Dehydration of butanol to butene over solid acid catalysts in high water environments

Article abstract of DOI:10.1016/j.jcat.2008.12.009

The effects were studied of high water concentrations on the kinetics of alcohol dehydration, as encountered in aqueous-phase processing of biomass-derived oxygenated hydrocarbons. These studies were carried out for dehydration of an aqueous solution of 10 wt% 2-butanol at 513 K, and at a total pressure of 52 bar to maintain the water in the liquid state. Under these high pressures of water, silica-alumina, niobium phosphate and niobic acid are found to be stable and active for the dehydration of butanol. These three catalysts showed an increase in rate after contact with liquid water, caused by an increase in the concentration of Bronsted acid sites. Zeolite catalysts (Beta, USY, H-ZSM-5) and zirconia based catalysts (WO_x/ZrO₂, MoO_x/ZrO₂, and MgO/ZrO₂) were ineffective due to deactivation or low catalytic activity. The flow rate of inert gas at constant aqueous flow rate had a significant effect on the rate of butene production, due to vaporization of butanol and water. At low flow rates of gas, increasing the gas flow rate causes the preferential vaporization of butanol, leading to a decrease in the butanol pressure in the reactor and a corresponding decrease in the rate of dehydration. Above a critical gas flow rate, the liquid feed becomes completely vaporized in the reactor, and increasing the gas flow rate further leads to a decrease in the pressure of water and a corresponding increase in the rate of dehydration. In the vapor-liquid equilibrium regime, kinetic models predict that most of the catalyst is covered with multiple layers of water, and dehydration takes place by reaction of hydrated-adsorbed butanol with a hydrated surface site. In the vapor-only regime, kinetic models suggest that the fraction of vacant active sites increases with increasing gas flow rate, and dehydration takes place by reaction of adsorbed butanol with a vacant surface site.

Full text of DOI:10.1016/j.jcat.2008.12.009

Journal of Catalysis 262 (2009) 134–143
Contents lists available at ScienceDirect
Journal of Catalysis
www.elsevier.com/locate/jcat
Dehydration of butanol to butene over solid acid catalysts in high water
environments
∗
Ryan M. West, Drew J. Braden, James A. Dumesic
University of Wisconsin–Madison, Department of Chemical and Biological Engineering, Madison, WI 53706, USA
a r t i c l e i n f o
a b s t r a c t
Article history:
The effects were studied of high water concentrations on the kinetics of alcohol dehydration, as
encountered in aqueous-phase processing of biomass-derived oxygenated hydrocarbons. These studies
were carried out for dehydration of an aqueous solution of 10 wt% 2-butanol at 513 K, and at a total
pressure of 52 bar to maintain the water in the liquid state. Under these high pressures of water, silica-
alumina, niobium phosphate and niobic acid are found to be stable and active for the dehydration
of butanol. These three catalysts showed an increase in rate after contact with liquid water, caused
by an increase in the concentration of Brønsted acid sites. Zeolite catalysts (Beta, USY, H-ZSM-5) and
zirconia based catalysts (WOx/ZrO2, MoOx/ZrO2, and MgO/ZrO2) were ineffective due to deactivation or
low catalytic activity. The flow rate of inert gas at constant aqueous flow rate had a significant effect
on the rate of butene production, due to vaporization of butanol and water. At low flow rates of gas,
increasing the gas flow rate causes the preferential vaporization of butanol, leading to a decrease in the
butanol pressure in the reactor and a corresponding decrease in the rate of dehydration. Above a critical
gas flow rate, the liquid feed becomes completely vaporized in the reactor, and increasing the gas flow
rate further leads to a decrease in the pressure of water and a corresponding increase in the rate of
dehydration. In the vapor–liquid equilibrium regime, kinetic models predict that most of the catalyst is
covered with multiple layers of water, and dehydration takes place by reaction of hydrated–adsorbed
butanol with a hydrated surface site. In the vapor-only regime, kinetic models suggest that the fraction
of vacant active sites increases with increasing gas flow rate, and dehydration takes place by reaction of
adsorbed butanol with a vacant surface site.
Received 28 August 2008
Revised 12 December 2008
Accepted 16 December 2008
Available online 22 January 2009
Keywords:
2-Butanol
Dehydration
Water
Aqueous phase processing
Silica-alumina
Niobium phosphate
Niobic acid
Butene
© 2008 Elsevier Inc. All rights reserved.
1. Introduction
(APDH) leads to the formation of straight-chain alkanes, such as
butane, pentane and hexane [3]. Olefinic species are not typically
observed during APDH processing at temperatures near 520 K and
pressures near 50 bar over catalysts consisting of Pt (e.g., 4 wt%)
supported on acidic supports, such as silica-alumina or niobium
phosphate [3,4], suggesting that dehydration is the rate limiting
step. In the present paper, we report results of reaction kinetics
studies of dehydration reactions over various solid acids catalysts
in the presence of liquid water as well as water vapor. Sec-butanol
was chosen as the reactant for studies of aqueous-phase dehydra-
tion, because it readily undergoes intramolecular dehydration to
form butene products. To ensure that dehydration was, in fact, the
rate limiting step in APDH processing, 2-butanol was converted to
butane over the same Pt/silica-alumina catalyst used in our previ-
ous work [3], and 2-butanol was also converted to butenes over
the silica-alumina support under the identical reaction conditions.
The rates of production of butane and butenes were the same on
these two catalysts, indicating that dehydration was indeed the
rate limiting step.
Renewable sources for fuels and chemicals must be developed
as reserves of non-renewable petroleum feed stocks diminish, and
biomass resources are promising alternatives to meet these de-
mands. Considerable research has focused on sugars obtained from
biomass, such as fructose and glucose, and their derivatives such
as sorbitol [1–5]. Carbohydrates contain high extents of oxygenated
functional groups that must be selectively removed or modified to
create desired products. These oxygenated groups lead to high sol-
ubilities of carbohydrates in water, requiring aqueous-phase pro-
cessing of these compounds to fuels and chemicals.
A particularly useful reaction sequence for aqueous-phase pro-
cessing is dehydration followed by hydrogenation, in which oxy-
genated hydrocarbons, such as sorbitol, are first dehydrated over
solid acid sites followed by hydrogenation to form alkyl species.
Sequential operation of aqueous-phase dehydration/hydrogenation
Butanol dehydration to butenes has been studied by various au-
thors under conditions involving water. For example, it has been
Corresponding author.
E-mail address: dumesic@engr.wisc.edu (J.A. Dumesic).
*
0021-9517/$ – see front matter © 2008 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2008.12.009

R.M. West et al. / Journal of Catalysis 262 (2009) 134–143
135
reported that water vapor can either increase or inhibit 2-butanol
dehydration over zirconia-supported tungsten and silicon based
solid acids [6]. A decrease in the activity of WO3/ZrO2 in the
presence of water was investigated by others [7]. However, the
performance of solid acid catalysts has not been studied system-
atically for a wide range of water concentrations, especially high
water concentrations and in the presence of liquid water.
flowed for an additional hour before introducing the butanol/water
mixture. After changing conditions, the system was allowed a min-
imum of 5 h to reach steady state before sampling the phases.
2.3. Catalyst characterization
The surface areas of all catalysts were determined from BET
isotherms of N2 adsorption at 77 K. The concentration of acid
sites per gram of each catalyst was quantified by temperature
programmed desorption of ammonia. Catalyst samples (∼100 mg)
were loaded into a glass flow through cell, ammonia was ad-
sorbed onto the catalyst for 1.5 h at room temperature, physically
adsorbed ammonia was then desorbed at 423 K, and TPD exper-
iments commenced using a temperature ramp of 10 K/min. The
desorbed ammonia was quantified on-line using a mass spectrom-
eter.
The distribution of Brønsted and Lewis acid sites was deter-
mined from infrared measurements of adsorbed pyridine. Approx-
imately 10 mg of catalyst was placed in a 1.2 cm dye and pressed
into a pellet, which was then placed in a treatment/sampling cell
where it was heated to 473 K under flowing dry N2 (Linde) for 2 h.
A reference spectrum of the catalyst was then taken. Pyridine was
introduced into the cell for 30 min at room temperature, followed
2. Experimental
2.1. Catalysts
Silica-alumina, MCC 25 (SiAl), was obtained from Grace David-
son with a Si/Al ratio of 4. Beta-zeolite with a Si/Al ratio of 25
and USY-zeolite with a Si/Al ratio of 5 [8] were obtained from
Engelhard and calcined at 773 K before use. H-ZSM-5 catalyst
with a Si/Al ratio of 14 was obtained from Engelhard. Tungstated-
zirconia, XZO1251/01 16% WO3, was obtained from MEI Chemicals.
MoOx/ZrO2, with 20.3% MoO3, was prepared in a similar manner as
reported in the literature [9], by incipient wetness impregnation of
(NH4)6Mo7O24·H2O (Aldrich) on ZrO2 XZO 880/01 purchased from
MEI Chemicals. The impregnated solids were dried overnight in
air and then treated in a flowing gas mixture of 20% O2 in He
at 723 K. MgO/ZrO2 was prepared as detailed elsewhere [10,11].
Niobic acid, HY-340 (Nb2O5), was obtained from CBMM in Brazil.
Niobium phosphate (NbOPO4) was prepared in a similar manner
to previous reports [12]. Briefly, 4 g of NbCl5 was reacted 1:2
with phosphoric acid. After thorough mixing, the resulting paste
was diluted in 70 mL water and stirred for 40 min. The pH was
then adjusted to 4.9 and the mixture was stirred for an additional
30 min. The NbOPO4 was then filtered and washed until the silver
by heating at 473 K overnight under flowing dry N2. The areas of
−1
the pyridine peaks at 1455 and 1545 cm
(Lewis and Brønsted
sites, respectively) were then determined by subtracting the spec-
tra of the sample before and after this exposure to pyridine.
2.4. Methods
The activity of each catalyst was tested under the same re-
action conditions, consisting of a flow equal to 0.1 mL/min of
feed solution (10 wt% butanol in water) and a gas flow of 200–
215 cm3(STP)/min. After this initial screening, the three zirconia-
based catalysts, MgO/ZrO2, WOx/ZrO2, and MoOx/ZrO2 were elimi-
nated from further investigation, because the rates of butene pro-
duction for these three catalysts were 1.3, 2.6 and 5.3 μmol/g/min,
respectively; these values are an order of magnitude lower than
the activities of the other catalysts. Beta-zeolite was also removed
from further consideration, because although it initially demon-
strated a high rate of butene production (1600 μmol/g/min), it
showed a continual decrease in activity under the reaction con-
ditions of this study, as shown in Fig. 1.
−
nitrate test showed no additional Cl ions. The catalyst was dried
overnight in air and then treated in a flowing gas mixture of 20%
O2 in He at 723 K.
For characterization purposes, SiAl, Nb2O5 and NbOPO4 were
additionally modified to study the effect of liquid water on acidity.
In this treatment, approximately 0.5 g of catalyst was placed in
a glass beaker with 50 mL of water. The beaker was placed in a
sealed Parr reactor and heated to the reaction temperature of 513 K
for 3 h. The reactor was cooled, the water was decanted, and the
catalyst was then placed in an oven overnight to dry at 393 K.
2.2. Reactor setup
The rates of dehydration of the remaining five catalysts are
shown in column 1 of Table 1. H-ZSM-5 and USY showed the high-
est rates of 2403 and 962 μmol/g/min, respectively. SiAl, Nb2O5
Catalysts were mixed with crushed silica and packed with
quartz wool end-plugs in tubular quarter-inch stainless steel reac-
tors. Except where noted otherwise, a ten weight percent solution
of 2-butanol (Aldrich) in deionized water was fed with an HPLC
pump in an up-flow direction over the catalyst bed. The efflu-
ent was collected in a gas–liquid separator. Inert gas (He or H2
from Linde) was flowed through the catalyst bed by means of a
mass flow controller, and this flow was used to control the water
concentration within the reactor. In addition, inert gas was flowed
though the separator to direct gaseous species to a gas chromato-
graph for on-line analysis. The pressure of the system was held
constant at 52 atm by a back pressure regulator. The reactor was
held at a constant temperature of 513 K as measured by a thermo-
couple attached to the exterior of the reactor and surrounded by
aluminum blocks within an oven. The effluent gas was analyzed by
an online GC, Varian GC-MS (Saturn 3) using an FID detector and
a GS-Q capillary column (J&W Scientific). The liquid effluent was
collected and analyzed by a Shimadzu GC2010 equipped with an
FID detector and a DB 5 ms column (J&W Scientific). Total carbon
material balances on individual points typically closed within 5%.
In a typical experimental run, the reactor was heated to the
reaction temperature while inert gas was flowed over the cata-
lyst. After reaching the reaction temperature, the inert gas was
Fig. 1. Normalized rate versus time on stream in high water environment for Beta-
zeolite.

136
R.M. West et al. / Journal of Catalysis 262 (2009) 134–143
Table 1
Reactivity and properties of solid acid catalysts for the dehydration of 2-butanol in aqueous media.
Rate at 200–215
cm3(STP)/min of He
(μmol/min g cat)
Rate after exposure
to liquid water
(μmol/min g cat)
Acid site
density
Brønsted
area/Lewis
areaa
BET surface
area
(μmol/g)a
(m2/g)a
SiAl
153
81
40
2403
962
181
262
147
1373
315
578/756
265/536
135/22
542/–
0.15/12.5
2.6/2.1
1.1/3.0
20.7/–
9.6/–
498/169
185/85
118/17
–/–
NbOPO4
Nb2O5
H-ZSM-5
USY
1015/–
–/–
a
Untreated catalyst/water treated catalyst.
262 μmol/g/min, respectively. Moreover, the rates were observed
to increase sharply at lower flow rates of gas through the reactor.
Upon returning to higher gas flow rates (under conditions leading
to complete vaporization of water), a higher rate of butanol dehy-
dration was observed for each of the three catalysts (see Fig. 3).
The acid site density calculated from ammonia TPD, the ratio
of the Brønsted/Lewis areas from pyridine IR, and the BET surface
area for each catalyst are given in columns 3 through 5 of Table 1.
Further studies were performed for the three catalysts that showed
an increase in activity after contact with liquid water. These cat-
alysts were treated with water as described in the experimental
section. The acid site density, Brønsted/Lewis ratio and BET surface
area after this treatment are also given as the second number in
columns 3 through 5.
3. Discussion
Fig. 2. Normalized rate (rate at given He flow rate/ flow rate at 200–215 cm3(STP)/
min) versus inert gas flow rate. SiAl (2); Nb2O5 ("); NbOPO5 (e); H-ZSM-5 (E);
USY (Q).
3.1. Liquid water/water vapor equilibrium
The vapor pressures of water and 2-butanol at 513 K are 32.9
and 30.4 atm, respectively. With a total system pressure of 52 atm,
and in the absence of a flowing gas, water should remain as a
liquid. However, if a gas is bubbled through the water, the gas bub-
bles will carry water vapor through and out of the reactor. At the
reaction conditions of this study, the butene produced by the de-
hydration reaction is present only in the gas phase. Therefore the
rate of butene production will also cause the sparging of liquid wa-
ter. If no butene is formed, then all of the 2-butanol and water are
sparged into the gas phase at a gas flow (FHge) through the reactor
of 74 cm3(STP)/min. As a second limit, one can consider the case
where all the 2-butanol is converted to butene. Under these con-
ditions, a flow of only 68 cm3(STP)/min of inert gas is necessary to
vaporize all the water present. Thus, regardless of the reactivity of
a catalyst, liquid water will always be present for a gas flow below
68 cm3(STP)/min, and the liquid feeds of 2-butanol and water will
be completely vaporized at a gas flow above 74 cm3(STP)/min.
A more quantitative description of the extents of vaporization
of water and butanol can be determined from thermodynamic ex-
pressions for species in vapor–liquid equilibrium. Literature sources
[13] indicate that dilute solutions of butanol in water behave non-
ideally. The partial pressures, Pi, of vapor phase species are given
by Eq. (1).
and NbOPO4 all had similar rates of 153, 40, and 81 μmol/g/min,
respectively.
At constant liquid flow of 0.1 mL/min of an aqueous solution of
10 wt% butanol into the reactor, the flow rate of an additional gas
through the reactor determines the concentrations of the reactant
and water. In particular, at vapor–liquid equilibrium, the addition
of a sweep gas leads to vaporization of liquid water and butanol,
and the extent of vaporization is determined by the flow rate of
the sweep gas. We calculate that all of the liquid water in the
reactor will be vaporized at a gas flow rate above approximately
70 cm3(STP)/min, as addressed in more detail below. The activities
of the five remaining catalysts were next tested in the absence of
liquid water, as the flow rate of the gas through the reactor was
varied between 70 and 250 cm3(STP)/min, where higher flow rates
lead to lower partial pressures of water vapor. As the gas flow rate
was increased, the rate of butene production was observed to in-
crease. The nature of this increase was similar for the five catalysts.
Fig. 2 shows the normalized rate (rate at given flow divided by the
rate at 200–215 cm3(STP)/min given in Table 1, column 1) versus
the gas phase flow through the reactor.
Liquid water was next allowed to contact each of the remain-
ing catalysts by decreasing the gas flow through the reactor below
70 cm3(STP)/min. The rate of butanol dehydration decreased for
flow rates below 70 cm3(STP)/min for both USY and H-ZSM-5.
Upon returning to a higher flow rate (200 cm3(STP)/min), both USY
and H-ZSM-5 showed a sharp decrease in activity, as shown in Ta-
ble 1, column 2. Exposure of these catalysts to liquid water thus
led to deactivation, and these two catalysts were not studied fur-
ther.
Pi = xiγi PVap = yi Ptot
.
(1)
i
In this equation xi and yi are mole fractions in the liquid and va-
por phases respectively, and γi are activity coefficients. Substitut-
ing Eq. (1) into an expression for the total system pressure results
in Eq. (2). The saturation pressures can be obtained from the An-
toinne equation using the appropriate constants for each species.
The remaining three catalysts demonstrated an increase in ac-
tivity when liquid water was present. SiAl showed a modest in-
crease in activity from 153 to 181 μmol/g/min, while Nb2O5 and
NbOPO4 both showed significant increases, 40 to 147 and 81 to

R.M. West et al. / Journal of Catalysis 262 (2009) 134–143
137
In the expression for the total pressure of the system, Eq. (2), the
total flow rate of the vapor phase, Ftot,gas, is given by:
Ftot,gas = Fhelium + Fbutene + Fwater,gas + Fbutanol,gas
.
(3)
While the helium and butene species are present only in the va-
por phase, butanol and water are in equilibrium with the liquid
and vapor phases. Using the vapor–liquid equilibrium expression
for butanol, it is possible to express the total flow rate of butanol
in the system as Eq. (4).
Fbutanol,tot = Fbutanol,gas + Fbutanol,liq
x
butanolγbutanol PbVuatpanol
=
(Fwater,tot − Fwater,liq
)
(1 − xbutanol)γwater PVap
+
water
xbutanol
Fwater,liq
.
(4)
(1 − xbutanol
)
Using Eqs. (2) and (4), the two unknown variables Fwater,liq and
xbutanol can be determined, and these two variables are then used
to calculate Fbutanol,liq. It is then possible to calculate Fwater,gas
and Fbutanol,gas in terms of Fwater,tot and Fbutanol,tot, and the over-
all materials balances are then written in terms of these two latter
quantities.
(A)
(B)
(C)
The initial values for the activity coefficients for the butanol and
water were chosen based on literature [13] as well as numerical
simulation of the binary mixture using Aspen. It has been reported
that as the mole fraction of butanol in water approaches zero, the
activity coefficient for butanol increases significantly. For the mole
fractions of butanol used in the current reaction system, approxi-
mately 0.03, an activity coefficient of at least 20 is expected for the
vapor–liquid regime. Increasing the activity beyond 20 did not have
a significant effect on model results, and hence a value of 20 was
used for all subsequent calculations. The activity coefficient of wa-
ter is close to one at the given reaction conditions, and is assumed
to be one for the model calculations. Additionally for all model cal-
culations, vapor liquid equilibrium is assumed for all points within
the reactor; hence the activities of liquid and gaseous species are
the same.
3.2. Catalyst performance
SiAl, NbOPO4 and Nb2O5 all showed high catalytic activity and
stability throughout the course of reaction kinetics studies. Impor-
tantly, these catalysts did not deactivate after exposure to liquid
water, and in fact, the activities of all three catalysts increased
after the introduction of liquid water, with Nb2O5 increasing to
the greatest extent (270% increase) followed by NbOPO4 (220% in-
crease) and SiAl (18%). To understand the origin of this increase,
the catalysts were characterized before and after treatment with
liquid water under high temperature and pressure in a Parr reactor
as shown in Table 1. The SiAl catalyst showed a modest increase
of about 30% in concentration of acid sites per gram, as measured
by ammonia TPD. The NbOPO4 catalyst showed a large increase of
100%, while Nb2O5 showed an 84% decrease in acid sites concen-
tration. All three catalysts showed a decrease in surface area upon
treatment with liquid water. SiAl and NbOPO4 showed modest de-
creases in surface area of 66% and 54% respectively. Nb2O5 showed
a large decrease in surface area of 86%. For Nb2O5, the decrease in
acid sites concentration (84%) is attributed to the large decrease in
surface area under the conditions of this water treatment (86%).
For SiAl the rate increased approximately 20% after treatment
in liquid water. From Fig. 3A, it is clear that this rate increase oc-
curred at all gas flows. Water treatment of SiAl increased the acid
surface site density by 30% and shifted the ratio of Brønsted/Lewis
acids from 0.15 to 12. Because Brønsted acid sites are significantly
more active sites for dehydration [7], it appears that the increase in
Fig. 3. (A) Rate of 2-butanol dehydration versus inert flow rate over SiAl. Initial
activity (!); after exposure to liquid water ("). (B) Rate of 2-butanol dehydration
versus inert flow rate over NbOPO4. Initial activity (1); after exposure to liquid
water (2). (C) Rate of 2-butanol dehydration versus inert flow rate over Nb2O5.
Initial activity (E); after exposure to liquid water (F).
ꢀ
Ptot
=
Ps = xbutanolγbutanol PSat
+ (1 − xbutanol)γwater PwSaatter
butanol
s
Ptot
+
(Fbutene + Fhelium).
(2)
Ftot,gas

138
R.M. West et al. / Journal of Catalysis 262 (2009) 134–143
1
θv =
,
(8)
(9)
[1 + KbuOH PbuOH + KH O PH O + Kbutene Pbutene
]
2
2
ꢁ
ꢂ
Pbutene PH2
PbuOH Keq
O
KbuOHkf,surf.rxn. PbuOH 1 −
[1 + KbuOH PbuOH + Kbutene Pbutene + KH O PH
rate =
.
2
]
2O
2
The second mechanism examined incorporates multi-layer ad-
sorption of water on the catalyst. The derivation of the multi-
layer water adsorption coverage is based on the BET adsorption
isotherm, as outlined below. The fractional coverage for the first
adsorbed water molecule on the catalyst surface can be written as
Eq. (10) for quasi-equilibrated adsorption. In this equation KH is
2O
the equilibrium constant for the adsorption of water onto the cat-
alyst surface and θH is the fractional coverage of mono-adsorbed
2O
water. The fractional coverage of a two-water stack on the site can
be written as Eq. (11), where KW1 is the equilibrium constant for
the adsorption of water on an already adsorbed water molecule.
The KW1 parameter corresponds to the equilibrium constant for
liquefaction of water and is set as the inverse of the saturation
pressure of water at the reaction temperature. Similarly, the frac-
tional coverage of an n-layer stack of water molecules can be writ-
ten in the form of Eq. (12).
Fig. 4. Butene TOF versus inert flow rate. SiAl initial activity (E); SiAl post liquid
water exposure (Q); NbOPO4 initial activity (e); NbOPO4 post liquid water exposure
(2); Nb2O5 initial activity (").
Table 2
Simple four step surface reaction mechanism.
θH O = KH O PH Oθv,
(10)
(11)
(12)
2
2
2
Steps
Surface reactions
∗
θ(H
= KH O PH O KW1 PH Oθv,
1.
2.
3.
4.
C4H9OH + ∗ ꢀ C4H9OH
2O)2
2
2
2
C4H9OH∗ + ∗ → C4H8∗ + H2O
∗
∗
θ(H
= K O P O(KW1
P
)
n−1θv.
H2O
C4H8 + ∗ ꢀ C4H8
H2
H2
2O)n
∗
H2O + ∗ ꢀ H2O
According to the BET adsorption isotherm, water can adsorb in an
infinite number of layers. Equation (13) shows the site balance,
where the sum of all species ranges from values of n from 1 to in-
finity, leading to a geometric Maclaurin series. The repeated term
raised to a power in the Maclaurin series is equal to the equilib-
rium constant for the water liquefaction multiplied by the partial
pressure of water.
catalytic activity of SiAl upon treatment with liquid water is caused
by an increase in the surface concentration of Brønsted acid sites.
NbOPO4 showed a large increase in both catalytic activity (in-
crease of 220%) and a large increase in the density of surface acid
sites (increase of 100%) upon contact with liquid water, see Fig. 3B.
The ratio of Brønsted/Lewis area after this treatment remained
about constant. Thus, the increase in catalytic activity of NbOPO4
upon treatment with liquid water is again caused primarily by an
increase in the surface concentration of Brønsted acid sites. The
reactivity change upon exposure to liquid water for Nb2O5 is sim-
ilar to that for NbOPO4, and we again suggest that the surface
chemistry is controlled by the number of Brønsted acid sites. Fig. 4
shows the rates reported as turnover frequencies. It is clear from
Fig. 4 that the TOF values are similar for all catalysts before water
treatment. The TOF values for SiAl are similar after water treat-
ment, whereas NbOPO4 appears to show a modest increase in TOF
after water treatment.
1 = θv + KH O PH Oθv + KH O PH O KW1 PH Oθv
2
2
2
2
2
+ KH O PH O(KW1 PH O)2θv
2
2
2
+ ··· + KH O PH O(KW1 PH O)nθv,
2
2
2
ꢃ
ꢄ
K
O P
H2
H2
O
1 = θv 1 +
,
α = KW1 PH
.
(13)
2O
1 − α
A similar approach can be used to derive expressions for the
hydration of butanol adsorbed on the surface. In this case, we
include butanol adsorbed on dry sites, with equilibrium constant
equal to KbuOH, and we include adsorption of water onto an ad-
sorbed butanol molecule, with an equilibrium constant equal to
as KW2. The value of KW2 can be set equal to the constant of
liquefaction, KW1, or it can be an adjustable parameter. The re-
sulting fractional surface coverages for adsorbed butanol on a dry
site, θbuOH, multilayer water adsorption, θhyd, and multilayer water
adsorption on butanol, θbuOH_hyd, are shown below.
3.3. Kinetic model formulation
Two adsorption models were used to understand how the
flow rate of inert gas through the reactor influences the rates
of dehydration in the water-stable systems of SiAl and NbOPO4.
The first mechanism is a Langmuir–Hinshelwood (L–H) mecha-
nism and involves the surface reaction of adsorbed butanol with
a vacant site to form adsorbed butene and water. Table 2 shows
the 4 step simple mechanism. Steps (1), (3), and (4) are the
adsorption–desorption steps of butanol, butene, and water, respec-
tively. Step (2) is the surface reaction where the dehydration of
butanol occurs. The adsorption–desorption steps are assumed to
be quasi-equilibrated. Following L–H kinetics, the surface coverage
of species and forward rate are as follows:
θbutanol = KbuOH PbuOHθv,
KH O PH Oθv
(14)
(15)
2
2
θhyd
=
,
(1 − α)
=
KbuOH PbuOH KW2 PH Oθv
2
θbuOH_hyd
,
(16)
(17)
(1 − α)
1
θv =
.
ꢅ
ꢆ
KH2O PH2O+KbuOH PbuOH KW2 PH2
O
1 + KbuOH PbuOH
+
(1−α)
θbuOH = KbuOH PbuOHθv,
θbutene = Kbutene Pbuteneθv,
θH O = KH O PH Oθv,
(5)
(6)
(7)
The standard state (i.e., 1 atm and 298 K) entropy and en-
thalpy values are obtained from standard handbooks. The con-
stant pressure heat capacities are used to adjust the enthalpies
2
2
2

R.M. West et al. / Journal of Catalysis 262 (2009) 134–143
139
and enthalpies to the reaction temperature. The gas phase three-
Table 3
Summary of reaction schemes for butanol dehydration.
dimensional entropies are calculated for all species using the rela-
tion:
Entry
Adsorption
Surface reactions
Parameters
# Par.
ꢃ
ꢄ
ꢇ
ꢇ
1
2
L–H
Multilayer
buOH∗ + ∗ →
Ea1, BEbuOH, BEH
Ea1, BEbuOH, BEH
Ea2
Ea1, BEbuOH, BEH
Ea2
3
4
(2πmkBT )3/2
kBT
h
5
2
2O
Storans,3D = R ln
+ ln
+
(18)
buOH∗ + ∗ →
2O
h3
∗
buOHhyd∗ + hyd
buOH∗ + ∗ →
→
→
3
Multilayer
5
2O
where m is the mass of the molecule of interest, h is Plank’s
constant, kB is Boltzmann’s constant and T is the temperature in
Kelvin. The thermodynamic properties of the surface species are
then determined by adjusting the gas phase thermodynamics ac-
cordingly. The enthalpy of formation of an adsorbed species is
calculated by adding the binding energy of the molecule to its
gas phase enthalpy. The local entropy values (i.e., rotational and
vibrational entropies) of the adsorbed species are calculated by
subtracting the three-dimensional translational entropy from the
gas phase entropy. It is assumed that all of the local entropy was
retained by the adsorbed species.
buOHhyd∗ + hyd
∗
→
∗
buOH∗ + hyd
buOH∗ + ∗ →
buOHhyd∗ + hyd
buOH∗ + ∗ →
Ea3
Ea1, BEbuOH, BEH
Ea2, BE
5
6
4
5
Multilayer
Multilayer
2O
∗
→
→
buOH–H2
O
Ea1, BEbuOH, BEH
Ea2, BE
2O
buOHhyd∗ + hyd
∗
→
buOH–H2
O
∗
buOH∗ + hyd
Ea3
data for the SiAl and NbOPO4 catalysts (after exposure to liquid
water) are then used to fit the catalyst specific parameters for
the binding energy of water, binding energy of butanol, activation
energies of the surface reactions and the energy for hydration of
surface adsorbed butanol. The parameter estimation is performed
in Matlab using the nonlinear parameter estimation function ‘nlin-
fit’. The nlinfit function uses the Gauss–Newton algorithm with
Levenberg–Marquardt modifications for determining the optimum
parameters for the model fit. The nonlinear fitting algorithm uses
the butene flow rate out of the reactor for comparing the exper-
imental data with the predicted values. Following the successful
convergence of the nonlinear parameter estimation function, the
residuals and Jacobian are then used to determine the confidence
intervals for the parameters using the ‘nlparci’ function built into
Matlab, which calculates the confidence intervals using a statistical
method based on the asymptotic normal distribution for the pa-
rameters estimates. Two adsorption schemes and multiple reaction
schemes are considered. A summary of the schemes that predict
the correct trends is given in Table 3. The individual entries in the
table are discussed below.
The overall equilibrium constant for each step is calculated
from the change in enthalpy and entropy upon reaction. Each equi-
librium constant is calculated using Eq. (19).
ꢃ
ꢄ
−ꢀHio
ꢀSoi
Ki,eq = exp
+
.
(19)
RT
R
In this equation, ꢀHio and ꢀSoi are the change in enthalpy and
change in entropy, respectively, for reaction i. To ensure thermo-
dynamic consistency, we adjust the forward rate constant to fit
the experimental data, and the reverse rate constant is calculated
using the expression:
ki,for
ki,rev
=
,
(20)
Ki,eq
where ki,for and ki,rev are the forward and reverse rate constants,
respectively. The binding energies of various species to the sur-
face are taken as adjustable parameters. The specific species under
consideration depends upon the type of adsorption and allowed
surface reactions as discussed below.
3.4. Kinetic model results
The rate constants for the surface reactions are calculated from
the Arrhenius expression in Eq. (21),
The simple Langmuir–Hinshelwood mechanism involves fitting
3 parameters: the activation energy of reaction, Ea1, and the bind-
ꢃ
ꢄ
−Ei
ki = A exp
(21)
ing energies of butanol and water, BEbuOH, and BEH O (Table 3,
2
RT
Entry 1). The best fit for this model does not accurately predict
the behavior of the liquid/water system, as shown in Fig. 5 for SiAl
and NbOPO4. The general trend of increasing rate at gas flow rates
above and below 70 cm3(STP)/min of inert flow is predicted qual-
itatively. At gas flow rates above 70 cm3(STP)/min, increased inert
gas flow decreases the partial pressure of water, thus increasing
the rate. However, the effect of the changing partial pressures of
reactants was not sufficiently pronounced in the simple mecha-
nism to predict the rates at high inert flow. At gas flow rates below
70 cm3(STP)/min, the non-ideality of the butanol water system, as
incorporated by the butanol activity coefficient, predicts that the
rate will decrease with increasing gas flow rate. In this range of gas
flow rates where liquid water is present in the reactor, increasing
the gas flow rate leads to the preferential vaporization of butanol
because of the high value of the activity coefficient. In particular,
molar flow rates of butanol and water in the gas and liquid phases
are given by:
where A is the pre-exponential factor and Ei is the activation en-
ergy barrier in the forward direction. The pre-exponential factors
are assumed for simplicity to be equal to the normal value of
−1
1013
s
for adsorbed species reacting on a surface. Therefore, the
surface reactions are parameterized by the forward activation en-
ergy barrier for each respective step.
The initial step in developing the kinetic model is to determine
the sensitivity of the parameters to be estimated. The sensitivity is
calculated by increasing the model parameter, P, by 1% and record-
ing the effect on the overall rate, r, for the reaction. The effect on
the rate is then normalized by the initial rate and initial parameter
value according to equation
ꢀr P0
σp =
(22)
ꢀP r0
where ꢀr and ꢀP are the change in rate and change in parame-
ter value, respectively and r0 and P0 are the initial values for the
reaction rate and parameter, respectively. In all cases, the binding
energy for butene is found to be insensitive.
The kinetic models are fit to the experimental observations us-
ing Matlab software. The model solves a differential equation for
the change in butanol flow rate as a function of the length of a
packed bed plug flow reactor. The model also includes the alge-
braic equations to calculate the catalyst surface coverage by ad-
sorbed species as a function of reactor length. The reaction kinetics
FbuOH,gas
x
buOHγbuOH PbuOH
=
,
F
(1 − xbuOH)P
H2O,gas
H2O
FbuOH,liq
xbuOH
=
,
FH
2O,liq
(1 − xbuOH
)
and because the value of γbuOH PbuOH/PH
is greater than 1
2O
(γbuOH = 20), the composition of the vapor phase is enriched in
butanol compared to the composition of the liquid phase. Thus,

140
R.M. West et al. / Journal of Catalysis 262 (2009) 134–143
Fig. 6. Calculated partial pressures at the outlet of the reactor for water (solid line)
and butanol (dashed line) under reactive conditions as a function of inert gas flow.
(A)
react with adsorbed–hydrated butanol to form butene and water.
The addition of this step adds an additional parameter, namely the
activation energy of this second surface reaction. The reaction of
butanol (without water adsorbed on it) with a hydrated site, how-
ever, fails to predict the experimentally observed trends.
The addition of a third surface reaction to the model also pre-
dicts the correct trends. The three surface reactions are adsorbed
butanol with a vacant site, adsorbed butanol with a hydrated site,
and adsorbed–hydrated butanol with a hydrated site (Table 3, En-
try 3). The addition of this third reaction adds a third activation
energy as an adjustable parameter to the model.
The final adjustable parameter under consideration is the bind-
ing energy of water on adsorbed butanol. In Entries 2 and 3 of
Table 3, this binding energy was constant. Entry 4 in Table 3 uses
the same two surface reaction steps as Entry 2; however, the bind-
ing energy of water on butanol is adjustable. Likewise, Entry 5 uses
the same three surface reaction steps as Entry 3, with the binding
energy of water on butanol being adjustable.
To determine the model that best fits the experimental ob-
servations with the minimum number of parameters, the Akaike
Information Criterion (AIC) is calculated for each proposed varia-
tion of the multilayer adsorption model [14]. The AIC is calculated
using the second order equation
(B)
Fig. 5. (A) Experimental data (") for 2-butanol dehydration over SiAl modeled
with Langmuir–Hinshelwood mechanism at varying activity coefficients. (B) Experi-
mental data (") for 2-butanol dehydration over NbOPO4 modeled with Langmuir–
Hinshelwood mechanism at varying activity coefficients.
the partial pressure of butanol in the reactor decreases with an in-
crease in the gas flow rate, as shown in Fig. 6, while the partial
pressure of water remains constant provided that all of the liquid
water is not vaporized. Increasing the activity coefficient beyond a
value of 20 does not significantly impact the model as shown in
Fig. 5 for SiAl and NbOPO4.
2k(k + 1)
AICc = AIC +
,
(23)
(24)
n − k − 1
ꢃ
ꢄ
ꢇ
2πRSS
AIC = 2k + n ln
+ 1 ,
n
For the multilayer adsorption model, there are multiple surface
reactions and multiple binding energies that can be modeled when
considering the different surface species present on the catalyst,
including the reaction of butanol, or hydrated butanol, with vacant
sites, hydrated sites or with both. The butanol and water binding
energies with the surface are included in every model. Additionally
in some models, the binding energy of water on butanol is allowed
to vary from its initial set value, the heat of liquefaction of water.
The simplest surface reaction scheme involves the L–H mecha-
nism previously described, where adsorbed butanol reacts with a
vacant site to produce adsorbed butene and water. This reaction
mechanism fails to describe the experimental data for all mul-
tilayered adsorption simulations. In this scheme a vacant site is
necessary for reaction, and there are no vacant sites in the liquid
water regime due to the multilayered adsorption of water on the
surface.
where k is the number of parameters, n is the number of obser-
vations and RSS is the residual sum of squares. The best model
corresponds to the minimum value of AICc. This analysis indicates
that the model that best fits the experimental observations with
the minimum number of parameters contains only two surface re-
actions, the reaction between a monolayer adsorbed butanol and
a vacant site, and the reaction between a butanol molecule that
has multiple layers of water adsorbed with a catalyst site that has
multiple layers of water adsorbed (Table 3, Entry 2). The model fit
for SiAl and NbOPO4 can be seen in Fig. 7, while a graph for the
AIC analysis is shown in Fig. 8.
The optimized parameter values and corresponding confidence
intervals for the best model are shown in Table 4, column 1. The
confidence intervals are within a few percent of the parameter
value and are in good agreement between the two catalysts tested,
NbOPO4 and SiAl.
If a second surface reaction step for reaction is included, the
multilayered adsorption correctly predicts the trends seen in ex-
periments (Table 3, Entry 2). In this model, a hydrated site can also
A plot of the predicted catalyst coverage as a function of he-
lium flow rate is shown in Fig. 9. In the regime containing liquid

R.M. West et al. / Journal of Catalysis 262 (2009) 134–143
141
Table 4
Optimized kinetic and thermodynamic parameters with corresponding 95% confi-
dence intervals.
Varied inert gas flow
(kJ/mol)a
Varied water vapor
(kJ/mol)a
NbOPO4
Ea1
Ea2
127
4
1
2
144
−76
−56
BEbuOH
BEH
2
2O
SiAl
Ea1
Ea2
BEbuOH
BEH
125
144
−70
−59
5
4
5
1
118 48
135 48
−70 49
−66
1
2O
a
95% confidence interval.
(A)
Fig. 9. Calculated fractional surface coverage for SiAl data: vacant sites, θv (solid
line); mono-adsorbed butanol, θbuOH (dotted line); multilayer water adsorbed on
butanol, θbuOH hyd (dashed line); and multilayer adsorbed water, θhyd (dash–dot
line).
(B)
Fig. 7. (A) Model fit (dashed line) for BET multilayer adsorption mechanism with 2
surface reactions for 2-butanol dehydration over SiAl ("). (B) Model fit (dashed line)
for BET multilayer adsorption mechanism with 2 surface reactions for 2-butanol
dehydration over NbOPO4 (").
studied. The fraction of multilayer adsorbed water decreases from
approximately 95% to almost 40% over the range studied. The sur-
face coverage by butanol species containing multilayers of water is
predicted to decrease from 5% to a negligible amount in the vapor-
only regime.
The effects of the changing catalyst coverage on the surface
reaction rates in the model are shown in Fig. 10. The dominant
reaction surface reaction in the liquid water regime is the reac-
tion between an adsorbed–hydrated butanol molecule and a hy-
drated vacant site. This surface reaction is predicted to decrease
steadily with increasing helium sparging. The surface reaction be-
tween mono-adsorbed butanol and a vacant site is predicted to be
negligible throughout the regime containing liquid water, and it
increases significantly in the vapor-only regime.
To test the model further, the flow rate of water and inert were
experimentally varied over SiAl such that the total molar flow and
butanol flow rate are held constant. The two-reaction multilayered
adsorption model is applied to this data set using the previously
determined optimized parameters. The resulting model and exper-
imental points can be seen in Fig. 11. The model predictions are in
good agreement with the experimental data collected at the vary-
ing water molar flow rates. Additionally, the 4 model parameters
were readjusted to this new data and are included in the right col-
umn of Table 4. The fit values of this second set are just outside
the 95% confidence intervals of the previous model for SiAl, how-
ever, the 95% confidence intervals for the two models overlap.
The estimated kinetic parameters from Table 4 compare fa-
vorably with values found in both experimental and theoretical
Fig. 8. Akaike Information Criterion calculated values (") for proposed reaction
mechanisms.
water, the dominant species predicted for the model is the mul-
tilayer adsorbed water species, with a small fraction of adsorbed
butanol with multiple layers of adsorbed water. In the vapor-only
regime, the fraction of vacant sites is predicted to increase from
a negligible amount to approximately 60% vacancy over the range

142
R.M. West et al. / Journal of Catalysis 262 (2009) 134–143
silica-aluminas could not be found; however, alcohol binding ener-
gies on amorphous silicates cover a range of values: 42.8 kJ/mol
[19] and 60–67 kJ/mol [18] for methanol; 55 kJ/mol [19] for
ethanol; and 67–77 kJ/mol [18] for tert-butyl alcohol. These re-
ported values compare closely to the estimated binding energy for
butanol on NbOPO4, 76 2 kJ/mol, and on SiAl, 70 5 kJ/mol, in
Table 4.
The reaction temperature for all experiments performed was
held constant at 513 K. Therefore, the activation energies for the
surface reactions cannot be distinguished from the pre-exponential
factors used. For this reason, absolute comparisons cannot be made
between the estimated activation energies and values reported in
literature. However, because the same pre-exponential factor is
used for all surface reactions in the model, relative comparisons
can be made for the differences in the estimated activation ener-
gies. The activation energy for the hydrated site reaction (Ea2) is
approximately 17 to 19 kJ/mol higher than the corresponding acti-
vation energy for the dry site reaction (Ea1). Both theoretical and
experimental studies have shown that acidic proton sites can be
covered with either one or two water molecules [20,23]. Upon ad-
sorption of a second water molecule, the binding energy per water
molecule is decreased by 11–17 kJ/mol; a theoretical study pre-
dicted a binding energy decrease from 48 to 37 kJ/mol [23] while
an experimental study measured a binding energy decrease from
51 to 34 kJ/mol [20]. Therefore, a hydrated site is expected to bind
a species less strongly than a non-hydrated site. Because the ap-
parent activation energy is a sum of the intrinsic activation energy
plus the heat of adsorption [24], our observed increase of 17 to
19 kJ/mol in the value of the activation energy (Ea2 vs. Ea1) com-
pares favorably with the expected difference in binding energy of
11 to 17 kJ/mol.
Fig. 10. Surface reaction rates for mono-adsorbed butanol with a vacant site (solid
line) and surface reaction rate for multilayer water on butanol site with a water
multilayer site (dashed line) using the best-fit kinetic model and optimized param-
eters for NbOPO4.
4. Conclusions
Under the conditions of this study, we find that SiAl, NbOPO4
and Nb2O5 are suitable for dehydration of butanol in environ-
ments containing high concentrations of water. These three cat-
alysts show increased activity after treatment with liquid water,
due to the formation of Brønsted acid sites. The rate of butene pro-
duction from an aqueous solution of butanol (10 wt%) is strongly
influenced by the flow rate of sweep gas through the reactor, due
to vaporization of butanol and water. At low flow rates of gas, in-
creasing the gas flow rate causes the preferential vaporization of
butanol because of the high activity coefficient of dilute alcohol in
water. Above a critical gas flow rate, the liquid feed becomes com-
pletely volatilized in the reactor. Increasing the gas flow rate in
the complete vaporization regime causes a decrease in the partial
pressure of both water and butanol.
Various kinetic models including multilayer adsorption of wa-
ter on the catalyst surface can be used to describe the experi-
mental observations. These kinetic models are able to seamlessly
predict the transition from a completely vaporized regime to a
mixed vapor–liquid regime. In the vapor–liquid equilibrium regime
the models predict that most of the catalyst is covered with mul-
tiple layers of water and dehydration takes place by reaction of
hydrated–adsorbed butanol with a hydrated surface site. In the
vapor-only regime the model predicts that the fraction of vacant
active sites increases with increasing gas flow rate as the fraction
of sites covered by multiple layers of water decreases and dehy-
dration takes place by reaction of adsorbed butanol with a vacant
surface site.
Fig. 11. Kinetic model results using the best-fit model for varying molar fraction of
water fed over SiAl catalyst. The experimental data (") are fit using the previously
optimized parameters (dashed line) as well as optimized parameters specific to the
data set (dotted line).
studies. It is typically observed that the heat of adsorption for a
molecule decreases from a relatively high value on a completely
dehydrated surface to the enthalpy of liquefaction as the surface
becomes saturated with multiple layers of the molecule [15–19].
Therefore, for the current study it is anticipated that the estimated
binding energy for water should be between the binding energy
on a clean surface and the enthalpy of liquefaction for water. For
water binding on the zeolite ZSM-5, the binding energy of water
on the proton form, H-ZSM-5 has been reported as 51 4 kJ/mol
for the binding of one water molecule to one site [20], while the
binding to various sodium forms has been reported in the range
of 50–60 kJ/mol [15]. Values for water adsorption on amorphous
silica-aluminas could not be found; however, values for amorphous
silicates are reported as 90–50 kJ/mol [16], 47–49 kJ/mol [17], and
40–53 kJ/mol [18]. These values compare closely to the estimated
values for water adsorption of 56 2 for NbOPO4 and 59 2 or
66 1 for SiAl in Table 4.
Reported binding energies for alcohols are comparable to the
estimated binding energy for butanol on the catalysts studied. Al-
cohols such as methanol bind more strongly to acid sites than
water [20]. Binding energies from 50–65 kJ/mol [21] and 58–
83 kJ/mol ([22] and references therein) have been reported for
alcohols on H-ZSM-5. Values for alcohol adsorption on amorphous
Acknowledgments
Supported by U.S. Department of Energy Office of Basic Energy
Sciences and the National Science Foundation Chemical and Trans-

R.M. West et al. / Journal of Catalysis 262 (2009) 134–143
143
port Systems Division of the Directorate for Engineering. We also
thank E. Kunkes and E. Codner for technical assistance.
[10] M.A. Aramendia, V. Borau, C. Jimenez, A. Marinas, J.M. Marinas, J.A. Navio, J.R.
Ruiz, F.J. Urbano, Colloids Surf. A 234 (2004) 17.
[11] R.R. Soares, D.A. Simonetti, J.A. Dumesic, Angew. Chem. Int. Ed. 45 (2006) 3982.
[12] N.K. Mal, A. Bhaumik, P. Kumar, M. Fujiwara, Chem. Commun. 7 (2003) 872.
[13] P.G. Whitehead, S.I. Sandler, Fluid Phase Equilib. 157 (1999) 111.
[14] G. Glatting, P. Kletting, S.N. Reske, K. Holm, C. Ring, Med. Phys. 34 (2007) 4285.
[15] K. Tsutsumi, K. Mizoe, Colloids Surf. 37 (1989) 29.
[16] V. Bolis, B. Fubini, L. Marchese, G. Marta, D. Costa, J. Chem. Soc. Faraday
Trans. 87 (1991) 497.
[17] E.O. Pedram, A.L. Hines, J. Chem. Eng. 28 (1983) 11.
References
[1] Y. Roman-Leshkov, J.N. Chheda, J.A. Dumesic, Science 312 (2006) 1933.
[2] Y. Roman-Leshkov, C.J. Barrett, Z.Y. Liu, J.A. Dumesic, Nature 447 (2007) 982.
[3] G.W. Huber, R.D. Cortright, J.A. Dumesic, Angew. Chem. Int. Ed. 43 (2004) 1549.
[4] R.M. West, Z.Y. Liu, M. Peter, J.A. Dumesic, ChemSusChem 1 (2008) 417.
[5] E.L. Kunkes, D.A. Simonetti, R.M. West, J.C. Serrano-Ruiz, C.A. Gärtner, J.A.
Dumesic, Science 322 (2008) 417.
[18] B. Fubini, V. Bolis, A. Cavenago, E. Garrone, P. Ugliengo, Langmuir 9 (1993) 2712.
[19] B.J. Stanley, G. Guiochon, Langmuir 11 (1995) 1735.
[6] L. Li, Y. Yoshinaga, T. Okuhara, Phys. Chem. Chem. Phys. 4 (2002) 6129.
[7] C.D. Baertsch, K.T. Komala, Y.H. Chua, E. Inglesia, J. Catal. 205 (2002) 44.
[8] M.A. Sanchez-Castillo, R.J. Madon, J.A. Dumesic, J. Phys. Chem. B 109 (2005)
2164.
[20] A. Ison, R.J. Gorte, J. Catal. 89 (1984) 150.
[21] B. Hunger, S. Matysik, M. Heuchel, W. Einicke, Langmuir 13 (1997) 6249.
[22] P.E. Sinclair, R.A. Catlow, J. Chem. Soc. Faraday Trans. 93 (1997) 333.
[23] M. Krossner, J. Sauer, J. Phys. Chem. 100 (1996) 6199.
[9] H. Liu, P. Cheung, E. Iglesia, J. Phys. Chem. B 107 (2003) 4118.
[24] M.I. Temkin, Adv. Catal. (1979) 173.