1997) 269–287.

Y. Önal et al. / Journal of Catalysis 223 (2004) 122–133
133
[16] T. Mallat, A. Baiker, Catal. Today 19 (1994) 247–284.
[17] M. Besson, P. Gallezot, Catal. Today 57 (2000) 127–141.
[18] I. Nikov, K. Paev, Catal. Today 24 (1995) 41–47.
[19] (a) W. Reschetilowski, in: Technisch-Chemisches Praktikum, Wiley–
VCH, 2002, pp. 116, 117;
[20] C. Mohr, H. Hofmeister, J. Radnik, P. Claus, J. Am. Chem. Soc. 125
(2003) 1905–1911.
[21] C. Mohr, H. Hofmeister, P. Claus, J. Catal. 213 (2003) 86–94.
[22] C.L. Bianchi, S. Biella, A. Gervasini, L. Prati, M. Rossi, Catal. Lett. 85
(2003) 91–96.
[23] D. Pantea, H. Darmstadt, S. Kaliaguine, Ch. Roy, Appl. Surf. Sci. 217
(2003) 181–193.
(b) Perry’s Chemical Engineers’ Handbook, McGraw–Hill, New York,
1999, pp. 2-217, 2-125.