Modifying structure-sensitive reactions by addition of Zn to Pd

Article abstract of DOI:10.1016/j.jcat.2014.07.016

Silica-supported Pd and PdZn nanoparticles of a similar size were evaluated for neopentane hydrogenolysis/isomerization and propane hydrogenolysis/ dehydrogenation. Monometallic Pd showed high neopentane hydrogenolysis selectivity. Addition of small amounts of Zn to Pd lead Pd-Zn scatters in the EXAFS spectrum and an increase in the linear bonded CO by IR. In addition, the neopentane turnover rate decreased by nearly 10 times with little change in the selectivity. Increasing amounts of Zn lead to greater Pd-Zn interactions, higher linear-to-bridging CO ratios by IR and complete loss of neopentane conversion. Pd NPs also had high selectivity for propane hydrogenolysis and thus were poorly selective for propylene. The PdZn bimetallic catalysts, however, were able to preferentially catalyze dehydrogenation, were not active for propane hydrogenolysis, and thus were highly selective for propylene formation. The decrease in hydrogenolysis selectivity was attributed to the isolation of active Pd atoms by inactive metallic Zn, demonstrating that hydrogenolysis requires a particular reactive ensemble whereas propane dehydrogenation does not.

Full text of DOI:10.1016/j.jcat.2014.07.016

Journal of Catalysis 318 (2014) 75–84
Contents lists available at ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Modifying structure-sensitive reactions by addition of Zn to Pd
a
b
c
c
d,
⇑
David J. Childers , Neil M. Schweitzer , Seyed Mehdi Kamali Shahari , Robert M. Rioux , Jeffrey T. Miller ,
Randall J. Meyer
a
a,
⇑
Department of Chemical Engineering, University of Illinois at Chicago, United States
Center for Catalysis and Surface Science, Northwestern University, United States
Department of Chemical Engineering, The Pennsylvania State University, United States
Division of Chemical Sciences and Engineering, Argonne National Laboratory, United States
b
c
d
a r t i c l e i n f o
a b s t r a c t
Article history:
Silica-supported Pd and PdZn nanoparticles of a similar size were evaluated for neopentane hydrogenol-
ysis/isomerization and propane hydrogenolysis/dehydrogenation. Monometallic Pd showed high neopen-
tane hydrogenolysis selectivity. Addition of small amounts of Zn to Pd lead Pd–Zn scatters in the EXAFS
spectrum and an increase in the linear bonded CO by IR. In addition, the neopentane turnover rate
decreased by nearly 10 times with little change in the selectivity. Increasing amounts of Zn lead to greater
Pd–Zn interactions, higher linear-to-bridging CO ratios by IR and complete loss of neopentane conversion.
Pd NPs also had high selectivity for propane hydrogenolysis and thus were poorly selective for propylene.
The PdZn bimetallic catalysts, however, were able to preferentially catalyze dehydrogenation, were not
active for propane hydrogenolysis, and thus were highly selective for propylene formation. The decrease
in hydrogenolysis selectivity was attributed to the isolation of active Pd atoms by inactive metallic Zn,
demonstrating that hydrogenolysis requires a particular reactive ensemble whereas propane dehydroge-
nation does not.
Received 9 May 2014
Revised 17 July 2014
Accepted 18 July 2014
Keywords:
Structure-sensitive reaction
Intermetallic PdZn
Propane dehydrogenation
CleanCat
DRIFTS
CO calorimetry
EXAFS
Ó 2014 Elsevier Inc. All rights reserved.
Neopentane conversion
1
. Introduction
The catalytic performance of metal nanoparticles can be modi-
often lead to changes in electronic properties. For example, as
the size of a metal cluster is reduced, the cluster exhibits quantum
confinement effects that perturb the electronic structure and in
some cases, can even give rise to a band gap [2–4]. The reduced
coordination of the cluster often results in changes in the energy
of the valence orbitals that alter the bond strength of adsorbates
[5–7].
fied by changes in the surface geometry due to modification of the
particle size or changes in the electronic properties by the addition
of promoters or alloy formation. Boudart was the first to divide
reactions into two groups based on sensitivity of the catalytic
activity to particle size [1]. Structure-sensitive reactions are those
for which the kinetics is dependent on the particle size (due to
changes in the coordination of surface atoms with particle size),
while structure-insensitive reactions are independent of the parti-
cle size. For structure-sensitive reactions, the turnover rate (TOR),
or rate per surface atom, changes with size. This geometric effect
results because the active site for structure-sensitive reactions
requires an ensemble of active atoms and the number of these
ensembles changes with particle size. For structure-insensitive
reactions, every surface atom is an active site, and thus, the rate
is directly proportional to the dispersion.
Alloying with another metal can also affect the electronic prop-
erties of a catalyst [8]. Two effects appear simultaneously when
metals are alloyed. First, charge transfer may occur between the
alloying elements due to differences in the level of filling and rel-
ative energies of the valence orbitals. Second, changes in hybridiza-
tion of the bonding between metals may result due to changes in
the bond distances between metal atoms. At the same time, the
orbital extent or the size of the valence orbitals of the alloying ele-
ments is different, resulting in changes in the overlap between the
bonding orbitals. Therefore, alloying can result in both electronic
effects (degree of charge transfer and hybridization) and geometric
effects due to the creation of specific reaction ensembles. Although
it is well known that these two aspects of the catalyst are impor-
tant, it is often unknown whether one factor contributes more than
another with respect to selectivity control for a particular reaction.
However, understanding which factor is dominant for a particular
reaction can be used to design improved catalysts.
Electronic effects alter the chemical reactivity of the metal
nanoparticle due to changes in the electronic structure. These
two effects are often interrelated as changes in surface geometry
⇑
Corresponding authors.
E-mail addresses: millerjt@anl.gov (J.T. Miller), rjm@uic.edu (R.J. Meyer).
http://dx.doi.org/10.1016/j.jcat.2014.07.016
021-9517/Ó 2014 Elsevier Inc. All rights reserved.
0

7
6
D.J. Childers et al. / Journal of Catalysis 318 (2014) 75–84
Previous work on Pt and Pd catalysts of different particle sizes
were synthesized by two different synthesis methods – sequential
and co-incipient wetness impregnation (co-IWI) under controlled
pH conditions [23,24]. During IWI, the support is contacted with
just enough metal precursor solution to fill the pore volume. All
of the metal precursor solution contacts the surface due to the
incipient amount of liquid used. Sequential impregnation requires
the two metals be added in separate incipient wetness impregna-
tion steps, while co-impregnation combines the metals into a sin-
gle solution. The 2% Pd catalyst was synthesized using a procedure
described in previous work [9].
yielded a correlation between neopentane isomerization selectiv-
ity and the initial CO heat of adsorption [9]. Isomerization selectiv-
ity increased with decreasing CO heat of adsorption and this
correlation appears to be independent of metal, nanoparticle size,
and the configuration of adsorbed CO. Large Pd particles have pri-
marily bridge-bound CO, while small Pt particles have primarily
linear-bound CO, but they have similar catalytic behavior. This pro-
vides strong evidence that the selectivity is dependent on electro-
nic effects and the strength of adsorbate chemisorption. Extending
this correlation, high selectivity to isomerization should be
achieved by materials which bind CO even less strongly than Pt
terrace sites. From previous DFT calculations, an intermetallic 1:1
PdZn (111) surface was shown to have a lower CO adsorption
energy compared to Pt(111) [10]. Sarkrany et al. found experimen-
tally that the CO heat of adsorption on a PdZn alloy was 67 kJ/mol
lower than that of monometallic Pd [11,12]. Using our correlation
between the heat of adsorption and the neopentane isomerization
selectivity, it is anticipated that alloy formation between Pd and Zn
would lead to a lower heat of adsorption and also improve neopen-
tane isomerization selectivity [10].
PdZn alloys have been studied for a number of different reac-
tions, including methanol steam reforming, water–gas shift, alkene
hydrogenation and auto-thermal reforming [11,13–19]. Pd and Zn
form a large number of bimetallic structures each with their own
unique crystal structure and stoichiometry. At a 1:1 ratio of Pd:Zn,
a face centered tetragonal structure forms, which is present for a
wide range of Pd compositions (30–70%) [13,17,18,20]. Improved
selectivity and stability were the main benefits that PdZn catalysts
exhibited for these various reactions. These improvements were
credited to both geometric changes caused by zinc altering the cat-
alyst surface by expanding the palladium bond distance and elec-
tronic effects observed by the weakening of the Pd–CO bond
2.1.1. 2%Pd–10%Zn
This catalyst was synthesized by sequential impregnation. Zinc
was first added to silica using the IWI method. 18.1 g of Zn(NO
ꢀ6H O was dissolved in 15 mL of H O. 15 M NH OH was then added
to this solution to initially form a white precipitate which dis-
solved when additional NH OH was added bringing the total vol-
3 2-
)
2
2
4
4
ume to 50 mL. This solution was added dropwise to 40 g of silica
and stirred. The catalyst was dried overnight at 125 °C and then
calcined at 300 °C for 3 h. Palladium was also added by the IWI
method. 2.81 g of 10% Pd(NH
added dropwise to 5 g of the 10% Zn/SiO
was then dried overnight at 125 °C, calcined at 500 °C for 3 h and
3
)
4
3
(NO )
2
solution from Aldrich was
2
catalyst. This catalyst
2
reduced at 550 °C in 4% H /He at 30 cc/min for 30 min.
2.1.2. 3%Pd–1.8%Zn
The catalyst was synthesized by co-IWI with both zinc and pal-
ladium precursors added simultaneously. 0.42 g of Zn(NO
ꢀ6H
was dissolved in 1.5 mL of 15 M NH OH and this solution was
added to 4.21 g of 10% Pd(NH (NO solution in H O. This solu-
3
)
2
2
O
4
3
)
4
3
)
2
2
tion was then added dropwise to silica (5 g) and mixed between
drops. The catalyst was dried overnight at 125 °C, calcined at
225 °C for 3 h and then reduced at 300 °C at 30 cc/min in
[
12,18]. Chen et al. performed DFT studies on the surface structure
4%H
2.2. STEM
The STEM images were taken at UIC’s Research Resources Cen-
ter facility using the JEOL-ARM 200CF aberration-corrected micro-
scope (70 pm spatial resolution and 300 meV energy resolution).
Samples were dispersed in isopropyl alcohol and sonicated for
20 min. A drop of the solution was added to a holey-carbon copper
grid and dried under a heat lamp for 20 min. Images were taken
using the high angle angular dark field (HAADF) mode and particle
size was counted using the Particule2 program. A minimum of 100
particles were counted to get an accurate representation of the
particle size distribution for each catalyst.
2
/He for 30 min.
of PdZn alloys and found that the (111) surface was the most ener-
getically favorable [21]. Calculations also suggested that the Pd d-
band valence width was significantly reduced compared to pure
Pd which was linked to improvements in methanol steam reforming
performance. Many of these studies used ZnO as a support or sup-
3 2
ported ZnO, e.g., from a Zn(NO ) precursor, as a source of zinc for
alloy formation. PdZn intermetallic structures are typically synthe-
sized with excess zinc and not in nominal molar ratios close to 1:1.
Iwasa et al. were unable to form the PdZn intermetallic structure on
silica at temperatures up to 700 °C due to a lack of zinc reduction in
ZnO [22]. Fottinger et al. performed methanol steam reforming over
Pd/ZnO catalysts and found that reaction conditions caused PdZn
intermetallic formation which resulted in improvements to selec-
tivity [19]. Through in situ XAS it was also shown that the PdZn inter-
metallic structure was also reversible upon exposure to oxygen.
Here, we report on the synthesis, characterization, and testing
of two PdZn nanoparticle catalysts supported on silica with differ-
ent Pd:Zn molar ratios. The catalysts were characterized by elec-
tron microscopy, diffuse reflectance infrared Fourier transform
spectroscopy (DRIFTS) of adsorbed CO, X-ray absorption spectros-
copy (XAS), CO chemisorption, and isothermal calorimetry.
Neopentane hydrogenolysis/isomerization and propane hydrogen-
olysis/dehydrogenation reactions were used to evaluate the influ-
ence of Zn on the hydrogenolysis selectivity of Pd.
2.3. X-ray absorption spectroscopy (XAS)
X-ray absorption spectroscopy (XAS) measurements for the Pd
K (24350 eV) edge were made on the bending magnet beamline
of the Materials Research Collaborative Access Team (MRCAT) at
the Advanced Photon Source (APS), Argonne National Laboratory.
Measurements were taken in transmission mode. A palladium foil
spectrum was acquired through a third ion chamber simulta-
neously with each measurement for energy calibration.
Samples were prepared by grinding the catalysts into a fine
powder and pressing them into the sample holder. The sample
holder is a metal cylinder capable of holding up to six individual
samples. The sample holder is then placed in a quartz tube with
ports containing Kapton windows on each end to flow gases or iso-
late the sample after treatment. The sample thickness was chosen
to give a total absorbance at the Pd K-edge between 1 and 2
absorption lengths and edge steps around 0.3–0.5. The XAS spectra
were obtained following reduction at 275 °C and 550 °C at
2
. Experimental methods
2
.1. Catalyst synthesis
PdZn bimetallic catalysts supported on silica (Davisil 646 silica
2
gel from Sigma–Aldrich, 300 m /g and 1.15 mL/g pore volume)

D.J. Childers et al. / Journal of Catalysis 318 (2014) 75–84
77
3
atmospheric pressure in a 4% H
2
/He mixture at 50 cm /min flow
tane hydrogenolysis. The reactor was purged with He for 5 min
before each run, and the catalyst was reduced in 4% H /He as the
rate. After reduction, the samples were purged with He at
00 cm /min at the reduction temperature and cooled to room tem-
2
3
1
temperature was increased to the reaction temperature,
550 ± 2 °C. Once the reaction temperature stabilized, the pre-
mixed reactant feed gas consisting of 2% propane and balance Ar
was passed through the reactor. The flow rate of the feed gas
perature in He flow. Trace oxidants in He were removed by passing
through a Matheson PUR-Gas Triple Purifier Cartridge containing a
Cu trap. All spectra were obtained at room temperature in He.
WINXAS 3.1 software was used to fit the XAS data. The EXAFS
coordination parameters were obtained by a least-squares fit in
k-space of the k -weighted Fourier transform data from 2.6 to
2.1 Å , and the first shell fit of the magnitude and imaginary
parts were performed between 1.6 and 3.0 Å. Because of the lim-
ited data range and number of allowed fit parameters of the two-
shell fit in the bimetallic nanoparticles, the error in the fits was
3
was set to 50 cm /min and held constant throughout the test so
that the deactivation of each catalyst could be determined. The test
was run until a steady-state conversion was reached. Each experi-
mental run was completed within 6 h for consistency and multiple
runs for each catalyst were performed. The maximum relative
error of any selectivity measurements was 6%, with most of the
data being reproducible within 2%.
2
ꢁ1
1
2
determined by fixing
D
r
at values typical of 2.5–3 nm nanoparti-
cles, that is, 0.001–0.002 greater than metallic foils. The error in N
was ±10% and in R was ±0.02 Å, within the typical fitting errors of
EXAFS. Fits were performed by altering the coordination number
2.6. Diffuse reflectance infrared Fourier transform spectroscopy
(DRIFTS)
2
2
(
CN), bond distance (R),
r
0
, and energy shift (E ). The
r
value
Infrared spectra were obtained using a Thermo Scientific Nico-
let 6700 FTIR spectrometer equipped with a Harrick Scientific Pray-
ing Mantis diffuse reflectance in situ cell at the Northwestern Clean
Catalysis (CleanCat) Core Facility. Samples were ground to a fine
powder using a mortar and pestle and packed into the sample
chamber to create a uniform surface. The chamber was purged
was kept constant through all sample fits, and CN and R were
allowed to vary in turn to determine the correct fit. Although CN
is determined for each catalyst, they used only to determine the
coordination environment of the Pd. Although coordination num-
bers can be used to determine the particle size [25] and morphol-
ogy in monometallic NPs [26], the complex structure, i.e., Pd–Zn
alloy and multiple phases, e.g., Pd core with Pd–Zn alloy shell,
makes determination of the particle size, for example, unreliable.
2
with Ar, then the gas was switched to 10% H /N2, and the temper-
ature was raised to 250 °C and held for 15 min. After reduction of
the catalyst, the gas was switched back to Ar and the temperature
was reduced to 25 °C. A background scan was then recorded, which
was averaged over 100 scans (2-min observation time) with
2.4. Neopentane hydrogenolysis and isomerization
ꢁ
1
4
cm resolution. The sample was then exposed to 1.02% CO/N
2
Neopentane hydrogenolysis and isomerization kinetics and
and another scan was taken once equilibrium was reached, at
which point the flow was changed back to Ar and a final scan
was taken once the intensity of the adsorbed CO peak was invari-
ant with time. The linear-to-bridge-bound ratios reported here do
not take into account the differences in extinction coefficients
between the adsorption sites and therefore do not represent quan-
titative coverages, but rather reflect qualitative differences
between catalysts [27].
selectivity were determined using 0.05–0.15 g of catalyst diluted
00
with 0.9 g of silica and loaded into a 0.5 O.D. quartz plug-flow
reactor. Glass wool was used for the bottom 2 cm of the bed. A
0
.5-cm silica layer was placed on top of the glass wool before the
catalyst and silica mixture was added to the reactor resulting in
a catalyst bed height of 3 cm. The reactor was purged with He
for 5 min before each run, and the catalyst was reduced in 4% H
He as the temperature was increased to the reaction temperature,
73 ± 2 °C. This temperature allowed for all the catalysts to be
2
/
2
2.7. CO heats of adsorption
tested at differential conversion. A K-type thermocouple was
inserted from the bottom of the reactor into the lower portion of
the catalyst bed. Once the reaction temperature stabilized, the
pre-mixed reactant feed gas consisting of 0.35% neopentane and
Determination of the initial heat of adsorption of carbon mon-
oxide on the two Pd–Zn catalysts (60–70 mg) reduced at two dif-
ferent temperatures (300 and 550 °C) was conducted utilizing a
Setaram Sensys EVO differential scanning calorimetry interfaced
with a plug-flow reactor. The plug-flow reactor was connected to
a mass spectrometer. After reduction at the specified temperature
in 5.11% H /Ar (both gases 99.999%) for 2 h, the catalyst was cooled
2
down to 35 °C in the same gas, and the catalyst was exposed to a
mixture of 1% CO (99.999% research grade) in He (99.9999%,
3
2
.5% H balanced in He was passed through the reactor. The flow
3
rate of the feed gas was varied from 25 to 100 cm /min to obtain
differential conversions (0.5–6%). Each flow rate was run for at
least 1 h to ensure steady-state conversion had been reached. An
Agilent 6890N gas chromatograph (J&W Scientific GS-Alumina col-
umn) with an FID detector was used to analyze the products and
was equipped with a back pressure regulator at the outlet to hold
the system at a constant pressure of 9 psig. Each experimental run
was completed within 6 h for consistency, and multiple runs for
each catalyst were performed. No appreciable deactivation during
neopentane hydrogenolysis was observed in any of the catalysts
over this period of time. The maximum relative error of any selec-
tivity measurements was 6%, with most of the data being repro-
ducible within 2%. Turnover rates were calculated based upon
the moles of neopentane converted divided by the number of
active sites determined by the dispersion calculated from CO
chemisorption.
research grade) pulsed into the 5% H
2
/Ar stream from a ten-way
switching value with a 1000 L sample loop. The number of moles
l
ꢁ4
of CO per pulse calculated from the ideal gas law was ꢂ4 ꢃ 10
-
mmol. Twenty pulses of carbon monoxide were employed; the ini-
tial heat of adsorption was determined by only considering
injections in which the entire pulse of carbon monoxide was con-
2
sumed. The 3Pd–1.8Zn/SiO adsorbed entirely 5–6 pulses depend-
ing on the reduction temperature, while the 2Pd–10Zn/SiO
2
adsorbed entirely ꢂ1 pulse of carbon monoxide.
2.8. CO chemisorption
2
.5. Propane dehydrogenation
The CO chemisorption measurements were conducted at the
Northwestern University Clean Catalysis (CleanCat) Core Facility
using an Altamira Instruments AMI-200. Catalysts (0.05–0.2 g)
were loaded into a U-shaped quartz reactor tube, which was
weighed before and after sample addition to ensure an accurate
Propane dehydrogenation selectivity and rate were determined
using 0.2–0.5 g of catalyst diluted with 0.9 g of silica and loaded
into the same 0.5 O.D. quartz plug-flow reactor used for neopen-
00

7
8
D.J. Childers et al. / Journal of Catalysis 318 (2014) 75–84
weight measurement. The loaded tube was then loaded into the
furnace, and the catalysts were reduced in H /Ar at 300 °C or
50 °C for 2 h (10 °C/min ramp rate) and then flushed for 30 min
in He. Using a six-way valve, 5% CO/He was then pulsed (595
Table 1
Particle sizes determined by STEM.
2
5
Sample
Particle size by STEM (nm)
lL
2%Pd
2.5 ± 1.1
2.2 ± 0.8
2.7 ± 0.8
2.6 ± 0.7
loop volume) into the system 15 times at 30 °C to ensure the sur-
face was saturated. Each pulse peak was integrated to find the vol-
ume of CO remaining following adsorption. Surface saturation was
typically reached within 10 pulses. The dispersion determined by
CO chemisorption also included average stoichiometric factor
determined from the uncorrected DRIFTS linear-to-bridging ratios
for each catalyst. CO bound linearly account for one CO molecule
per surface atom while bridge-bound CO implies there is one CO
molecule per two surface atoms. The equation used to get the dis-
persion is given below.
2%Pd–10%Zn 300 °C
3%Pd–1.8%Zn 300 °C
3%Pd–1.8%Zn 550 °C
shown in Fig. 2b. XANES were analyzed to determine the Pd K-edge
position of each catalyst and evaluate the oxidation state of palla-
dium. The edge energy is defined as the inflection point of the lead-
ing edge and is determined by taking the maximum of first
derivative of the spectra in Fig. 2a. The edge energy for all three
mol CO adsorbed
catalysts is summarized in Table 2. A K-edge energy of
Dispersion ¼
ꢃ ððlinear CO fraction ꢃ 1Þ
mol metal
þ ðbridge CO fraction ꢃ 2ÞÞ
24.350 keV corresponds to metallic Pd and the PdZn catalysts are
slightly shifted as seen in Fig. 2b (to 24.349 keV) from Pd which
is consistent with the presence of bimetallic nanoparticles [8].
The intensity of the first absorption peaks is similar indicating that
the palladium is metallic rather than oxidized which would cause
the peak intensity to increase. The small differences in the spectra
following the first peak, changes in intensities and peak shifts, also
indicate that non-Pd neighbors are near to Pd.
3
. Results
3
.1. Particle size analysis
STEM imaging was used to determine metal nanoparticle size. A
Fig. 3 shows the EXAFS spectra for the Pd foil reference and all
three catalysts reduced at 275 °C, and Table 2 summarizes the
parameters calculated from the fitting of the spectra. Coordination
numbers and bond distances were determined by fitting the mag-
representative STEM image for the 3%Pd–1.8%Zn catalyst and cor-
responding size distribution are shown in Fig. 1. The particle sizes
for the three catalysts tested are reported in Table 1 and show that
the average particle sizes are similar. The 3%Pd–1.8%Zn catalyst
was also reduced at the propane dehydrogenation reaction tem-
perature (550 °C) to determine if any sintering occurred. The parti-
cle size did not appreciably change, between the two
temperatures. Having similar particle sizes between catalysts
enables the changes caused by alloying to be studied without hav-
ing to account for changes in particle size which are known to alter
the kinetics of structure-sensitive reactions. From our previous
studies, there are results for similar sized monometallic Pd cata-
lysts for comparison with these PdZn bimetallic catalysts [9].
2
nitude and imaginary parts of the Fourier transform of the k -
weighted first shell spectra. Although the total coordination num-
bers for each of the samples are identical, the identity of the neigh-
bors and their distances change between the three catalysts. The
monometallic Pd catalyst shows two prominent peaks at 2.05
and 2.50 Å (phase uncorrected distances) at a ratio of 2:1, typical
of metallic Pd and also observed for the Pd foil spectra seen in
Fig. 3a [18]. The Pd–Pd bond distance of 2.75 Å in Table 2 also cor-
responds to metallic Pd. Fig. 3b shows both bimetallic PdZn cata-
lysts; the 3%Pd–1.8%Zn spectra look similar to metallic Pd;
however, a change in the peak ratio to 1.45:1 is consistent with a
small number of non-Pd scattering atoms. The fit confirms there
are Pd–Zn bonds with a coordination number of 2.3 and a bond dis-
tance of 2.56 Å. The coordination number for the Pd–Pd scatters is
5.3 with a bond distance of 2.73 Å (similar to metallic Pd). The dif-
ferences in the coordination numbers between the Pd–Pd and Pd–
Zn scatterers indicate that the majority of the palladium atoms in
this catalyst are still surrounded by other palladium atoms with a
3
.2. X-ray absorption spectroscopy (XAS)
Extended X-ray absorption fine structure (EXAFS) and X-ray
absorption near-edge (XANES) spectra were gathered for each cat-
alyst following reduction at 275 °C and fitting done to determine
bond distances and coordination numbers (CN). The XANES spectra
are shown in Fig. 2a and an expanded view to the leading edge is
8
6
4
2
0
0
0
0
0
0
1
2
3
4
5
6
Particle Diameter (nm)
Fig. 1. 3%Pd–1.8%Zn sample image and particle size distribution determined by STEM.

D.J. Childers et al. / Journal of Catalysis 318 (2014) 75–84
79
1
1
0
0
0
0
0
.2
.0
.8
.6
.4
.2
.0
0.6
(
a)
(b)
2
3
2
Pd
Pd-1.8Zn
Pd-10Zn
2Pd
3Pd-1.8Zn
2Pd-10Zn
0
.4
.2
0
24.26 24.28 24.30 24.32 24.34 24.36 24.38 24.40
24.345
24.350
24.355
Photon Energy (keV)
Photon Energy (keV)
Fig. 2. Pd XANES K-edge spectra for catalysts reduced at 275 °C. (a) Full edge and (b) expanded view of the leading edge of the XANES near the inflection point.
3
.3. Diffuse reflectance infrared Fourier transform spectroscopy
Table 2
(
DRIFTS)
XANES and EXAFS fitting parameters following 275 °C reduction.
Catalyst
Edge energy
keV)
Coordination
number
Bond
distance
DRIFTS of CO adsorption was used to evaluate changes in sur-
(
face composition and structure of the catalysts. Fig. 4 shows the
DRIFTS spectra for the PdZn catalysts as well as the monometallic
Pd catalyst with a similar particle size. Bridge-bound CO has previ-
(
Å)
2
3
2
Pd
24.3500
24.3496
24.3494
Pd–Pd
Pd–Zn
7.5
–
2.75
–
ꢁ1
ously been assigned to the peaks in the range of 1800–2000 cm
.
Pd–1.8Zn
Pd–10Zn
Pd–Pd
Pd–Zn
5.3
2.3
2.73
2.56
ꢁ1
Within this range, peaks between 1800 and 1900 cm are attrib-
uted to CO bridge-bound on terrace and hollow sites, while peaks
Pd–Pd
Pd–Zn
2.7
5.2
2.72
2.56
ꢁ
1
between 1900 and 2000 cm are assigned to CO bridge-bound to
corner and edge sites [28,29]. Linear bound CO is characterized by
ꢁ1
two peaks typically appearing around 2080 and 2050 cm
in
which Lennon et al. have assigned to corner and edge adsorption
sites respectively [28].
few Zn neighbors. The 2%Pd–10%Zn catalyst also has two promi-
nent peaks, but these peaks deviate from the pure Pd spectra sig-
nificantly. Although the peak positions are similar, the ratio of
the peaks has changed, suggesting that the bonding environment
of Pd has changed with the addition of zinc. The fits in Table 2 con-
firm this shift from Pd-rich particles to a catalyst with significant
interaction between Pd and Zn as the Pd–Zn coordination number
increases to 5.3 and the Pd–Pd coordination number decreases to
The 2%Pd catalyst, with an average particle size of 2.5 nm, has
ꢁ1
one very prominent bridge-bound CO peak at 1936 cm , a shoul-
ꢁ
1
der centered around 1835 cm , and one linear CO peak at
ꢁ1
2084 cm . The presence of two bridge-bound CO peaks is typical
of pure Pd catalysts [28,29]. The 3%Pd–1.8%Zn catalyst has one
ꢁ1
bridge-bound CO peak located at 1975 cm . The loss of the
ꢁ1
bridge-bound CO peak around 1835 cm
peak suggests zinc
2
.7. The bond distances have not changed significantly in the two
breaks up palladium terraces on the surface typically responsible
for bridge-bound CO. Since the tail of the peak extends into the ter-
race site range, there may be a small percentage of Pd ensembles
Pd–Zn catalysts and these bond distances are also similar to those
found in the literature [18]. Although the average coordination
environment in these catalysts is different, EXAFS is a bulk charac-
terization; therefore, the surface composition and structure, which
is responsible for the catalytic activity, is uncertain.
ꢁ1
remaining. The shift of the linear peak to 2075 cm
from
ꢁ1
2084 cm indicates there is a small fraction of CO linearly bound
to edge sites since most is still bound to corner sites. These shifts
0
0
0
0
0
0
.05
.04
.03
.02
.01
.00
0
0
0
.03
.02
.01
(a)
(
b)
2
Pd
3
2
Pd+1.8Zn
Pd+10Zn
Pd Foil
0.00
0
1
2
3
4
0
1
2
3
4
R (angstroms)
R (angstroms)
Fig. 3. Pd K-edge spectra and fitting parameters following 275 °C reduction. (a) 2%Pd (black) and Pd foil (red) and (b) 3%Pd–1.8%Zn (red) and 2%Pd–10%Zn (blue). (For
interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

8
0
D.J. Childers et al. / Journal of Catalysis 318 (2014) 75–84
0
0
0
0
.3
.2
.1
.0
was preferred on the intermetallic surface past ½ ML CO coverage
31].
2
3
2
Pd
[
Pd-1.8Zn
Pd-10Zn
3.4. Heat of adsorption of CO determined by isothermal calorimetry
The initial heat of adsorption determined previously for a pair of
Pd/SiO
ꢂ130 kJ mol [9]. The heat of CO adsorption on Pd was previously
determined for the catalysts reduced in H , evacuated at the reduc-
2
catalysts with a mean TEM particle size of 2.5–3 nm was
ꢁ1
2
tion temperature, followed by cooling in an inert gas to the adsorp-
tion temperature (35 °C). In the previous work, the differential
(
initial) heat of adsorption was determined using a volumetric
adsorption instrument (Micromeritics ASAP 2020C) interfaced
with a Setaram Sensys EVO differential scanning calorimeter. For
the Pd–Zn catalysts studied in this work, calorimetric measure-
ments of CO adsorption led to abnormally large heat flow which
was associated with the oxidation of CO by small quantities of oxy-
gen in He (99.999% UHP) led to the oxidation of CO. Enthalpy mea-
surements in He were unreliable since the oxidation reaction is
highly exothermic and contributed significantly to the measured
heat. All attempts to eliminate the CO oxidation reaction were
2
200
2100
2000
1900
1800
1700
-
1
Wavenumber (cm )
Fig. 4. Normalized DRIFTS spectra of CO adsorption on: 2%Pd (black), 2%Pd–10%Zn
blue) and 3%Pd–1.8%Zn (red) reduced at 250 °C. (For interpretation of the
references to color in this figure legend, the reader is referred to the web version
(
of this article.)
unsuccessful including utilizing
research grade He (99.9999%). As a result, the PdZn bimetallic cat-
alysts were cooled in a 5% H /Ar mixture to the adsorption temper-
a CO/He mixture containing
have also been reported for other PdZn catalysts [18,30]. The
2
2
1
%Pd–10%Zn catalyst has a small bridge-bound CO peak at
960 cm corresponding to corner and edge sites. The peak posi-
ature of 35 °C. Due to the small particle size and alloy formation
with Zn, the formation of a hydride is not expected, but there will
be an adsorbed layer of hydrogen on the surface of these PdZn cat-
alysts [32–35]. Chemisorbed hydrogen will lower the heat of
adsorption compared to an adsorbate-free surface of Pd [36]. The
results in Table 3 also show that increasing reduction temperature
slightly lowered the heat of adsorption. For comparison the heat of
CO adsorption of monometallic Pd catalyst was obtained using the
same procedure.
ꢁ1
tion has shifted from the other PdZn catalyst; however, it is still in
the range of bridge-bound CO on corner and edge sites and there
are no longer peaks for CO in the terrace region. The linear region
has two distinct peaks at 2080 cm and 2050 cm , which indi-
cates that CO is linearly bound to edge sites as well as corner sites
suggesting further zinc addition continues to reduce the number of
surface palladium ensemble sites. Two peaks in the linear region
have also been observed for PdZn catalysts [18].
ꢁ1
ꢁ1
The linear-to-bridge-bound ratio calculated by comparing the
raw areas of the different peaks is presented in Table 3. The addi-
tion of zinc causes a dramatic shift in the CO adsorption resulting
in greatly reduced concentration of bridge-bound CO. While the
monometallic Pd catalyst has mostly bridge-bound CO, the addi-
tion of zinc causes a significant increase in linear-bound CO. The
addition of excess zinc in the 2%Pd–10%Zn catalyst leads to a minor
increase in the overall ratio of linear-to-bridge-bound CO (2.5:1)
compared to the lower Zn sample (2.1:1), but there are differences
in the peak positions as discussed above. The shift in the linear-to-
bridge ratio indicates that zinc breaks up palladium ensembles on
the surface and decreases the number of Pd ensemble sites capable
of adsorbing bridge-bound CO. Weilach et al. used DFT to study CO
adsorption on PdZn surfaces and found that adsorption on top sites
3
.5. Neopentane isomerization and hydrogenolysis
Neopentane conversion was conducted at 273 ± 2 °C, 9 psig and
varying flow rates in order to vary conversion within range of dif-
ferential conversion (<10%). There are two possible reaction path-
ways: hydrogenolysis and isomerization. Isomerization products
include isopentane produced from an initial isomerization reaction
and n-pentane if isopentane undergoes a second isomerization
reaction before desorption from the surface. A single hydrogenoly-
sis reaction will produce isobutane and methane, and any subse-
quent steps will produce additional methane and lighter
hydrocarbons (propane and ethane). All selectivities given in
Table 4 are extrapolated to 0% conversion. Calculated turnover
rates (TOR) using the same method described previously are also
given in Table 4 [9]. Products with non-zero selectivity at 0% con-
version are considered primary products while products that have
Table 3
0
% selectivity are considered to be secondary products [37]. The
isomerization selectivity is defined as the selectivity to isopentane
and hydrogenolysis products are any molecule lower than C
Initial heat of CO adsorption values (±5 kJ/mol) in the presence of chemisorbed
hydrogen and linear-to-bridge ratios from DRIFTS.
Samplea
Reduction
temperature
CO (initial) heat of
adsorption with
Linear-to-
bridge
ratio
5
shown in Table 4. The 2%Pd catalyst has an isomerization and
hydrogenolysis selectivity of 9% and 91%, respectively. The 3%Pd–
1.8%Zn catalyst has a similar isomerization selectivity (13%) and
hydrogenolysis selectivity (87%). The product distributions are also
similar for the various hydrogenolysis products. If a single hydrog-
enolysis event occurs per converted neopentane molecule, equal
amounts of methane and isobutane should be identified in the
product distribution. An excess of methane can be seen for both
catalysts in Table 4 which indicates that further hydrogenolysis
is occurring, as evidenced by the presence of small amounts of
ethane and propane at 0% conversion.
(°C)
chemisorbed H
b
(
kJ/mol CO)
Pd/SiO c
Pd–10Zn/SiO
Pd–1.8Zn/SiO
2
2
3
2
225
300
300
92
102
99
0.2:1
2.5:1
2.1:1
1.8:1
2
2
550
93
a
Catalysts were initially reduced under 5.11% H
2
/Ar (both gases, 99.999% UHP)
for 2 h at the specified reduction temperature, followed by subsequent cooling in
2
H .
b
Initial heat of CO adsorption determined by microcalorimetry on a reduced Pd
or PdZn surface containing chemisorbed hydrogen.
c
Since zinc does not adsorb CO, the amount of surface palladium
was determined by chemisorption and used to calculate the TOR
Monometallic Pd catalyst from previous work studied using the modified
procedure.

D.J. Childers et al. / Journal of Catalysis 318 (2014) 75–84
81
Table 4
Neopentane hydrogenolysis TOR and product selectivity.
Catalyst
Dispersion
by CO chemisorption)
TOR (mol conv/metal site/s)
Initial product distribution (%)
(
CH
4
C
2
H
6
C
3
H
8
n-C
4
H
10
i-C
4
H
10
5 12
i-C H
ꢁ
3
4
2
3
2
%Pd
%Pd–1.8%Zn
%Pd–10%Zn
0.23
0.35
0.12
1.0 ꢃ 10
1.3 ꢃ 10
0
49
46
1
1
5
5
2
2
33
32
9
13
ꢁ
No measurable products
Table 5
Propane dehydrogenation product selectivity data.
ꢁ1
Catalyst
Dispersiona
Steady-state TOR (s
Rapid deactivation
)
Product distribution at 15% conversion (%)
CH
4
C
2
H
6
C
2
H
4
3 6
C H
2
3
2
Pd
Pd–1.8Zn
Pd–10Zn
0.23
0.27
0.08
75
0.3
1.2
6.3
0
0
8.1
1.7
0.7
10.7
98.3
98.4
ꢁ
ꢁ
3
2
8.8 ꢃ 10
1.9 ꢃ 10
a
Determined by CO chemisorption.
values in Table 4 [38–40]. Despite the similar selectivity distribu-
tion and particle size, the TOR of 3%Pd–1.8%Zn is an order of mag-
nitude lower than that of the 2%Pd catalyst. The 2%Pd–10%Zn
catalyst had negligible activity for neopentane conversion despite
reduced palladium on the nanoparticle surface as observed by
DRIFTS, CO chemisorption and calorimetry.
3%Pd–1.8%Zn differs from the selectivity measured during neopen-
tane conversion on the same catalyst. The selectivity for dehydro-
genation over both PdZn catalysts is comparable to those reported
for PtSn and chromia, the two industry standard catalysts [41–45].
Although both PdZn catalysts have similar selectivity, there are
some differences in their time-on-stream performance (Fig. 5a).
The 2%Pd catalyst begins with an initial conversion of 54%, but
deactivates to 2% conversion within 1 h. Due to the rapid catalyst
deactivation and large loss in conversion, TORs were not deter-
mined for the Pd catalyst. The 2%Pd–10%Zn (35%) and 3%Pd–
1.8%Zn (31%) catalysts start with lower initial conversions and
deactivate to a steady-state conversion of 7% and 15%, respectively.
The TORs of the PdZn catalysts calculated at steady state are given
in Table 5. These have not been corrected for deactivation and thus
are only approximate values.
Propane dehydrogenation is an endothermic reaction and equi-
librium-limited to a conversion of 40% at 550 °C [45]. The high con-
version observed initially for Pd and 3%Pd–1.8%Zn is a result of the
hydrogenolysis pathway; while the steady-state conversion for
dehydrogenation is below the equilibrium conversion.
Fig. 5b shows the change in propylene selectivity vs. conversion.
At high conversion, which also corresponds to early time-on-
stream, the 2%Pd catalyst has low selectivity to propylene (0%)
and methane is the most abundant product. The selectivity
improves as the catalyst deactivates; although the 2%Pd catalyst
does not begin to improve until about 10% conversion. The
3.6. Propane dehydrogenation
The catalysts were also evaluated for propane dehydrogenation
at 550 °C. In addition to propane dehydrogenation to form propyl-
ene, hydrogenolysis may also occur under these conditions produc-
ing methane and ethylene. The latter hydrogenolysis product can
also be hydrogenated to ethane. In order determine differences in
catalyst performance, all catalysts are compared at a similar level
of conversion. The product selectivities in Table 5 are all reported
at a propane conversion of 15%. The 2%Pd catalyst has a propylene
selectivity of 11% and an overall hydrogenolysis selectivity of 89%.
Within the hydrogenolysis product distribution, the selectivity to
methane is 75%, 8% to ethylene and the remainder (6%) to ethane.
The high hydrogenolysis selectivity is consistent with the results
for neopentane.
The two PdZn catalysts have a similar dehydrogenation selec-
tivity of 98%. The very low propane hydrogenolysis selectivity of
2
%Pd–10%Zn is consistent with the low selectivity for hydrogenol-
ysis of neopentane; however, the low hydrogenolysis selectivity of
6
5
4
3
2
1
0
0
0
0
0
0
0
(
a)
(b)
1
0
0
0
0
.0
.8
.6
.4
.2
0.0
0
2
4
0
10
20
30
40
50
60
Time on Stream (h)
Propane Conversion (%)
Fig. 5. Plots of (a) propane dehydrogenation conversion vs. time and (b) propylene selectivity vs. propane conversion for the 2%Pd (black triangles), 2%Pd–10%Zn (blue
squares) and 3%Pd–1.8%Zn (red circles). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

8
2
D.J. Childers et al. / Journal of Catalysis 318 (2014) 75–84
2
%Pd–10%Zn catalyst had a stable propylene selectivity at 98%,
Table 6
EXAFS fitting parameters after reduction 550 °C for both PdZn catalysts.
while the 3%Pd–1.8%Zn catalyst increases in propylene selectivity
from 79% as the conversion decreases.
Catalyst
Edge energy
(keV)
Coordination
number
Bond
distance
The 3%Pd–1.8%Zn catalyst has a low selectivity toward propane
hydrogenolysis at 550 °C, which is in contrast to the high selectiv-
ity to hydrogenolysis for neopentane at 275 °C. This difference in
selectivity suggests a change in the surface composition may be
occurring between these two reaction temperatures. Additional
characterization was done after reduction at this elevated temper-
ature in order to evaluate these changes. The XANES spectra (not
shown) do not indicate any change in Pd K-edge energy position,
but there are small changes in the peaks following the edge, consis-
tent with a change in the nearest neighbors to Pd.
(
Å)
3Pd–1.8Zn
2Pd–10Zn
24.3495
24.3492
Pd–Pd
Pd–Zn
2.7
4.9
2.72
2.56
Pd–Pd
Pd–Zn
2.6
8.0
2.84
2.57
Fig. 6 shows the differences in the magnitude of the Fourier
transform of the EXAFS region of the XAS spectra taken at both
reaction temperatures (275 and 550 °C) for the two PdZn catalysts.
The EXAFS of the 3%Pd–1.8%Zn catalyst shown in Fig. 6a undergoes
a change in the EXAFS from a Pd-rich PdZn bimetallic catalyst to
one that looks similar to the 2%Pd–10%Zn catalyst reduced at
0
0
0
0
.3
.2
.1
.0
3Pd-1.8Zn 550R
Pd-1.8Zn 300R
3
2
75 °C, i.e., a structure with an increased number of Pd–Zn bonds.
The fits in Table 6 confirm the number of Pd–Pd scatters decreases
form 5.3 to 2.7 and the number of Pd–Zn increases from 2.3 to 4.9
with increasing reduction temperature. After high-temperature
reduction, the structure of the metallic nanoparticles on the
3
%Pd–1.8%Zn catalyst has very similar composition to those on
the 2%Pd–10%Zn catalyst reduced at 275 °C. Fig. 6b also shows a
change in the composition of the 2%Pd–10%Zn nanoparticles
reduced at 550 °C with the appearance of a single peak centered
at 2.19 Å (phase uncorrected distance). This single peak spectrum
is characteristic of a PdZn intermetallic alloy [18,20]. The coordina-
tion numbers and bond distances given in Table 6 show a majority
of Pd–M scatterers in the catalyst are now Pd–Zn interactions and
the Pd–Pd bond increases from 2.75 Å to 2.84 Å. A similar change in
EXAFS has been observed in the literature which used increasing
amounts of Zn to form the intermetallic structure [18,46]. After
high-temperature reduction, the structure of the 2%Pd–10%Zn
nanoparticles is similar to that of a 1:1 PdZn intermetallic alloy
2
400 2300 2200 2100 2000 1900 1800 1700 1600
-1
Wavenumber (cm )
Fig. 7. Normalized DRIFTS spectra taken of the 3%Pd–1.8%Zn catalyst pre-reduced
at 300 °C (red) and 550 °C (blue). (For interpretation of the references to color in
this figure legend, the reader is referred to the web version of this article.)
culated from the areas of each peak did not change significantly at
the higher reduction temperature, from 2.1:1 at 275 °C to 1.8:1 at
5
50 °C.
[
18].
DRIFTS of the 3%Pd–1.8%Zn catalyst pre-reduced at 300 and
50 °C is shown in Fig. 7. The linear CO peak shifts from
4. Discussion
5
2
ꢁ1
ꢁ1
075 cm
to 2070 cm
at the higher reduction temperature,
The 2%Pd catalyst had a high hydrogenolysis selectivity during
neopentane (91%) and propane (89%) conversion. Results for both
reactions were expected as palladium is well known to have high
hydrogenolysis selectivity [9,47–52]. For propane, the initial con-
version of a 2%Pd catalyst is high, but deactivates quickly due to
coking (Fig. 5) with high selectivity to methane during this period.
Pd ensemble sites, as evidenced by bridge-bound CO, promote
which indicates a slight increase in linear-bound CO at edge sites.
The bridge-bound CO peak also shifts when the reduction temper-
ꢁ
1
ꢁ1
ature is increased from 1975 cm to 1965 cm
and the peak
broadens. There is also a shift in the linear CO peak which brings
the peak position in line with the linear CO peak of the 2%Pd–
1
0%Zn spectra seen in Fig. 4. The overall linear-to-bridge ratio cal-
Fig. 6. EXAFS spectra following reduction at both reaction temperatures 275 °C (red) and 550 °C (blue) of (a) 3%Pd–1.8%Zn and (b) 2%Pd–10%Zn (right plot). (For
interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

D.J. Childers et al. / Journal of Catalysis 318 (2014) 75–84
83
hydrogenolysis and coking reactions, rather than dehydrogenation
of propane.
For the 3%Pd–1.8%Zn catalyst reduced at low temperature, the
selectivity during neopentane conversion is very similar to mono-
metallic Pd nanoparticles. The lack of significant change in neopen-
tane selectivity suggests there are active Pd ensembles similar to
those in the Pd nanoparticles. While the selectivity is similar, the
TOR of 3%Pd–1.8% Zn is 10 times lower than that of Pd. If the sur-
face of the 3%Pd–1.8% Zn is high fraction of (unreactive) ordered
PdZn intermetallic alloy and a small fraction of the surface covered
with Pd ensemble sites, then the IR would look very similar to the
2%Pd–10%Zn catalyst, and the selectivity would look similar to that
of Pd, but with a reduced TOR. This also implies that even when
there is a small amount of reduced Zn, i.e., at low temperature,
the surface layer forms an ordered PdZn intermetallic alloy leading
to isolation of the active Pd atoms. This geometric effect is primar-
ily responsible for the changes in catalytic performance.
For propane dehydrogenation, it might be expected that the Pd
and 3%Pd–1.8%Zn catalysts would have similar catalytic perfor-
mance; however, the latter has high propylene selectivity with lit-
tle hydrogenolysis selectivity similar to that of the 2%Pd–10%Zn
catalyst. The EXAFS shows there is an increase in the number of
PdZn neighbors, while there is only a small increase in the lin-
ear-to-bridge ratio of adsorbed CO, which indicates a change in
the structure of the bimetallic nanoparticle at elevated tempera-
tures. The high propylene selectivity similar to the intermetallic
catalyst suggests there are very few Pd ensembles left on the sur-
face. In order to verify this conclusion, the 3%Pd–1.8%Zn catalyst
was first reduced at 550 °C and reactions of neopentane at 275 °C
were conducted. Although the catalyst was active after reduction
at low temperature, there is no neopentane conversion after reduc-
tion at high temperature.
While the Pd catalyst performed as expected, the 2%Pd–10%Zn
PdZn catalyst exhibited behavior not characteristic of monometal-
lic Pd. It proved to be unreactive toward neopentane despite evi-
dence for the presence of surface (metallic) Pd. Although the
PdZn intermetallic alloy was inactive for neopentane conversion,
it was highly active for propane dehydrogenation with high pro-
pylene selectivity and stability. From Table 5, propane dehydroge-
nation is a structure-insensitive reaction and the high selectivity
results from the ability to catalyze dehydrogenation but not the
structure-sensitive propane hydrogenolysis reaction.
The EXAFS and DRIFTS spectra are consistent with the forma-
tion of a PdZn intermetallic alloy. The changes in performance
are likely related to differences in the surface structure of Pd and
PdZn, i.e., geometric effects. Hydrogenolysis is a well-known exam-
ple of a structure-sensitive reaction, where a certain ensemble size
or geometry is required for the reaction to occur [48,53–55]. From
previous literature, it has been proposed that the hydrogenolysis
pathway requires a minimum of two active sites [53,55,56].
Fig. 8 shows the structure and bond distances for the bulk PdZn
intermetallic alloy. In this intermetallic composition, Pd has only
Zn neighbors and the Pd–Pd bond distance is 0.15 Å larger than
in monometallic Pd nanoparticles. In the ideal structure there are
no Pd ensemble sites since Pd atoms are at a non-bonding distance.
The lack of adjacent Pd atoms results in no catalytic activity for
neopentane conversion. While hydrogenolysis is a well-known
structure-sensitive reaction, the lack of neopentane isomerization
suggests this is also a structure-sensitive reaction. This conclusion
is consistent with the mechanism postulated by Anderson and
Avery where neopentane isomerization requires ensembles with
more than one attachment point for the neopentane molecule [57].
The above results are summarized by the scheme shown in
Fig. 9. The structure of the 3%Pd–1.8%Zn catalyst reduced at
2
75 °C has a Pd core with a small number of surface Zn (metallic)
atoms. The high ratio of linear-bound CO peaks in the DRIFTS sug-
gests that this PdZn film has an intermetallic alloy structure. The
catalytic performance suggests that approximately 90% of the sur-
face is the Pd–Zn alloy with the remaining 10% of the surface con-
sisting of Pd ensembles. At 550 °C, there is an increase in the
surface coverage of the PdZn intermetallic alloy with few remain-
ing Pd ensembles, although the nanoparticle core is still Pd-rich
[
58]. The inability to form a fully intermetallic PdZn alloy nanopar-
ticle is due to a limited amount of Zn interacting with the Pd. Datye
et al. and van Bokhoven et al. have also shown that PdZn interme-
tallic formation increases with reduction temperature [46,59].
For the 3%Pd–1.8%Zn catalyst, there is a high surface coverage of
the PdZn alloy with a Pd-rich core at 275 °C. For both Pd–Zn cata-
lysts, it appears that even under conditions which lead to a small
amount of reduced Zn, the PdZn surface forms an ordered alloy
structure rather than a random distribution of metallic Pd and Zn
Fig. 8. Structure of PdZn intermetallic alloy, coordination numbers and bond
distances of the bulk alloy. Pd atoms are blue and Zn atoms are purple. (For
interpretation of the references to color in this figure legend, the reader is referred
to the web version of this article.)
Fig. 9. Simplified cross section of Pd and PdZn catalysts showing structure change as reduction temperature increased.

8
4
D.J. Childers et al. / Journal of Catalysis 318 (2014) 75–84
surface atoms. At 550 °C, there is little change in the DRIFTS spec-
tra, but the EXAFS shows an increase in the number of PdZn bonds
with near complete formation of the PdZn alloy nanoparticles. For
each PdZn catalyst with complete surface coverage of the PdZn
intermetallic alloy, the isolated Pd atoms are active for structure-
insensitive reactions, but are inactive for structure-sensitive reac-
tions. The large change in catalytic performance is due to a geo-
metric effect, i.e., isolation of the active Pd atoms, rather than an
electronic effect due to alloy formation.
[7] F.M. Poveda, A. Sierraalta, J.L. Villaveces, F. Ruette, J. Mol. Catal. A – Chem. 106
1996) 109–118.
8] N. Schweitzer, H.L. Xin, E. Nikolla, J.T. Miller, S. Linic, Top. Catal. 53 (2010) 348–
56.
[9] D. Childers, A. Saha, N. Schweitzer, R.M. Rioux, J.T. Miller, R.J. Meyer, ACS Catal.
(2013) 2487–2496.
10] H.J. Wei, C. Gomez, R.J. Meyer, Top. Catal. 55 (2012) 313–321.
11] A. Sarkany, Z. Zsoldos, B. Furlong, J.W. Hightower, L. Guczi, J. Catal. 141 (1993)
566–582.
(
[
3
3
[
[
[
[
12] J.A. Rodriguez, J. Phys. Chem. 98 (1994) 5758–5764.
13] M. Nilsson, K. Jansson, P. Jozsa, L.J. Pettersson, Appl. Catal. B – Environ. 86
(
2009) 18–26.
[
14] N. Iwasa, T. Akazawa, S. Ohyama, K. Fujikawa, N. Takezawa, React. Kinet. Catal.
Lett. 55 (1995) 245–250.
5
. Conclusion
[
15] N. Iwasa, S. Masuda, N. Ogawa, N. Takezawa, Appl. Catal. A – Gen. 125 (1995)
145–157.
Bimetallic PdZn on silica catalysts can form an intermetallic
[16] N. Iwasa, S. Masuda, N. Takezawa, React. Kinet. Catal. Lett. 55 (1995) 349–353.
[
17] M. Lenarda, E. Moretti, L. Storaro, P. Patrono, F. Pinzari, E. Rodríguez-Castellón,
A. Jiménez-López, G. Busca, E. Finocchio, T. Montanari, R. Frattini, Appl. Catal.
A: Gen. 312 (2006) 220–228.
alloy with isolated active palladium atoms. At low reduction tem-
perature and with limited amounts of zinc near the Pd nanoparti-
cles, there are small regions of the surface with Pd ensembles and
catalytic selectivity similar to that of monometallic Pd. At higher
Zn loadings, or higher reduction temperatures, the PdZn nanopar-
ticle surface is fully covered by the PdZn alloy, which is inactive for
structure-sensitive reactions like neopentane hydrogenolysis/
isomerization and propane hydrogenolysis. The PdZn intermetallic
surface alloy, however, has high catalytic activity for structure-
insensitive reactions like propane dehydrogenation. The high pro-
pane dehydrogenation to propylene selectivity results from the
geometric isolation of the active Pd atoms, which have only metal-
lic Zn neighbors. From this work, catalytic reactions where there
are both structure-sensitive (hydrogenolysis) and structure-insen-
sitive (propane dehydrogenation) reactions and the structure-
insensitive reaction leads to the desired products, the ideal catalyst
will have single metallic sites with few ensemble sites.
[18] L. Bollmann, J.L. Ratts, A.M. Joshi, W.D. Williams, J. Pazmino, Y.V. Joshi, J.T.
Miller, A.J. Kropf, W.N. Delgass, F.H. Ribeiro, J. Catal. 257 (2008) 43–54.
[
19] K. Fottinger, J.A. van Bokhoven, M. Nachtegaal, G. Rupprechter, J. Phys. Chem.
Lett. 2 (2011) 428–433.
[20] O.P. Tkachenko, A.Y. Stakheev, L.M. Kustov, I.V. Mashkovsky, M. van den Berg,
W. Grunert, N.Y. Kozitsyna, Z.V. Dobrokhotova, V.I. Zhilov, S.E. Nefedov, M.N.
Vargaftik, I.I. Moiseev, Catal. Lett. 112 (2006) 155–161.
[
[
[
21] Z.-X. Chen, K.M. Neyman, A.B. Gordienko, N. Rösch, Phys. Rev. B 68 (2003)
075417.
22] N. Iwasa, T. Mayanagi, W. Nomura, M. Arai, N. Takezawa, Appl. Catal. A: Gen.
248 (2003) 153–160.
23] J.R. Regalbuto, in: Ryan Richards (Ed.), Surface and Nanomolecular Catalysis,
CRC Press, 2006, pp. 161–194.
[24] T. Komatsu, K. Inaba, T. Uezono, A. Onda, T. Yashima, Appl. Catal. A – Gen. 251
2003) 315–326.
(
[
[
25] D.C. Bazin, D.A. Sayers, J.J. Rehr, J. Phys. Chem. B 101 (1997) 11040–11050.
26] G. Agostini, A. Piovano, L. Bertinetti, R. Pellegrini, G. Leofanti, E. Groppo, C.
Lamberti, J. Phys. Chem. C 118 (2014) 4085–4094.
[
[
27] M.A. Vannice, C.C. Twu, J. Chem. Phys. 75 (1981) 5944–5948.
28] T. Lear, R. Marshall, J.A. Lopez-Sanchez, S.D. Jackson, T.M. Klapotke, M. Baumer,
G. Rupprechter, H.J. Freund, D. Lennon, J. Chem. Phys. 124 (2006).
Acknowledgments
[29] J.L. Lu, B.S. Fu, M.C. Kung, G.M. Xiao, J.W. Elam, H.H. Kung, P.C. Stair, Science
35 (2012) 1205–1208.
30] V. Lebarbier, R. Dagle, T. Conant, J.M. Vohs, A.K. Datye, Y. Wang, Catal. Lett. 122
2008) 223–227.
[31] C. Weilach, S.M. Kozlov, H.H. Holzapfel, K. Fottinger, K.M. Neyman, G.
Rupprechter, J. Phys. Chem. C 116 (2012) 18768–18778.
3
[
JTM and NS were supported as part of the Institute for Atom-
Efficient Chemical Transformations (IACT), an Energy Frontier
Research Center funded by the United States Department of
Energy, Office of Science, Office of Basic Energy Sciences. R.J.M.
and D.C. gratefully acknowledge funding for this work from the
National Science Foundation (CBET Grant No. 0747646). Partial
funding for DC was provided by the Chemical Sciences and Engi-
neering Division at Argonne National Laboratory and the Office of
the Vice Chancellor for Research at the University of Illinois at
Chicago. S.M.K.S. and R.M.R. acknowledges funding from the
Department of Energy, Office of Basic Energy Sciences, Chemical
Sciences, Geosciences, and Biosciences Division, Catalysis Sciences
Program under Grant No. DE-FG02-12ER16364. R.M.R. acknowl-
edges financial support provided through a 3M Non-Tenured Fac-
ulty Grant (NTFG). The STEM work was performed at the UIC
Research Resource Center. Use of the Advanced Photon Source
was supported by the U.S. Department of Energy, Office of Science,
Office of Basic Energy Sciences, under Contract No. DE-AC02-
(
[
[
32] M. Boudart, H.S. Hwang, J. Catal. 39 (1975) 44–52.
33] M.W. Tew, J.T. Miller, J.A. van Bokhoven, J. Phys. Chem. C 113 (2009) 15140–
15147.
[
[
34] S. Penner, B. Jenewein, H. Gabasch, B. Klotzer, D. Wang, A. Knop-Gericke, R.
Schlogl, K. Hayek, J. Catal. 241 (2006) 14–19.
35] W. Palczewska, in: Z. Paal, P.G. Menon, (Eds.), Hydrogen Effects in Catalysis.
Fundamentals and Practical Applications, N. Y.: Marcel Dekker Inc, 31 (1988)
373–398.
[
[
[
36] G. Rupprechter, M. Morkel, H.J. Freund, R. Hirschl, Surf. Sci. 554 (2004) 43–59.
37] N.A. Bhore, M.T. Klein, K.B. Bischoff, Ind. Eng. Chem. Res. 29 (1990) 313–316.
38] Y.C. Huang, W.P. Ding, Z.X. Chen, J. Chem. Phys. 133 (2010).
[39] A. Tamtogl, M. Kratzer, J. Killman, A. Winkler, J. Chem. Phys. 129 (2008).
[
40] H. Gabasch, A. Knop-Gericke, R. Schlögl, S. Penner, B. Jenewein, K. Hayek, B.
Klötzer, J. Phys. Chem. B 110 (2006) 11391–11398.
[
41] H. Liu, Y.M. Zhou, Y.W. Zhang, L.Y. Bai, M.H. Tang, Ind. Eng. Chem. Res. 48
(2009) 5598–5603.
[
[
[
42] F.M. Ashmawy, J. Appl. Chem. Biotechnol. 27 (1977) 137–142.
43] O.A. Barias, A. Holmen, E.A. Blekkan, J. Catal. 158 (1996) 1–12.
44] Y.W. Zhang, Y.M. Zhou, A.D. Qiu, Y. Wang, Y. Xu, P.C. Wu, Catal. Commun. 7
(2006) 860–866.
[
45] M.M. Bhasin, J.H. McCain, B.V. Vora, T. Imai, P.R. Pujado, Appl. Catal. A – Gen.
0
6CH11357. MRCAT operations are supported by the Department
221 (2001) 397–419.
of Energy and the MRCAT member institutions. The CleanCat Core
facility acknowledges funding from the Department of Energy (DE-
FG02-03ER15457 and DE-AC02-06CH11357) used for the purchase
of the DRIFTS system and the AMI-200, respectively.
[
46] M.W.Tew,H.Emerich,J.A.vanBokhoven,J.Phys.Chem.C115(2011)8457–8465.
[47] G.C. Bond, Appl. Catal. A – Gen. 191 (2000) 23–34.
[
[
48] J.H. Sinfelt, CRC Crit. Rev. Solid State Sci. 4 (1973) 311–332.
49] A.J. Sandee, T.J. Chintada, C. Groen, J.G. Donkervoort, R. Terorde, Chim. Oggi-
Chem. Today 31 (2013) 20–23.
[
50] R. Burch, Z. Paal, Appl. Catal. A – Gen. 114 (1994) 9–33.
References
[51] W. Juszczyk, D. Lomot, Z. Karpinski, J. Pielaszek, Catal. Lett. 31 (1995) 37–45.
[
52] Z. Karpinski, W. Juszczyk, J. Pielaszek, J. Chem. Soc.-Faraday Trans. I 83 (1987)
293–1305.
1
[
[
1] N. Musselwhite, G.A. Somorjai, Top. Catal. 56 (2013) 1277–1283.
2] O.Z. Didenko, G.R. Kosmambetova, P.E. Strizhak, J. Mol. Catal. A – Chem. 335
[
[
[
[
[
[
53] D.W. Blakely, G.A. Somorjai, J. Catal. 42 (1976) 181–196.
54] J.H. Sinfelt, J. Catal. 27 (1972) 468.
55] D.W. Flaherty, E. Iglesia, J. Am. Chem. Soc. 135 (2013) 18586–18599.
56] J.H. Sinfelt, J.L. Carter, D.J.C. Yates, J. Catal. 24 (1972) 283–296.
57] J. Anderson, N. Avery, J. Catal. 5 (1966) 18.
(
2011) 14–23.
[
[
[
3] M.S. Chen, D.W. Goodman, Catal. Today 111 (2006) 22–33.
4] D.W. Goodman, J. Catal. 216 (2003) 213–222.
5] A.A. Peterson, L.C. Grabow, T.P. Brennan, B.G. Shong, C.C. Ooi, D.M. Wu, C.W. Li,
A. Kushwaha, A.J. Medford, F. Mbuga, L. Li, J.K. Norskov, Top. Catal. 55 (2012)
58] T. Conant, A.M. Karim, V. Lebarbier, Y. Wang, F. Girgsdies, R. Schlogl, A. Datye, J.
Catal. 257 (2008) 64–70.
1
276–1282.
[
59] A. Karim, T. Conant, A. Datye, J. Catal. 243 (2006) 420–427.
[
6] M. Casarin, C. Maccato, A. Vittadini, Inorg. Chem. 37 (1998) 5482–5490.