From Active-Site Models to Real Catalysts: Importance of the Material Gap in the Design of Pd Catalysts for Methane Oxidation

Article abstract of DOI:10.1002/cctc.201601333

Rapid computational screening to aid novel catalyst design has evolved into an important and ubiquitous tool in modern heterogeneous catalysis. A possible shortcoming of this approach, however, is the material gap, that is, simplified computational models used for catalyst screening do not always capture the complexity of real catalytic systems. Here we investigate the importance of the material gap for complete methane oxidation over supported Pd/γ-Al₂O₃ catalysts using a combination of DFT simulations and temperature-programmed oxidation experiments. The Pd/γ-Al₂O₃ active site was approximated by four models of increasing complexity, namely Pd(1 0 0), Pd(2 1 1), PdO(1 0 1), and Pd₁₀/γ-Al₂O₃(1 1 0), and each was also modified with metal promoters to discover reactivity trends. Although the unpromoted Pd model surfaces exhibit different methane activation activities, our DFT results indicate that an experimentally verified performance trend can be predicted for their promoted counterparts irrespective of the active-site representation. We attribute the robustness of the trend predictions in this particular system to localized changes in the electron density during methane activation. Overall, our work supports the commonly practiced active-site model simplifications during computational catalyst screening and provides fundamental insight into the qualitative agreement between theory and experiment for methane oxidation over promoted Pd catalysts.

Full text of DOI:10.1002/cctc.201601333

Accepted Article
Title: From Active Site Models to Real Catalysts: Importance of the
Material Gap in the Design of Pd Catalysts for Methane Oxidation
Authors: Hieu A Doan, Munish K Sharma, William S Epling, and Lars C
Grabow
This manuscript has been accepted after peer review and appears as an
Accepted Article online prior to editing, proofing, and formal publication
of the final Version of Record (VoR). This work is currently citable by
using the Digital Object Identifier (DOI) given below. The VoR will be
published online in Early View as soon as possible and may be different
to this Accepted Article as a result of editing. Readers should obtain
the VoR from the journal website shown below when it is published
to ensure accuracy of information. The authors are responsible for the
content of this Accepted Article.
To be cited as: ChemCatChem 10.1002/cctc.201601333
Link to VoR: http://dx.doi.org/10.1002/cctc.201601333
A Journal of
www.chemcatchem.org

10.1002/cctc.201601333
ChemCatChem
FULL PAPER
From Active Site Models to Real Catalysts: Importance of the
Material Gap in the Design of Pd Catalysts for Methane Oxidation
Hieu A. Doan[a], Munish K. Sharma[a], William S. Epling[a,b] and Lars C. Grabow*[a]
Rapid computational screening to aid novel catalyst design has
evolved into an important and ubiquitous tool in modern
reactivity trends. Although the unpromoted Pd model surfaces
exhibit different methane activation activities, our DFT results
heterogeneous catalysis. A possible shortcoming of this approach, indicate that an experimentally verified performance trend can be
however, is the material gap, where simplified computational
models used for catalyst screening do not always capture the
complexity of real catalytic systems. Here we investigate the
importance of the material gap for complete methane oxidation
over supported Pd/γ-Al2O3 catalysts using a combination of
Density Functional Theory (DFT) simulations and Temperature
Programmed Oxidation (TPO) experiments. The Pd/γ-Al2O3 active
site was approximated by four models of increasing complexity,
namely Pd(100), Pd(211), PdO(101), and Pd10/γ-Al2O3(110), and
each was also modified with metal promoters in order to discover
predicted for their promoted counterparts irrespective of the active
site representation. We attribute the robustness of the trend
predictions in this particular system to localized changes in the
electron density during methane activation. Overall, our work
supports the commonly practiced active site model simplifications
during computational catalyst screening and provides
fundamental insight into the qualitative agreement between
theory and experiment for methane oxidation over promoted Pd
catalysts.
Introduction
Recent advances in density functional theory (DFT) and
computing power have made computational catalyst screening a
viable alternative to traditional trial-and-error experimental
approaches.[1–5] Computational screening typically reduces the
number of parameters that describe catalytic behaviors in terms
of activity and selectivity to a few energy descriptors that can be
easily calculated from DFT. This reduction in complexity to the
descriptor-based method is achieved through scaling
relationships, which relate the energy of intermediates and
transition states to the adsorption energies of a small set of
intermediates that serve as descriptors.[6–10]
Although heterogeneous catalysts have complex or even
multiple active sites, computational catalyst screening
approaches routinely assume that simplified active site
representations, such as fcc(100), fcc(111) or fcc(211) facets, are
sufficiently accurate to obtain reliable catalytic trends at
maximized computational efficiency. Although such simplification
definitely affects the absolute values and quantitative outcomes,
trend predictions are assumed to be independent of the material
gap between oversimplified computational model systems and
real catalyst surfaces. Yet, other than anecdotal evidence, there
has not been a systematic study questioning the validity of this
assumption. Given the success and increasing popularity of
computational catalyst screening,[11,12] the important question
regarding the role of the material gap remains: is it always
necessary to close the material gap by using increasingly complex
catalyst models to predict qualitative reactivity, and more
importantly, selectivity trends, or is the current practice and use of
simplified models sufficient in certain cases?
In this study, we attempt to answer the above question by
investigating the importance of the material gap in catalytic trend
studies for complete methane oxidation over Pd/γ-Al2O3 catalysts
using a combination of DFT calculations and temperature
programmed oxidation (TPO) experiments. This catalytic system
is a good benchmark because the reaction is structure-
sensitive,[13,14] the activity depends on the oxidation state of Pd in
response to the reaction environment,[15–18] and the choice of
support affects catalytic performance.[19] Furthermore, methane
combustion catalysis is of practical importance as unburnt
methane slip from natural gas vehicles contributes to detrimental
greenhouse gas emissions. To examine the effects of the material
gap, we approximated the Pd/γ-Al2O3 active site with four popular
computational models of increasing complexity: Pd(100), Pd(211),
PdO(101), and Pd10/γ-Al2O3(110). Among these, Pd(100) and
Pd(211) distinguish between terrace and step sites, respectively,
of a metallic Pd catalyst nanoparticle. Pd(100) was chosen over
Pd(111) to represent the terrace site, because methane activation
was shown to be noticeably more favorable on the former [20] and
the active oxide phase of Pd, PdO, was suggested to
preferentially form on the less packed Pd(100) than the dense
hexagonal Pd(111).[21] The PdO(101) thin film, on the other hand,
has been shown to be the most stable termination that grows
epitaxially on Pd(100)[22] or Pd(111)[23] under oxidizing conditions.
Since the structure and activity of PdO(101) thin film do not
deviate much from those of bulk-terminated PdO(101),[20,23] the
latter model was employed in this study. The more complex
[a]
Hieu A. Doan, Munish K. Sharma, William S. Epling and Lars C.
Grabow
Department of Chemical and Biomolecular Engineering
University of Houston
Houston, TX 77204-4004, USA
E-mail: grabow@uh.edu
[b]
William S. Epling
Department of Chemical Engineering
University of Virginia
Charlottesville, VA 22904-4741, USA
Supporting information for this article is given via a link at the end of
the document.
This article is protected by copyright. All rights reserved.

10.1002/cctc.201601333
ChemCatChem
FULL PAPER
Pd10/γ-Al2O3(110) model, which consists of a 10-atom Pd
nanocluster supported on γ-Al2O3, was used to represent the
metal/support interface. To define a tractable problem scope we
limit our study to reaction conditions, under which the
unimolecular CH4 activation reaction is rate-limiting,[24] and we
consider only γ-Al2O3 as support, which is not expected to actively
participate, but only indirectly alter the properties of the supported
metal nanoparticle. Furthermore, in order to predict reactivity
trends for each active site representation, we modified each active
site model with cobalt (Co), copper (Cu), nickel (Ni), and platinum
(Pt) metal promoters under consideration of their thermodynamic
most favorable substitution site.
sites on metallic systems (Figure 1b-d), whereas lattice oxygen
(Olat) accepts the H atom in PdO(101) (Figure 1a). For our most
complex catalyst model, Pd10/γ-Al2O3 that represents the
metal/support interface, the d-band center calculations show that
the most active Pd atoms, i.e., those with the highest d-band
center, are not the interfacial ones, but the metallic Pd atoms
away from the interface (Figure S1). Indeed, subsequent NEB
calculations confirmed that the methane activation energy barrier
at Pd sites away from the interface is about 1.0 eV smaller than
the energy barrier required for the same process at the interface
(1.64 eV vs. 0.55 eV). These results are consistent with the
previous study by Baldwin and Burch, which concluded that
alumina supports only affect methane oxidation activity indirectly
by altering the morphology of Pd nanoparticles.[26]
The promoted Pd and PdO surfaces were created by
replacing a surface Pd atom with a metal promoter within one unit
cell (see Methods and Table S1 for more details). Such a
replacement scheme translates to a ca. 3:1 ratio of Pd to promoter
in the uppermost layer of the surface. As shown in Figure 1,
methane preferentially activates on top of a Pd atom on the un-
promoted surfaces. Thus, we considered two distinct pathways
for methane activation on promoted systems and differentiated
between CH3-H bond cleavage over either a Pd or the promoter
atom. The first pathway results in the methyl group bound to a Pd
atom, whereas the methyl group bonds to the promoter in the
second pathway. Although both pathways were examined for
methane activation on all promoted systems, only Ea values
associated with the more favorable pathway are reported here. A
summary of these results is shown in Figure 2, showing the
methane activation energy barrier (ordinate) calculated for four Pd
surface models promoted with Co (CoprPd), Cu (CuprPd), Ni
(NiprPd) and Pt (PtprPd) (abscissa). The data points for Pd (red)
correspond to the Ea values of the un-promoted systems reported
earlier in Figure 1 and serve as references. With the Cu promoted
systems being an exception, the most interesting feature in Figure
2 is the consistent trend in promoted activity across all surface
models. For example, Pt (blue) and Ni (green) promoted systems
always show higher activities (lower activation energy barriers)
than those of the pure surfaces. In contrast, the presence of Co
(purple) leads to the highest activation energy barrier, or lowest
activity, regardless of the surface model. Besides the notable
agreement in the qualitative trend across all models, quantitative
discrepancies are also apparent. The larger range of calculated
activation energy barriers of the low-coordinated Pd surfaces, i.e.
Results and Discussion
Methane activation was first investigated on un-promoted Pd and
PdO surface models, which are considered reference systems.
Figure 1 shows transition state geometries as well as activation
energy barriers (Ea) associated with methane activation on pure
PdO(101), Pd(100), Pd(211) and Pd10/γ-Al2O3. The Ea values
obtained for Pd(100), Pd(211) and PdO(101) are in good
agreement with previously reported data.[20,25] As shown in Figure
1, the calculated activation barriers depend on the catalyst models
and the material gap certainly influences their absolute
magnitudes. Among our investigated systems, the terrace model,
Pd(100), activates methane with the highest energy barrier of 0.78
eV, whereas methane activation on the supported Pd10/γ-Al2O3
requires an energy barrier of only 0.55 eV. For Pd(100), Pd(211),
and PdO(101) surfaces, methane was found to adsorb weakly
(Eads ~ -0.1 eV); hence, the expected activation mechanism is
dissociative adsorption of methane gas (CH4 (g) + 2* → CH3* +
H*, where * represents an unoccupied Pd site). On the other hand,
methane adsorption is stronger on Pd10/γ-Al2O3 (Eads = -0.21 eV),
which may indicate adsorption and pre-activation of methane
molecules prior to H-CH3 bond cleavage (CH4(g) + 2* → CH4* + *
→ CH3* + H*). Indeed, methane adsorbed on Pd10/γ-Al2O3 is
distorted with a noticeably larger H-C-H bond angle (∠114°) for H
atoms close to the surface, compared to that on Pd(100) (∠111°)
and any H-C-H angle of the methane gas molecule (∠109.5°). In
the final state after CH4 dissociation the resulting methyl (CH3)
group preferentially binds on top of a (low-coordinated) Pd atom
for all surface models, but the final position of the H atom differs.
Specifically, bridge or hollow sites serve as hydrogen abstraction
Figure 2. Calculated energy barriers for methane activation (in eV) on pure
and promoted surface models of Pd and PdO catalysts. Squares and
triangles indicate activation pathways over Pd or the promoter atom,
respectively.
Figure 1. Transition state geometries and calculated energy barriers for
methane activation on (a) PdO and (b-d) Pd catalyst models. Atomic color
codes: blue (Pd), red (O), grey (Al), black (C) and white (H).
This article is protected by copyright. All rights reserved.

10.1002/cctc.201601333
ChemCatChem
FULL PAPER
Pd(211) (coordination number, cn = 7) and Pd10/γ-Al2O3 (cn = 5),
compared to the terrace Pd(100) (cn = 8) suggests a stronger
promotional or inhibitive effect. By comparing the slopes of Figure
2b-d, we may conclude that the lower the coordination of the Pd
surface is, the larger the extent of promotion will be. Again, the
sensitivity of the absolute values for methane activation on
modified Pd and PdO surfaces speaks to the existence of
structure sensitivity and the material gap. In particular, these
results indicate that the effect of promoters is not only electronic,
but also geometric in nature. Nevertheless, all models are equally
reliable when differentiating between activity promoters and
inhibitors.
of Cu atoms may be responsible for this phenomenon. Indeed,
our DFT results indicate that Cu substitution is ca. 0.1 eV more
stable on the lower coordinated Pd(211) step site than on the
Pd(100) terrace. Therefore, it is plausible that Cu atoms migrate
from high to low coordinated surface sites, or even between the
bulk and surface of nanoparticles[28] after every cycle, and
eventually reduce the catalyst activity. The data in Figure 2 shows
that the effect of Cu depends on the structural model of the
catalyst site and is consistent with this interpretation. Other
phenomena such as sintering or inhibition by H2O may also
contribute, but a detailed investigation of the deactivation
mechanism is outside the scope of this study.
Figure 3. Temperature programmed oxidation of methane on pure and promoted Pd/γ-Al2O3 catalysts. a) Effects of promoters on methane combustion light-off
curves and T50 values (where the black dashed line intersects) during the first cycle. b) T50 values of methane combustion over three cycles (0.2% CH4, 6.0% O2,
N2 balance, heating/cooling rate 5°C min-1). T50 value indicates the temperature at which methane conversion reaches 50%.
Temperature programmed oxidation (TPO) tests were
conducted to compare CH4 oxidation activities over a series of
pure and promoted Pd/γ-Al2O3 catalyst samples. Three heating-
cooling cycles and T50 (οC) values, i.e. the temperature when
methane conversion reaches 50%, were measured. A 0.2 vol%
CH4, 6 vol% O2, and balance N2 feed composition was used, for
which the activation of methane is known to be rate-limiting.[13]
Additional information is provided in the Methods section. Figure
3a shows light-off curves and lists the T50 for each catalyst for the
first cycle. The T50 values show a consistent trend with activation
barrier estimates using TPO data below 20 % conversion, which
are given in Figure S2. A complete summary of the T50 (οC) values,
which are indicative of catalytic activity, is given in Figure 3b for
all three cycles. The relative order of the overall promotional effect
extracted from Figure 3 is Pt > Ni > Cu > Co and is identical to our
theoretical predictions. Certainly, Pt stands out as the best
methane oxidation promoter, and the consistent activity of the Pt-
promoted catalyst for all cycles suggests that Pt also reduces
catalyst deactivation. Ni also enhanced the activity over all three
cycles, and the addition of Co consistently leads to lower methane
oxidation activity. For Cu-substituted systems, the experimental
results do not follow such a consistent trend, since Cu acts as a
promoter in the first two cycles (decreased T50), and as a weak
inhibitor in the third cycle (increased T50). In the past, a Cu-Pd
alloy has been shown unfavorable for methane combustion over
Pd-Cu/SiO2 catalysts, where increasing the Cu loading led to a
decrease in activity.[27] Since the Cu content in our catalyst
remains constant over three cycles, we speculate that the mobility
Results obtained from our computational and experimental
study exhibit a consistent trend for metal promoters. The trend is
largely independent of the computational model representation of
the active site, but the extent of promotion is affected by the model
choice. Notably, theory and experiment both show that Pt-
promoted Pd catalysts are superior for methane oxidation.
Table 1. Calculated dipole moments of CH4 in the transition states of
methane activation on Pd(100), Pd(211), PdO(101), and Pd/γ-Al2O3
surfaces.
ꢀꢁꢂ(ꢃ × Å)
Promoter
PdO(101)
Pd(100)
Pd(211)
Pd10/γ-Al2O3
Pt
Ni
0.25
0.17
0.23
0.12
0.25
0.13
0.36
0.12
Cu
Pd
Co
0.19
0.21
0.18
0.13
0.14
0.14
0.13
0.13
0.14
0.15
0.12
0.17
In order to explain why the active site geometry influences
the magnitude, but not the direction of promotion/inhibition, we
examine the C-H bond breaking transition state more closely. For
methane activation over similar metallic systems such as Ni(100)
and Ni(111), Bengaard and co-workers showed that the activation
energy barriers are largely influenced by the electrostatic
interactions between CH4 and these surfaces.[29] Hence, dipole
moments (in the direction perpendicular to the surface) of
This article is protected by copyright. All rights reserved.

10.1002/cctc.201601333
ChemCatChem
FULL PAPER
methane in the transition state, µTS, on the systems of interest
were calculated with DFT and summarized in Table 1.
As shown in Table 1, pure metallic Pd(100), Pd(211), and
Pd/γ-Al2O3 surfaces seem to polarize the transition state of CH4
equally with µTS values of 0.14, 0.13, and 0.12 (e×Å), respectively.
For PdO(101), however, the induced dipole moment on the
methane activation transition state is significantly larger.
Considering the fact that the calculated activation barrier on
PdO(101) lies between the barrier for Pd(211) and Pd10/γ-Al2O3,
we can conclude that the dipole moments in Table 1 cannot be
directly related to the corresponding activation barriers. Moreover,
Ni, Cu and Co-promoted Pd systems produce similar dipole
moments of methane in the transition state compared their un-
promoted counter parts, yet differences in activation barriers exist.
When Pt is used as the promoter, however, the magnitudes of µTS
for promoted and pure systems show considerable differences.
Interestingly, when the change in activation energy barrier is
plotted against the change in dipole moment for Pt-promoted
surfaces (Figure 4), two observations can be made. First, the
promoting effect of Pt differs across all surface models, with the
supported Pd10/γ-Al2O3 system having the most significant
improvement and largest dipole change. Furthermore, since the
ratio of Pt to Pd is similar on all promoted surfaces, the variance
in activity gain must primarily come from geometric effects.
Second, the straight line with a slope of ca. -1.32 and x-intercept
Figure 5. Charge density difference for the transition states of methane
activation (top view) on pure and promoted Pd(100) surfaces. The promoter
atoms on PtprPd(100) and NiprPd(100) surfaces are directly underneath the
methyl group, while Co in CoprPd(100) accepts the H atom. The Cu atom in
CuprPd(100) is further away and not visible in the inset (See Table S1 for more
details). Red/blue indicates charge depletion/accumulation. Atomic color code:
C = black, H = white, Pd = blue, and Co = magenta.
The H-CH3 bond cleavage transition state on pure Pd(100)
is characterized by charge depletion between CH3 and H (red
area) and charge accumulation between H and the surface (blue
area). While these features are also present on the modified
Pd(100) systems, their distribution differs from one surface to
another. According to Figure 5, the charge density difference
isosurface for PtprPd(100) is confined around the Pt atom at which
CH4 binds to the surface. On the other hand, the same isosurface
is more delocalized on the Co-promoted Pd(100) system.
of -0.08 eV indicates
a linear relationship between the
enhancement in activity (measured by the change in activation
energy barrier) and the dipole moment induced by Pt atoms. A
simple rule of thumb that can be deduced from this relationship is
that the activation energy barrier is lowered by 1.32 eV for every
e×Å gained in the dipole moment of methane in the transition state
on PtprPd surfaces.
By visual inspection, one can rank Pd(100) surfaces in
terms of the spatial extent of charge density difference distribution
in the following ascending order: PtprPd(100) < NiprPd(100) <
CuprPd(100) < CoprPd(100). Since this is the same order we
obtained for DFT-derived activation energy barriers (Figure 2) or
experimental T50 values (Figure 3), we propose that the
localization of charge density changes in the transition state,
rather than the dipole moment projected onto the surface normal,
is responsible for the improvement in catalytic activity. Assuming
that charge is redistributed to cause a similar potential difference
∆ꢄ over a shorter distance d, then the C-H bond is subject to a
stronger local electric field E, which can be approximated as ꢅ =
⁄
− ∆ꢄ ꢆ .This stronger local electric field aids in the C-H scission
Figure 4. Gain in activity (measured as decrease in energy barrier) as a function
of change in transition state dipole moment of methane on PtprPd surfaces.
step and suggests that electrochemical routes with controllable
field effects combined with thermal activation are promising
approaches for selective CH4 conversion.[30] This argument is in
agreement with calculations shown in Figure S3 for Pd(100),
CoprPd(100), and PtprPd(100) in the presence of an externally
applied field. The PtprPd(100) system with the strongest local field
exhibits the highest sensitivity to changes in an externally applied
field. Furthermore, as the promoted transition states are more
confined, they become less sensitive to the geometric
arrangement and property of the surrounding atoms. Hence, we
speculate that the localization of the transition state charge
density changes contributes to the robustness of the activity trend
predictions across promoted model systems of varied complexity.
Although electrostatic interactions, as quantified by surface-
induced dipole moments in the transition state of methane
activation, can provide some possible explanation for the
promoting effects of Pt on Pd catalysts, they do not capture the
consistent promotional or inhibitive effects of the other metal
modifiers, namely Ni, Cu, and Co. Since the transition state dipole
moments discussed so far result from a projection of charge
density distribution changes onto the surface normal, we further
investigated the full charge density difference to gain more
insights into the transition state of methane activation. In particular,
the full charge density difference map for methane activation on
modified Pd(100) surfaces is shown in Figure 5.
This article is protected by copyright. All rights reserved.

10.1002/cctc.201601333
ChemCatChem
FULL PAPER
Conclusions
Al2O3 system, both interfacial and low-coordinated Pd atoms away from
the interface are important, and the most active atom of each type,
according to d-band center calculations, was promoted.
In conclusion, the activity of a series of promoted Pd catalysts for
methane activation has been investigated with DFT calculations
and TPO experiments. Four active site representations of varied
complexity were evaluated. DFT calculations predict a consistent
trend for metal modification and identified Pt as the best promoter
regardless of the surface model. TPO experiments performed on
promoted Pd/γ-Al2O3 catalysts yield the same activity trend as
predicted by DFT. The strong promotional effect of Pt was
attributed to a strong electrostatic interaction between methane in
the transition state and the Pt promoted surface. In general, we
propose that the localization of charge density changes is
ultimately responsible for promoting or inhibiting CH4 activation on
Pd-based catalysts. Furthermore, the confined charge density
distribution in the transition state of methane activation
contributes to the robustness of the predicted trend of promoter
efficacy. Overall, this work supports the common use of simplified
active site models for computational catalyst screening of reaction
systems with negligible metal/support interaction, and a well-
defined, unimolecular rate-determining step with a localized
transition state.
Gas phase methane molecules were calculated in a cubic box of 103
Å3, and the Brillouin zone was sampled at the Gamma point.[40] Adsorption
energies for CH4 (Eads) are provided with reference to the corresponding
relaxed surfaces and the gas-phase species CH4. Activation energy
barriers (Ea) were calculated using the climbing image nudged elastic band
(CI-NEB) method[44] and all transition states were confirmed via vibrational
analysis, showing a single imaginary mode along the reaction pathway.
Catalyst preparation
Pure Pd (1 wt%) catalysts and Pd catalysts doped with a second metal
(0.4 wt%) were prepared using the incipient wetness impregnation method.
The catalysts were prepared using impregnation of metal salt precursors
on γ-Al2O3 (high purity activated alumina Al2O3, Puralox TH 100/150 Sasol
Germany GmbH). The γ-Al2O3 has a packed bulk density of 700-950 (g/l),
particle size D50 = 40 µm, pore volume = 0.8-1 (ml/g), pore radius=11 nm,
surface area = 20 m2/g. The metal precursors used were water soluble
nitrate salts such as palladium (II) nitrate dihydrate Pd(NO3)2·2H2O
(Aldrich, 40% metal basis), tetraammineplatinum (II) nitrate
Pt(NH3)4(NO3)2 (Aldrich, 99.99 % trace metal basis), nickel (II) nitrate
hexahydrate Ni(NO3)2·6H2O (Aldrich, 99.99 % trace metal basis), copper
(II) nitrate hydrate Cu(NO3)2·xH2O (Aldrich, 99.99 % trace metal basis),
cobalt (II) nitrate hexahydrate Co(NO3)2·6H2O (Aldrich, 99.99 % trace
metal basis). All promoted catalysts were prepared using co-impregnation,
the Pd and dopant metal atoms were both present in solution. Excess
water was removed from the samples though overnight heating at 100οC.
The final calcination step was a 550 οC exposure in air, held for 5 hrs.
Calcined catalysts were sieved to 40/60 mesh size. Silica beads with
similar size were mixed with the catalyst particles to form the catalyst bed.
Glass wool was used to hold the particles in place. A K-type thermocouple
(Omega Engineering, Inc.) was placed at the middle of the bed.
Methods
Density functional theory
Periodic density functional theory (DFT) calculations were performed using
the Vienna ab-initio simulation package (VASP)[31–34] and the python
modeling interface provided by the atomic simulation environment
(ASE).[35] The core and valence electrons were represented by the
projected augmented wave (PAW) method[36,37] with a kinetic energy cut-
off of 400 eV. Exchange and correlation were described by the Perdew-
Wang (GGA-PW91) functional.[38] Gaussian smearing with Fermi
temperatures of kbT = 0.1 eV was employed and the total energies were
subsequently extrapolated to kbT = 0.0 eV.[39] For geometry optimizations,
forces were converged below 0.05 eV/Å.
Catalysts testing
Catalytic tests were carried out in a quartz tube reactor at atmospheric
pressure. A catalyst weight of 29.3 mg was used with a total gas flow rate
of 200 mL/min. The gas hourly space velocity (GHSV) for all the
experiments was fixed at 50,000 hr-1 (this is based on monolith loading of
2 g/inch3). The feed gas composition was 0.2 vol% CH4, 6 vol% O2, and
balance N2. The TPO experiments consisted of three cycles, where the
temperature was ramped up and down between 25οC to 500οC at a
heating/cooling rate of 5 οC /min using a Thermo Scientific, Lindberg Blue
Mini-Mite tube furnace. A pretreatment in 6 vol% O2 and balance N2 (total
gas flow rate of 200 mL/min) at 500 οC for 1 hr (heating/cooling rate of 5
οC /min) was done before recording the first light-off curve. The CH4 and
CO2 compositions were measured at the outlet of reactor using a Pfeiffer
OMNIstar mass spectrometer.
The Pd(100) and PdO(101) surfaces were modeled as slabs with a
(2×2) unit cell, whereas slabs with a (2×1) unit cell were employed for the
stepped Pd(211) surface. Each slab consists of an equivalent of four layers
in which the top two were allowed to relax and the bottom two were fixed
to their bulk positions. The Brillouin zone for these systems was sampled
using a 6 × 6 × 1 Monkhorst-Pack k-point set.[40] The calculated lattice
constants are a = 3.958 Å for Pd and a = 3.100 Å and c/a = 1.755 for PdO.
The Pd/γ-Al2O3 interface was represented by a 10-atom Pd cluster
connected to the γ-Al2O3(110) support via its largest (111) facet. The γ-
Al2O3(110) surface was cleaved from Digne’s γ-Al2O3 model[41] to make a
three-layer slab with a (2x2) unit cell, where the top layer was relaxed and
the bottom two layers were fixed in their bulk position. The calculated
lattice parameters for bulk γ-Al2O3 are a = 5.577 Å, b/a = 1.506, c/a = 1.447,
α = γ = 90° and β = 90.54°, and in good agreement with the previous works
of Digne[41] and Wang.[42] All slabs were separated with a vacuum of 14 Å
along the normal direction to the surface. A dipole correction to the
electrostatic potential[43] was included to separate adjacent unit cell images.
Gaussian smearing with Fermi temperature of 0.01 eV was employed and
the total energies were subsequently extrapolated to kbT = 0.0 eV. [39]
The promoted systems were created by replacing surface Pd atoms,
one at a time, with a promoter and allowed to relax to their optimized
geometries. On the Pd(100) surface one of the identical terrace atoms was
replaced, while for Pd(211) an undercoordinated step atom was
substituted. For PdO(101), there are two types of Pd atoms: one
coordinated with four oxygen atoms and the other coordinated with only
three oxygen atoms; both were considered for substitution. For the Pd10/γ-
Acknowledgements
The authors acknowledge Shell Oil Company for financial support.
LCG also acknowledges the NSF CAREER award number
1454384. Computational resources were provided by the uHPC
cluster managed by the University of Houston and acquired
through NSF award number 1531814. This work used the
Extreme Science and Engineering Discovery Environment
(XSEDE), which is supported by National Science Foundation
grant number ACI-1053575, and HPC resources provided by the
Texas Advanced Computing Center (TACC) at The University of
Texas at Austin. We also acknowledge the use of computational
This article is protected by copyright. All rights reserved.

10.1002/cctc.201601333
ChemCatChem
FULL PAPER
resources provided by the National Energy Research Scientific
Computing (NERSC) Center, a DOE Office of Science User
Facility supported by the Office of Science of the U.S. Department
of Energy under contract number DE-AC02-05CH11231. Use of
the computational resources at the Center for Nanoscale
Materials, an Office of Science user facility, was supported by the
U.S. DOE, Office of Science, Office of Basic Energy Sciences,
under contract number DE-AC02- 06CH11357. Finally, we thank
the Center of Advanced Computing and Data Systems (CACDS)
at the University of Houston for access to the Maxwell/Opuntia
HPC clusters.
[20]
A. Hellman, A. Resta, N. M. Martin, J. Gustafson, A. Trinchero, P. A.
Carlsson, O. Balmes, R. Felici, R. Van Rijn, J. W. M. Frenken, et al.,
J. Phys. Chem. Lett. 2012, 3, 678–682.
[21]
[22]
E. Garbowski, C. Feumi-Jantou, N. Mouaddib, M. Primet, Appl.
Catal. A, Gen. 1994, 109, 277–292.
N. Seriani, J. Harl, F. Mittendorfer, G. Kresse, J. Chem. Phys. 2009,
131, 54701.
[23]
[24]
H. H. Kan, J. F. Weaver, Surf. Sci. 2008, 602, L53–L57.
A. Trinchero, A. Hellman, H. Grönbeck, Surf. Sci. 2013, 616, 206–
213.
[25]
Y.-H. Chin, C. Buda, M. Neurock, E. Iglesia, J. Am. Chem. Soc.
2013, 135, 15425–15442.
Keywords: Material gap • Methane oxidation • Palladium
catalysis • Density Functional Theory • Temperature
programmed oxidation
[26]
[27]
T. R. Baldwin, R. Burch, Appl. Catal. 1990, 66, 337–358.
P. Reyes, A. Figueroa, G. Pecchi, J. L. G. Fierro, Catal. Today
2000, 62, 209–217.
[28]
K. A. Goulas, S. Sreekumar, Y. Song, P. Kharidehal, G. Gunbas, P.
J. Dietrich, G. R. Johnson, Y. C. Wang, A. M. Grippo, L. C. Grabow,
et al., J. Am. Chem. Soc. 2016, 138, 6805–6812.
H. S. Bengaard, I. Alstrup, I. Chorkendorff, S. Ullmann, J. R.
Rostrup-Nielsen, J. K. Nørskov, J. Catal. 1999, 187, 238–244.
F. Che, S. Ha, J. S. McEwen, Appl. Catal. B Environ. 2016, 195,
77–89.
[1]
M. Andersson, T. Bligaard, A. Kustov, K. Larsen, J. Greeley, T.
Johannessen, C. Christensen, J. K. Norskov, J. Catal. 2006, 239,
501–506.
[29]
[30]
[2]
[3]
[4]
[5]
[6]
J. K. Nørskov, T. Bligaard, J. Rossmeisl, C. H. Christensen, Nat.
Chem. 2009, 1, 37–46.
F. R. Negreiros, E. Aprà, G. Barcaro, L. Sementa, S. Vajda, A.
Fortunelli, Nanoscale 2012, 4, 1208–19.
[31]
[32]
[33]
[34]
[35]
[36]
[37]
[38]
G. Kresse, J. Hafner, Phys. Rev. B 1993, 47, 558–561.
G. Kresse, J. Hafner, Phys. Rev. B 1994, 49, 14251–14269.
G. Kresse, J. Furthmüller, Phys. Rev. B 1996, 54, 11169–11186.
G. Kresse, J. Furthmüller, Comput. Mater. Sci. 1996, 6, 15–50.
S. R. Bahn, K. W. Jacobsen, Comput. Sci. Eng. 2002, 4, 56–66.
P. E. Blöchl, Phys. Rev. B 1994, 50, 17953–17979.
G. Kresse, Phys. Rev. B 1999, 59, 1758–1775.
J. Perdew, J. Chevary, S. Vosko, K. Jackson, M. Pederson, D.
Singh, C. Fiolhais, Phys. Rev. B 1993, 48, 4978–4978.
M. J. Gillan, J. Phys. Condens. Matter 1989, 1, 689–711.
H. J. Monkhorst, J. D. Pack, Phys. Rev. B 1976, 13, 5188–5192.
M. Digne, J. Catal. 2004, 226, 54–68.
L. C. Grabow, in Comput. Catal. (Eds.: A. Asthagiri, M.J. Janik),
RSC Publishing, 2013, pp. 1–58.
C. A. Wolcott, A. J. Medford, F. Studt, C. T. Campbell, J. Catal.
2015, 330, 197–207.
F. Abild-Pedersen, J. Greeley, F. Studt, J. Rossmeisl, T. Munter, P.
Moses, E. Skúlason, T. Bligaard, J. K. Nørskov, Phys. Rev. Lett.
2007, 99, 4–7.
[7]
E. M. Fernandez, P. G. Moses, A. Toftelund, H. A. Hansen, J. I.
Martinez, F. Abild-Pedersen, J. Kleis, B. Hinnemann, J. Rossmeisl,
T. Bligaard, et al., Angew. Chemie Int. Ed. 2008, 47, 4683–4686.
G. Jones, F. Studt, F. Abild-Pedersen, J. K. Nørskov, T. Bligaard,
Chem. Eng. Sci. 2011, 66, 6318–6323.
[39]
[40]
[41]
[42]
[8]
[9]
H. Wang, L. Chen, Y. Lv, R. Ren, Appl. Surf. Sci. 2014, 290, 154–
160.
S. Wang, B. Temel, J. Shen, G. Jones, L. C. Grabow, F. Studt, T.
Bligaard, F. Abild-Pedersen, C. H. Christensen, J. K. Nørskov,
Catal. Letters 2010, 141, 370–373.
[43]
[44]
L. Bengtsson, Phys. Rev. B 1999, 59, 12301–12304.
G. Henkelman, B. P. Uberuaga, H. Jonsson, J. Chem. Phys. 2000,
́
[10]
[11]
H. Wang, Y. Liu, M. Li, H. Huang, H. M. Xu, R. J. Hong, H. Shen,
Optoelectron. Adv. Mater. Rapid Commun. 2010, 4, 1166–1169.
J. Sehested, K. E. Larsen, A. L. Kustov, A. M. Frey, T.
Johannessen, T. Bligaard, M. P. Andersson, J. K. Nørskov, C. H.
Christensen, Top. Catal. 2007, 45, 9–13.
113, 9901.
[12]
F. Studt, I. Sharafutdinov, F. Abild-Pedersen, C. F. Elkjær, J. S.
Hummelshøj, S. Dahl, I. Chorkendorff, J. K. Nørskov, Nat. Chem.
2014, 6, 320–324.
[13]
[14]
Y.-H. Chin, D. Resasco, in R. Soc. Catal., 1999, pp. 1–39.
R. F. Hicks, H. Qi, M. L. Young, R. G. Lee, J. Catal. 1990, 122, 280–
294.
[15]
[16]
J. G. McCarty, Catal. Today 1995, 26, 283–293.
M. Monai, T. Montini, C. Chen, E. Fonda, R. J. Gorte, P. Fornasiero,
ChemCatChem 2015, 7, 2038–2046.
[17]
[18]
[19]
G. Zhu, J. Han, D. Y. Zemlyanov, F. H. Ribeiro, J. Am. Chem. Soc.
2004, 126, 9896–9897.
T. P. Senftle, A. C. T. van Duin, M. J. Janik, ACS Catal. 2015, 5,
6187–6199.
T. R. Baldwin, R. Burch, Appl. Catal. 1990, 66, 359–381.
This article is protected by copyright. All rights reserved.

10.1002/cctc.201601333
ChemCatChem
FULL PAPER
Entry for the Table of Contents (Please choose one layout)
FULL PAPER
Hieu A. Doan, Munish K. Sharma,
William S. Epling, and Lars C. Grabow*
Page No. – Page No.
From Active Site Model to Real
Catalysts: Importance of the Material
Gap in the Design of Pd Catalysts for
Methane Oxidation
Consistent trends for methane activation over a series of modified Pd/γ-Al2O3 catalyst models in agreement with experimental activity
measurements support the common practice of using simplified catalyst representations for computational catalyst screening.
This article is protected by copyright. All rights reserved.