Propylene oxidation on copper oxide surfaces: Electronic and geometric contributions to reactivity and selectivity

Article abstract of DOI:10.1021/ja981579s

Cu₂O is an efficient catalyst in the partial oxidation of propylene to acrolein, while propylene oxidation on CuO leads to complete combustion. The interaction of propylene at elevated temperature (> 300 K) and elevated pressure (5 Torr) with cuprous and cupric oxide has been investigated with core level XPS, resonant photoemission, and temperature-programmed desorption. Reduction of the copper oxide surfaces was examined as a function of temperature and revealed that cupric oxide has a greater reactivity toward propylene oxidation than cuprous oxide (E(a) = 5.9 versus 11.5 kcal/mol for cuprous oxide (24.7 and 48.1 kJ/mol)). This variable temperature oxidation of propylene was also monitored via core level and resonant photoemission and was found to occur by a similar mechanism on both surfaces. Reaction at lower temperature produces a surface intermediate which exhibits carbon 1s XPS peaks at 284.0 and 285.5 eV binding energy in a 2:1 intensity ratio. This is consistent with an allyl alkoxide surface species, indicating a reaction mechanism involving an initial H atom abstraction from propylene followed by rapid oxide insertion. The relative surface reactivities are related to the redox potential of the metal ion and the pK(a) of the protonated surface oxide. The presence of a significant amount of this surface alkoxide is consistent with a relatively slow alkoxide decomposition step. This decomposition occurs more readily on the cuprous oxide surface (E(a) (decomposition) = 24.5 kcal/mol (102.6 kJ/mol) versus 28.7 kcal/mol (120.1 kJ/mol) on cupric oxide) and involves a hydride elimination mechanism. At elevated temperatures a new carbon 1s peak at ~288 eV binding energy is observed which is consistent with the formation of further oxidized surface species (RCO(x)). The CuO surface is found to be more reactive in forming these nonselective highly oxidized products. The observed differences in reactivity, rates of reaction steps, selectivity, and product distribution are addressed and provide insight into the factors which influence the reactivity and selectivity of the copper oxides toward the heterogeneous oxidation of propylene.

Full text of DOI:10.1021/ja981579s

J. Am. Chem. Soc. 1998, 120, 11467-11478
11467
Propylene Oxidation on Copper Oxide Surfaces: Electronic and
Geometric Contributions to Reactivity and Selectivity
John B. Reitz and Edward I. Solomon*
Contribution from the Department of Chemistry, Stanford UniVersity, Stanford, California 94305
ReceiVed May 6, 1998. ReVised Manuscript ReceiVed August 27, 1998
Abstract: Cu2O is an efficient catalyst in the partial oxidation of propylene to acrolein, while propylene oxidation
on CuO leads to complete combustion. The interaction of propylene at elevated temperature (>300 K) and
elevated pressure (5 Torr) with cuprous and cupric oxide has been investigated with core level XPS, resonant
photoemission, and temperature-programmed desorption. Reduction of the copper oxide surfaces was examined
as a function of temperature and revealed that cupric oxide has a greater reactivity toward propylene oxidation
than cuprous oxide (Ea ) 5.9 versus 11.5 kcal/mol for cuprous oxide (24.7 and 48.1 kJ/mol)). This variable
temperature oxidation of propylene was also monitored via core level and resonant photoemission and was
found to occur by a similar mechanism on both surfaces. Reaction at lower temperature produces a surface
intermediate which exhibits carbon 1s XPS peaks at 284.0 and 285.5 eV binding energy in a 2:1 intensity
ratio. This is consistent with an allyl alkoxide surface species, indicating a reaction mechanism involving an
initial H atom abstraction from propylene followed by rapid oxide insertion. The relative surface reactivities
are related to the redox potential of the metal ion and the pKa of the protonated surface oxide. The presence
of a significant amount of this surface alkoxide is consistent with a relatively slow alkoxide decomposition
step. This decomposition occurs more readily on the cuprous oxide surface (Ea (decomposition) ) 24.5 kcal/
mol (102.6 kJ/mol) versus 28.7 kcal/mol (120.1 kJ/mol) on cupric oxide) and involves a hydride elimination
mechanism. At elevated temperatures a new carbon 1s peak at ∼288 eV binding energy is observed which
is consistent with the formation of further oxidized surface species (RCOx). The CuO surface is found to be
more reactive in forming these nonselective highly oxidized products. The observed differences in reactivity,
rates of reaction steps, selectivity, and product distribution are addressed and provide insight into the factors
which influence the reactivity and selectivity of the copper oxides toward the heterogeneous oxidation of
propylene.
1. Introduction
12% conversion to acrolein with 60-85% selectivity.3 Research
on understanding the nature of the copper oxide catalyst has
focused on two major areas: (1) the oxidation state and form
of the catalytically active copper oxide and (2) the mechanism
for the formation of acrolein from propylene including the rates
of the various reaction steps.
The catalytic oxidation of olefins by metals and metal oxides
is a constantly growing and evolving field of study. As most
hydrocarbon reactants derive from petroleum, a nonrenewable
resource, extensive work revolves around discovering new, more
efficient catalysts which have enhanced product yield and
selectivity. Despite significant research effort directed toward
understanding the general features and reaction steps of oxida-
tion catalysis mechanisms on surfaces, many systems can still
be regarded as poorly defined. Reactants are injected and
products are formed in good yield and with good selectivity,
but a complete understanding of the origin of the catalytic
behavior is not yet available. In particular, fundamental ques-
tions exist concerning the mechanism of reaction as well as the
factors associated with the substrate, reactants, and reaction
conditions which contribute to effective, selective catalytic
behavior.
Heterogeneous oxidation catalysis on copper oxide surfaces
is one class of reactions that has received significant attention,
yet there are still fundamental questions concerning the mech-
anism of reaction and the factors which influence rates and
selectivity. Hearne and Adams were the first to discover that
cuprous oxide is a selective catalyst for the oxidation of un-
saturated hydrocarbons.1,2 One of the more useful reactions over
copper oxide is the formation of acrolein from propylene. On
cuprous oxide, the reaction is performed at temperatures of 300-
400 °C with an olefin:oxygen ratio of ∼5:1, and has a 10-
The oxidation state and structure of the copper oxide catalyst
is highly dependent on reactant gas composition as well as
reaction temperature. Propylene is a reducing gas, while di-
oxygen is an oxidizing agent, and thus one can envision a
range of possible oxidation states and substrate surface com-
positions depending on the relative concentrations of the reac-
tant gases. Wood, Ylles, and Wise4 have determined, via
electrical conductivity measurements, that selectivity to acro-
lein reaches a maximum at a catalyst composition of nearly
stoichiometric cuprous oxide, while Inui, Ueda, and Suehiro
proposed that a slightly copper-rich catalyst (Cu2.17O) was
the most effective at selective acrolein formation.5 In con-
trast to the high selectivity of cuprous oxide, Holbrook and
Wise determined that cupric oxide, CuO, produces predomi-
nately complete oxidation products (COx) and very little
acrolein.6
(1) Hearne, G. W.; Adams, M. L. U.S. Patent 2,451,485, 1985, to Shell
Development Co.
(2) Voge, H. H.; Adams, C. R. AdV. Catal. 1967, 17, 151.
(3) Adams, C. R.; Jennings, T. J. J. Catal. 1964, 3, 549.
(4) Wood, B. J.; Yolles, R. S.; Wise, H. J. Catal. 1969, 15, 355.
(5) Inui, T.; Ueda, T.; Suehiro, M. J. Catal. 1980, 65, 166.
10.1021/ja981579s CCC: $15.00 © 1998 American Chemical Society
Published on Web 10/27/1998

11468 J. Am. Chem. Soc., Vol. 120, No. 44, 1998
Reitz and Solomon
Scheme 1
oxygen, it has been determined that the oxygen incorporated
into the final acrolein product comes from oxide associated with
the substrate.14,15
Following the second H atom abstraction and the oxygen
insertion, the aldehyde is formed and the acrolein product is
desorbed. The formation of acrolein results in a four-electron
reduction of the surface (O2- f O + 2e-; 2H f 2H+ + 2e-),
and this reduced catalyst is reoxidized by gas-phase dioxygen.
The kinetics of the catalytic reaction have been found to be
first order in oxygen and zero order in propylene,16 indicating
that reoxidation and diffusion of oxide to reduced sites is rate-
determining in the catalytic reaction.
In the present study, surface science techniques have been
applied toward understanding the reaction of propylene with
the cuprous and cupric oxide surfaces. While past research has
established that the oxidation of propylene to acrolein over
cuprous oxide and the related BiMo catalysts8,17,18 occurs via
the mechanism presented in Scheme 1, there are still funda-
mental questions concerning the relative order of the reaction
steps and the identity and electronic and geometric structure of
surface intermediates formed in the course of the reaction. Also,
in contrast to the large body of research on cuprous oxide, there
is relatively little data on the mechanism of propylene oxidation
with cupric oxide which results in, almost exclusively, complete
combustion products. Thus, core and valence band photoemis-
sion, Auger electron spectroscopy, and temperature-programmed
desorption (TPD) have been applied toward determining the
relative reactivity of the two surfaces and the nature of the
surface intermediates formed following reaction with propylene
under varied temperature (135-623 K) and pressure (10-6-5
Torr) conditions. These results yield important insight into the
mechanism and relative order of the reaction steps and the
contributions of the substrate and reactant to the reactivity and
selectivity in propylene oxidation on the two surfaces.
Scheme 2
The currently proposed mechanism for the oxidation of
propylene to acrolein over cuprous oxide is summarized in
Scheme 1. The first step involves activation of propylene by
abstraction of an H atom to yield the allylic intermediate. This
intermediate next undergoes either another H atom abstraction
(2a) or oxygen insertion (2b), with the third step being whichever
of these has not yet occurred. This results in the formation of
the aldehyde product following CdO bond formation. In this
reaction, the incorporated oxygen is a lattice oxide, O2-, rather
than a surface species originating from dioxygen. Product
formation leaves the catalyst surface in a multielectron reduced
state, and it is reoxidized by gas-phase dioxygen.
The initial H atom abstraction has been extensively studied,
primarily through isotopic labeling experiments, and its viability
as a reaction pathway for propylene oxidation is due, in part,
to the resonance-stabilized, weak C-H bond on the methyl
group of propylene (dissociation energy: CH2CHCH2-H, ∼87
kcal/mol (364.2 kJ/mol)). Voge, Wagner, and Stevenson7
prepared propylene with a 13C-labeled methyl group and
monitored the isotopic distribution in the products evolved
following reaction with cuprous oxide. Their results revealed
that only half of the original 13C content in the propylene reactant
was found in the carbonyl group of the acrolein product. Similar
results were reported by Adams and Jennings8 from deuterated
propylene (96% 1-propene-3-d), where the deuterated carbon
was incorporated into both end carbons of the acrolein product.
This scrambling of the terminal carbons was postulated to result
from the formation of a symmetric allylic intermediate (Scheme
2). From deuterium-substituted propylene, this initial H atom
abstraction was found to be the rate-determining step of the
oxidation of propylene to acrolein (kinetic isotope effect (kH/
kD ) 1.73 at 350 °C).3
2. Experimental Section
The cuprous oxide sample used in this study was prepared from a
Cu(111) single crystal. The Cu crystal was aligned by Laue back-
scattering to within (1° and was polished to 1 µm with an alumina
grit. Sample cleaning was completed in a vacuum through argon
sputtering cycles (2000 V) and annealing (900 K). The sample
cleanliness was monitored by carbon 1s XPS, and surface order was
checked by LEED. The cleaned Cu(111) sample was exposed to 1
Torr of oxygen while at 623 K, and this resulted in the formation of
both Cu(I) and Cu(II) sites. These oxygen exposures were repeated
until no Cu(0) was detectable by X-ray-induced Cu LMM Auger. The
oxidized sample was annealed in a vacuum to 900 K to yield cuprous
oxide, and the oxidation state, composition, and cleanliness of this
cuprous oxide was checked by Cu LMM Auger and core level (Cu,
oxygen 1s, and carbon 1s) XPS. LEED produced no discernible pattern,
indicating that this cuprous oxide surface possessed no long-range
surface order.
The CuO sample was prepared by cooling the Cu2O sample prepared
above from 573 K in 250 mTorr of oxygen. The oxidation state,
composition, and cleanliness of the resulting sample were verified by
Cu 2p, oxygen 1s, and carbon 1s XPS.
Spectroscopic data were obtained on two different instruments, one
using conventional radiation sources (Vacuum Generators ESCALAB
Mk II) and the other employing synchrotron radiation (Perkin-Elmer
The next step of the reaction of propylene on cuprous oxide
has been the subject of much debate. While early research
favored a second hydrogen abstraction,3,8,9 more recent studies
using IR10,11 and TPD and XPS12,13 have proposed oxygen
insertion and the formation of an allyl alkoxide (C3H5O-)
intermediate prior to the second hydrogen abstraction. Despite
the relative order of these two steps, through the use of labeled
(6) Holbrook, L.; Wise, H. J Catal. 1971, 20, 367.
(7) Voge, H. H.; Wagner, C. D.; Stevenson, D. P. J. Catal. 1963, 2, 58.
(8) Adams, C. R.; Jennings, T. J. J. Catal. 1963, 2, 63.
(9) Keulks, G. W.; Krenske, L. D.; Notermann, T. M. AdV. Catal. 1978,
27, 183.
(14) Keulks, G. W. J. Catal. 1970, 19, 232.
(15) Akimoto, M.; Akiyama, M.; Echigoya, E. Bull. Chem. Soc. Jpn.
1976, 49, 3367.
(10) Davydov, A. A.; Mikhaltchenko, V. G.; Sokolovskii, V. D.;
Boreskov, G. K. J. Catal. 1978, 55, 299.
(11) Mikhalchenko, V. G.; Sokolovskii, V. D.; Filpora, A. A.; Davydov,
A. A. Kinet. Katal. 1973, 14, 253.
(12) Schulz, K.; Cox, D. F. Surf. Sci. 1992, 262, 318.
(13) Schulz, K. H.; Cox, D. F. J. Catal. 1993, 143, 464.
(16) Isaev, O. V.; Margolis, L. Ya. Kinet. Katal. 1960, 1, 237.
(17) Bielanski, A.; Haber, J. Oxygen in Catalysis; M. Dekker: New York,
1991; see also references therein.
(18) Veatch, F.; Callahan, J. L.; Milberger, E. C.; Forman, R. Proceedings
of the 2nd International Congress on Catalysis; Editions Techniq: Paris,
1961.

Propylene Oxidation on Copper Oxide Surfaces
J. Am. Chem. Soc., Vol. 120, No. 44, 1998 11469
Table 1. Reduction of Cuprous and Cupric Oxide Surfaces with
Propylene
Cu2O
% surface
CuO
relative
% surface
reduction
relative
reduction
temp (K)
reduction
reduction
300
353
423
473
523
573
623
0
0
0
0
1.7
4.5
10
0
0
0
0
0.17
0.45
1
0
3
18
0
0.24
1.4
43
72
3.4
5.7
Copper oxidation states can be determined by the Auger
parameter,19 R,
R ) KE(LMM) + BE(2p3/2)
(3.1)
which is the sum of the kinetic energy of the core Auger peak
and the photoemission binding energy of that same core level.
From the experimentally determined peak positions, the cuprous
oxide Auger parameter is 1848.9 eV and that of the new feature
following exposure to propylene at elevated temperatures is
1850.7 eV. This 1.8 eV difference, as well as the absolute value
of the Auger parameter, is consistent with assignment of the
new Auger feature as Cu(0).19 Thus, following exposure of
the cuprous oxide surface to propylene at T > 473 K, surface
reduction is observed. Table 1 lists the percent surface reduction
of cuprous oxide with increasing temperature of reaction. In
addition to surface reduction, the Cu:O ratio was found to
increase by 12% following reaction with propylene at 623 K,
indicating the loss of surface oxide.
As discussed above, propylene oxidation over cuprous oxide
is a redox process which, in the absence of oxygen, results in
reduction of the copper oxide surface. Therefore, surface
reduction and loss of oxide is indicative of the oxidation of
propylene, and the amount of reduction with time can be related
to the reactivity and rate of propylene oxidation. Figure 2A
plots ln(amount of reduction) versus 1/T. The resulting data
are fit well with a straight line (R2 ) 0.999), and, through the
application of the Arrhenius equation,
Figure 1. Cu LMM Auger region of cuprous oxide before and after
exposure to 5 Torr of propylene at variable temperature.
PHI system). Both the Al KR (1486.6 eV) and Mg KR (1253.6 eV)
anodes on the ESCALAB were used for XPS data collection, and the
hemispherical analyzer was operated at a pass energy of 20 eV. The
base pressure was ∼8 × 10-11 Torr in the analytic chamber. Data
taken with photon energies other than those listed above were obtained
with synchrotron radiation utilizing Beamlines 1-2 and 8-1 at the
Stanford Synchrotron Radiation Laboratory (SSRL). All beamlines
were operated at energy resolutions of better than 0.20 eV. The ion-
pumped Perkin-Elmer PHI vacuum chamber, equipped with a cylindri-
cal mirror analyzer (operated at a pass energy of 25 eV), used at SSRL
had a base pressure of better than 1 × 10-10 Torr.
High-pressure propylene exposures were conducted by cooling the
heated copper oxide sample in 5 Torr of propylene for 2 min. For
experiments utilizing the ESCALAB, exposures were conducted in a
separate, valved-off preparatory chamber while those using the PHI
system were conducted in the main chamber. Pump out of the
propylene, for both systems, was achieved in 5 min (to 10-7 Torr),
following which the samples were cooled to room temperature for data
collection. Experimental errors in exposure pressure and time are
estimated to be (10%, while sample temperature error is (25 K.
Temperature-programmed desorption (TPD) data were collected
using the ESCALAB equipped with a UTI Model 100C residual gas
analyzer connected to a computerized data collection system capable
of monitoring multiple masses simultaneously. A heating rate of 2
K/s was used for all desorption experiments.
k ) Ae-E /RT
(3.2)
a
an activation energy, Ea, of 11.5 kcal/mol can be extracted from
the slope of the line.
All PES spectra were signal averaged until a sufficient signal-to-
noise ratio had been obtained. Additionally, all synchrotron data were
normalized to the incident flux by monitoring the photoelectron signal
emitted from a gold-coated stainless steel mesh located in the path of
the monochromatized radiation.
All gases used were of research grade purity (oxygen, 99.997%;
propylene, 99.6%) and were introduced into the main UHV chamber
through a separate, turbo-pumped manifold (base pressure <10-8 Torr).
3.A.2. Cupric Oxide. Figure 3 shows the copper 2p XPS
spectra of cupric oxide following exposures to 5 Torr of
propylene for 120 s under variable temperature conditions. The
Cu 2p spectrum of cupric oxide is comprised of intense spin-
orbit split 2p3/2 and 2p1/2 lines at 933.5 and 953.3 eV binding
energy which correspond to a 2p ionization plus a charge-
transfer transition to the Cu(II) valence hole to yield a 2p53d10L
final state (where L refers to a ligand hole) and a broad feature
at ∼9 eV to deeper binding energy (only the one associated
with the 2p3/2 line is shown). The feature at ∼942 eV is a
Cu(II) satellite20,21 which arises from a Cu 2p core ionization
and results in a 2p53d9 final state. This process results in the
emission of a photoelectron with a kinetic energy lower than
3. Results and Analysis
3.A. Reaction with Propylene: Surface Reduction. 3.A.1.
Cuprous Oxide. Figure 1 shows the X-ray-induced Cu
L3M4,5M4,5 (LMM) Auger spectra for cuprous oxide following
exposure to 5 Torr of propylene for 120 s at variable temper-
atures. In this spectral region, cuprous oxide is characterized
by a peak maximum at 916.4 eV kinetic energy, and following
exposure to propylene at temperatures greater than 473K, a new
peak is observed at ∼918.2 eV. No change in the copper 2p
XPS region is observed over the entire temperature range.
(19) Wagner, C. D.; Riggs, W. M.; Davis, L. E.; Moulder, J. F.;
Muilenberg, G. E. Handbook of X-ray Photoelectron Spectroscopy; Perkin-
Elmer Corporation, 1979.
(20) Ghijsen, J.; Tjeng, L. H.; van Elp, J.; Eskes, H.; Weterink, J.;
Sawatzky, G. A.; Czyzyk, M. T. Phys. ReV. B 1988, 38, 11322.
(21) Parmigiani, F.; Pacchiuni, G.; Illas, F.; Bagus, P. S. J. Electron
Spectrosc. Relat. Phenom. 1992, 59, 255.

11470 J. Am. Chem. Soc., Vol. 120, No. 44, 1998
Reitz and Solomon
Figure 3. Copper 2p XPS spectra of CuO following exposure to 5
Torr of propylene at variable temperature. The top spectrum is cuprous
oxide.
from the observed reduction of the CuO sample, and the slope
of the fitted line (R2 ) 0.98) yields an activation energy, Ea, of
5.6 kcal/mol.
3.A.3. Comparison. The observed reduction for each oxide
surface is measured as a percent change in the Cu(I) Auger and
Cu(II) 2p signals for the cuprous and cupric oxide samples,
respectively. To compare the reduction of the two surfaces with
propylene, the relative reduction values in Table 1 have been
corrected for the different sampling depths of the detection
techniques which have been approximated by the escape depth
of the ejected electrons (18 Å for ∼920 eV kinetic energy of
the Cu LMM Auger electrons and 7 Å for the ∼320 eV energy
of the Mg KR ejected Cu 2p electrons)22 and for the total copper
content of both samples which was obtained experimentally
from the total area of the copper 2p XPS peaks. Including these
factors allows for the quantitative comparison of the reaction
on the two surfaces.
From the relative reduction of the two surfaces, which is
proportional to the rate of propylene oxidation, it is observed
that at comparable reaction temperatures (523 and 573 K) CuO
has 8-10 times the reduction of Cu2O, indicating a significantly
enhanced reactivity for cupric oxide. This enhanced reactivity
toward propylene oxidation is further supported by the Arrhenius
plots which reveal a lower activation energy for the oxidation
of propylene on the CuO surface.
3.B. Reaction with Propylene: Surface Intermediates and
Their Relation to the Reaction Mechanism. 3.B.1. Cuprous
Oxide. Figure 4 shows the carbon 1s XPS spectra following
exposure of the Cu2O surface to propylene at variable temper-
atures and exposures. The low signal-to-noise ratio of the
carbon 1s XPS peaks is due to two main factors. First, the
amount of adsorbate on the surface (obtained from adsorbate
intensities and substrate peak attenuation) is estimated to be
<0.5 monolayer (ML). This low coverage, particularly coupled
with the use of Al KR radiation, results in a weak adsorbate
XPS signal. Also, the reaction of propylene with the copper
oxide substrates at the data collection temperature (300 K) was
Figure 2. (A) Arrhenius plot of the reduction of the cuprous oxide
surface following exposure to 5 Torr of propylene at variable temper-
atures. (B) Arrhenius plot of CuO.
that of the main peak due to final state relaxation. In contrast,
the closed-shell 3d10 Cu(I) oxide has a 2p XPS spectrum
comprised of only 2p3/2 and 2p1/2 features at 932.5 and 952.3
eV, respectively. Thus, the deeper binding energy satellite is
characteristic of Cu(II) and can be used to determine both the
presence and amount of Cu(II) in the sample. In contrast, the
Cu(II) LMM Auger feature is broadened and shifted by only
+0.8 eV versus the Cu(I) line and, due to significant over-
lap, is not a useful experimental probe of the concentration of
Cu(II) sites. From Figure 3, it is observed that following
reaction with propylene at increasing temperature, the Cu(II)
2p satellite decreases in intensity, and the main photoemission
lines are observed to narrow and shift lower in energy. These
spectral changes indicate reduction of the Cu(II) sites, and by
examining the Cu LMM Auger region (and corresponding Auger
parameter) following reaction at 573 K, it is determined that
the Cu(II) sites are being reduced to Cu(I).
Since CuO was observed to thermally decompose to cuprous
oxide in a vacuum, control experiments without propylene were
conducted, and a correction, which amounts to a maximum of
a 20% decrease in the Cu(II) signal at 573 K, has been applied
to the percent surface reduction quantitated in Table 1. As in
the Cu2O case, an Arrhenius plot (Figure 2B) can be constructed
(22) Lindau, I.; Spicer, W. E. J. Electron Spectrosc. Relat. Phenom. 1974,
3, 409.

Propylene Oxidation on Copper Oxide Surfaces
J. Am. Chem. Soc., Vol. 120, No. 44, 1998 11471
the Cu LMM Auger region at the lower reaction temperatures
(Figure 1, 300 K). Due to the finite sampling depth of
photoelectrons, adsorption of significant amounts of hydrocarbon
species on the cuprous oxide surface will cause attenuation of
the substrate photoemission peak intensities, and this attenuation
can be used to estimate the amount of allyl alkoxide formed
and thus the amount of surface reduction. From the experi-
mentally determined attenuation of the Cu 2p3/2 XPS peak (Mg
KR radiation), and assuming a two-electron reduction for every
propylene oxidized to an allyl alkoxide (H atom abstraction, H
f H+ + 1e-; C-O bond formation, O2- f C-O- + 1e-),
the amount of allyl alkoxide at 300 K would only correspond
to a ∼3% (( 2%) change in the total Cu(I) LMM Auger signal.
This small amount of reduction is below the detection limit in
this experiment. It should be noted that this 3% reduction is a
lower limit on the amount of substrate reduction, since any
oxygenated products formed and desorbed from the surface will
contribute to reduction, but not to the carbon 1s signal. A
similar assignment of an allyl alkoxide surface species has been
made following exposure of cuprous oxide single crystals to
high pressures (1 atm) of propylene at 300 K.13
Significant changes occur in the carbon 1s spectra following
exposures to 5 Torr of propylene in the 573-623 K range
(Figure 4). The 573 and 623 K spectra can be fit with three
peaks at 284.4, 286.0, and ∼288.0 eV (Figure 5A,II). The peak
at 286.0 is consistent with the previous assignment of a
R-C*-O carbon, and the peak at 284.4 is once again assigned
as vinylic (or aliphatic) carbons. However, the intensity ratio
of the 284.4 and 286.0 peaks now deviates from 2:1.25 The 288.0
eV binding energy of the new peak is consistent with the
oxygenated carbon of an aldehyde or ketone (RsC*dO),23
possibly associated with acrolein formation. The carbon 1s XPS
spectrum of acrolein physisorbed at low temperature on Cu2O-
(100) consists of two peaks, one at 285.8 eV assigned to the
vinylic carbons and one at ∼288.4 eV assigned to the carbonyl
carbon.26 Chemisorbed carboxylates (R-C*OO-) on cuprous
oxide have also been shown to have carbon 1s peaks at
comparable binding energies (284.9 and 288.0 eV).27 Thus, the
288.0 peak in Figure 5A,II could be associated with an aldehydic
carbon such as that contained in acrolein or a surface carboxy-
late. Assignment as a carboxylate is attractive since acrolein
is not expected to remain adsorbed to the surface at these
elevated temperatures.26 Carboxylate species would likely
originate from attack on carbonyl carbons by surface oxide28,29
and have been shown to be more strongly bound to the surface
(Tdesorp > 500 K).27
Figure 4. Carbon 1s XPS spectra following exposure of the cuprous
oxide surface to propylene at variable temperature conditions.
found to be relatively fast and results in a changing carbon 1s
XPS spectrum. Thus, a smaller number of signal-averaged scans
were collected in order to obtain an accurate picture of the
surface adsorbate species immediately following propylene
exposures.
In Figure 4, the bottom spectrum was obtained following a
10 Langmuir (1 Langmuir ) 10-6 Torr s) exposure at 135 K
on a Cu2O(111) single crystal. The spectrum, collected using
hν ) 370 eV synchrotron radiation, shows one peak at 284.2
eV. Propylene adsorption on cuprous oxide has previously been
studied and was found to adsorb in its molecular state following
low-temperature and low-pressure exposures.13 In contrast,
exposure to 5 Torr of propylene at 300 K for 2 min results in
a carbon 1s spectrum (Figure 4, fit in Figure 5A,I) which can
be fit with two peaks (Gaussian/Lorentzian, fwhm ) 1.8 eV)
at 284.0 and 285.5 eV. In the 300-473 K temperature range
(Figure 4) the two-peak pattern is retained and a 284.0:285.5
peak intensity ratio of ∼2.2 is maintained even though there
are significant changes in total carbon 1s intensity (Table 2).
This coupled behavior of the two carbon 1s peaks indicates that
they are associated with the same surface species.
From the reaction mechanism of the oxidation of propylene
to acrolein presented in Scheme 1, as well as the binding
energies and peak intensity ratio of ∼2:1, the low-temperature
species is tentatively assigned as arising from a surface allyl
alkoxide (CH2dCHsCH2sO-) species.23 A close analogue
to this oxygenated surface species is allyl alcohol (C3H5OH),
whose carbon 1s spectrum when chemisorbed on cuprous oxide-
(100) as an alkoxide, has been shown to consist of two peaks
in a ratio of ∼2:1 at 285.1 and 286.4 eV, respectively.24 The
peaks in Figure 4 are shifted down in absolute binding energy,
but have a comparable intensity ratio and ∆ΒΕ of 1.5 (vs 1.3)
eV. On the basis of this initial assignment as an allyl alkoxide,
the 285.5 eV peak is assigned as originating from the oxygenated
carbon while the peak at 284.0 is assigned to the vinylic carbons.
The formation of an allyl alkoxide surface species requires
reduction of the surface; however, no reduction is observed in
In summary, the carbon 1s XPS data following propylene
exposures at 300-523 K on cuprous oxide indicate the presence
of an allyl alkoxide surface species. Reaction with propylene
at 573-623 K produces a new, higher binding energy peak
which is assigned as a carbonyl or carboxylate species and
(25) The relative peak intensities in the 573-623 K temperature range
(Table 2) indicate that there is excess carbon intensity (∼15-30%) in the
284.5 eV region (associated with a CxHy species). This carbon intensity
cannot be derived solely from the allyl alkoxide and acrolein or carboxylate
surface species which would be expected to have a 2:1 intensity ratio
between vinylic and oxygenated carbons. Similar increases in hydrocarbon-
based carbon 1s intensity have been observed following heating of cuprous
oxide surfaces dosed with acrolein26 and acrylic acid,27 and have been
proposed to originate from hydrocarbon fragments remaining from nonse-
lective oxidation and/or decomposition of oxygenated surface species.
(26) Schulz, K.; Cox, D. J. Phys. Chem. 1993, 97, 3555.
(27) Schulz, K.; Cox, D. F. J. Phys. Chem. 1992, 96, 7394.
(28) Gorshkov, A. P.; Kolchin, I. K.; Gribov, A. M.; Margolis, L. Ya.
Kinet. Katal. 1968, 9, 1086.
(29) Keulks, G. W.; Rosynek, M. P.; Daniel, C. Ind. Eng. Chem. Prod.
Res. DeV. 1971, 10, 140.
(23) Bakke, A. A.; Chen, H.-W.; Jolly, W. L. J. Electron Spectrosc. Relat.
Phenom. 1980, 20, 333.
(24) Schulz, K.; Cox, D. J. Phys. Chem. 1993, 97, 647.

11472 J. Am. Chem. Soc., Vol. 120, No. 44, 1998
Reitz and Solomon
Figure 5. (A) (I) Fitted carbon 1s spectrum of cuprous oxide exposed to 5 Torr of propylene at 300 K. (II) Exposed to 5 Torr of propylene at 623
K. (B) (I) Fitted carbon 1s spectrum of cupric oxide exposed to 5 Torr of propylene at 300 K. (II) Exposed to 5 Torr of propylene at 623 K.
Table 2. Relative Intensities of Carbon 1s Peaks Following
Exposure of Copper Oxide Surfaces to 5 Torr of Propylene
temp 284.0-284.5 eV 285.5-286.0 eV 288.2 eV
CxHy:COz
(error (0.2)
(K)
(CxHy)
(COz)
(COz)
A. Cu2O
1.0
1.0
1.0
1.0
1.0
3.6
2.86
300
373
423
473
523
573
623
1.81
2.46
1.92
2.12
2.30
13.5
8.90
0
0
0
0
0
1.0
1.0
1.81
2.46
1.92
2.12
2.30
2.93
2.31
B. CuO
1.0
1.14
1.31
0.82
0.78
300
353
423
523
573
1.92
2.26
2.68
1.88
2.19
0
1.92
1.06
1.16
1.03
1.23
1.0
1.0
1.0
1.0
increased hydrocarbon (CxHy) intensity from byproducts of
surface oxidation reactions.
3.B.2. Cupric Oxide. Figure 6 shows the carbon 1s XPS
spectra following 2 min, 5 Torr propylene exposures on CuO
at variable temperatures. As discussed for cuprous oxide, the
signal-to-noise ratio of the carbon 1s peaks is due to low
coverage of adsorbate species and a reduced number of signal-
averaged scans due to the relatively quick reaction of propylene
with the copper oxide surfaces at 300 K. Additionally, for CuO,
which was found to react faster, the energy step size was
increased in order to rapidly collect accurate data.
Figure 6. Carbon 1s XPS spectra following exposure of the cupric
oxide surface to propylene at variable temperatures.
estimated to be too small to unambiguously detect (∼3%
decrease in the Cu(I) LMM signal), and it seems likely that the
amount of reduction on the cupric surface also falls below the
detection limit of the Cu 2p spectrum. However, in the CuO
case, we can take advantage of the M-edge resonance photo-
emission process to determine the presence and magnitude of
surface reduction associated with formation of the allyl alkoxide
surface intermediate. M-edge resonance photoemission effects
arise from interference between the two channels of photoemis-
sion intensity shown in Scheme 3.30-32
In the first process (a), at photon energies corresponding to
the Cu 3p absorption edge (∼76 eV), a metal 3p electron is
(30) Davis, L. C.; Feldkamp, L. A. Phys. ReV. B 1981, 23, 6239.
(31) Davis, L. C.; Feldkamp, L. A. Phys. ReV. Lett. 1980, 44, 673.
(32) Thuler, M. R.; Benbow, R. L.; Hurych, Z. Phys. ReV. B 1982, 26,
669.
In Figure 6 the 300 K spectrum contains two peaks in a 2.4:1
intensity ratio at 284.0 and 285.5 eV (Figure 5B,I). On the
basis of the binding energies and intensity ratio, as in the cuprous
oxide case, this surface species is assigned as an allyl alkoxide
intermediate. Following the 300 K exposure, the carbon 1s
intensity of alkoxide on cupric oxide is equivalent to that on
the cuprous oxide surface.
As discussed above, the formation of an allyl alkoxide in-
termediate would require reduction of the surface, but no
reduction is observed in the Cu(II) 2p XPS region at 300 K. In
the cuprous oxide case, the amount of surface reduction was

Propylene Oxidation on Copper Oxide Surfaces
J. Am. Chem. Soc., Vol. 120, No. 44, 1998 11473
Scheme 3
promoted into the hole in the 3d orbital. This intermediate state
decays via an Auger process whereby a 3d electron fills the 3p
hole and a second 3d electron is ejected. This Auger process
involves a super-Coster-Kronig transition (SCK) in which both
the initial and final state holes reside in the same principal
quantum shell (n ) 3). This final state is the same as that
obtained from the second channel, direct 3d photoemission (b),
and thus, at the Cu 3p absorption edge, the two channels of
intensity produce resonance and interference effects. This
process requires a hole in the 3d band and does not occur for
the 3d10 Cu(I) configuration. Therefore, just as was the case
for the Cu(II) 2p satellite, these resonance enhancements are
characteristic of Cu(II) and can be used to probe the presence
and amount of Cu(II) in the sample. An added benefit of this
technique is the low kinetic energy (∼64 eV) of the ejected
electrons which provides a good surface probe (5 Å escape
depth)22 and thus higher sensitivity for detecting reduction of
the Cu(II) surface sites.
Figure 7 shows the satellite region of the CuO valence band
photoemission spectrum (on-resonance, hν ) 76 eV) before and
after exposure to 1 Torr of propylene for 2 min at 300 K. The
observed decrease in the resonance-enhanced peak intensity
indicates a decrease in the Cu(II) content of the sample and
thus surface reduction. This amount of reduction would
correspond to only a small decrease in the Cu(II) peaks of the
2p XPS spectrum and would not be easily detectable by XPS.
Reaction of the CuO surface with propylene at elevated
temperature (350-573 K) results in the appearance of a new
carbon 1s peak at 288.4 eV (Figure 5B,II), with the balance of
the region being fit with peaks at 284.5 and 286.1 eV which
correlate with aliphatic carbons and an allyl alkoxide surface
species, respectively. Following these elevated temperature
exposures, at comparable temperatures the total adsorbate peak
intensities on CuO are less than those on the cuprous oxide
surface.
As in the cuprous oxide case, the position of this new, higher
binding energy peak is consistent with assignment as an
aldehydic or carboxylate species, and on CuO it appears at a
much lower temperature (350 K) than observed on the cuprous
oxide surface (573 K). From the CxHy:COz ratios listed in Table
2, it can be seen that the reaction of propylene with CuO results
in a higher degree of oxidation of the remaining surface species
versus the cuprous oxide case. The low ratio of ∼1.1 indicates
that the oxygenated surface species remaining after reaction with
propylene cannot be simply identified as alkoxide, acrolein, or
carboxylate species which should yield CxHy:COz ratios of 2.0.
This increased degree of oxidation versus cuprous oxide at
comparable amounts of surface reduction indicates that cupric
oxide is more active at incorporating oxygen into surface
hydrocarbon species. This observation is consistent with studies
conducted on the catalytic behavior of CuO toward oxidation
of propylene which indicate that CuO forms mainly complete
oxidation products (CO, CO2).4-6
Figure 7. hν ) 76 eV (on-resonance) valence band satellite spectra
of CuO before (solid) and after (dashed) exposure to propylene at
300 K.
for a given number of oxidized propylene molecules. An upper
limit on this contribution can be estimated from the two limiting
cases of propylene to acrolein conversion on Cu2O versus the
combustion of propylene to three CO2 molecules on CuO.
Propylene to acrolein is a four-electron reduction of the surface
while oxidation of one propylene to three carbon dioxides would
be an eighteen-electron reduction resulting in CuO having a
4.5-fold increase in reduction for the same amount of propylene
oxidized. This cannot fully account for the 8-10-fold increase
in reduction (Table 1) and indicates fundamental differences in
the rates of reaction and possibly the nature of the reaction
mechanism of propylene oxidation on cuprous versus cupric
oxide.
In summary, as in the Cu2O case, the oxidation of propylene
on CuO proceeds through an allyl alkoxide intermediate. Higher
temperature reactions (>350 K) result in the formation of
chemisorbed aldehydic or carboxylate species, which were only
observed on the cuprous oxide surface following reaction at
573 K. Additionally, the surface species on cupric oxide have
a higher degree of oxidation than would be expected from the
formation of acrolein (or carboxylate) and alkoxide species,
indicating a greater extent of reaction than on the cuprous oxide
surface.
3.C. Temperature-Programmed Desorption. 3.C.1. Cu2O.
Temperature-programmed desorption (TPD) studies of propy-
lene on cuprous oxide have been reported.12,13 Of particular
interest in this study are the similarities and differences in the
desorption behavior and reactivity of the alkoxide species
formed following reaction of the cuprous and cupric oxide
surface with 5 Torr of propylene at 300 K. Figure 8A shows
the TPD plots for the cuprous oxide surface for the parent
masses of acrolein (m/e ) 56) and CO2 (m/e ) 44). Interest-
ingly, both oxidation products are observed to desorb at ∼365
K, and application of the Redhead equation33 with a preexpo-
nential factor of 1013 s-1 yields an activation energy of
desorption of 24.5 kcal/mol (102.6 kJ/mol).
Similar to the carbon 1s XPS data, the TPD data have a low
signal-to-noise ratio due to the low adsorbate coverage. Also,
the geometry of the experimental chamber results in contribu-
tions to the TPD data from sources other than the sample (i.e.,
In light of this more extensive oxidation of propylene on CuO
versus Cu2O it is necessary to further evaluate the observed
differences in reduction observed in section 3.A. If oxidation
of propylene proceeds to a greater extent on CuO (i.e., formation
of COx species), then more surface reduction will be observed
(33) Redhead, P. A. Vacuum. 1962, 12, 203.

11474 J. Am. Chem. Soc., Vol. 120, No. 44, 1998
Reitz and Solomon
Scheme 4
By applying the Redhead equation,33 a 425 K desorption
corresponds to an activation energy of desorption of 28.7 kcal/
mol (120.1 kJ/mol). As discussed above for Cu2O, the
formation and desorption of these products at comparable
temperatures occurs through a reaction-limited process which
is again assigned as alkoxide decomposition to an aldehyde.
The higher energy of desorption on CuO indicates that this
decomposition step is more facile on the cuprous oxide surface.
3.D. Kinetics of the Propylene Oxidation Reaction. The
carbon 1s XPS and TPD results reveal some important informa-
tion about the relative order and rate of the various steps in the
oxidation of propylene on the copper oxide surfaces. As
discussed previously, the mechanism of the oxidation of
propylene can be broken down into the four steps defined in
Scheme 4. These steps consist of an initial H atom abstraction
to form the symmetric allylic intermediate (1), oxide incorpora-
tion (2), a second hydrogen abstraction and formation of the
carbonyl bond (3), and finally further oxidation to nonselective
oxidation products (4).
From the surface reduction and carbon 1s XPS data, we can
draw some conclusions about the mechanism of reaction on both
surfaces as well as the relative rates of the individual steps.
Extensive research3,8 involving kinetic isotope experiments has
been conducted to determine the rate-determining step (RDS)
of the catalytic oxidation of propylene on cuprous oxide. The
work revealed that the initial H atom abstraction is likely the
RDS of the reaction and hence relatively slow. The next step,
oxygen incorporation, is expected to be a relatively fast step,
and the absence of detectable levels of allyl on the surface in
the present study supports this expectation.
The observation of significant amounts of surface alkoxide
reveals some important information about the relative rate of
the third step. Figure 9 shows the concentration of surface
reaction species with time (propylene, allyl, alkoxide, acrolein)
if further oxidation is ignored. In this study, it is assumed that
propylene gas is in excess and quickly results in a monolayer
of adsorbed propylene which then undergoes reaction. The rate
and overall effect of reverse reactions are assumed negligible,
and the rate of each step is assumed first order in reactants.
While these are significant approximations which are not likely
to be strictly correct, the aim of this analysis is an explanation
of the presence and relative amount of the experimentally
observed surface species. With these assumptions and having
a time axis which parallels the experimental conditions (2 min,
5 Torr propylene exposure and 5 min pump out), it is found
that the rate of step 3 (k3 in Scheme 4) can only be marginally
faster than the rate of the initial H atom abstraction (k1). In
fact k3/k1 must be less than 2 and could, in fact, be less than 1
in order to produce the observed amount of surface alkoxide at
the data collection time. Thus, its relative rate is on the same
order as that of the initial H atom abstraction, i.e., slow.
The final step in the oxidation of propylene (k4) can be
regarded as further oxidation of the alkoxide decomposition
product. As discussed in the TPD results, CO2 was observed
to evolve from the surface as a reaction-limited process which
was assigned as the decomposition of the surface alkoxide. Thus,
the steps following alkoxide decomposition are fast.
Figure 8. (A) Temperature-programmed desorption (TPD) curves of
the 5 Torr, 300 K propylene-exposed cuprous oxide surface. (B) TPD
curves of the CuO surface following exposure to 5 Torr of propylene
at 300 K.
the sample holder). The desorption behavior of the sample
holder without a copper oxide sample was determined, and it
was found that the shoulder peak at ∼400 K in the m/e ) 44
curve (Figure 8A) is due to the sample holder. No m/e ) 56
desorption signal could be observed without the copper oxide
sample.
34
Acrolein26 and CO2 have been observed to desorb much
lower than 300 K on cuprous oxide, thus indicating that the
observed desorption of these products at 365 K likely arises
from a reaction-limited process. From the mechanism of
acrolein formation over cuprous oxide, as well as the extensive
data available on alkoxide decomposition on surfaces,24,35 this
rate-limiting step is assigned as the formation of acrolein from
the allyl alkoxide species. This acrolein can desorb or be further
oxidized to CO2 and other nonselective oxidation products in a
subsequent fast step.
3.C.2. CuO. Figure 8B shows the results of TPD experi-
ments following a 5 Torr, 300 K propylene exposure on the
CuO surface. Acrolein and CO2 again desorb together, though
at 425 K, which is 60 K higher than the desorption on the
cuprous oxide surface. The intensities of the m/e ) 44 and 56
desorption curves from cupric oxide were found to be greater
than those from the cuprous oxide surface. Also, as in the
cuprous oxide case, there is a small contribution to the m/e )
44 signal from the sample holder (∼15% of the total intensity,
centered at 400 K).
(34) Reitz, J. B.; Solomon, E. I. Unpublished results.
(35) Bowker, M.; Madix, R. J. Surf. Sci. 1982, 116, 549.

Propylene Oxidation on Copper Oxide Surfaces
J. Am. Chem. Soc., Vol. 120, No. 44, 1998 11475
Scheme 5
bond dissociation energy (BDE) for the methyl C-H bond of
propylene is ∼87 kcal/mol (364.2 kJ/mol), and the pKa is
∼48.17,37 This relatively weak C-H bond, which is a conse-
quence of resonance stabilization, makes its cleavage an
attractive process for activation of the propylene molecule.
Mayer38-41 has recently shown, through the application of a
thermodynamic cycle originally developed by Brauman42 and
adapted by Bordwell43,44 for determining bond dissociation
energies in solutions, that the driving force for the C-H bond
cleavage can be related to the C-H bond strength and the
thermodynamic affinity of the surface for an H atom. Abstrac-
tion of an H atom can be thought of as the simultaneous transfer
of a H+ to the surface oxide (in the current system) and an
electron to a surface cation. Mayer and co-workers have shown
that, for C-H bond cleavage catalyzed by permanganate41 and
chromyl chloride,39,40 the high bond strength of the resulting
O-H bond, as well as the affinity of the substrate for an electron
(Mn+ + e- f M(n-1)+), provides a thermodynamic driving force
for formation of the activated hydrocarbon. Bordwell has
derived the equation
Figure 9. Concentration of reaction intermediates with time assuming
k1 ) 1.0, k2 ) 100, and k3 ) 1.5.
In summary, the relative rates of the various steps in the
oxidation of propylene on the copper oxide surfaces (Scheme
4) are k1, slow; k2, fast; k3, slow; and k4, fast.
BDE(M(n-1)+O-H) (kcal/mol) ) 1.37pKa + 23.1E° + C
4. Discussion
(4.1)
4.A. Propylene Oxidation on Copper Oxides: Contribu-
tions to Reactivity. The results presented reveal important
differences between the copper oxides in terms of their reactivity
in propylene oxidation. Oxidation of propylene in the absence
of reoxidizing dioxygen results in reduction of the copper oxide
surface, and from the reduction data presented in section 3.A,
the reaction with CuO was found to produce ∼10 times the
amount of surface reduction as detected on Cu2O at comparable
reaction temperatures (573 K). This temperature-dependent
reduction, which is measured following a constant reaction time
and constant propylene pressure, is related to the rate of the
propylene oxidation reaction. Both the greater degree of reduc-
tion and lower activation energy obtained from the Arrhenius
plots indicate an enhanced rate of reaction for the cupric oxide
surface.
As discussed above, it has been well established that the
oxidation of propylene on cuprous oxide proceeds through an
initial rate-determining step3,36 which consists of cleavage of a
C-H bond on the methyl group to form the symmetric allylic
intermediate. To compare the rate of propylene oxidation on
the two copper oxides, the reaction mechanism on CuO must
proceed through the same rate-limiting H atom abstraction step.
This similar reaction mechanism appears likely from the carbon
1s data presented in section 3.B which shows the presence of
the same alkoxide intermediate on both surfaces. This inter-
mediate is indicative of an H atom abstraction followed by oxide
insertion into the resulting allyl intermediate.
which relates the BDE of the resulting surface O-H bond to
its pKa and the redox potential.
From this equation the affinity of the substrate for a proton
is related to the acid dissociation constant, pKa, of the resulting
one-electron-reduced, protonated M(n-1)+O-H species, while
the affinity for a electron is represented by the potential of the
M
n+O/M(n-1)+O redox couple. For different surfaces, a stronger
BDE means a greater thermodynamic driving force. Application
of this equation to our systems and adaptation to the scheme
developed by Mayer38-41 results in the expressions shown in
Scheme 5. The last two equations involve aqueous values and
are thus not directly applicable to the heterogeneous system;
however, these differences should be relatively constant between
the Cu2O and CuO substrates, and thus the H atom affinities
and corresponding relative BDEs can be compared.
While experimentally determined redox potentials have been
reported for cuprous oxide,45 they have not been reported for
solid cupric oxide. Theoretical calculations based on energies
(37) Kemp, D. S.; Vellaccio, F. Organic Chemistry; Worth Publishers:
New York, 1980.
(38) Gardner, K.; Kuehnert, L.; Mayer, J. Inorg. Chem. 1997, 36, 2069.
(39) Cook, G. K.; Mayer, J. M. J. Am. Chem. Soc. 1995, 117, 7139.
(40) Cook, G. K.; Mayer, J. M. J. Am. Chem. Soc. 1994, 116, 1855.
(41) Gardner, K. A.; Mayer, J. M. Science 1995, 269, 1849.
(42) Janousek, B.; Reed, K.; Brauman, J. J. Am. Chem. Soc. 1980, 102,
3125.
(43) Bordwell, F.; Cheng, J.-P.; Ji, G.-Z.; Satish, A.; Zhang, X. J. Am.
Chem. Soc. 1991, 113, 9790.
(44) Bordwell, F.; Zhang, X.-M.; Satish, A. V.; Cheng, J.-P. J. Am. Chem.
Soc. 1994, 116, 6605.
The overall rate of the homolytic cleavage of a C-H bond
is directly proportional to the strength of the C-H bond.37 The
(36) Imachi, M.; Kuczkowski, R.; Groves, J. T.; Cant, N. J. Catal. 1983,
82, 355.
(45) Rickert, H. Electrochemistry of Solids; Springer-Verlag: Berlin,
1982.

11476 J. Am. Chem. Soc., Vol. 120, No. 44, 1998
Reitz and Solomon
Scheme 6
Scheme 7
of formation are available46 and yield values of 0.758 and 0.550
eV for the one-electron reductions of Cu2O and CuO at 25 °C,
respectively. The relative order of these redox potentials,
pKa and binding energies are also available for permangan-
ate38,53 as well as protonated Fe2O3 and TiO2 surfaces19,54
(Scheme 7), which are potentially more applicable to the current
system. However, the iron and titanium metal oxide values were
obtained by using oxygen 1s XPS to monitor the ion exchange
of K+ with the hydrated surfaces, and these surfaces were found
to have significant contamination by oxygen-containing hydro-
carbons, and thus the absolute pKa values are subject to possible
error. As can be seen, both TiO2 and Fe2O3, which have
comparable binding energies, have similar pKa values, indicating
a reasonable correlation between oxygen 1s binding energy and
pKa. These values result in a correlation of +0.335 log unit
per 0.1 eV shift to lower binding energy.
From these two approximations, a ∆pKa (){pKa(CuO) - pKa-
(Cu2O)}) of +9 to +3.0 log units is obtained based on the
oxygen 1s binding energy difference (+0.9 eV) for cuprous
relative to cupric oxide. An intermediate value of +6 log units
is used below for the difference in pKa of the two protonated
surfaces. Furthermore, the pKa(Cu2O-H) will be approximated
as 11.5 which correlates to the experimentally determined values
for Fe2O3 and TiO2 which have similar oxygen 1s binding
energies for the unprotonated surface.
Thus, the redox potentials indicate that Cu2O has a greater
affinity for an electron, while the pKa values obtained from the
oxygen 1s binding energies indicate that CuO has a greater
affinity for a proton due to the increased basicity of its oxides.
Substituting these pKa values and redox values into eq 4.1 gives
relative BDEs of 90.2 kcal/mol (377.6 kJ/mol) for cuprous oxide
and 93.7 kcal/mol (392.2 kJ/mol) for CuO, with the difference,
∆BDE, of 3.5 kcal/mol (14.6 kJ/mol) being the important
quantity. Thus, CuO will form a stronger O-H bond, and
therefore has an overall higher affinity for the H atom. This
high affinity provides a greater thermodynamic driving force
for the H atom abstraction reaction on cupric versus cuprous
oxide. Comparison of the BDE(O-H) and rates of reaction
for RO• radicals reveals that a ∆BDE of +3 kcal/mol can result
in an ∼102 increase in k.39 Thus, the calculated difference in
BDEs for cuprous and cupric oxide is consistent with a
significant difference in the rate of the H atom abstraction
reaction on the two surfaces.
In summary, the rate of oxidation of propylene on the copper
oxide surfaces is proportional to the rate of the initial, rate-
determining H atom abstraction step. The increased reduction
on CuO (∼10× at 573 K and ∆Ea of 5.9 kcal/mol (24.7 kJ/
mol)) indicates a more favorable H atom abstraction step. The
rate of the H atom abstraction has been shown to be related to
the affinity of the substrate for the electron and proton, where
a higher affinity provides a larger thermodynamic driving force
for the reaction. While redox potentials, which are a measure
of electron affinity, favor H atom abstraction on cuprous oxide,
the increased base strength of the oxides of the Cu(II) surface
increases the proton affinity of CuO. This stronger Bronsted
base character facilitates proton transfer from the propylene
methyl group and therefore increases the rate of the H atom
abstraction step relative to cuprous oxide.
namely, that the Cu1+/Cu0 couple is more favorable than Cu2+
Cu1+ is consistent with the aqueous values47 of 0.522 (Cu1+
/
/
Cu0) and ∼0.160 (Cu2+/Cu1+) eV, as well as experimentally
determined values for molten salts of copper halides.46 Thus,
the redox potential term of the Bordwell equation favors the
cuprous oxide substrate.
In determining the pKa of the protonated copper oxides, the
acidity of the protonated oxide in the one-electron-reduced case
(pKa of Cu(n-1)+O-H) is taken to be similar to that of the
original, oxidized substrate (pKa of Cun+O-H). This ap-
proximation is valid since in both copper oxides each oxide
center is surrounded by four cations, and the reduction of only
one of these should have a minor effect on the acidity of the
protonated oxide.
No experimental data have been reported for the pKa of
protonated copper oxide surfaces; however, it has been observed
by Shirley,48,49 through XPS studies of alcohols, that the binding
energy of the oxygen 1s orbital, which is a direct measure of
the atomic charge on the oxygen, can be directly related to the
basicity of the oxygen atom. A more negative oxygen, charac-
terized by a lower binding energy, will be a stronger Bronsted
base. Thus, the oxygen 1s binding energies of the oxides of
cupric and cuprous can be used to estimate the pKa values.
To compare the binding energies of the oxygen 1s orbital of
cuprous and cupric oxide, the observed binding energies must
be relaxation-corrected.50-52 For the same atom in different
chemical environments, the relative difference in the relaxa-
tion term, ∆R, can be obtained from the Auger parameter, R,
where ∆R ) 2∆R. In this case the Auger parameter is given
by R(oxide) ) BE(O 1s) + KE(O KLL Auger). The experi-
mentally determined Auger parameters for cuprous and cupric
oxide are 1042.1 and 1041.9 eV, respectively. This corresponds
to ∆R (R(Cu2O) - R(CuO)) ) 0.1 eV, and applying this
correction to experimentally determined oxygen 1s binding
energies yields relaxation-corrected oxygen 1s binding energies
of 530.3 eV for cuprous oxide and 529.4 eV for cupric oxide.
Both the relative difference in the relaxation term and the
binding energies are in good agreement with previously reported
values.20 The lower oxygen 1s binding energy for the cupric
case indicates that its oxide has a more negative charge than in
cuprous oxide and is therefore a stronger base with a cor-
respondingly higher pKa for the protonated species.
Determining the relationship between absolute pKa values and
differences in oxygen 1s binding energies requires a correlation
to known oxide systems. Sulfur acids19,37 are one class of
compounds which have known, accurate pKa values and binding
energies (Scheme 6). From these values, we obtain a correlation
of +1.0 pKa unit per every 0.1 eV shift to lower binding energy
of the oxygen 1s XPS peak.
(46) Bertocci, U.; Turner, D. In Encylcopedia of Electrochemistry of the
Elements; Bard, A. J., Ed.; Marcel Dekker Publishing: New York, 1974.
(47) CRC Handbook of Chemistry and Physics, 70th ed.; Weast, R. C.,
Lide, D. R., Eds.; CRC Press: Boca Raton, 1990.
(48) Martin, R. L.; Shirley, D. A. J. Am. Chem. Soc. 1974, 96, 5299.
(49) Shirley, D. A. J. Electron Spectrosc. Relat. Phenom. 1974, 5, 135.
(50) Thomas, T. J. Electron Spectrosc. Relat. Phenom. 1980, 20, 117.
(51) Chasse, T.; Franke, R.; Streubel, P.; Meisel, A. Phys. Scr. 1992,
T41, 281.
(53) Reinert, F.; Steiner, P.; Blaha, P.; Claessen, R.; Zimmerman, R.;
Hufner, s. J. Electron Spectrosc. Relat. Phenom. 1995, 76, 671.
(54) Simmons, G. W.; Beard, B. C. J. Phys. Chem. 1987, 91, 1143.
(52) Wagner, C. D. Faraday Discuss. Chem. Soc. 1975, 60, 291.

Propylene Oxidation on Copper Oxide Surfaces
J. Am. Chem. Soc., Vol. 120, No. 44, 1998 11477
4.B. Alkoxide Decomposition. As determined from the
TPD experiments which are a direct probe of the reactivity of
the two copper oxide surfaces toward decomposition of the
surface alkoxide, the cuprous oxide surface has a lower
activation energy and consequently faster rate of reaction. To
determine the origin of this enhanced activity on the cuprous
oxide surface for this subsequent reaction step, it is useful to
evaluate two possible mechanisms for the decomposition of a
surface alkoxide, namely, H atom abstraction and hydride
elimination from the R-carbon.
In evaluating the appropriateness of an H atom abstraction
step from the alkoxide, two factors must be considered: the
bond dissociation energy of the C-H bond to be homolytically
cleaved and the affinity of the surface toward the abstracted H
atom. While the BDE of the R-C-H bond of allyl alkoxide is
not known accurately, the BDE for allyl alcohol is known and
is ∼5 kcal/mol (20.9 kJ/mol) weaker than that of the methyl
C-H bond of propylene.55 If we assume the surface alkoxide
to be similar to the parent free alcohol, then this H atom
abstraction mechanism is certainly viable from a BDE stand-
point. In section 4.A we evaluated the affinity of the copper
oxide surface for an H atom and found that the cupric oxide
surface has a greater affinity for an H atom by virtue of the
greater Bronsted base strength of its oxides. Thus, one would
predict that the CuO surface would be more facile at the alkoxide
decomposition reaction through an H atom mechanism. This
is at odds with the TPD results which show that the cuprous
oxide surface is in fact more facile. Thus, it seems unlikely
that an H atom abstraction mechanism is active in alkoxide
decomposition.
The hydride elimination mechanism has been considered in
the literature35 to be the relevant pathway for the decomposition
of surface alkoxides. The propensity for a surface to perform
a hydride elimination can be related to the affinity of the surface
for a hydride, in this case represented by the strength and
stability of the resulting Cun+-H- complex. A review of the
literature reveals that Cu(I)-H- complexes are well-known and
characterized,56 while those of Cu(II) are not. From this
standpoint, it would be predicted that the stability of the
Cu(I)-hydride complex could serve as a driving force for
hydride elimination and thus result in a faster rate of alkoxide
decomposition as observed in the TPD data. The absolute value
of the contribution this stronger copper-hydride complex could
have on the overall rate of the reaction step is unknown. Other
factors such as the nature of the alkoxide-surface bond and
the availability of surface sites for reaction can also have a
significant role in the rate of alkoxide decomposition.
4.C. Nonselective Oxidation. The carbon 1s XPS data in
section 3.B revealed that the surface species formed on CuO
following reaction with propylene have a much greater extent
of oxidation than those on the cuprous oxide surface at similar
reaction temperature and comparable amounts of surface reduc-
tion. This higher proportion of more fully oxidized surface
species indicates that the oxidation of propylene proceeds to a
greater extent on the cupric oxide surface. To understand the
origin of this increased reactivity toward nonselective oxidation
products, it is useful to consider possible mechanisms for the
formation of these products.
Scheme 8
formation of complete oxidation products from propylene on
bismuth molybdate partial oxidation catalysts likely proceeds
through a consecutive oxidation reaction mechanism,28,29 the
formation of higher oxidation products (CO, CO2) has also been
proposed to occur through a parallel mechanism17 involving
partially reduced surface oxygen species. These electrophilic
species can attack adsorbed hydrocarbons in regions of high
electron density (i.e., π-bonds), resulting in C-C cleavage and
nonselective oxidation. Two likely sources of surface electro-
philic oxygen species include the partial reduction of adsorbed
molecular oxygen and metal oxide decomposition. In the case
of cuprous oxide, both sources can be ruled out since the reaction
with propylene occurs in the absence of gas-phase oxygen and
the cuprous oxide lattice is stable to decomposition at the
relevant reaction temperatures. Thus, contributions to product
distributions from a parallel electrophilic reaction mechanism
will be negligible. CuO on the other hand is unstable in a
vacuum at the reaction temperatures, decomposing to Cu2O. The
possible contribution from an electrophilic oxidation reaction
mechanism can be evaluated from the surface species distribu-
tions at two reaction temperatures. At 350 K, CuO is relatively
stable toward decomposition (20 h of heating results in a 10%
decrease in Cu(II) content), and thus one would expect only a
small contribution to reactivity from an electrophilic oxygen
species. At 573 K, thermal decomposition during propylene
exposure results in a 20% decrease in the Cu(II) content and a
higher amount of electrophilic species, and thus a greater con-
tribution to the overall reaction would be expected. If a parallel
electrophilic reaction mechanism were occurring, the differences
in amounts of electrophilic species at different reaction condi-
tions would be expected to produce differences in the distribu-
tion of reaction products on the surface. However, from the
carbon 1s data in Table 2 the product distribution is virtually
identical at 350 and 573 K, indicating no alteration in the relative
contributions from competing reaction mechanisms. Thus, no
significant contribution from an electrophilic mechanism appears
to be present in copper oxides in the absence of O2.
The formation of complete combustion products originates
primarily from the further oxidation of acrolein, and the CuO
surface is found to be more active as evidenced by its greater
distribution of more fully oxidized surface species. This
formation of complete combustion products from aldehydes has
been proposed to occur through the formation of carboxylate
surface intermediates.28,29 There are two possible mechanisms
for their formation (Scheme 8).
Reaction a in Scheme 8 proceeds via an H atom abstraction
from the R-carbon of acrolein, followed by oxide insertion to
form the carboxylate. The BDE of this C-H bond is ∼87 kcal/
mol55 (364.2 kJ/mol), and thus this reaction is reasonably
accessible for both surfaces. As discussed earlier, the relative
ability of each of the copper oxide surfaces to abstract an H
atom is related to the thermodynamic affinity for the H atom.
In section 4.A it was determined that the CuO surface has a
greater affinity for the H atom by virtue of a greater negative
charge on its surface oxides and thus a greater Bronsted base
strength. Thus, this surface should be more capable of further
oxidizing acrolein by an H atom abstraction mechanism.
Reaction b in Scheme 8 proceeds by an initial nucleophilic
attack by a surface oxide on the R-carbon to directly yield a
dialkoxide species which can then eliminate a hydride to form
In this discussion it is assumed that higher oxidation products
originate via a consecutive mechanism involving the further
oxidation of acrolein. While it has been determined that the
(55) McMillen, D. F.; Golden, D. M. Annu. ReV. Phys. Chem. 1982, 33,
493.
(56) Libowitz, G. G. The Solid State Chemistry of the Binary Metal
Hydrides; W. A. Benjamin: New York, 1965.

11478 J. Am. Chem. Soc., Vol. 120, No. 44, 1998
Reitz and Solomon
a carboxylate. As in reaction a in Scheme 8, this mechanism
would be facilitated by a surface oxide with a greater negative
charge and thus would also be expected to proceed faster on
the cupric oxide surface.
As in the case of the other mechanistic steps of propylene
oxidation on copper oxides examined here, other factors can
also be expected to have a significant effect on the rate of the
further oxidation of acrolein. These include the nature of the
acrolein-surface bond and the availability of surface oxide.
From the unit cells and stoichiometry of cuprous and cupric
oxide,57 one would expect cupric oxide to have a greater
concentration of surface oxides, resulting in a positive effect
on the rate.
Thus, in the further oxidation of acrolein to complete combus-
tion products on copper oxide surfaces, both H atom abstraction
and nucleophilic addition would be affected by the oxide charge.
The more negative charge on the surface oxides of CuO results
in greater nucleophilic and Bronsted base strength and thus an
enhanced rate for the nonselective oxidation reaction.
4.D. Product Distribution and Selectivity. The observed
differences in selectivity, with cupric oxide favoring more
complete oxidation, originates from either differences in reaction
mechanism or differences in reaction rates. From the carbon
1s XPS and TPD data, there are similar reaction intermediates
under the present experimental conditions. Thus, a similar
mechanism of propylene oxidation appears to occur on both
surfaces. The observed differences in selectivity and product
distribution therefore likely originate from differences in relative
rates of reaction.
Bronsted base strength of its oxides. In contrast, the alkoxide
decomposition step is favored on the cuprous oxide surface,
and this likely originates from the more stable Cu(I)-hydride
complex formed following a hydride elimination reaction.
Further oxidation of acrolein proceeding through a H atom or
nucleophilic reaction is favored by cupric oxide once again due
to the greater negative charge on its oxides which increases both
Bronsted base and nucleophilic strength.
As discussed in section 3.D, the rate of the initial H atom
abstraction can only be marginally slower than the rate of the
alkoxide decomposition to account for the observed relatively
high levels of surface alkoxide. In cupric oxide, the initial pro-
pylene activation step is accelerated, but the alkoxide decom-
position step is slowed; thus, there is a shift in the slow step in
the reaction from the initial H atom abstraction to the alkoxide
decomposition. It is the relative rates of these two steps plus
the accelerated further oxidation reaction which contribute to
the observed differences in selectivity of cuprous and cupric
oxide.
Acknowledgment. This research is supported by the NSF
MRL program at the Center for Materials Research at Stanford
University. Support for the work performed at the Stanford
Synchrotron Radiation Laboratory, which is operated by the
Department of Energy Division of Chemical Sciences, is
acknowledged. Brad Reitz also thanks the NSF for a graduate
fellowship.
JA981579S
As discussed in section 4.A the initial H atom abstraction is
favored by the cupric oxide surface by virtue of the greater
(57) Wyckoff, R. W. G. Crystal Structures; Interscience Publishers: New
York, 1948.