Mechanism for the Direct Synthesis of H2O2 on Pd Clusters: Heterolytic Reaction Pathways at the Liquid-Solid Interface

Article abstract of DOI:10.1021/jacs.5b10669

Direct synthesis (H₂ + O₂ → H₂O₂) is a promising reaction for producing H₂O₂, which can replace chlorinated oxidants in industrial processes. The mechanism of this reaction and the reasons for the importance of seemingly unrelated factors (e.g., Pd cluster size and solvent pH) remain unclear despite significant research. We propose a mechanism for H₂O₂ formation on Pd clusters consistent with steady-state H₂O₂ and H₂O formation rates measured as functions of reactant pressures and temperature and the interpretations of proton concentration effects. H₂O₂ forms by sequential proton-electron transfer to O₂ and OOH surface intermediates, whereas H₂O forms by O-O bond rupture within OOH surface species. Direct synthesis, therefore, does not proceed by the Langmuir-Hinshelwood mechanism often invoked. Rather, H₂O₂ forms by heterolytic reaction pathways resembling the two-electron oxygen reduction reaction (ORR); however, the chemical potential of H₂ replaces an external electrical potential as the thermodynamic driving force. Activation enthalpies (ΔH^?) for H₂O formation increase by 14 kJ mol^-1 when Pd cluster diameters increase from 0.7 to 7 nm because changes in the electronic structure of Pd surface atoms decrease their propensity to cleave O-O bonds. ΔH^? values for H₂O₂ remain nearly constant because barriers for proton-electron transfer depend weakly on the coordinative saturation of Pd surface atoms. Collectively, these results provide a self-consistent mechanism, which clarifies many studies in which H₂O₂ rates and selectivities were shown to depend on the concentration of acid/halide additives and Pd cluster size. These findings will guide the rational design of selective catalysts for direct synthesis.

Full text of DOI:10.1021/jacs.5b10669

This is an open access article published under an ACS AuthorChoice License, which permits
copying and redistribution of the article or any adaptations for non-commercial purposes.
Article
pubs.acs.org/JACS
Mechanism for the Direct Synthesis of H O on Pd Clusters:
2
2
Heterolytic Reaction Pathways at the Liquid−Solid Interface
Neil M. Wilson and David W. Flaherty*
Department of Chemical and Biomolecular Engineering University of Illinois Urbana−Champaign, Urbana, Illinois 61801, United
States
*
S Supporting Information
ABSTRACT: Direct synthesis (H + O → H O ) is a
2
2
2
2
promising reaction for producing H O , which can replace
2
2
chlorinated oxidants in industrial processes. The mechanism of
this reaction and the reasons for the importance of seemingly
unrelated factors (e.g., Pd cluster size and solvent pH) remain
unclear despite significant research. We propose a mechanism
for H O formation on Pd clusters consistent with steady-state
2
2
H O and H O formation rates measured as functions of
2
2
2
reactant pressures and temperature and the interpretations of
proton concentration effects. H O forms by sequential
2
2
proton−electron transfer to O and OOH surface inter-
2
mediates, whereas H O forms by O−O bond rupture within
2
OOH surface species. Direct synthesis, therefore, does not
proceed by the Langmuir−Hinshelwood mechanism often invoked. Rather, H O forms by heterolytic reaction pathways
2
2
resembling the two-electron oxygen reduction reaction (ORR); however, the chemical potential of H replaces an external
2
⧧
−1
electrical potential as the thermodynamic driving force. Activation enthalpies (ΔH ) for H O formation increase by 14 kJ mol
when Pd cluster diameters increase from 0.7 to 7 nm because changes in the electronic structure of Pd surface atoms decrease
their propensity to cleave O−O bonds. ΔH values for H O remain nearly constant because barriers for proton−electron
transfer depend weakly on the coordinative saturation of Pd surface atoms. Collectively, these results provide a self-consistent
2
⧧
2
2
mechanism, which clarifies many studies in which H O rates and selectivities were shown to depend on the concentration of
2
2
acid/halide additives and Pd cluster size. These findings will guide the rational design of selective catalysts for direct synthesis.
4
,6
1
. INTRODUCTION
H O . Direct synthesis, however, suffers from low H O
2 2 2 2
selectivity because significant amounts of H O form by primary
and secondary reactions that irreversibly cleave the O−O bond
of surface intermediates chemisorbed on transition metal cluster
2
Hydrogen peroxide (H O ) is an effective and environmentally
benign alternative to chlorinated oxidizers (e.g., HClO and
ClO ) commonly used to selectively oxidize organic molecules
2
2
7
2
5,8−11
1
catalysts.
The greatest H O selectivities are achieved on
2 2
(e.g., propylene), bleach pulp and paper, and treat wastewater.
12,13
catalysts that cleave O−O bonds at low rates but add
hydrogen to O * at high rates. Pd catalysts have higher H O
2
The primary byproducts of the reactions between H O and
organic compounds are H O and O ; therefore, H O is an
appealing oxidant, which unlike chlorinated oxidizers does not
produce toxic chloroform, dioxins, and polychlorinated hydro-
carbons. However, the current market cost of H O is not
2
2
10
2
2
2
2
2
2
14
selectivities (60−75%) than Pt catalysts (5−20%) and give
H O formation rates nearly 100-fold greater than those for
2
2
1
5,16
2
,3
Au.
Yet Pd only gives such high selectivities in strongly acidic
2
2
14
solvents (e.g., ethanol, 0.12 M H SO ) containing halides (e.g.,
2
4
competitive with chlorinated oxidizers because the dominant
method for H O production (anthraquinone auto-oxidation
−
−
Cl , Br ), which modify the reactivity of the Pd surfaces in ways
2
2
14,17−20
1
not fully understood.
that give greater selectivities toward H O without the need for
solvents containing mineral acids and halides in order for direct
synthesis to be more viable.
Rational design of such catalysts require fundamental,
molecular-level knowledge of the mechanism for H O
2
formation on metal clusters; however, the direct synthesis
reaction is poorly understood. H O formation rates on Pd
Researchers must develop catalysts
(AO)) inherently requires energy-intensive separation and
concentration steps, and thus has high capital and operating
costs. For H O to be more widely used, costly separation steps
must be eliminated; therefore, its production cannot involve
anthraquinones. Rather, H O needs to be produced by a
2
2
2
2
1
,19
4
2
2
radically different catalytic chemistry.
2
The direct synthesis of H O (H + O → H O ) appears to be
2
2
2
2
2
2
4
,5
the most promising alternative to AO. Direct synthesis does
not involve organic substrates (i.e., anthraquinones), uses
greener solvents (e.g., H O, CH OH), and consequently
2
2
Received: October 12, 2015
2
3
requires fewer separation steps and much less energy to produce
©
XXXX American Chemical Society
A
DOI: 10.1021/jacs.5b10669
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
Table 1. Characterization Results for Pd Catalysts
temperature (K)
a
b
c
d
e
sample
Pd content (% wt)
oxidative treatment
reductive treatment
<dCHEM> (nm)
<dTEM> (nm)
0
3
7
3
.7 nm Pd−SiO2
nm Pd−SiO2
nm Pd−SiO2
.8 wt % Pd−SiO2
0.05
0.04
0.05
3.8
n/a
573, 4 h
573, 4 h
573, 4 h
573, 4 h
0.7
3
n/a
673, 4 h
973, 4 h
n/a
4 ± 1
7 ± 1
5 ± 1
7
n/a
a
Metal content of samples determined by inductively coupled plasma optical emission spectroscopy following digestion in a 2:1 volume ratio of
b
3
−1
c
3
−1
HNO (Fisher Scientific, 67−70 wt %) to HF (Fisher Scientific, 48−51 wt %) 100 cm min dry air. 20 kPa H , 81 kPa He, 100 cm min .
3
2
d
Average cluster diameter determined by CO chemisorption (<d >) was calculated using total CO uptake obtained over the range of 0.03−0.4
chem
3
9,40 e
kPa CO at 303 K with the assumption of hemispherical clusters and an adsorption stoichiometry of two Pd to one CO molecule.
averaged mean cluster diameter from transmission electron microscopy analysis using <dTEM> = (Σ n d /Σ n d ) of >100 clusters.
Surface-
s
3
i
2
i
i
i i i
clusters depend strongly on several seemingly disparate factors
kPa) and O (25−400 kPa), the protic or aprotic nature of the
2
21−23
including the pH of the solvent,
the presence (or absence)
and the size of the metal
The manner by which these factors collectively
determine H O selectivities and rates have not been resolved in
reaction solvent, and the concentration and identity of proton
donors over a range of relevant temperatures (273−305 K).
Formation rates of both H O and H O are sensitive to the H
2
2−25
of halide coadsorbates,
clusters.
2
6,27
2
2
2
2
2
pressure but show no dependence on the O pressure. H O
selectivities and formation rates require that H are present in the
solvent and increase slightly with the addition of H donors (e.g.,
2
2
2
2
+
+
previous studies. Protons (H ) have been suggested to affect
surface chemistry through indirect (i.e., electronic interactions)
or by directly participating in O reduction.
mineral acids (e.g., HCl,
HNO3) to solvents for direct synthesis increase the
concentrations of H and H O selectivities, but they also
introduce strongly binding anions (e.g., Cl , SO , PO4 , and
NO ) that modify the electronic structure of metal clusters
2
1
+
2
4,28
The additions of
H SO , HCl, or H CO ), whereas rates for H O formation are
2 4 2 3 2 2
2
2
1,29
14,21
22,24
3
H SO ,
H PO ,
and
immeasurable (>10 times lower) in aprotic solvents. Collec-
tively, these rate measurements are inconsistent with previously
proposed Langmuirian reaction mechanisms. Instead, these data
2
4
3
4
2
2
+
2
2
−
2−
3−
show that H
solvent molecules)electron (provided by heterolytic hydrogen
↔ 2H + 2e ) transfer to hydroperoxy surface
2
O is formed by kinetically relevant proton (from
4
2 2
−
3
+
−
directly. To the best of our knowledge, there are no direct
oxidation, H
intermediates. This mechanism resembles that for the two-
+
measurements that implicate H in the direct synthesis of H O .
2
2
37
electron oxygen reduction reaction (ORR); however, both
half-reactions occur on a single Pd cluster in the absence of an
The addition of halides to Pd clusters in acidic ethanol (0.12 M
21
H SO ) increases H O selectivities from 60 to 80%. These
2
4
2
2
electrical potential. Rather, gaseous H provides a chemical
changes have been attributed to halide adsorption either blocking
2
1
2,24,30,31
potential which drives O2 reduction. These mechanistic
conclusions agree also with measured activation enthalpies
sites that promote O−O bond scission
or reducing the
density of states near the Fermi level and consequently making
metal surfaces less reactive for O−O cleavage.
⧧
3
1−35
(
ΔH ) and the differences among their values on 0.7, 3, and 7 nm
H O2
2
⧧
Pd clusters. Values of ΔH for H O formation rates remain
selectivities do increase when halides and acids are added to
solvents for direct synthesis, yet these species are difficult to
remove from the product stream. As such, there is significant
interest in using other strategies to achieve high selectivities
2
2
−1
nearly constant (9−14 kJ mol ) as the mean size of the Pd
⧧
clusters increase; however, ΔH for H O formation increases
2
−1
noticeably (18 to 32 kJ mol ). These comparisons show that
processes that form H O (proton−electron transfer) are less
8
,19
including alloying Pd with other metals (e.g., Au)
controlling the size of the Pd clusters.
and
PdAu alloy
nanoclusters supported on carbon give H O selectivities
2
2
2
6,27
sensitive to changes in the electronic structure of Pd surfaces than
those that create H O (i.e., O−O bond dissociation), which is
2
2
2
consistent with prior research of the ORR and our mechanistic
conclusions. The data and interpretations suggest that greater
H O selectivities can be achieved when metal clusters, which
(
80%) much greater than those on Pd (42%) and H O
2 2
19
formation rates 100 times larger than those on Au. In addition,
studies over a narrow range of Pd cluster sizes (3.4−4.2 nm) on
SiO show that H O selectivity increases modestly with
2
2
heterolytically dissociate H and bind O tightly enough to
2
2
2
2
2
+
2
7
facilitate electron transfer, contact solvents with high H
increasing cluster size. The mechanisms by which alloying or
concentrations. These findings will guide the development of
changes in the cluster size affect H O selectivities remain
2
2
productive catalysts for the direct synthesis of H O and help to
2
2
unknown but are likely related to either ensemble effects (e.g.,
1
2,36
advance the use of H O for the production of commodity
2 2
decreased number of sites that rupture O−O bonds)
or
chemicals.
electronic effects (e.g., reduced electron back-donation from Pd
2
6,32
into π* orbitals of the O−O bond).
The uncertain
2
. MATERIALS AND METHODS
explanations for the effects of acids and cluster size as well as
the poorly understood connections between these factors and
the mechanism for H O formation impede rational design of
2.1. Synthesis of Supported Pd Catalysts. Pd catalysts were
prepared by strong electrostatic adsorption (SEA) of Pd precursors onto
2
2
3
8
improved catalysts for direct synthesis.
In this study, we propose a mechanism for direct synthesis of
silica, followed by oxidation and reduction to form metallic clusters. All
gases used were 99.999% pure and supplied by S. J. Smith Co. unless
otherwise stated. Silica (15 g, Sigma-Aldrich, Davisil 646, 35−60 mesh)
H O on Pd clusters that is consistent with our measurements
2
2
3
was added to 300 cm of deionized (DI) water (17.8 MΩ) followed by
and with previously published observations and that provides
guiding principles for the design of metal catalysts that will give
greater selectivities and formation rates for H O . Here, we
3
the addition of 30 cm of 14.5 M NH OH (Macron, 28−30 wt %) in
4
order to obtain a solution pH greater than 11. In a separate beaker, 17.4
3
2
2
mg of Pd(NH ) Cl (Sigma-Aldrich, ≥99.99%) was added to 15 cm of
DI water; subsequently, this solution was added to the basic solution
containing silica. The resulting solution was stirred for 3 h and then
3
4
2
report steady-state H O and H O formation rates on Pd clusters
2
2
2
(
0.7−7 nm) and their dependence on the pressure of H (5−400
2
B
DOI: 10.1021/jacs.5b10669
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
vacuum filtered to recover the solids. The wet solids were rinsed with an
3
additional 500 cm of DI water, vacuum filtered, and then allowed to dry.
−
1
The dried solids were heated to 573 K at 3 K min and then held at 5733
K for 4 h in a flowing mixture of 20 kPa H and 81 kPa He (100 cm
2
−
1
min ) with the intent to reduce the Pd to metallic nanoclusters.
Oxidative treatments in flowing dry air (Table 1) were used after the
initial reduction to increase the average size of the Pd clusters. These
were followed by a second reductive treatment (573 K, 4 h, 20 kPa H ,
2
3
−1
8
1 kPa He, 100 cm min ) to form metallic Pd nanoclusters. After
reduction, all samples were passivated at ambient temperature by
3
−1
exposure to a flowing (500 cm min ) mixture of 4 kPa O and 97 kPa
2
He for 0.5 h.
A 3.8 wt % Pd−SiO catalyst was prepared by incipient wetness
2
3
impregnation of 200 mg of washed SiO with 0.21 cm of 0.45 M
2
K PdCl (Sigma-Aldrich, 99.99%). The catalyst was then dried under
2
4
ambient conditions for 4 h before reduction and passivation treatments
identical to those described above.
2.2. Mean Pd Cluster Sizes and Metal Loading. The average
diameters of clean Pd clusters were estimated from the volumetric
uptake of CO (<dCHEM>) and from cluster size distributions obtained by
transmission electron microscopy (TEM). Volumetric CO uptakes were
measured at 303 K using a custom-built borosilicate-glass chemisorption
cell. Loaded catalysts were degassed by heating to 323 K at 3 K min−1
Figure 1. Representative TEM image of the 7 nm Pd−SiO catalyst with
2
cluster size distribution (inset). More than 100 clusters were measured
and holding for 12 h while evacuating the system with a turbo pump
to determine the value of <dTEM>, the surface area averaged diameter.
(
Edwards, EXT250), after which the sample was cooled and the cell was
3
checked for leaks. Samples were then treated in flowing H (50 cm
2
−
1
−1
min ) while heating to 523 K at 3 K min and holding at 523 K for 0.5
h. The sample was then evacuated at 523 K for 0.5 h using the turbo
pump, and the cell was allowed to cool to 303 K under dynamic vacuum.
The total amount of CO (S.J. Smith, 99.99%) adsorbed onto the Pd
clusters within each sample was determined by measuring volumetric
uptakes between 0.03 and 0.4 kPa and extrapolating the linear portion of
(
Neslab, Endocal). The temperature of the catalyst bed was monitored
directly using a K-type thermocouple contained within the cooling
jacket and in firm contact with the stainless-steel wall surrounding the
catalyst bed. Certified reactant gas mixtures (25% H , balance N , and
2
2
5
% O , balance N ; or 5% H , balance N , and 25% O , balance N ) and
2
2 2 2 2 2 2
CO (S.J. Smith, 99.99%) were introduced to the system upstream of the
the isotherm to zero pressure. The number of exposed Pd atoms (Pd )
s
catalyst bed using digital mass-flow controllers (MFC; Porter, 601
3
9,40
was determined by assuming a ratio of two Pd to one CO molecule.
s
series). Gas mixtures were chosen to contain sufficient N to ensure that
2
The mean cluster diameter for each sample was estimated from Pd and
41
s
flammable mixtures of H and O never formed. The reaction solvent
2
2
the total Pd loading by assuming a hemispherical geometry and an
atomic radius of 0.14 nm for Pd.
The distribution of cluster diameters was measured by bright-field
(
aqueous CH OH (Macron, ≥ 99.8%) solutions, 20−100% v/v) were
3
introduced to the system downstream of the MFCs using a high-
performance liquid chromatography (HPLC) pump (Waters, 515). The
molecular species used to control the pH of the solution (H SO
TEM imaging (JEOL 2010 LaB ) of more than 100 clusters. Samples
6
2
4
were prepared by grinding the catalyst to a fine powder (<200 mesh),
which was then suspended in ethanol (Decon Laboratories Inc., 100%)
by gentle stirring and loaded onto a Cu holey-carbon TEM grid (200
mesh, Ted Pella Inc.). The wet grids were dried under ambient
conditions for at least 4 h. The surface area normalized average cluster
diameter (<dTEM>) for each catalyst was calculated using
(
Macron, 95−98%), H PO (Fisher Scientific, 88.9%), HCl (Macron,
3
4
3
6.5−38%), and NaHCO (Fisher Scientific, > 99.7%)) were combined
3
with the solvent within the HPLC pump reservoir, and the solvent pH
was measured with a digital pH meter (Omega, PHH22). The gas and
liquid streams contacted and mixed within 120 cm of 1.6 mm I.D. tubing
before flowing through the catalyst bed in an upflow configuration. The
reactor pressure (0.1−3.1 MPa) was controlled using an electronic
pressure reducer (EPR; Proportion Air, QB1S) and a back-pressure
regulator (BPR; Equilibar, EB1LF1-SS316) equipped with a Kapton
diaphragm. The pressure at the inlet of the reactor was measured with a
digital pressure gauge (Omega, DPG8001-1K), and the pressure at the
reactor outlet was monitored using the EPR.
3
i
∑i nid
<
dTEM >=
∑i nidi2
(1)
where n is the number of clusters with diameter d . Figure 1 shows a
i
i
representative TEM image of the 7 nm Pd clusters, with the cluster
diameter distribution as inset, and TEM images for the other Pd catalyst
are given in Figure S1.1. The diameters of the clusters formed using only
reductive treatments were too small (<1 nm) to be measured accurately
by TEM. The values of <dTEM> and <dCHEM> agree closely for the 3 and
The collection and characterization of the liquid and gaseous effluent
streams was automated and operated continuously. The liquid and
gaseous portions of the reactor effluent were separated using an acrylic
gas−liquid separator (GLS) located downstream of the BPR. The gas
stream passed through a check valve (Swagelok, 1.7 kPa cracking
pressure) before flowing into the automated gas sampling valve for a gas
chromatograph (Agilent, 7890). The components of the gas stream
were separated using a packed column (Sigma-Aldrich, 3 m length × 2.1
mm ID, Molecular Sieve 5A) and their concentrations were analyzed
using a thermal conductivity detector. Ar was used as both the carrier
and reference gas during gas chromatography. Liquid samples were
withdrawn from the GLS using an automated valve (Vici, 10 port valve),
which was configured to inject 0.4 cm of the effluent liquid and also 1
cm of a colorimetric indicator solution (12 mM neocuproine, Sigma-
Aldrich, ≥ 98%; 8.3 mM CuSO , Fisher Scientific, ≥ 98.6%; 25% v/v
ethanol, Decon Laboratories Inc., 100%) into glass vials held within an
automated fraction collector (Biorad, 2110). The concentration of H O
7
nm Pd clusters. The Pd content of each sample was measured by
inductively coupled plasma optical emission spectroscopy (PerkinElm-
er, Optima 2000DV). The characterization results for all prepared
catalysts are shown in Table 1.
2
.3. Steady-State H O and H O Formation Rate Measure-
2 2 2
ments. Steady-state reaction rates were measured using a packed-bed,
continuous-flow reactor (0.32 cm I.D., 11 cm long) contained within a
stainless-steel cooling jacket (3.8 cm O.D.). The reactor was loaded with
3
1
50−300 mg of Pd−SiO , which was held between plugs of glass wool
2
3
(
6−10 mg), each supported by borosilicate glass rods (2 mm diameter).
These, in turn, were secured between silver-coated fritted VCR gaskets
Swagelok, SS-4-VCR-2-60M), which were also used to seal the reactor
4
4
2
(
to the tubing of the gas−liquid handling manifold. The temperature of
the reactor (273−305 K) was controlled by flowing aqueous ethylene
glycol through the cooling jacket using a refrigerated recirculating bath
2
2
in each vial was determined from its absorbance at 454 nm, measured
using a spectrophotometer (Spectronic, 20 Genesys). The absorbance
C
DOI: 10.1021/jacs.5b10669
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
values were calibrated using solutions of known colorimetric indicator
solution with known H O concentrations.
2
2
Primary rates of H O2 and H O formation were measured by
2
2
avoiding artifacts introduced by mass transport limitations and
uncertainties from secondary H O decomposition. To ensure that
2
2
4
3
the Madon−Boudart criterion was satisfied, H2 conversion was
−1
measured as a function of gas residence time ((mol Pd s) (mol H ) )
s
2
on catalysts with different metal loadings (0.025−0.5 wt %) at high H2
pressure (400 kPa H ) where H O and H O formation rates were the
2
2
2
2
greatest (Figure S2.1). The H conversion at a given gas residence time
2
did not depend on metal loading for catalysts with Pd contents ≤0.05 wt
%
, indicating that intrapellet mass transport limitations did not exist in
this range of metal loading. H O formation rates were measured as a
2
2
function of liquid residence time (2.1−4.2 s) over a range of
temperatures (266−305 K) to examine the significance of secondary
decomposition of H O (Figure S2.2). The H O formation rate
2
2
2
2
depended strongly on the liquid residence time at temperatures greater
than 281 K, indicating significant secondary decomposition of H O .
2
2
Mass transport restrictions and secondary decomposition of H O were
2
2
avoided throughout this study by using 0.05 wt % Pd catalyst, a solvent
3
−1
flow rate of 10 cm min , and a reaction temperature of 277 K unless
otherwise stated. The catalyst was changed after each data set, typically
on a daily basis, which prevented slow leaching of the Pd from affecting
the rate measurements significantly. After the catalyst was loaded, it was
allowed to stabilize for at least 6 h under reaction conditions prior to data
collection. Data was collected for at least 1.5 h under each set of
conditions to acquire multiple rate measurements at steady state.
Reported rates comprise the mean values of at least six measurements,
and error bars represent the standard deviation of these measurements.
2
.4. Rate Measurements in Semibatch Reactors. H O
2 2
formation rates were measured in protic and aprotic solvents using a
3
6
0 cm borosilicate glass semibatch reactor (43.5 mm I.D.). The reactant
gases entered through a fine glass frit at the bottom of the reactor and
were highly dispersed as they rose up through a turbulent suspension of
the catalyst and the solvent. The outlet at the top of the reactor was
made to be 3.5 mm I.D. to decrease the rate of solvent loss during each
3
experiment. The reactor was filled with 40 cm of either pure DI water
(
9
Figure 2. (a) H O and (b) H O turnover rates as functions of H
2
2
2
2
17.8 MΩ), methanol (Macron, ≥99.8%), acetonitrile (Fisher Scientific,
9.9%), dimethyl sulfoxide (Fisher Scientific, 99.9%), or propylene
pressure at 60 kPa O (black, ■) and O pressure at 60 kPa H (red, ●)
2
2
2
3
−1
on 0.7 nm Pd clusters (277 K, 10 cm min CH OH). Lines are
3
carbonate (Sigma-Aldrich, 99.7%), which were all used as received. The
reactant gas mixture (4.2 kPa H , 4.2 kPa O , balance N ) was
intended to guide the eye.
2
2
2
3
−1
introduced at a total flow rate of 60 cm min and controlled using
variable area rotameters (Omega, FL-1461-S). The 3.8 wt % Pd−SiO2
catalyst (20 mg) was added to the reactor to initiate the reaction after the
2
1
intermediates such as O ** (where ** and * denote η and η
2
adsorbed Pd surface intermediates, respectively) or hydroperoxy
3
gas flow rate became stable. Liquid samples (0.4 cm ) were taken every
(
OOH**) over the full range of (O ) tested here (25−400 kPa).
2
2
−5 min in order to measure the concentration of H O by the
2
2
This interpretation is consistent with high heats of adsorption for
colorimetric titration procedure described earlier (section 2.3), and the
reaction was conducted at ambient temperature (∼295 K). The SEA-
prepared Pd catalysts were not used in the semibatch reactor because the
−1 10,44
molecular oxygen onto H O-saturated Pd (48−75 kJ mol )
2
−1
45,46
and Pt (39−47 kJ mol )
surfaces (i.e., 2H O* + O → O **
2 2 2
−1
−2
+ 2H O) even within liquid water. Presumably, reactive sites on
long Pd−liquid contact times (46 s g
L
compared to 1.5 × 10 s gPd
in the plugged-flow reactor) caused significant amounts of Pd to
leach into the solution. Control experiments showed that these
dissolved Pd species interact with and perhaps oxidize the colorimetric
titrant so that the H O concentration could not be measured accurately.
2
Pd
−1
Pd surfaces would be covered by H O* and rates would depend
2
L
on the value of (O ) at sufficiently low (O ). Calculated changes
2
2
in the Gibbs free energy for H O* displacement by O (−39 kJ
2
2
−1
45
2
2
mol ) suggest, however, that metal surfaces are covered by
Therefore, the semibatch experiments were conducted using the 3.8 wt
Pd−SiO2 catalyst prepared by incipient wetness impregnation
section 2.1), which contains trace amounts of Cl that may reduce
leaching.
O ** even at pressures orders of magnitude less than the lowest
2
%
(
(O ) tested here (25 kPa O ; Supporting Information section
2
2
S3).
Figure 2 shows that the formation rates of H O and H O
2
2
2
depend strongly on H pressure ((H )), even though they do not
2
2
3
. RESULTS AND DISCUSSION
.1. Dependence of Rates on Reactant Pressures and
change with (O ). The H O formation rate initially increases
2
2
2
linearly with (H ) from 5 to 150 kPa (Figure 2a) and reaches a
3
2
1/2
half-order dependence on (H ) (i.e., rH2 ∼ (H ) ) at partial
Mechanistic Interpretation. Figure 2 shows turnover rates for
2
O
2
2
the formation of H O (Figure 2a) and H O (Figure 2b) from
pressures near 400 kPa. Under these same conditions, H O
2
2
2
2
turnover rates are proportional to (H2)1 between 5 and 150 kPa
and become constant at greater pressures (Figure 2b). These
data (Figure 2a) are consistent with the previously reported first-
order dependence of H O formation rates on (H ), measured
/2
the reduction of O on 0.7 nm Pd clusters as a function of H2
2
pressure (5−400 kPa H , 60 kPa O , 277 K) and O pressure
2
2
2
(
25−400 kPa O , 60 kPa H , 277 K). The H O and H O
2
2
2
2
2
formation rates do not depend on O pressure ((O )), likely
2
2
2
2
2
21
because the reactive surface sites are saturated with O -derived
from 5 to 24 kPa H on silica supported Pd clusters, yet that
2
2
D
DOI: 10.1021/jacs.5b10669
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
work did not observe changes in the (H ) dependence over the
Scheme 1. Proposed Series of Elementary Steps for H O and
2
2
2
small range of pressures tested. Mechanistic interpretations of
those results could not disprove or prove any of the proposed
H O Formation during Direct Synthesis on Supported Pd
Clusters
2
a
10,11,28,47
mechanisms for H O formation,
in part because the
2
2
range of H pressures was limited but also because formation
2
rates for H O were not reported.
2
H O and H O formation rates depend on the prevalent H
2
2
2
2
and O pressures because changes in the partial pressure of the
2
reactants change either the concentrations of species involved in
kinetically relevant steps, the identity of the most abundant
32
reactive intermediates (MARI) or both. The data shown in
Figure 2 implies that the reactive sites are saturated with O2-
derived intermediates whose composition changes as (H2)
increases. Oxygen isotope labeling experiments show that H O
2
2
7
cannot be formed once the O−O bond of molecular O cleaves;
2
therefore, the rate data (Figure 2), together with this observation,
suggest that the MARI may be a species that contains an intact
O−O bond (i.e., O **, OOH**, or H O **). The saturation of
2
2
2
metal surfaces by such intermediates is consistent with in situ
infrared spectroscopy studies showing large coverages of O **
2
4
8
on Pt during the ORR at low potentials and interpretations of
49−51
the effects of anions and bromide on ORR rates.
In addition,
the dependence of both H O and H O formation rates on (H )
2
2
2
2
shows that H -derived intermediates participate in the kinetically
2
relevant steps that form each product. Finally, the simultaneous
changes in the dependence of H O and H O formation rates on
2
2
2
(
H ) at ∼150 kPa H suggest that the average composition of the
2
2
MARI increases by a single H atom.
Langmuir−Hinshelwood mechanisms proposed for H O
2
2
1
0,11,47
formation,
which involve direct reduction of O ** by
2
sequential reactions with H* atoms bound to adjacent surface
sites, fail to describe these data (Figure 2). Specifically, such
a
*
denotes an empty site, X* represents an adsorbate bound to a
2
single Pd atom, X** signifies an intermediate adsorbed in an η
configuration, the equilibrium arrows (⇄) with a circle indicate that an
mechanisms predict H O formation rates that are independent
2
2
of (H ), depend inversely on (O ), or have a first-order
elementary step is quasi-equilibrated, and k is the rate constant for
elementary step x.
2
2
x
dependence on (O ) depending on the identity of the MARI
2
(derivation, Supporting Information section S4), which are
inconsistent with measured reaction rates (Figure 2a). A two-site
55,56
near saturation coverages
and facile decomposition of H O
2 2
Langmuirian model, which involves distinct sites for O and H
5,12
2
on Pd surfaces, which suggest that these species readily desorb
and adsorb at rates greater than measured turnover rates (1 ×
0 −7 × 10 (Pd ·s) ). Subsequently, O ** undergoes either
adsorption could describe the data in Figure 2 (Supporting
Information section S5); however, it would require that the sites
capable of binding H* atoms remain empty over the entire range
−3
−1
−1
1
s
2
quasi-equilibrated proton−electron transfer (Scheme 1, step 4)
to form hydroperoxy (OOH**) or irreversibly cleaves the O−O
bond (Scheme 1, step 7) to form chemisorbed oxygen atoms
(O*). Adsorbed OOH** then reacts by kinetically relevant
proton−electron transfer (Scheme 1, step 5) to form H O ** or
of H pressures (5−400 kPa). This assumption seems unlikely
2
−
1
because the heat of adsorption of a H atom on Pd (72 kJ mol )
from calculated energies of H adsorption on Pd at low H
coverage and the homolytic bond dissociation energy of H is
5
2
53
2
2
2
significantly higher than that for H O on Pd(111) (39 kJ
2
−1
52
dissociates by irreversible O−O rupture (Scheme 1, step 8).
mol ). Additional evidence, which shows that H O forms
2
2
Finally, H O ** desorbs to the liquid-phase (Scheme 1, step 6)
2
2
only in protic solvents (section 3.2), further suggests that this
or dissociates (Scheme 1, step 12) to form two hydroxyls (OH*).
reaction occurs by a non-Langmuirian mechanism. Thus, H O2
2
Following irreversible O−O bond rupture in O ** (Scheme 1,
2
appears to form by pathways that involve reactive intermediates
not present on the surface of Pd clusters.
step 7), OOH** (Scheme 1, step 8), or H O ** (Scheme 1, step
2
2
1
2), O* and OH* species hydrogenate (Scheme 1, steps 9 and
Scheme 1 shows a proposed series of elementary steps that are
consistent with the effects of H and O pressure on product
1
0) to form H O*, which then desorbs (Scheme 1, step 11).
2
2
2
Scheme 1 suggests net H O formation rates (r O2) that
increase with the number of hydroperoxy surface intermediates
([OOH**]) and with the concentration of H in solution and
decrease with the number of adsorbed H O2 molecules
2
2
H
2
formation rates on Pd clusters (Figure 2) and on other transition
2
1,47,54
metal clusters
within liquid solvents. These steps involve
+
+
bimolecular reactions between liquid-phase species, namely H ,
and chemisorbed surface intermediates. Dissociative H2
adsorption (Scheme 1, step 1) and heterolytic H* oxidation
Scheme 1, step 2) steps as well as molecular adsorption of O2
Scheme 1, step 3) and desorption of H O (Scheme 1, step 6)
2
(
[H O **]).
2 2
(
(
+
−
r
= k [OOH**][H ][e ] − k [H O **]
H O
5
12
2
2
(2)
2
2
2
2
and H O (Scheme 1, step 11) are assumed to be quasi-
where k is the rate constant for proton−electron transfer to
+ + −
2
5
equilibrated under all conditions. These assumptions are based
on free energies for O and H adsorption that become negligible
OOH**, [H ] is the concentration of H in the solvent, [e ] is
−
the number of free e on the cluster (provided by heterolytic
2
2
E
DOI: 10.1021/jacs.5b10669
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
rH2O2
oxidation of H ; Scheme 1, steps 1 and 2), and k is the rate
k K K 1/2(H2)1/2
5 2 1
2
12
=
constant for O−O rupture within H O **. Applying the pseudo-
2
2
[L]
(8)
steady-state hypothesis (PSSH) to [OOH**] and other reactive
This expression agrees with H O formation rates that remain
independent of (O ) and which increase as (H ) in the range
intermediates involved in H O formation and decomposition
2
2
2
2
1
/2
gives the rate equation
2
2
of 200−400 kPa H (Figure 2a). The mechanism described for
2
⎛
⎞
k12
K6
H O formation proposed in Scheme 1 and the derived rate
2
2
2
r
= ⎜k K K K K (O )(H ) −
(H O )⎟[**]
H O
5
4
3
2
1
2
2
2
2
2
2
expression (eq 6) are therefore consistent with measured H O
formation rates over the full range of conditions tested.
Scheme 1 shows that the rate of H O formation (r ) equals
2
2
⎝
⎠
(3)
in which k and K are rate and equilibrium constants respectively
x
x
2 H O
2
12
for each step x, (H O ) is the liquid-phase concentration of
2
2
the summed rates of irreversible O−O bond scission within all
H O , and [**] is the number of unoccupied sites available to
2
2
dioxygen-containing surface species (i.e., O **, OOH**, and
2
2
bind species containing dioxygen (i.e., −O−O−) in η
configurations.
summing the numbers of likely surface intermediates
H O **) at steady-state. The rate of O−O bond scission in these
5
7−59
2
2
An expression for [**] is given from first
three species is given by
r
= k [O **] + k [OOH**] + k [H O **]
H O
7
2
8
12
2
2
(9)
2
[
L] = [**] + [O **] + [OOH**] + [H O **]
(4)
2
2
2
which takes a new form after accounting for the quasi-
equilibrated steps in Scheme 1
where [L] and [O **] are the number of available surface sites
2
and O ** intermediates, respectively. Formation rates of H O
2
2
2
and H O do not depend on (O ) over the pressure range tested
⎛
⎝
2
2
1
/2
1/2
r
= ⎜k K (O ) + k K K K K (H2) (O2)
(
Figure 2), indicating that the MARI contains oxygen and thus
H O
7
3
2
8 4 3 2 1
2
that H* coverages are insignificant. Moreover, coverages of
dissociation products (i.e., O*, OH*) appear to be low on active
sites because H O and H O rate expressions derived after
⎞
k12
K6
+
(H O )⎟[**]
2
2
2
2
2
⎠
(10)
assuming a MARI of either O* or OH* (Supporting Information
section S6) were inconsistent with measured rate dependencies
Implementing the site balance yields the rate equation for H O
formation:
2
on H and O pressures (Figure 2). Equation 4 can be re-
2
2
expressed in terms of the rate and equilibrium constants as well as
reactant pressures and concentrations:
1
1
/2
1/2
k12
K6
1
K6
k K (O ) + k K K K K (H2) (O ) + (H O )
r
7
3
2
8
4
3
2
2
2
2
H O
2
=
1
1
/2
1/2
[
L] = [**] + K (O )[**] + K K K K /2(H2)1/2(O )[**]
1
[L]
1 + K (O ) + K K K K (H2) (O ) + (H O )
3
2
4
3
2
2
2
2
3
2
4
3
2
1
2
1
(11)
+
(H O )[**]
2
2
K6
(5)
The rate of H O formation by H O decomposition (Scheme 1,
2 2 2
step 12; (k /k )(H O )) is negligible in comparison to H O
12
6
2
2
2
The combination of eqs 3 and 5 yields the rate expression for
H O formation:
formation by O ** and OOH** decomposition, as already
2
2
2
discussed. Density functional theory calculations show that O−O
−1
2
k
K
1/2
12
bond dissociation barriers are ≥40 kJ mol lower in OOH**
k K K K K (H )(O ) − (H O )
r
5
4
3
2
1
2
2
2
2
H O
6
10,11,44
2
2
=
than for O
2
** on Pd surfaces,
which agrees qualitatively
1
/2
1
[
L]
1 + K (O ) + K K K K (H2) (O ) + (H O )
K6
3
2
4
3
2
1
2
2
2
with homolytic bond dissociation energies (BDE) that are 230 kJ
mol greater for the O−O bond in O (g) than for that in ·
OOH(g) radicals. Together these results imply that rate
−1
(6)
2
60
The complete derivation of this rate expression and alternative
rate expression that utilize other assumptions (and do not match
the experimental data) are provided in Supporting Information
section S6. H O selectivities do not change significantly with H
2
constants and likely rates for OOH** dissociation may be much
higher than those for O ** on Pd clusters. The combination of
2
these assumptions yields the following simplified rate expression
2
2
1
conversion (or residence time) under the conditions used here
<10% H conversion, Figure S2.3), which shows that the rate of
r
H O
k K K K K /2(H2)1/2(O2)
2
8 4 3 2 1
=
(
1 + K (O ) + K K K K 1/2(H2)1/2(O ) + (H O )
1
K6
2
[L]
3
2
4
3
2
1
2
2
2
H O decomposition (Scheme 1, step 12, (k /k )(H O )) is
2
2
12
6
2
2
(12)
negligible compared to the rate of H O formation (Scheme 1,
2
2
step 5). Thus, H O decomposition can be neglected, and the
At low (H ) (<150 kPa H ), O ** is the MARI, and eq 12
2 2 2
becomes
2
2
forward rate for H O formation (r ) is equal to the net rate
2
2
5
(
rH2 2). Active sites become saturated with molecular oxygen
O
r
H O
2
1/2
k K K K
(H2)
8 4 2 1
1/2
=
(i.e., O ** is the MARI) in the limit of low (H ), and the rate
2
2
[
L]
(13)
equation simplifies to
and as seen for H O formation rates, OOH** is the MARI at
r
2
2
H O
2
2
2
=
k K K K (H )
higher (H ) (200−400 kPa H ), giving
5
4
2
1
2
2
2
[
L]
(7)
r
H O
2
=
k8
Equation 7 is consistent with H O formation rates that increase
2
2
[
L]
(14)
in proportion to (H ) at the lowest values of (H ) (Figure 2a, 5−
2
2
1
4
50 kPa H ), and which do not depend on (O ) (Figure 2a, 25−
These simplified rate expressions (i.e., eqs 13 and 14) are
2
2
00 kPa). The coverage of OOH** increases with (H ) such that
consistent with H O formation rates that are constant for all (O )
2
2
2
OOH** becomes the MARI when (H ) is greater than 150 kPa,
and that transition from a half- to zero-order dependence on
2
which causes eq 6 to take the form
(H ) at ∼150 kPa (Figure 2b). These comparisons show that O−
2
F
DOI: 10.1021/jacs.5b10669
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
2
1,63,64
O bond rupture occurs primarily within OOH** surface
ratios;
however, those studies did not discuss precautions
intermediates, which agrees also with mechanistic predictions
taken to measure primary formation rates for H O and H O and
may include contributions from H O ** decomposition, which
2 2
would complicate analysis.
In contrast to the previously proposed Langmuir−Hinshel-
wood mechanisms,
2 2 2
61,62
for the ORR reaction on Pt surfaces.
The rate equations for H O (eq 6) and H O (eq 12)
2
2
2
formation suggest that the selectivity towards H O will increase
2
2
10,11,47
with (H ) while remaining constant for all (O ) values. These
the findings of this work implicate a
2
2
expectations are shown clearly when H O selectivity is
proton−electron transfer mechanism (Scheme 1, steps 4 and 5)
2
2
expressed as the ratio of H O and H O formation rates (χ):
as the primary pathway for H O formation. Notably, this
2
2
2
2
2
mechanism resembles proposals for the two-electron ORR to
k K K /2(H2)1/2
1
r
46,65
H O
5 2 1
2
2
H O ,
however, without an applied electrical potential.
χ =
=
2
2
r
k8
Instead, the catalytic direct synthesis of H O on silica-supported
H O
(15)
2
2
2
Pd clusters is driven by the chemical potential of the supplied
The value of χ does not depend on the identity of the MARI,
H (g). This mechanism requires that each Pd cluster catalyzes
2
because pathways for H O and H O formation occur on the
2
2
2
both electrochemical half-reactions in order to conserve charge
(Scheme 2). To further confirm the role of proton and electron
1/2
same active sites. Figure 3 shows that χ values increase as (H2)
Scheme 2. Pd Clusters Catalyze Both Heterolytic H2
Oxidation and O Reduction Steps in Order to Form H O at
2
2
2
Steady-State while Conserving Charge
transfer pathways in the direct synthesis of H O , we next
2
2
+
describe how H O formation rates require that H are present in
2
2
the solvent.
.2. Effects of H Concentration on H O and H O
+
3
2
2
2
Formation Rates. The proposed mechanism for H O
2
2
formation (Schemes 1 and 2) contains steps identical to
46,65
charge-transfer processes for the ORR.
These steps are
well-accepted when an electrical potential is applied between two
metal electrodes separated by an electrolyte; however, they are
not commonly invoked to describe reactions at the surface of
nanoclusters supported on insulators (e.g., Pd−SiO ). Figure 4
2
shows that increases in the H O concentration over time within
2
2
a semibatch reactor are much greater in protic solvents (e.g.,
methanol and water) than in aprotic solvents (e.g., acetonitrile,
dimethyl sulfoxide, and propylene carbonate) (4.2 kPa H , 4.2
2
kPa O , 295 K, 3.8 wt % Pd−SiO ). Primary formation rates for
2
2
H O are calculated from the change in H O concentration with
2
2
2
2
time at the early stages of each experiment (0−0.17 h), and the
Figure 3. Ratio of the formation of H O to that for H O (χ) as a
2
2
2
results are shown in Table 2 along with the pK values and
function of (a) H pressure at 60 kPa O and (b) O pressure at 60 kPa
a
2
2
2
H on 0.7 nm (black ■), 3 nm (red ●), and 7 nm (blue ▲) Pd clusters
dielectric constant for each solvent. These data (Figure 4 and
Table 2) show that the H O turnover rates in protic solvents are
2
3
−1
(
277 K, 10 cm min CH OH). Lines are intended to guide the eye.
3
2
2
3
at least 10 times greater than those in aprotic solvents (whose
H O turnover rates fall below the detection limit of the
(
Figure 3a) and remain constant with changing (O ) (Figure 3b)
2
2
2
−6
−1
on 0.7 nm Pd clusters. These comparisons show that the
expression for χ is consistent with the rate data (Figure 2). The
interpretation of these results shows that O−O bonds cleave at
negligible rates in O ** and H O ** intermediates and that both
experiment, <1.7 × 10 (mol H O ) (mol Pd s) ). Turnover
2 2 s
rates depend on the protic−aprotic nature of the solvent but do
not correlate with the dielectric constants of the solvents (Table
2), which suggests that H O formation is not sensitive to solvent
2
2
2
2
2
H O and H O form on the same sites. The values of χ increase
polarity. These data (Table 2) are completely consistent with our
proposed mechanism (Schemes 1 and 2); moreover, these
results cannot be explained by mechanisms for direct synthesis
that only involve elementary steps describing reactions between
2
2
2
by a factor of 2 on 0.7 nm Pd clusters as (H ) increases over the
2
range 5−400 kPa (Figure 3a) and demonstrate that significant
increases in selectivity can be achieved by increasing (H ). These
2
findings (Figure 3) differ from previous studies that reported
coadsorbed surface intermediates (e.g., O * + H* → OOH* +
2
+
H O selectivities that decreased with increasing H /O reactant
*), which are not influenced by H . We next discuss rate
2
2
2
2
G
DOI: 10.1021/jacs.5b10669
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
Figure 4. H O concentrations as functions of time during direct
2
2
synthesis in a well-mixed semibatch reactor using protic (methanol
black ■) and water (red ●)) or aprotic (dimethyl sulfoxide (green ▲),
acetonitrile (blue ▼), and propylene carbonate (magenta ⧫)) solvents
(
3
(
4.2 kPa H , 4.2 kPa O , 40 cm solvent, 20 mg of 3.8 wt % Pd−SiO , 295
2
2
2
K). Lines are intended to guide the eye.
+
measurements conducted as a function of H concentration
+
using multiple H donors and acceptors and interpret these data
using the previously derived rate expressions (eqs 6 and 12) to
clarify the role of mineral acids commonly used to increase H O
2
2
2
1,22
selectivities during direct synthesis.
Figure 5 shows that steady-state H O formation rates (Figure
a) and χ (Figure 5b) depend weakly on the H concentration,
2
2
+
66
Figure 5. (a) H
O turnover rates and (b) ratio of the formation of H O
2 2 2 2
5
+
to that for H O (χ) on 0.7 nm Pd clusters as a function of H
2
which is changed by adding H SO , H PO , HCl, NaHCO , or
2
4
3
4
3
concentration, which was controlled by the addition of a mineral acid
H CO (via dissolved CO ) to the solvent (50 kPa H , 60 kPa O ,
2
3
2
2
2
or base including H SO (black ■), H PO (blue ▲), HCl (magenta
2
4
3
4
2
0% v CH OH, 277 K). Figure 5a demonstrates that formation
3
▼), NaHCO (green ⧫), or H CO (red ●, by applying 0−0.7 MPa
3
2
3
+
rates for H O increase with increasing H concentration (r
3
−1
2
2
2
H O
2
CO ) (50 kPa H , 60 kPa O , 277 K, 10 cm min 20% v/v CH OH).
2 2 2 3
+
0.1
+
∼
[H ] ) and are similar at any given H concentration. These
Empty symbols represent measurements taken prior to the addition of
each acid or base. Lines are intended to guide the eye.
data are qualitatively consistent with our proposed mechanism
for H O formation (Scheme 1), which is shown by restating the
2
2
r
H O
2
2
+ 2 − 2
simplified rate expressions for H O formation (eqs 7 and 8) in
2
2
= (k K )[H ] [e ]
5
4
+
terms of the H concentration. The quasi-equilibrated steps for
heterolytic H oxidation (Scheme 1, steps 1 and 2) show that the
H pressure and the concentrations of H and free e are related
[L]
(17)
2
+
−
+
−
and a first-order dependence on [H ] and [e ] when OOH** is
2
the MARI:
by
r
+
2
− 2
2
H O
2
2
1/2
+
−
[
H ] [e ] = K K (H ) = α
(16)
2
1
2
= (k K K )[H ][e ]
5
2 1
[
L]
(18)
+
−
where [H ] and [e ] are the proton and electron concentrations,
−
respectively, and α is a constant equal to the product of the
H O formation rates are undetectable in the absence of an e
2 2
2
equilibrium constants (K K ) and the H pressure under the
source (i.e., 0 kPa H , Figure 2), even in solutions with a high
[H ] (pH 3, 0.5 mM H SO in H O), which is in agreement with
eqs 17 and 18 and previous measurements. The addition of
proton donors (i.e., mineral acids) to the solvent gives a
2
1
2
2
+
conditions of these experiments (Figure 5). Substitution of eq 16
into the rate equations for H O formation yields a second-order
2
4
2
21
2
2
+
−
dependence on both [H ] and [e ] when O ** is the MARI:
2
a
Table 2. Initial Rates of H O Formation in Protic and Aprotic Solvents with Different Dielectric Constants
2
2
solvent
methanol
protic/aprotic
pKa
16
dielectric constant (ε)
H O formation rate ((mol H O ) (mol Pd s)−1)
2 2 2 2 s
1.5 × 10−
1.3 × 10−
<1.7 × 10
<1.7 × 10−
3
protic
32.6
78.5
47.0
37.5
64.0
3
6
6
water
protic
16
−
dimethyl sulfoxide
acetonitrile
aprotic
aprotic
aprotic
35
25
b
−6
propylene carbonate
40−50
<1.7 × 10
a
Initial rates of H O formation were calculated by a linear fit of measured H O concentrations within a semibatch reactor over the range of 0−0.17
2
2
2
2
b
h. Data was collected on the 3.8 wt % Pd−SiO catalyst at 4.2 kPa H and 4.2 kPa O . Estimated from standard pK values for alkyl groups as well as
induction effects from oxygen atoms within the molecule
2
2
2
a
72
H
DOI: 10.1021/jacs.5b10669
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
+
measurable increase in [H ] due to the dissociation of these
species:
reflect significant differences between the electronic structures of
Pd clusters caused, in turn, by changes to the degree of
coordinative saturation of surface atoms. Figure 3 shows that
values of χ increase monotonically with the size of Pd clusters at
all H or O pressures, and 7 nm Pd gives χ values that are more
K
HX
+
−
[
HX] ←→ [H ] + [X ]
(19)
2
2
where KHX is the equilibrium constant for acid dissociation and
HX] and [X ] are the concentrations of mineral acids and
counterions in the solution, respectively. H O formation rates
increase only slightly when proton donors are added to the
solvent (rH2 ∼ [H ] , Figure 5a) because [H ] and [e ] are
inversely proportional to one another as dictated by the
than three times greater than those for 0.7 nm Pd under all
conditions, which agrees qualitatively with comparisons between
−
[
27
2
2
H O selectivities for 3.4 and 4.2 nm Pd clusters on SiO . In
2
2
2
addition, H O turnover rates are much larger on 7 nm Pd
2
2
+
0.1
+
−
O
2
clusters than those on 0.7 nm Pd (Figures S9.2a and 2a,
respectively). The rate ratios for all Pd clusters (Figure 3)
equilibrium for heterolytic H oxidation (eq 16). Thus, the
increase with a half-order dependence on (H ) and do not
2
2
+
addition of acid increases [H ] by a certain factor but
depend on (O ) (Figure S8.1 shows linearized rates). The
2
−
simultaneously decreases [e ] by the same amount. Con-
change in H O and H O formation rates with changes in (H )
2
2
2
2
sequently, we attribute the small increase in H O formation
and (O ) on 3 and 7 nm clusters (Figures S9.1 and S9.2) are
2
2
2
rates to electronic modification of the solution at the liquid−solid
described accurately by the mechanism (Scheme 1) and rate
equations (eqs 6 and 12) given earlier. These differences in
turnover rates and selectivities between Pd clusters of different
sizes result, therefore, from changes in the stability of surface
intermediates and transition states and not from differences
−
interface induced by the presence of counterions (i.e., Cl ,
SO42 , HCO , and H PO ), which increase with the amount of
−
−
−
3
2
4
acid added to the solvent. The H O turnover rates show also a
2
+
+ −0.05
weak inverse dependence on [H ] (rH2O ∼ [H ]
; Figure
between the mechanisms for H O and H O formation on
S7.1). The H O formation rate does not directly depend on
2
2
2
2
+
clusters of different sizes. The changes in the stability of reactive
intermediates are reflected in the values of the apparent
[
H ]; therefore, these small changes should only reflect the
adsorption of counterions onto the Pd, which are likely present at
⧧
⧧
6
7−69
activation enthalpies (ΔH ) and entropies (ΔS ) for these
reactions, determined from measurements of rates as a function
of temperature.
low coverages.
These results agree with previous observa-
tions that strongly binding anions decrease H O formation rates
2
during ORR, perhaps because the anions must be displaced from
4
9−51
Transition state theory proposes that reactant species (e.g.,
the surface prior to O−O bond rupture.
Additional
+
−
OOH* and a H −e pair) exist in equilibrium with the transition
measurements that probe counterion surface coverages, perhaps
by in situ infrared spectroscopy, are needed in order to develop a
deeper understanding of the manner by which rates and
selectivities are influenced specifically by counterions. Overall,
the addition of proton donors increases the selectivity toward
H O (i.e., χ, Figure 5b) by both increasing H O formation rates
+
⧧
72
state (H −OOH** ) for a given elementary step. Close
agreement between measured H O and H O formation rates
2
2
2
(
Figures 2, S9.1, and S9.2) and rate expressions (eqs 6 and 12)
suggest that these reactant species also exist in quasi-equilibrium
with gaseous H and O reactants. These multiple equilibrated
2
2
2
2
2
2
steps and the form of eq 7, therefore, suggest that the transition
states for H O formation are equilibrated also with chemisorbed
(
Figure 5a) and decreasing H O formation rates (Figure S7.1).
2
These observed trends are similar among all mineral acids tested;
2
2
oxygen (O **) and H , as shown in Scheme 3 and expressed in
therefore, χ can be increased significantly (10-fold) by adding
2
2
−
9
−2
+
the following reaction:
acids to the system (10 −10 M H ) with little dependence on
the identity of the counterion.
Collectively, these data (Figures 4 and 5) show that both H
and e are required to produce H O , which helps to clarify the
previously vague role of acids for H O formation.
⧧
K
H2O2
+
+
⧧
O ** + H (g) ←⎯⎯⎯ →⎯ H − OOH**
(20)
2
2
−
2
2
where K⧧
21,24
is the transition state equilibrium constant for
O
2
H
2
These
2
2
results and their interpretation are consistent with the proposed
mechanism for H O formation (Schemes 1 and 2). Moreover,
H
2
O
2
formation. By using conventions of transition state theory,
the rate of H formation can be expressed in terms of the
number of transition state species ([H −OOH** ]):
O
2
2
2
2
+
⧧
our measurements cannot be explained by Langmuirian
mechanisms that involve homolytic hydrogenation steps,
occurring at one or even two distinct surface sites. These results
r
H O
k T
h
2
2
B
+
⧧
=
[H −OOH** ]
[
L]
(21)
also explain the reported benefits of using CO as a diluent during
2
16,70,71
direct synthesis of H O .
H CO , formed following
dissolution of CO in protic solvents, provides HCO and
CO3 counterions that may bind to Pd clusters or alter the
properties of the solution near the surface of these clusters, and
increase H O selectivities, as shown above. Thus, high-pressure
2
2
2 3
where T is the temperature in Kelvin, and h and k are Planck’s
B
−
2
3
and Boltzmann’s constants, respectively. The equilibrated nature
2−
of eq 20, along with the observation that H O formation is
2
2
independent of (O ) (Figure 2a) allow eq 21 to be re-expressed
2
⧧
2
2
in terms of (H ) and K O2:
2
H
2
CO can replace strong mineral acids (e.g., H SO and HCl) that
2
2
4
r
H O
k T
h
are less environmentally benign. However, to be competitive
with AO, direct synthesis will require higher selectivities than
2
2
B
⧧
=
K
(H2)
H O
2 2
[
L]
(22)
those achieved here on 0.7 nm Pd−SiO from the addition of
2
Supporting Information section 10 contains equivalent ex-
CO alone (31% H O selectivity). In the next section, we show
2
2
2
pressions for H O formation. Figure 6 shows calculated values
that higher H O selectivities are achieved when we decrease the
2
2
2
⧧
⧧
extent of e− back-donation to adsorbed oxygen species by
for K
and K
(eqs 22 and S10.3) as a function of inverse
2
H O
2
2
H O
increasing the size of the Pd clusters.
temperature, and these values reflect only primary formation
3.3. Effects of Pd Cluster Size on Selectivities and
rates, calculated at zero H conversion by extrapolating rates
2
Activation Enthalpies. Changes in the size of Pd clusters lead
measured as a function of residence time at every temperature
(Figure S2.2).
to changes in both H O selectivities and turnover rates that
2
2
I
DOI: 10.1021/jacs.5b10669
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
⧧
⧧
Scheme 3. Thermochemical Cycle Showing the Changes in
Free Energy at Each Step for the Direct Synthesis of H O
Values of ΔH and ΔS are determined from the temperature
dependence of the overall change in free energy for H O
2
2
2
2
a
⧧
⧧
(
Blue) and H O (Red) on Pd Clusters
(ΔG
) and H O (ΔG 2O) formation (Scheme 3).
equals the sum of energies for the dissociation of
2
H
2
O
2
2
H
⧧
ΔG
2 2
H O
+
−
H (g) to H and to e localized on adsorbed O ** (ΔG ),
2
2
H
2
protonation of O22 ** to hydroperoxy (ΔGOOH), and
−
−
protonation of OOH ** to form the transition state
(
ΔGH−OOH):
ΔG⧧
= ΔGH2 + ΔGOOH + ΔGH−OOH
H O
(23)
2
2
whereas ΔG⧧
includes the free energy change for activating
2
H O
the O−O bond in OOH** (ΔGO−OH):
ΔG⧧
= ΔGH2 + ΔGOOH + ΔGO−OH
H O
(24)
2
⧧
Lastly, transition state theory allows values of K (Figure 6) to
x
⧧
⧧
⧧
⧧
be expressed in terms of ΔG as well as ΔH and ΔS (from ΔG
⧧
⧧
=
ΔH − TΔS ):
⧧
⧧x
⧧
a
⧧
= e−ΔG
x/RT
= e−ΔH /RTeΔS
x/R
Changes in free energy are shown for the heterolytic dissociation of
K
(25)
x
H (ΔG ) and the protonation of O ** (ΔGOOH) and OOH**
2
H
2
2
where R is the ideal gas constant and the subscript x indicates that
the associated variable corresponds to the formation of product
(
ΔGH−OOH). The sum of these free energy differences corresponds to
the measured activation free energy for H O formation on O2-
2
2
⧧
x.
saturated Pd clusters (ΔG O2). The measured activation free energy
for H O formation (ΔG
ΔG , and the change in free energy for O−O bond scission in
OOH** (ΔG
H
2
Table 3 shows values of ΔH⧧
nm Pd clusters calculated from K H2O2and K
and ΔH
⧧
for 0.7, 3, and 7
2
O
⧧
H
O
2 2
H
) corresponds to the sum of ΔG ,
2
H
2
O
H
2
⧧
⧧
(Figure 6 and eq
2
H O
OOH
25). ΔH⧧
values increase from 9 to 14 ± 2 kJ mol−1 as the
).
O−OH
H O
2 2
⧧
mean sizes of Pd clusters increase, which shows that ΔH
is
2 2
H O
⧧
weakly sensitive to cluster size. These ΔH H2O2values are similar
to barriers for proton−electron transfer steps on Pt electro-
−1
73,74
catalysts determined from DFT calculations (∼10 kJ mol )
−
1
75
and measured experimentally (<15 kJ mol ) as well as
−1
75−77
estimates (10−20 kJ mol ) for other ORR electrocatalysts.
Such comparisons indicate that these charge-transfer processes
depend weakly on the coordinative saturation and the elemental
identity of group 8 metal surfaces. The strong agreement
73−77
between this work and these previous studies
provides
further evidence that the mechanisms for direct synthesis of
H O are comparable to those for the electrochemical ORR.
2
2
⧧
Measured ΔH H2O2 values are significantly lower than calculated
barriers for hydrogenation of molecular oxygen (77−109 kJ
−
1
10,11
mol ) on Pd,
which shows that direct synthesis of H O
2 2
should proceed predominantly by proton−electron transfer and
⧧
⧧
Figure 6. Transition state equilibrium constants K for H O (solid
not by hydrogenation. Measured values for ΔH
, in contrast,
2
H O
2
2
symbols) and H O (empty symbols) formation as a function of inverse
temperature on 0.7 nm Pd (black ■ and □), 3 nm Pd (red ● and ○),
and 7 nm Pd (blue ▲ and Δ) clusters (50 kPa H , 60 kPa O , 20% v/v
CH OH). Turnover rates used to calculate K values were determined
at zero H conversion by extrapolating turnover rates measured at three
different conversions, each less than 10%. Lines represent fits to the
−1
2
increase from 18 to 32 ± 2 kJ mol as the mean diameters of Pd
⧧
clusters increase from 0.7 to 7 nm (Table 3), and ΔH
clearly
H2O
2
2
⧧
depends on the coordinative saturation of the exposed Pd atoms.
Notably, ΔH H2O on 7 nm Pd clusters, whose surfaces primarily
expose close-packed terraces, match activation energies (∼30
kJ mol ) for O−O bond dissociation within chemisorbed O on
3
⧧
2
78
−1
Eyring equation.
2
a
Table 3. Catalyst Activation Enthalpies and Entropies for H O and H O Formation
2
2
2
sample
ΔH⧧
(kJ mol−1
)
ΔH⧧ (kJ mol−1
)
ΔS⧧
(J mol−1 K−1
)
ΔS⧧ (J mol−1 K−1
)
H O
H
2
O
2
H
2
O
H
2
O
2
2
0
3
7
.7 nm Pd−SiO2
nm Pd−SiO2
nm Pd−SiO2
9 ± 2
9 ± 2
14 ± 2
18 ± 2
24 ± 2
32 ± 3
−232 ± 23
−226 ± 23
−200 ± 20
−197 ± 20
−165 ± 17
−129 ± 13
a
⧧
Activation enthalpies and entropies were calculated from measured transition state equilibrium constants (K ) and eq 25. Error was estimated to be
⧧
⧧
1
0% of the measured value based on the maximum error recorded for the turnover rate measurements used to calculate ΔH and ΔS .
J
DOI: 10.1021/jacs.5b10669
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
7
9,80
Pd(111).
This quantitative agreement supports our
2
transition states gain more translational mobility than does O **
as the coordinative saturation of the Pd surface atoms increases.
These data and the interpretations presented here show that
2
conclusion that H O forms following kinetically relevant O−O
bond scission on Pd. The smallest Pd clusters (0.7 nm) give the
lowest ΔH
expose a greater fraction of undercoordinated surface sites that
possess a greater density of states near the Fermi level as a result
of less overlap between the d orbitals of Pd atoms. O−O bond
dissociation barriers that decrease with the decreasing
coordinative saturation of surface Pd atoms (Table 3) agree
⧧
−1
values (18 kJ mol ) because these clusters
increasing the coordination of Pd atoms at cluster surfaces
2
H O
⧧
78
increases the H O selectivity (χ) by increasing ΔH
while
2
2
2
H O
⧧
only slightly affecting ΔH
values. The weak dependence of
2 2
H O
32
⧧
ΔH
on the electronic structure of the surface is consistent
2 2
H O
with the proposed mechanism for H O formation (Schemes 1
2
2
79,80
and 2) because H O forms by proton−electron transfer that
2
2
with published comparisons for O dissociation on Pd(111),
2
8
1,82
83
involves liquid-phase reactants insensitive to the electronic
Pd(100),
and Pd(110) surfaces. Thus, the amount of
⧧
structure of the catalyst surface. In contrast, ΔH H2O depends on
charge back-donated to 2π* of O on 0.7 nm clusters is greater
than that for 7 nm Pd clusters, which is reflected in the ΔH H2O
values. Consequently, H O selectivities (i.e., χ values) increase
2
⧧
the coordination of Pd surface atoms because the propensity of
surfaces to cleave O−O bonds depends strongly on their
2
2
⧧
electronic structure. The increase in ΔS values with cluster size
with the size of the Pd clusters because O−O bond rupture
transition states are more sensitive to electronic changes in the
Pd surface than those for proton−electron transfer (Figure 3).
shows that the transition states for H O and H O formation
2
2
2
both become more mobile relative to O ** as the size of Pd
2
⧧
⧧
clusters increase.
Table 3 shows values for ΔS
and ΔS H2 on Pd clusters.
2
H O
2
O
⧧
These ΔS values are large and negative, which shows that the
4
. CONCLUSIONS
formation of the transition states for each pathway involve a
⧧
Turnover rates for H O and H O formation were measured on
2
2
2
significant loss of entropy. Here, ΔS H corresponds to the
difference in entropy of the product and reactant states of the
process depicted in Scheme 3 such that
2 2
O
silica-supported Pd clusters over a wide range of H and O
2
2
+
pressures, and solvent H concentrations in the absence of
artifacts from mass transport and secondary decomposition of
H O . These rates were interpreted in order to evaluate the
ΔS⧧
2
2
= S
⧧
− SH2 − SO2**
H O
H O
2
(26)
2
2
2
mechanism of direct synthesis. Experiments in solvents with
+
and a corresponding equation describes ΔS⧧H2O:
varying H concentrations (including aprotic solvents) demon-
strated that protons are essential for H O formation, whereas
2
2
− 1 SH2 − SO2**
the identity of the counterion is much less important. The
dependence of rates on H and O pressures were inconsistent
ΔS⧧
= S
⧧
H O
H O
2
2
(
27)
2
2
2
with Langmuirian mechanisms and instead provided evidence for
heterolytic reaction pathways similar to the two-electron ORR.
Together, these data and their interpretations provided a
complete mechanistic understanding of the direct synthesis of
H O that is consistent with the data and implicates proton−
⧧
⧧
where SH2
2
and SH2 are the entropies of the transition states
O
O
for H O and H O formation, respectively, whereas S and S
2
O **
2
2
2
H
2
are the entropies of H and O . Values of ΔS⧧
(−232 to
2
2**
2 2
H O
2
2
1
−1
⧧
−1 −1
−
200 J mol− K ) and ΔS
(−197 to −129 J mol K ) are
H
2
O
electron transfer elementary steps to form H O and O−O bond
2
2
⧧
shown in Table 3.
For H O formation, the difference between the entropies of
scission in OOH** to form H O. Calculated values for ΔH and
2
⧧
2
2
ΔS from measured H O and H O formation rates at varying
2
2
2
the reactants (O ** and H (g)) and that of the transition state
temperatures show that H O formation is more sensitive to the
2
2
2
+
⧧
(
H −OOH** ) can be estimated roughly as being equal to the
surface electronic structure than is H O ; therefore, selectivity
2
2
translational and rotational entropy loss from binding H (g) to
can be increased by increasing the ratio of coordinated to
undercoordinated surface atoms. Overall, these results can aid
the rational design of inexpensive catalysts that can selectively
2
−1
−1
84
O ** (i.e., S , 131 J mol
K
at 101 kPa H , 273 K ). The
2
2
H
2
⧧
values of ΔS
(Table 3), however, are much more negative
2
H O
2
produce H O by direct synthesis in an environmentally benign
by >70 J mol−1 K ) than the estimate from this simple
−1
2
2
(
manner.
approximation. These significant differences suggest that addi-
tional entropy is lost, perhaps, as the hydrogen-bonded solvent
reorganizes into a lower energy configuration about the
ASSOCIATED CONTENT
■
* Supporting Information
85
S
hydrophilic transition state, which forms at the interface
between water and the hydrophobic O **-covered metal
cluster. Such a process would also be consistent with values
that also reflect a much larger entropy loss (again by
70 J mol− K ) than estimated by the consumption of one-half
The Supporting Information is available free of charge on the
ACS Publications website at DOI: 10.1021/jacs.5b10669.
2
8
6
⧧
of ΔS
H
2
O
TEM images and cluster size distributions of Pd catalysts,
procedures used to avoid mass-transport limitations and
artifacts from secondary decomposition of H O , addi-
1
−1
>
of a H (g) molecule. Estimates for the exact entropic change
2
2
2
caused by reorganization of the solvating molecules falls beyond
tional rate expressions and their associated derivations,
the scope of this study. Values of both ΔS⧧H and ΔS H2O
⧧
plots showing H O turnover rates as a function of [H ],
+
O
2 2
2
and formation rates and selectivities for 3 and 7 nm Pd
clusters. (PDF)
become slightly more positive (i.e., less entropy is lost) as the size
of Pd clusters increase (Table 3). These changes show that
⧧
⧧
entropies of the transition states (i.e., SH2O2 and SH2O ) increase
more with the size of Pd clusters than the surface intermediate
from which they form (i.e., SO2**) because SH2 does not depend
on the size of Pd clusters. This comparison implies that the
AUTHOR INFORMATION
■
Corresponding Author
*Phone: (217) 244-2816. Email: dwflhrty@illinois.edu.
K
DOI: 10.1021/jacs.5b10669
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
Notes
(27) Kim, S.; Lee, D.-W.; Lee, K.-Y.; Cho, E. A. Catal. Lett. 2014, 144,
9
05.
The authors declare no competing financial interest.
(
28) Abate, S.; Centi, G.; Melada, S.; Perathoner, S.; Pinna, F.; Strukul,
G. Catal. Today 2005, 104, 323.
29) Krishnan, V. V.; Dokoutchaev, A. G.; Thompson, M. E. J. Catal.
000, 196, 366.
ACKNOWLEDGMENTS
(
2
■
We gratefully acknowledge Mr. Luke Greenfield, Ms. Rong Jiao,
and Ms. Christine Rovani for their help synthesizing catalysts and
collecting and analyzing rate data. D.W.F. acknowledges Drs.
Damien Guironnet and Prashant Jain for their helpful discussions
and feedback. Partial financial support for this work was provided
by a departmental graduate student fellowship, funded by the
Dow Chemical Company. This work was carried out in part in
the Frederick Seitz Materials Research Laboratory Central
Research Facilities, University of Illinois.
(30) Burch, R.; Ellis, P. R. Appl. Catal., B 2003, 42, 203.
(31) Hohenegger, M.; Bechtold, E.; Schennach, R. Surf. Sci. 1998, 412-
4
13, 184.
32) Chorkendorff, I.; Niemantsverdriet, J. W. Concepts of Modern
Catalysis and Kinetics, 2nd ed.; Wiley-VCH: Weinheim, Germany, 2007.
33) Erley, W. Surf. Sci. 1982, 114, 47.
(
(
(34) Erley, W. Surf. Sci. 1980, 94, 281.
(35) Fukushima, T.; Song, M.-B.; Ito, M. Surf. Sci. 2000, 464, 193.
(
(
36) Han, Y.-F.; Lunsford, J. H. J. Catal. 2005, 230, 313.
37) Song, C.; Zhang, J. In PEM Fuel Cell Electrocatalysts and Catalyst
Layers; Zhang, J., Ed.; Springer-Verlag: Vancouver, BC, Canada, 2008;
pp 89−134.
(38) Jiao, L.; Regalbuto, J. R. J. Catal. 2008, 260, 329.
(39) Fagherazzi, G.; Benedetti, A.; Polizzi, S.; Di Mario, A.; Pinna, F.;
Signoretto, M.; Pernicone, N. Catal. Lett. 1995, 32, 293.
(40) Canton, P.; Fagherazzi, G.; Battagliarin, M.; Menegazzo, F.;
Pinna, F.; Pernicone, N. Langmuir 2002, 18, 6530.
(41) Safety Standard for Hydrogen and Hydrogen Systems: Guidelines for
Hydrogen System Design, Materials Selection, Operations, Storage, and
Transportation; NSS 1740.16, prepared for National Aeronautics and
Space Administration (NASA), Office of Safety and Mission Assurance:
Washington DC, 1997.
REFERENCES
■
(
1) Hydrogen Peroxide. In Ullmann’s Encyclopedia of Industrial
Chemistry; Wiley-VCH: Weinheim, Germany, 2012; Vol. 18, pp 393−
4
(
27.
2) EPA Fact Sheet- The Pulp and Paper Industry, the Pulping Process, and
Pollutant Releases to the Environment; Technical Report EPA-821-F-97-
11 prepared for U.S. Environmental Protection Agency (EPA), Office
of Water: Washington, DC, 1997.
3) Paper and Pulp. In Ullmann’s Encyclopedia of Industrial Chemistry;
Wiley-VCH: Weinheim, Germany, 2005.
4) Campos-Martin, J. M.; Blanco-Brieva, G. B.; Fierro, J. L. G. Angew.
Chem., Int. Ed. 2006, 45, 6962.
0
(
(
(42) Baga, A. N.; Johnson, G. R. A.; Nazhat, N. B.; Saadalla-Nazhat, R.
(
(
5) Samanta, C. Appl. Catal., A 2008, 350, 133.
6) Centi, G.; Perathoner, S.; Abate, S. In Modern Heterogeneous
A. Anal. Chim. Acta 1988, 204, 349.
(43) Madon, R. J.; Boudart, M. Ind. Eng. Chem. Fundam. 1982, 21, 438.
(44) Hibbitts, D. D.; Neurock, M. J. Catal. 2013, 299, 261.
(45) Yeh, K.-Y.; Wasileski, S. A.; Janik, M. J. Phys. Chem. Chem. Phys.
2009, 11, 10108.
(46) Keith, J. A.; Jacob, T. Angew. Chem., Int. Ed. 2010, 49, 9521.
(47) Voloshin, Y.; Halder, R.; Lawal, A. Catal. Today 2007, 125, 40.
(48) Kunimatsu, K.; Yoda, T.; Tryk, D. A.; Uchida, H.; Watanabe, M.
Phys. Chem. Chem. Phys. 2010, 12, 621.
Oxidation Catalysis: Design, Reactions, and Characterization; Mizuno, N.,
Ed.; Wiley-VCH Verlag GmbH & Co. KGaA: Weinheim, Germany,
2
(
009; pp 253−287.
7) Dissanayake, D. P.; Lunsford, J. H. J. Catal. 2003, 214, 113.
8) Landon, P.; Collier, P. J.; Carley, A. F.; Chadwick, D.; Papworth, A.
(
J.; Burrows, A.; Kiely, C. J.; Hutchings, G. J. Phys. Chem. Chem. Phys.
003, 5, 1917.
9) Liu, Q.; Bauer, J. C.; Schaak, R. E.; Lunsford, J. H. Angew. Chem., Int.
Ed. 2008, 47, 6221.
10) Ford, D. C.; Nilekar, A. U.; Xu, Y.; Mavrikakis, M. Surf. Sci. 2010,
04, 1565.
2
(
(49) Anderson, A. B. Electrochim. Acta 2002, 47, 3759.
(50) Markovic,
Electroanal. Chem. 1999, 467, 157.
(51) Grgur, B. N.; Markovic, N. M.; Ross, P. N. Can. J. Chem. 1997, 75,
1465.
́
N. M.; Gasteiger, H. A.; Grgur, B. N.; Ross, P. N. J.
(
6
́
(
(
11) Todorovic, R.; Meyer, R. J. Catal. Today 2011, 160, 242.
12) Lunsford, J. H. J. Catal. 2003, 216, 455.
(52) Karlberg, G. S. Phys. Rev. B: Condens. Matter Mater. Phys. 2006, 74,
(
13) Ham, H. C.; Stephens, J. A.; Hwang, G. S.; Han, J.; Nam, S. W.;
153414.
Lim, T. H. Catal. Today 2011, 165, 138.
14) Liu, Q.; Bauer, J. C.; Schaak, R. E.; Lunsford, J. H. Appl. Catal., A
008, 339, 130.
15) Han, Y.-F.; Zhong, Z.; Ramesh, K.; Chen, F.; Chen, L.; White, T.;
Tay, Q.; Yaakub, S. N.; Wang, Z. J. Phys. Chem. C 2007, 111, 8410.
16) Edwards, J. K.; Thomas, A.; Carley, A. F.; Herzing, A. A.; Kiely, C.
J.; Hutchings, G. J. Green Chem. 2008, 10, 388.
17) Biasi, P.; Canu, P.; Menegazzo, F.; Pinna, F.; Salmi, T. O. Ind. Eng.
Chem. Res. 2012, 51, 8883.
18) Edwards, J. K.; Hutchings, G. J. Angew. Chem., Int. Ed. 2008, 47,
192.
19) Edwards, J. K.; Solsona, B.; Ntainjua, E. N.; Carley, A. F.; Herzing,
A. A.; Kiely, C. J.; Hutchings, G. J. Science 2009, 323, 1037.
20) Biasi, P.; Menegazzo, F.; Pinna, F.; Eranen, K.; Canu, P.; Salmi, T.
O. Ind. Eng. Chem. Res. 2010, 49, 10627.
(53) Dean, J. A. Lange’s Handbook of Chemistry, 15th ed.; McGraw-Hill,
Inc.: New York, 1999.
(54) Shi, L.; Goldbach, A.; Zeng, G.; Xu, H. Catal. Today 2010, 156,
118.
(55) Guo, X.; Hoffman, A.; Yates, J. T., Jr. J. Chem. Phys. 1989, 90, 5787.
(
2
(
(
́
(56) Johansson, M.; Skulason, E.; Nielsen, G.; Murphy, S.; Nielsen, R.
M.; Chorkendorff, I. Surf. Sci. 2010, 604, 718.
(57) Li, J.; Yoshizawa, K. Catal. Today 2015, 248, 142.
(58) Ham, H. C.; Hwang, G. S.; Han, J.; Nam, S. W.; Lim, T. H. J. Phys.
Chem. C 2010, 114, 14922.
(
(
9
(
(59) Keith, J. A.; Jerkiewicz, G.; Jacob, T. ChemPhysChem 2010, 11,
2779.
(60) McMillen, D. F.; Golden, D. M. Annu. Rev. Phys. Chem. 1982, 33,
493.
(61) Wang, Y.; Balbuena, P. B. J. Chem. Theory Comput. 2005, 1, 935.
(62) Jacob, T. Fuel Cells 2006, 6, 159.
(63) Voloshin, Y.; Lawal, A. Chem. Eng. Sci. 2010, 65, 1028.
(64) Choudhary, V. R.; Samanta, C.; Jana, P. Method for production of
hydrogen peroxide with improved yield and selectivity by direct oxidation of
hydrogen over palladium containing catalyst. U.S. Patent 7,288,240,
October 30, 2007.
(
(
(
21) Liu, Q.; Lunsford, J. H. Appl. Catal., A 2006, 314, 94.
22) Ntainjua N., E.; Piccinini, M.; Pritchard, J. C.; Edwards, J. K.;
Carley, A. F.; Moulijn, J. A.; Hutchings, G. J. ChemSusChem 2009, 2, 575.
(
(
(
23) Deguchi, T.; Iwamoto, M. J. Catal. 2011, 280, 239.
24) Choudhary, V. R.; Samanta, C. J. Catal. 2006, 238, 28.
25) Choudhary, V. R.; Samanta, C.; Gaikwad, A. G. Chem. Commun.
2
(
004, 2054.
26) Menegazzo, F.; Signoretto, M.; Frison, G.; Pinna, F.; Strukul, G.;
Manzoli, M.; Boccuzzi, F. J. Catal. 2012, 290, 143.
(65) Keith, J. A.; Jacob, T. In Theory and Experiment in Electrocatalysis,
50th ed.; Balbuena, P. B., Subramanian, V. R., Eds.; Springer: New York,
2010; pp 89−132.
L
DOI: 10.1021/jacs.5b10669
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
(
66) Gao, F.; Wang, Y.; Goodman, D. W. J. Am. Chem. Soc. 2009, 131,
734.
67) Markovic,
68) Faguy, P. W.; Markovic, N.; Adzic, R. R.; Fierro, C. A.; Yeager, E.
B. J. Electroanal. Chem. Interfacial Electrochem. 1990, 289, 245.
69) Lucas, C. A.; Markovic, N. M.; Ross, P. N. Phys. Rev. B: Condens.
Matter Mater. Phys. 1997, 55, 7964.
70) García-Serna, J.; Moreno, T.; Biasi, P.; Cocero, M. J.; Mikkola, J.-
P.; Salmi, T. O. Green Chem. 2014, 16, 2320.
5
(
(
́
N. M.; Ross, P. N. Surf. Sci. Rep. 2002, 45, 117.
(
́
(
(
(
71) Han
̂
cu, D.; Beckman, E. J. Green Chem. 2001, 3, 80.
72) Anslyn, E. V.; Dougherty, D. A. Modern Physical Organic
Chemistry; University Science Books: Sausalito, CA, 2006.
73) Janik, M. J.; Taylor, C. D.; Neurock, M. J. Electrochem. Soc. 2009,
56, B126.
74) Ou, L.; Yang, F.; Liu, Y.; Chen, S. J. Phys. Chem. C 2009, 113,
0657.
75) Anderson, A. B.; Roques, J.; Mukerjee, S.; Murthi, V. S.; Markovic,
N. M.; Stamenkovic, V. J. Phys. Chem. B 2005, 109, 1198.
76) Khan, A.; Lu, X.; Aldous, L.; Zhao, C. J. Phys. Chem. C 2013, 117,
8334.
77) Tang, W.; Lin, H.; Kleiman-Shwarsctein, A.; Stucky, G. D.;
McFarland, E. W. J. Phys. Chem. C 2008, 112, 10515.
(
1
(
2
(
́
(
1
(
(
78) van Hardeveld, R.; Hartog, F. Surf. Sci. 1969, 15, 189.
(
79) Kolasinski, K. W.; Cemic, F.; Hasselbrink, E. Chem. Phys. Lett.
1
(
1
994, 219, 113.
80) Nolan, P. D.; Lutz, B. R.; Tanaka, P. L.; Mullins, C. B. Surf. Sci.
998, 419, L107.
(
81) Nyberg, C.; Tengstal, C. G. Surf. Sci. 1983, 126, 163.
(
82) Stuve, E. M.; Madix, R. J.; Brundle, C. R. Surf. Sci. 1984, 146, 155.
(
83) Junell, P.; Honkala, K.; Hirsimaki, M.; Valden, M.; Laasonen, K.
Surf. Sci. 2003, 546, L797.
84) Burgess, D. R. In NIST Chemistry WebBook, NIST Standard
(
Reference Database Number 69; Linstrom, P. J., Mallard, W. G., Eds.;
National Institute of Standards and Technology: Gaithersburg, MD,
2
015. http://webbook.nist.gov (accessed on November 18, 2015).
(
(
85) Ben-Naim, A.; Marcus, Y. J. Chem. Phys. 1984, 81, 2016.
86) Chandler, D. Nature 2005, 437, 640.
M
DOI: 10.1021/jacs.5b10669
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX