Comparative study of the kinetics of methane steam reforming on supported Ni and Sn/Ni alloy catalysts: The impact of the formation of Ni alloy on chemistry

Article abstract of DOI:10.1016/j.jcat.2009.02.006

We report the results of detailed kinetic studies for methane steam reforming on supported Ni and Sn/Ni surface alloy catalysts. The kinetic data were interpreted in terms of mechanism-based overall rate expression. We show that the activation of C{single

Full text of DOI:10.1016/j.jcat.2009.02.006

Journal of Catalysis 263 (2009) 220–227
Contents lists available at ScienceDirect
Journal of Catalysis
www.elsevier.com/locate/jcat
Comparative study of the kinetics of methane steam reforming on supported Ni
and Sn/Ni alloy catalysts: The impact of the formation of Ni alloy on chemistry
∗
Eranda Nikolla, Johannes Schwank, Suljo Linic
Department of Chemical Engineering, 2300 Hayward Street, University of Michigan, Ann Arbor, MI 48109, USA
a r t i c l e i n f o
a b s t r a c t
Article history:
We report the results of detailed kinetic studies for methane steam reforming on supported Ni and
Sn/Ni surface alloy catalysts. The kinetic data were interpreted in terms of mechanism-based overall
rate expression. We show that the activation of C–H bonds in methane is the rate-controlling step on
both catalysts. Isotopic CH4/CD4 labeling studies were performed to independently verify the proposed
mechanism. The role of Sn is to displace Ni atoms from under-coordinated sites on Ni particles
and to move the critical reaction channels to more abundant well-coordinated sites. We show that
previously observed increased resistance to carbon deactivation of Sn/Ni compared to monometallic Ni in
hydrocarbon reforming reactions can be attributed to the Sn-induced lowering in the binding energy of
carbon on low-coordinate sites, which serve as carbon nucleation centers, and to an enhanced propensity
of Sn/Ni to oxidize carbon surface species. The conclusions derived from the experimental studies are in
agreement with DFT calculations.
Received 23 December 2008
Revised 8 February 2009
Accepted 9 February 2009
Available online 5 March 2009
Keywords:
SOFC
DFT
Steam reforming kinetics
Carbon deactivation
Ni alloys
Isotopic labeling
©
2009 Elsevier Inc. All rights reserved.
1
. Introduction
in Ref. [11]). Sn/Ni catalysts showed significantly improved carbon
tolerance compared to the monometallic Ni catalysts under identi-
cal experimental conditions [10–12].
While comprehensive kinetic studies on Sn/Ni catalysts have
not been performed to date, it has been proposed, based on Den-
sity Functional Theory (DFT) calculations, that there are two likely
mechanisms associated with the superior performance of Sn/Ni
Steam reforming of hydrocarbons (for example methane, CH4 +
H2O ↔ CO+3H2) is a crucial reaction for the production of synthe-
sis gas. The process is also important for the direct electrochem-
ical conversion of hydrocarbons in solid oxide fuel cells (SOFCs).
Commercial catalyst for this reaction is Ni supported on an oxide.
The process is run under a wide range of conditions with operat-
ing temperatures from ∼750 to 1200 K. One of critical problems
with long-term performance of Ni catalysts is the formation of
carbon deposits on the catalyst surface, which evolve into carbon
filaments, ultimately diminishing the performance of the catalyst
[10–12]. One mechanism is based on the kinetic control of the
chemistry of carbon atoms and CHx on the catalysts surface. This
mechanisms assumes that the difference in the rates of the oxida-
tion of carbon atoms and CHx fragments and the rates associated
with the formation of C–C bonds, which are critical for the de-
velopment of solid carbon deposits that deactivate the catalyst, is
significantly higher on Sn/Ni compared to Ni, i.e., the Sn/Ni cata-
lyst is more efficient than Ni in the preferential formation of C–O
rather than C–C bonds. The second mechanism assumes that car-
bon deactivation is governed by the thermodynamic control of the
nucleation of carbon at the low coordinated Ni sites. Previous in-
situ transmission electron studies (TEM) have demonstrated the
importance of the low coordinated sites for nucleation and growth
of carbon fibers [5,16]. This mechanism assumes that Sn atoms
displace low-coordinated Ni atoms, preventing the nucleation of
carbon deposits on these sites.
[1–9].
It has been shown previously that the tolerance of Ni catalysts
and electro-catalysts to carbon-induced deactivation in hydrocar-
bon steam reforming, partial oxidation, and electro-chemical ox-
idation reactions can be significantly improved by impregnating
Ni with small amounts of Sn (2–1 wt% with respect to Ni for
Ni particles with the diameter between 30 and 200 nm) [10–15].
For example, we have demonstrated that supported monometal-
lic Ni catalysts deactivate rapidly in steam reforming of isooc-
tane at stoichiometric steam to carbon ratio (see Fig. 8 in Ref.
[11]). Similar conclusions were obtained for methane steam re-
In this contribution we report the results of detailed kinetic
studies for methane steam reforming on supported Sn/Ni and Ni
catalysts. We focus on the analysis of the reaction kinetics at high
temperatures (973–1073 K). These conditions are important for
tubular methane steam reformer reactors and SOFCs [2]. The ki-
forming at sub-stoichiometric steam to carbon ratios (see Fig. 7
Corresponding author. Fax: +1 734 764 7453.
*
E-mail address: linic@umich.edu (S. Linic).
0
021-9517/$ – see front matter © 2009 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2009.02.006

E. Nikolla et al. / Journal of Catalysis 263 (2009) 220–227
221
netic data for supported Ni and Sn/Ni catalysts were interpreted
in terms of the mechanism-based overall rate expressions. Isotopic
CH4/CD4 labeling studies were performed to further investigate
the proposed catalytic sequence. These studies allowed us to in-
dependently identify the rate-controlling elementary steps on the
different catalysts. The results of the kinetic studies were used to
draw conclusions about the mechanisms responsible for the im-
proved carbon tolerance of Sn/Ni compared to monometallic Ni.
The conclusions of the experimental studies are supported by DFT
calculations.
size of approximately 50 nm for Ni and Sn/Ni samples. The crys-
tallite size was independently verified using transmission electron
microscopy (TEM) and X-ray diffraction (XRD). The TEM experi-
ments were performed with the Joel 2010F electron microscope at
−7
200 keV under a pressure of 1.5×10
Torr. Ni and Sn/Ni particles
were detected using energy X-ray dispersive spectroscopy. The size
of the crystallites was measured using Digital Micrograph. The XRD
experiments were conducted using a Cu-Kα source of a Philips
XRG5000 3 kW X-ray generator with crystal alignment stage and
a Rigaku thin film camera. The spectra were analyzed using the
Scherrer equation with Jade v.7 software.
The kinetics of methane steam reforming on supported mono-
metallic Ni catalysts have been studied by multiple researchers
It was established based on quantitative analysis of X-ray
photo-electron spectroscopy (XPS) results that for this Sn loading
and the size of active particles the Sn concentration in the surface
layers of Ni was between 15 and 20% [11]. Assuming that Sn and
Ni atoms are homogeneously distributed among different particles,
these results suggested that Sn almost exclusively segregates to the
surface layers of Ni crystallites. To ensure that Sn and Ni atoms are
in contact with each other in the Sn/Ni crystallites, we measured
the near edge electronic structure of Sn/Ni, showing that chemical
bonds between Sn and Ni atoms are formed in the surface layers
of the particles [25]. Temperature programmed reduction (TPR) ex-
periments showed that pure Ni and Sn materials reduce at ∼700
and 900 K respectively, which is below the temperatures at which
our experiments were performed [11,12]. The Sn/Ni catalysts re-
duced at ∼670 K [11,12]. These measurements suggested that the
catalytic materials used in these experiments are not expected to
be oxidized under the reducing reaction conditions.
Kinetic studies were conducted isothermally in a packed bed
quartz reactor operated at methane conversion from 2 to 10%. The
diameter of the quartz reactor was 0.375 inches. The set-up in-
cluded a set of mass flow controllers, a pair of thermocouples
(inside and outside the reactor), and a syringe pump for water
delivery. The steam lines were heated to 473 K to avoid conden-
sation. The reactor effluent was analyzed using a Varian gas chro-
matograph (Varian CP 3800) equipped with thermal conductivity
detectors (TCD) and a flame ionization detector (FID). Approxi-
mately 0.01 g of Ni/YSZ and Sn/Ni/YSZ catalysts with grain sizes
of 150–200 μm were diluted with 0.5 g of quartz powder (250–
400 μm) and loaded into the packed bed reactor. The bed height
was ∼0.5 cm. The catalyst powder was held in place by quartz
wool plugs. The catalysts were reduced at 1173 K under a stream
of 30% H2/N2 for 3 h prior to measuring reaction rates. The rates
of methane steam reforming were measured at total flow rates of
50–200 sccm as temperature was varied from 923 to 1073 K for Ni
and 973 to 1123 K for Sn/Ni.
[9,17–23]. Although there have been many controversies, recent
theoretical and experimental contributions have shown some con-
sistent results. For example, Bengaard et al. utilized a combined
theoretical and experimental approach to propose a mechanism for
steam reforming of methane on Ni [9]. It was shown that methane
steam reforming is a structure sensitive reaction and that under-
coordinated Ni surface sites are more active than close packed
Ni sites. It was also demonstrated that the rate-limiting step in
the process is the activation of C–H bonds in methane. These
conclusions are consistent with the recent experimental analy-
sis of methane steam reforming at temperatures of 823–1023 K
on Ni/MgO catalysts by Wei and Iglesia [20]. These authors also
demonstrated that the forward rate is first order with respect to
the methane partial pressure and that it does not depend on the
partial pressure of water or the reaction products. It was also ar-
gued that the rates of carbon filament formation are governed by
the chemical activity of carbon atoms chemisorbed on the catalyst
surface. Previous contributions have postulated that for lower op-
erating temperatures there is a switch in the rate-controlling step
from the activation of CH4 to the formation of CO on the catalyst
surface [24].
2
. Experimental methods and catalyst characterization
The Ni and the Sn/Ni catalysts were supported on (8 mol%)
yttria-stabilized zirconia (YSZ). YSZ was prepared via a standard
co-precipitation method. A mixture of yttrium nitrate (Y(NO3)3·
6
H2O) and zirconyl chloride (ZrOCl2·2H2O) dissolved in deionized
water was precipitated using a solution of ammonium hydroxide.
After filtration and drying for 12 h, the precipitate was calcined
at 1073 K for 2 h. Aqueous solutions of Ni nitrate were used to
impregnate the YSZ support via the incipient wetness technique.
The nominal Ni loading was about 15 wt% with respect to the
total catalyst. After impregnation the catalysts were calcined at
8
73 K for 2 h in dry air to convert the Ni nitrate to Ni oxide and
Thermal gravimetric analysis (TGA) was performed using a TA
instruments thermo-gravimetric analyzer (TGA Q500). Approxi-
mately 0.2 g of catalyst was reduced under 30% H2/N2 at 1073 K
for 2 h. Once reduced, the catalysts were cooled to room temper-
ature while exposed to pure nitrogen. The catalysts were then ex-
posed to dry methane and the temperature was ramped to 1073 K
at a set heating rate to measure the rate of methane decompo-
sition. The weight of the sample was measured as a function of
temperature to determine the amount of deposited carbon.
The Dacapo pseudo-potentials plane wave code (http://www.
camp.dtu.dk) was employed for all spin polarized Density Func-
tional Theory (DFT) calculations. The Ni surface was modeled using
two different model surfaces, Ni(111) and Ni(211). These model
systems allowed us to explore two representative active sites;
under-coordinated step-edge sites on the (211) surface and close
packed sites on the (111) surface. Our approach was to find a
set of computational parameters that ensured the convergence of
the relative energies of various structures with respect to each
other. The optimal lattice constant for Ni bulk was calculated to
be 3.52 Å. For model systems with the (111) surface termination,
then reduced at 1173 K for 3 h in a stream of 30% H2/N2. The
Sn/Ni/YSZ catalyst was synthesized by impregnating the NiO/YSZ
catalyst with aqueous solution of Sn chloride (SnCl2·4H2O) via the
incipient wetness technique. The nominal Sn loading was about
1
wt% with respect to the Ni content in the catalyst. The Sn/Ni/YSZ
catalyst was also reduced at 1173 K for 3 h in a stream of 30%
H2/N2.
Single point BET analysis was utilized to measure the surface
area of the YSZ support via N2 physisorption. The physisorption
experiments were conducted using Quatachome’s ChemBet 3000
equipped with thermal conductivity detectors (TCD). The active
surface area and dispersion of the Ni and Sn/Ni crystallites was
measured using H2 chemisorption at 313 K. The measurements
were conducted using Micromeritics ASAP 2020 instrument. Prior
to hydrogen chemisorption the samples were reduced under hy-
drogen for 2 h at 1073 K ex situ and for 1 h at 1073 K in situ.
The particle diameter and dispersion were calculated assuming a
stoichiometry of one hydrogen atom per one Ni atom and spheri-
cal Ni particle geometry. These experiments yielded the crystallite

222
E. Nikolla et al. / Journal of Catalysis 263 (2009) 220–227
we used a 3 × 3 supercell with 4 layers of metal atoms. The cal-
culations for the (211) surface termination were performed using
slabs with 9 layers of metal in 1 × 3 supercells. The convergence
for the (111) surface was obtained with 18 special Chadi–Cohen k-
points to sample the Brillouin zone. For the (211) surface, we used
Monkhorst–Pack mesh with a 3 × 3 × 1 k-point grid to sample the
Brillouin zone. Adsorbates were adsorbed on one side of a slab.
Approximately 15 Å of vacuum separated the slabs, and a dipole-
correction scheme was employed to electro statically decouple the
slabs. Electron exchange correlation effects were described using
the generalized gradient approximation (GGA) with the Perdew–
Wang 91 (PW91) functional [26–28]. Vanderbilt pseudo-potentials
were employed to describe core electrons [29]. The density of va-
lence electrons was determined self-consistently by iterative diag-
onalization of Kohn–Sham Hamiltonian using Pulay mixing of den-
sities. The plane wave basis set used to describe the one-electron
mass transport and diffusion limitations and not compromised by
reverse rates [20].
The overall reaction rate (raten) of a chemical reaction can be
expressed in terms of forward reaction rate (rate ) and the ap-
f
proach to equilibrium according to Eq. (1). The approach to equi-
librium (η) can be calculated based on Eq. (2) as the product of
partial pressures of the reaction products divided by the product
of the partial pressures of reactants and divided by the equilibrium
constant for the overall reaction. The approach to equilibrium in a
packed bad reactor varies from zero at the reactor inlet (if only re-
actants are fed into the reactor) to some finite value less than one
at the reactor outlet, and it can be manipulated by changing the
flow rate, mass of the catalyst, temperature, and other factors that
might impact the overall conversion. We have performed all our
experiments under conditions that yielded an approach to equilib-
rium less than 0.01 at the reactor outlet. Equation (1) shows that
if the approach to equilibrium (η) is much less than one, then the
states was cut off at 350 eV. An electronic temperature (k T ) of 0.1
b
was utilized during calculations with the final results extrapolated
to 0 K. In the geometry optimization calculations on the (111) sur-
face termination, the two top substrate layers and adsorbates were
allowed to relax. On the (211) termination, the top six substrate
layers and adsorbates were allowed to relax. The calculations were
assumed to be converged when the maximum force on every de-
gree of freedom was smaller than 0.01 eV/Å.
measured overall reaction rate (rate ) is approximately equal to the
n
forward reaction rate (rate ). The forward reaction rate is required
f
to obtain rigorous reaction orders and overall activation barriers.
raten
rate =
,
(1)
(2)
f
(
1 − η)
3
[H ] [CO]
2
1
η =
.
To model the Sn/Ni surface alloy we have employed identical
Ni(111) slabs with a number of Sn atoms, equivalent to 1/9 and
[
CH4][H2O] Keq
Aside from ruling out possible artifacts in our measurements
due to the influence of reverse rates, we have also eliminated ar-
tifacts due to mass transport limitations or possible catalyst deac-
tivation during the measurements. To make sure that the catalysts
were not deactivating or sintering during the experiments, we cy-
cled the temperature between the high and low limits multiple
times during the process and were able to reproduce consistent
rates. The absence of external mass transport limitations was veri-
fied by measuring the reaction rate at constant space time (W/Fao,
catalyst mass/molar flow rate) but changing the flow velocity.
These measurements showed that there was no change in the rate
and therefore no external mass transport limitations. We have also
changed the size of Ni/YSZ particles and mixed these with various
amounts of inert. In these experiments we observed linear mea-
sured rate as a function of the fraction of Ni/YSZ mixed in the
inert. The linearity to this relationship indicated that there were
no internal diffusion limitations. This procedure is similar to those
employed recently in the contributions of Iglesia and coworkers
and outlined previously by other contributors [20,37–40].
2
/9 ML, displacing Ni atoms in the surface layer of Ni. As described
above, we have previously established by X-ray photoelectron spec-
troscopy (XPS) and electron energy loss near edge spectroscopy
(ELNES) that Sn preferentially segregates to the surface layers of
Ni [10–12,25]. Furthermore, DFT calculations also showed that the
model surface alloy structures, with Sn atoms displacing Ni atoms
from the top Ni layer, have lower formation energies relative to Sn
displacing Ni atoms from the Ni bulk, Sn adsorbing on the surface
of Ni, or Sn assembling in segregated pure Sn phases [11,12].
The first-order transition states for one-atom diffusion were
identified by probing the high-symmetry sites between reactant
and product states. The diffusing atom was fixed in the x–y plane
on these sites and it was allowed to relax in the z direction. En-
ergies were calculated for all high symmetry sites (hollow, bridge,
and on-top) and the potential energy surfaces were constructed.
This procedure allowed us to identify the transition state geome-
tries as those associated with the maximum energy states along
the reaction coordinate and minimum with respect to other de-
grees of freedom [30–32]. The forces in these calculations were
minimized to 0.01 eV/Å. Transition states for the activation of
C–H bonds in CH4 and for the formation of C–O and C–C bonds
were identified using Climbing Nudged Elastic Band method [33–
3.1. Reaction orders and activation barriers
3
5]. The identity of the first-order transition states was validated
The reaction orders with respect to methane and water for
methane steam reforming on Sn/Ni/YSZ and Ni/YSZ were deter-
mined by measuring the response of the forward rate to a change
in the respective partial pressures of CH4 and H2O while keeping
everything else in the system constant. Figs. 1a and 1c show that
there is a linear increase in the forward rate for both Ni and Sn/Ni
catalysts as the partial pressure of methane is increased. In these
experiments, the partial pressure of methane was varied from 0.2
to 0.6 atm, while the partial pressure of water was kept constant
at 0.39 atm. The feed was balanced by argon to reach the total
pressure of 1 atm. Figs. 1b and 1d show that the forward reac-
tion rate is not affected by a change in the partial pressure of
water from 0.2 to 0.6 atm for a constant methane partial pres-
sure of 0.39 atm. The behavior was consistently observed over a
temperature range between 973 and 1073 K for Ni/YSZ and 1013
and 1123 K for Sn/Ni/YSZ.
by making sure that the forces acting on the system change sign
as the systems moves through the transition state geometry. We
have further validated the transition states by slightly changing the
geometry of the transition state along the reaction coordinate to-
wards the product or reactant geometry and allowing the system
to fully relax into the respective product and reactant states [36].
3
. Results and discussions
The steam reforming reaction is operated at fairly high temper-
atures, leading to large gradients in temperature and concentration
within the catalyst bed. These gradients along with contributions
due to significant rates of reverse reactions can compromise the
kinetic information obtained in reactor studies over a range of
temperature, flow rates, and partial pressures [20]. To obtain re-
liable and rigorous kinetic parameters it is important to establish
that the measurements are performed in reactor systems free of
Our results also show that as we varied the methane conver-
sion from 1 to 10% by changing the reactor temperature for a give

E. Nikolla et al. / Journal of Catalysis 263 (2009) 220–227
223
Fig. 1. (a) A plot of reaction rate as a function of the partial pressure of methane (PCH ) for a Sn/Ni/YSZ catalyst at 1073 K and a constant water partial pressure of 0.39 atm
4
diluted in argon. Total pressure was 1 atm. (b) A plot of reaction rate as a function of PH2O for a Sn/Ni/YSZ catalyst at 1073 K and a constant methane partial pressure of
0.39 atm diluted in argon. Total pressure was 1 atm. (c) Same as in (a) for Ni/YSZ and at 1013 K. (d) Same as (b) for Ni/YSZ and at 1013 K.
flowrate, the order of the forward rate with respect to H2O and
CH4 did not change on either catalyst. Since higher CH4 conversion
would lead to higher partial pressures of the reaction products,
these observations imply that, for the conditions explored in this
paper, the forward reaction rate is not sensitive to the changes in
the partial pressure of the reaction products for either catalyst, and
that the steady state coverage of product species (or surface inter-
mediates which are precursors for the reaction products) on active
sites does not change significantly. If this was not the case we
would have seen a deviation in the reaction orders as a function of
the change in conversion. Identical conclusions were made when
small amount of product species were added to the inlet stream
(up to 10 mol% with respect to methane). These conclusions are
consistent with the results of our DFT calculations, which showed
that at the relevant conditions (T ∼ 1000 K) the adsorption energy
of CO, H2O, CO2, and H2 in its lowest energy configurations (e.g.,
dissociated or molecular) on under-coordinated sites of the Ni(211)
surface are −164, −135, 19.2, and −77.2 kJ/mol, respectively. The
adsorption energies are even lower for close packed Ni surfaces
and for all sites on the Sn/Ni surfaces. These adsorption energies
are insufficient to keep these species on the surface in significant
concentrations under the high temperature conditions (T > 973 K),
i.e. the Gibbs free energy of adsorption is positive. Based on the
presented experimental results and the results of DFT calculations,
we suggest that for both catalysts, the catalyst surface is approx-
imately free of adsorbates under steady state conditions, and the
overall rate is controlled only by the rate of the dissociation of
CH4. Similar observations were made previously for Ni/MgO cata-
lysts [20].
Fig. 2. A plot of the turnover frequency for methane steam reforming as a function
of the inverse temperature for 1 wt% Sn/Ni/YSZ and Ni/YSZ catalysts when exposed
to 0.39 atm of methane, 0.39 atm of water diluted in argon. The total pressure
was 1 atm.
The apparent activation energy for methane steam reforming,
defined in terms of the response of the forward rate to changes
in operating temperature, was determined in an identical reactor
setup. Prior to each measurement, the absence of mass trans-
port limitations was verified. Fig. 2 shows an Arrhenius plot for
methane steam reforming on the Ni/YSZ and Sn/Ni/YSZ catalysts
for 0.39 atm of methane, 0.39 atm of water diluted in 0.22 atm
of argon. We note that the reported TOFs for the Sn/Ni/YSZ cat-
alyst were measured for the Sn loading of ∼1% and the particle

224
E. Nikolla et al. / Journal of Catalysis 263 (2009) 220–227
size of approximately 50 nm, which results in the Sn surface con-
centration between 15 and 20%. We find that as the loading of
Sn increases the TOFs (reported with respect to the total number
of surface Sn + Ni sites) decreases. The apparent activation bar-
rier was determined to be ∼101 ± 4 and 132 ± 4 kJ/mol for the Ni
and Sn/Ni catalysts, respectively. The measured turnover frequency
and activation barrier on Ni is consistent with previous reports for
supported Ni catalysts [20,22,41]. Further analysis of Fig. 2 suggests
that the overall pre-exponential factor, calculated from the y-axis
intercept in the Arrhenius plot, is by ∼10–15 times larger for the
Sn/Ni catalyst compared to the Ni catalyst. It is important to note
that the overall pre-exponential factors were calculated from our
experimental turnover frequencies based on the total number of
surface sites, which are approximately identical for the Sn/Ni and
Ni catalysts, i.e., it was not based on the number of active sites
since it is almost impossible to count the active sites. This means
that if the numbers of active surface sites (the sites that partici-
pate in the most frequent reaction channels) on Ni and Sn/Ni are
different from each other, the overall pre-exponential factors will,
in addition to describing the pre-exponential factor associated with
the elementary step that controls the overall rate, also reflect this
difference.
If one accepts that under the relevant conditions the surfaces
of both catalysts are nearly free of adsorbates and that the rate is
controlled by the dissociation of CH4, then the difference in the
overall activation barrier and pre-exponential factors between Ni
and Sn/Ni is a consequence of the change in the nature of surface
sites that dominate the catalytic process. It appears that on pure
monometallic Ni the active sites (i.e., the sites responsible for the
dominant reaction channel) are different in their chemical activity
and relative abundance compared to the Sn/Ni alloy catalyst. We
shed more light on this issue further below.
We have also calculated, based on the measured rates of the
decomposition of dry methane, the activation barriers for the dom-
inating pathways, i.e., the pathways corresponding to the temper-
ature where the rate reaches maximum values. To accomplish this
we have employed the Redhead analysis [42]. If we assume that
the rate of methane decomposition is first-order with respect to
the partial pressure of methane and nth order with respect to the
∗
concentration of empty sites [ ] we obtain the following decompo-
sition rate expression:
ꢀ
ꢁ
∗
d( )
−Ea
∗
n
rate = −
= A P
a
CH4
[ ] exp
,
(3)
dt
RT
where Aa is the pre-exponential factor for adsorption of methane,
∗
PCH4 is the partial pressure of methane, [ ] is the concentration
of the empty sites, Ea is the activation energy for methane dis-
sociative adsorption, R is the gas constant, and T is the tempera-
ture. In our experiments temperature was varied linearly with time
(T = T + βt). Assuming that the maximum rate occurs at a tem-
0
dr
dT
perature Tp when
= 0, we derive a relationship between the
temperature associated with the maximum rate and the activation
barrier associated with the reaction channel that yields the maxi-
mum rate:
ꢀ
ꢁ
−
Ea
βEa
RTp2
Aan[∗]n−1 PCH exp
=
.
(4)
4
RTp
If we assume that the pre-exponential factors for the dissocia-
tive adsorption step, the fraction of active sites that are empty at
T = Tp, and the reaction order with respect to empty sites are
identical for Ni and Sn/Ni, we find that the activation barrier for
methane decomposition is higher by ∼30 kJ/mol on Sn/Ni than on
∗
Ni. If we further assume that the maximum rate occurs at [ ] = 0.5
(this is a reasonable assumption since it provides a compromise
between the availability to free sites required for the activation
3
.2. Thermal gravimetric analysis of methane activation
of CH4 and the high rate constant which depends on T ), Aa is
5
10
1/s/atm for both catalysts (this assumption is based on an
We have also utilized thermal gravimetric analysis (TGA) to
approximation of the translational, rotational and vibrational par-
tition functions for the initial and transition states), and that the
reaction order with respect to the concentration of empty sites n
is 2 (two sites are required to activate methane), we obtain an
activation barrier for methane decomposition of ∼132 kJ/mol on
Sn/Ni/YSZ and 103 kJ/mol on Ni/YSZ. We note that these activa-
study the decomposition of methane on the Ni and Sn/Ni cata-
lysts. In these experiments, the change in weight of the catalyst
exposed to 1 atm of dry methane was measured as the tempera-
ture was ramped from 298 to 1173 K. Fig. 3a shows a plot of the
weight gain for the Sn/Ni/YSZ and Ni/YSZ catalysts as a function
of temperature. The increase in the weight of the catalysts is a
consequence of the formation of carbon deposits during methane
decomposition (rxn 1).
∗
tion barriers are not a strong function of n and [ ], and similar
activation barriers are obtained for significantly different values n
∗
and [ ]. These activation barriers are almost identical to the activa-
CH4 + ∗ → C∗ + 2H2.
tion barriers obtained in the Arrhenius analysis of the steady state
reactor data shown in Fig. 2. The similarity in the measured acti-
vation barriers is not surprising, and it reinforces our conclusions
that the rate of the overall process is governed by the activation of
methane, and it is not sensitive to partial pressures of other gas-
phase species.
Fig. 3a shows that methane decomposes at lower temperature
on Ni than on Sn/Ni. We have also plotted in Fig. 3b the measured
rate of weight gain as a function of temperature, which is a mea-
sure of the rate of methane decomposition on the catalyst surface.
Fig. 3b shows that the rate curve is broader on Sn/Ni compared
to Ni. A possible interpretation for this is that there are multi-
ple reaction channels (different sites) for methane decomposition
on Sn/Ni. This supports the conclusions of our previous charac-
terization studies which showed that in the Sn/Ni catalyst, the Sn
atoms are dispersed in the surface layers of Ni particles forming
a large number of geometrically and electronically diverse surface
sites [10–12]. Careful inspection of Fig. 3b shows that for Sn/Ni
there is a small shoulder in the rate curve, appearing at temper-
ature just below 500 K, which overlaps with the peak of the rate
curve for Ni. This suggests that there are a small number of sites
on Sn/Ni that are similar to the sites on Ni in their ability to ac-
tivate the C–H bonds in methane. However, the number of these
sites is very small, and these sites are not responsible for the main
methane decomposition channels, which are moved to higher tem-
peratures for Sn/Ni compared to monometallic Ni.
3.3. DFT calculations of methane activation on Ni and Sn/Ni model
systems
We have also utilized DFT calculations to study the dissociation
of methane on Ni and Sn/Ni surface alloy. We have focused on two
representative active sites, one containing closed packed terrace
sites and another containing under-coordinated step/edge sites. To
model the terrace and step/edge sites we have utilized the (111)
and (211) surface termination, respectively. We calculated that the
differences in the electronic energies between the activated com-
plex (transition state) and the reactant in the methane dissociation
reaction are 115, 105, and 79.5 kJ/mol on Sn/Ni(111), Ni(111) and
Ni(211), respectively. The calculated geometries of transition states
are shown in Fig. 4. The transition states geometries are very simi-
lar to those reported by Abild-Pedersen et al. in Ref. [43]. Previous

E. Nikolla et al. / Journal of Catalysis 263 (2009) 220–227
225
Fig. 3. (a) The percent weight gain of the catalyst as a function of temperature during methane decomposition on Sn/Ni/YSZ and Ni/YSZ catalysts. (b) The rate of methane
decomposition as a function of temperature for Sn/Ni/YSZ and Ni/YSZ.
Fig. 4. Potential energy barriers for the activation of C–H bond in methane on Ni(211), Ni (111) and Sn/Ni (111). The inserts show the geometries of the initial, transition, and
final states. The unit cell in the calculations was 3 × 3 for the (111) surface and 3 × 1 for the (211) surface. The concentration of Sn in the surface layer of Sn/Ni(111) was
1
/9 ML. The Sn atom is in purple, H is black, Ni is light blue, and C is gray. (For interpretation of the references to color in this figure legend, the reader is referred to the
web version of this article.)
calculations have reported the activation barriers for the dissocia-
tion of CH4 on Ni(111) ranging between 70 and 125 kJ/mol [44–49].
It has also been reported that the difference between the barri-
ers for methane activation on Ni(111) and Ni(211) is ∼20 kJ/mol,
which is close to our value of 25 kJ/mol. It can easily be shown
that if we express the rate constant based on the transition state
theory and evaluate the appropriate derivative of the rate constant
to obtain the activation barrier, defined in terms of Arrhenius rate
expressions as the parameter that describes the change in the rate
constant due to the change in T , the calculated activation barriers
for methane dissociation, defined in terms of Arrhenius rate ex-
pressions, at relevant temperatures are ∼123, 113, and 87.5 kJ/mol
for Sn/Ni(111), Ni(111), and Ni(211), respectively. The analysis sug-
gests that the dissociation of CH4 is more facile on low-coordinated
Ni sites. Reasonable agreement between the measured overall ac-
tivation barrier and calculated barrier on under-coordinated Ni
sites argues that the dominant reaction channels on monometal-
lic Ni involve the activation of CH4 in the rate-controlling step on
under-coordinated Ni sites. We note that similar observations have
been made by others based on DFT calculations or by compar-
ing the reaction rates on Ni particles with different concentrations
of under-coordinates sites [24,43]. The DFT studies also suggest
that on Sn/Ni the measured overall activation barrier associated
with the dominant reaction channel is close to the barrier cal-
culated for the dissociation of CH4 on the Ni(111) and Sn/Ni(111)
surfaces. The results suggest that Sn neutralizes the highly active
under-coordinated sites on Ni and moves the dominant pathways
to well-coordinated sites. This is consistent with the higher mea-
sured value of Arrhenius pre-exponential factor for Sn/Ni compared
to monometallic Ni, which is a consequence of a larger number
of well-coordinated than close packed sites on a catalytic particle.
Similar conclusions have been made for sulfur modified Ni cata-
lysts which also exhibit improved carbon tolerance compared to
monometallic Ni [9].
3.4. CH /CD kinetic isotope labeling studies
4
4
The analysis presented in the previous sections suggested that
the rate controlling step in methane steam reforming on Ni and
Sn/Ni is the activation of C–H bonds in methane. To independently
verify our conclusions regarding the nature of the rate-controlling
elementary step we have studied the kinetics of the process using
labeled methane (CD ). In these experiments, the reaction rates
4
were measured in an identical differential reactor setup, free of
mass transport limitations and artifacts due to reverse rates. Fig. 5
shows plots of the measured forward reaction rates as a function of
time for Ni/YSZ and Sn/Ni/YSZ catalysts as the feed was switched
from CH4 to CD4 and back to CH . Fig. 5 shows that as the feed is
4
switched from CH4 to CD4, the forward rate decreases for both cat-
alysts. Quantitative analysis of the reactions rates shows that the

226
E. Nikolla et al. / Journal of Catalysis 263 (2009) 220–227
Fig. 5. The rate of steam reforming is measured for CH4 and CD4 reactants as a function of time on (a) Ni/YSZ at 973 K and (b) Sn/Ni/YSZ at 1033 K. The dark lines depict
points in time when the reactant is switched from CH4 to CD4 and back to CH4. The steam-to-carbon ratio was 1.
drop in forward rate is approximately equal on both catalysts for
a given reaction temperature. The observed kinetic isotope effect
demonstrates that the overall rate is controlled to a similar de-
gree by the rate of C–H bond dissociation on both catalysts. Similar
observations were made by others for supported monometallic Ni
catalysts [20,50]. We note that the measured kinetic isotope effect
of ∼1.6 is slightly lower than the isotope effect predicted based on
the transition state theory, assuming that the rate-controlling step
involves the C–H and C–D vibration with a frequency identical to
those measured in the gas-phase for CH4 and CD4, respectively.
The reasons for this discrepancy might be in slightly different vi-
brational frequencies of normal modes involved in the reaction
coordinate on two surfaces compared to the normal modes associ-
ated with gas-phase CH4 and CD4.
ing paragraphs. The calculated barriers for the activation of CH
bonds in CH4 are shown in Fig. 4. The activation barrier for the
formation of C–O bonds was calculated as the energy difference
between the highest and lowest points on the lowest potential en-
ergy surface associated with the diffusion of C and O atoms and
the formation of the C–O bonds on the closed packed Ni(111) and
Sn/Ni(111) surfaces. Similarly, the activation barrier for the forma-
tion of C–C bonds was calculated as the energy difference between
the highest point and lowest point on the lowest potential energy
surfaces associated with the diffusion of C atoms and the forma-
tion of C–C bonds on the identical surfaces. It is important to stress
that the calculated activation barriers for the formation of C–O and
C–C bonds are very similar to those associated with the formation
of HC–O and HC–CH bonds respectively, which might be impor-
tant if the CH fragment is the most abundant carbon containing
adsorbate. The pre-exponential factors for these processes were ap-
3
.5. Mechanisms associated with the improved carbon-tolerance in
5
reforming reactions of Sn/Ni compared to Ni
proximated using the transition state theory to be 10 1/atm/s for
the dissociative adsorption of CH4 and 1013 1/s for the formation
of C–O and C–C bonds. In the rate calculations we have assumed
that the concentration of reactants in a particular step is unity
with respect to a standard state. This means that the obtained rate
values for the formation of C–O and C–C bonds represent the up-
per bound since the coverage of surface intermediates in steam
reforming of methane will be significantly lower than unity.
Based on the simple analysis in Fig. 6, it is clear that on
monometallic Ni the maximum rates of C–C bond formation are
higher than the rates of the C–O bond formation. This might be
an important reason for the deactivation of the Ni catalyst by solid
carbon deposits during hydrocarbon steam reforming at moderate
steam to carbon ratios, i.e., moderate ratio of oxygen to carbon ad-
sorbates on surface. It is also clear that on Sn/Ni the rate of C–C
bond formation is lower than the rate of the C–O bond formation,
which suggests that C atoms and CHx fragments are preferentially
oxidized on Sn/Ni. The reason for this is that the barrier for the
diffusion of C atoms and CHx fragments on Sn/Ni is significantly
larger than on pure Ni surface. We find that as the concentration of
Sn in the Ni surface layer is increased the activation barrier for the
diffusion of carbon species increases dramatically due to repulsion
between Sn and C atoms. For example, for an Sn surface concentra-
tion of 2/9 ML the impact on the rate of the C–C bond formation
is dramatic as corroborated by the results shown in Fig. 6. This
means that even a small number of Sn atoms dispersed in the top
layer of Ni would affect the rate of carbon-induced deactivation
significantly. This observation is in agreement with our experimen-
tal findings.
The kinetic studies allow also us to draw conclusions regarding
the underlying mechanisms responsible for the improved carbon-
tolerance of Sn/Ni compared to Ni. Our results show that Sn atoms
cause the rate-controlling CH4 activation step to move from low-
coordinated Ni sites to the well-coordinated sites. This is consistent
with our DFT calculations which showed that Sn preferentially dis-
places Ni atoms from the under-coordinated sites. The presence of
the Sn atoms increases the activation barriers for the dissociation
of CH4. Furthermore, it also weakens the binding of the carbon
atoms to the low-coordinated sites, which is a requirement for
the nucleation of carbon at these sites. In very simple terms, the
Sn atoms increase the chemical potential of carbon at the under-
coordinated sites and decrease the thermodynamic driving force
for the carbon atoms to approach these sites. This mechanism by
which Sn improves the tolerance of the catalyst to the carbon-
induced deactivation is very similar to the mechanism proposed
previously for Au and S additives [9].
We also find that under the explored conditions (above 950 K)
the formation of the CO adsorbate does not control the reaction
rate on either catalyst. Based on the experimental results, we can-
not rule out that the rate of formation of the C–C bonds (the
kinetic control of the formation of carbon deposits described in
the introduction section), which is a prerequisite for the formation
of carbon fibers, is not affected by the formation of the surface
alloy. To address this question we have calculated the maximum
rates for the dissociation of CH4, and the formation of the C–C
and C–O bonds on the Ni and Sn/Ni model systems. The activation
barriers for these processes were calculated in DFT calculations. In
these calculations we have assumed that the activation of CH4 on
monometallic Ni takes place on low-coordinated sites of the (211)
surface, while on Sn/Ni, the close-packed sites on the (111) sur-
face are involved in the process. These assumptions are supported
by the results of experimental analysis discussed in the preced-
The analysis suggests that in addition to lowering the thermo-
dynamic driving force associated with the nucleation of carbon
on low-coordinated Ni sites, the Sn/Ni catalysts also enhance the
rates of carbon removal by preferentially oxidizing carbon species
rather than forming C–C bonds. Furthermore, the analysis in Fig. 6
also suggests that at high temperatures, consistent with the experi-

E. Nikolla et al. / Journal of Catalysis 263 (2009) 220–227
227
Fig. 6. Calculated maximum rates of C–H bond activation in methane, C–C bond formation, and C–O bond formation are plotted as a function of temperature for Ni and Sn/Ni
surface alloy. It is assumed the C–H bond is activated on Ni(211) surface for Ni and on Sn/Ni(111) surface for Sn/Ni. The pre-exponential factor for this reaction is assumed
to be 105 1/s. The activation barrier for the formation of C–O and C–C bonds was calculated as the energy difference between the highest and lowest points on the potential
energy surfaces associated with the diffusion of C and O atoms and the formation of the C–O and C–C bonds respectively on the closed packed Ni (111) and Sn/Ni (111)
surfaces. The pre-exponential factors for these reactions are assumed to be 1013 1/s for both surfaces. The unit cell in the calculations was 3 × 3 for the (111) surface and
3
× 1 for the (211) surface. The concentration of Sn in the surface layer of Sn/Ni(111) was 2/9 ML.
ments performed herein, the rate-controlling step in the formation
of products on both surfaces is the activation of C–H bonds in
methane. We note that the rate controlling step is the step with a
lower rate between the activation of C–H bond and the formation
of C–O bond, i.e., the rate of C–C bonds cannot control the process
since it leads to a dead end in the reaction. Fig. 6 also shows that
at lower temperatures the rate-limiting step on both surfaces is the
formation of C–O bonds as has been postulated previously [24].
[8] J.R. Rostrupnielsen, L.J. Christiansen, Appl. Catal. A 126 (1995) 381.
[
9] H.S. Bengaard, J.K. Norskov, J. Sehested, B.S. Clausen, L.P. Nielsen, A.M. Molen-
broek, J.R. Rostrup-Nielsen, J. Catal. 209 (2002) 365.
[
10] E. Nikolla, A. Holewinski, J. Schwank, S. Linic, J. Am. Chem. Soc. 128 (2006)
11354.
[11] E. Nikolla, J. Schwank, S. Linic, J. Catal. 250 (2007) 85.
[12] E. Nikolla, J. Schwank, S. Linic, Catal. Today 136 (2008) 243.
13] D.L. Trimm, Catal. Today 49 (1999) 3.
[
[
[
[
14] D.L. Trimm, Catal. Today 37 (1997) 233.
15] C. Padeste, D.L. Trimm, R.N. Lamb, Catal. Lett. 17 (1993) 333.
16] S. Helveg, C. Lopez-Cartes, J. Sehested, P.L. Hansen, B.S. Clausen, J.R. Rostrup-
Nielsen, F. Abild-Pedersen, J.K. Norskov, Nature 427 (2004) 426.
4
. Conclusion
[
17] E. Achenbach, E. Riensche, J. Power Sources 52 (1994) 283.
[
[
18] J.G. Xu, G.F. Froment, AIChE J. 35 (1989) 88.
19] J.G. Xu, G.F. Froment, AIChE J. 35 (1989) 97.
A kinetic assessment of methane steam reforming on Sn/Ni/YSZ
and Ni/YSZ catalysts is obtained using a combination of kinetic
studies, isotopic-labeling experiments, and DFT calculations. We
found that the reaction rate on Sn/Ni/YSZ and Ni/YSZ is first-order
with respect to methane and it is independent of the water partial
pressure. Furthermore, we established that C–H bond activation
in methane is the rate-limiting step for both catalysts. We found
that while on monometallic Ni the active sites are the under-
coordinated Ni sites, on Sn/Ni the active sites are more abundant,
well-coordinated terrace sites. The observed decrease in carbon de-
activation of Sn/Ni was attributed to the Sn-induced lowering in
the binding energy of carbon on low-coordinate sites, serving as
carbon nucleation center and to an improved propensity of Sn/Ni
to oxidize carbon atoms and fragments.
[
20] J.M. Wei, E. Iglesia, J. Catal. 224 (2004) 370.
[21] J.W. Snoeck, G.F. Froment, M. Fowles, Ind. Eng. Chem. Res. 41 (2002) 4252.
[
[
[
22] K. Ahmed, K. Foger, Catal. Today 63 (2000) 479.
23] P. Munster, H.J. Grabke, J. Catal. 72 (1981) 279.
24] G. Jones, J.G. Jokobsen, S.S. Shim, M.P. Andersson, J. Rossmeisi, F. Abild-
Pedersen, T. Bligaard, S. Helveg, B. Hinnemann, J.R. Rostrup-Nielsen, I. Chork-
endorff, J. Sehested, J.K. Norskov, J. Catal. 259 (2008) 147.
[25] E. Nikolla, J. Schwank, S. Linic, J. Am. Chem. Soc. (2009), Article ASAP.
[26] J.P. Perdew, J.A. Chevary, S.H. Vosko, K.A. Jackson, M.R. Pederson, D.J. Singh, C.
Fiolhais, Phys. Rev. B 46 (1992) 6671.
27] J.P. Perdew, Y. Wang, Phys. Rev. B 46 (1992) 12947.
[
[28] J.P. Perdew, K. Burke, M. Ernzerhof, Chemical Applications of Density-Functional
Theory 629 (1996) 453.
[29] D. Vanderbilt, Phys. Rev. B 41 (1990) 7892.
[
[
30] S. Linic, J.W. Medlin, M.A. Barteau, Langmuir 18 (2002) 5197.
31] S. Laursen, S. Linic, Phys. Rev. Lett. (2006) 97.
[32] M. Enever, S. Linic, K. Uffalussy, J.M. Vohs, M.A. Barteau, J. Phys. Chem. B 109
(2005) 2227.
Acknowledgments
[33] G. Henkelman, B.P. Uberuaga, H. Jonsson, J. Chem. Phys. 113 (2000) 9901.
[34] G. Mills, H. Jonsson, G.K. Schenter, Surf. Sci. 324 (1995) 305.
[35] S. Linic, M.A. Barteau, J. Am. Chem. Soc. 126 (2004) 8086.
[36] S. Linic, M.A. Barteau, J. Catal. 214 (2003) 200.
We gratefully acknowledge the support of the US Depart-
ment of Energy DOE-BES, Division of Chemical Sciences (FG-02-
0
5ER15686), NSF (CBET 0756255), ONR (N00014-08-1-0122), DOE-
[37] H.S. Fogler, Elements of Chemical Reaction Engineering, 4th ed., Prentice–Hall,
2005.
NETL (FC26-05-NT-42516) and National Automotive Center (NAC).
[
[
[
38] R.L. Augustine, Heterogeneous Catalysis for Synthetic Chemist, CRC Press, 1996.
39] R.M. Koros, E.J. Nowak, Chem. Eng. Sci. 22 (1967) 470.
40] R.J. Madon, M. Boudart, Ind. Eng. Chem. Fundam. 21 (1982) 438.
References
[
41] S.B. Wang, G.Q. Lu, Energy Fuels 12 (1998) 1235.
[
1] C.M. Finnerty, N.J. Coe, R.H. Cunningham, R.M. Ormerod, Catal. Today 46 (1998)
37.
[42] R.I. Masel, Principles of Adsorption and Reaction on Solid Surfaces, John Wiley
& Sons, Inc., Canada, 1996.
[43] F. Abild-Pedersen, O. Lytken, J. Engbaek, G. Nielsen, I. Chorkendorff, J.K.
Norskov, Surf. Sci. 590 (2005) 127.
1
[2] J.R. Rostrup-Nielsen, in: Catalysis—Science and Technology, vol. 5, Springer,
Berlin, 1984.
[3] J.R. Rostrup-Nielsen, J. Sehested, J.K. Norskov, Adv. Catal. 47 (2002) 65.
[4] J. Rostrup-Nielsen, J.K. Norskov, Top. Catal. 40 (2006) 45.
[5] F. Abild-Pedersen, J.K. Norskov, J.R. Rostrup-Nielsen, J. Sehested, S. Helveg, Phys.
Rev. B (2006) 73.
[44] H. Burghgraef, A.P.J. Jansen, R.A. Vansanten, Surf. Sci. 324 (1995) 345.
[45] H. Burghgraef, A.P.J. Jansen, R.A. Vansanten, J. Chem. Phys. 101 (1994) 11012.
[46] H. Yang, J.L. Whitten, J. Chem. Phys. 96 (1992) 5529.
[47] H. Yang, J.L. Whitten, Surf. Sci. 255 (1991) 193.
[
[
6] N.C. Triantafyllopoulos, S.G. Neophytides, J. Catal. 217 (2003) 324.
7] T. Takeguchi, Y. Kani, T. Yano, R. Kikuchi, K. Eguchi, K. Tsujimoto, Y. Uchida, A.
Ueno, K. Omoshiki, M. Aizawa, J. Power Sources 112 (2002) 588.
[48] P.E.M. Siegbahn, I. Panas, Surf. Sci. 240 (1990) 37.
[49] P. Kratzer, B. Hammer, J.K. Norskov, J. Chem. Phys. 105 (1996) 5595.
[50] K. Otsuka, S. Kobayashi, S. Takenaka, J. Catal. 200 (2001) 4.