Time-resolved studies for the mechanism of reduction of copper oxides with carbon monoxide: Complex behavior of lattice oxygen and the formation of suboxides

Article abstract of DOI:10.1021/jp040366o

Temperature-programmed reduction (TPR), synchrotron-based time-resolved X-ray diffraction (TR-XRD), extended X-ray absorption fine-structure (EXAFS), and X-ray absorption near-edge structure (XANES) were employed in this work to systematically study the reaction of Cu₂O and Cu₂O with CO gas molecules. Both TPR and isothermal reduction results showed that CuO was easier to reduce than Cu₂U under the same reaction conditions. In situ measurements of XRD and XANES showed a direct transformation pathway for CuO reduction (CuO a?? Cu) when there was a large supply of CO, while they showed a sequential step pathway involving one intermediate (CuO a?? Cu ₂O a?? Cu) with a limited supply of CO. An induction period for CuO reduction was seen and increased with decreasing temperature. O vacancies in CuO were observed during the induction period that could foster formation of a nonstoichiometric metastable copper-oxide species. The metastable species can either react rapidly with CO to form metallic copper during high CO flow rates or can relax into a Cu₂O lattice when the supply of CO is limited. The Cu₂O-like intermediate showed extra disordered oxygen in the empty tetrahedral sites of Cu₂O. Based on these observations, a possible mechanism for CuO reduction is proposed. This study demonstrates that the mechanism for the reduction of an oxide by CO can vary considerably with the experimental conditions (gas flow rate, temperature, sample size, etc.), and its complex kinetics cannot be described by a single n^th-order expression over the entire range of reaction. The behavior of CuO-based catalysts is discussed in the light of these results.

Full text of DOI:10.1021/jp040366o

J. Phys. Chem. B 2004, 108, 13667-13673
13667
Time-resolved Studies for the Mechanism of Reduction of Copper Oxides with Carbon
Monoxide: Complex Behavior of Lattice Oxygen and the Formation of Suboxides
Xianqin Wang,† Jonathan C. Hanson,† Anatoly I. Frenkel,‡ Jaw-Yong Kim,†,§ and
Jose´ A. Rodriguez*,†
Chemistry Department, BrookhaVen National Laboratory, Upton, New York, 11973, Physics Department,
YeshiVa UniVersity, New York, New York 10016, and Beamline DiVision, Pohang Accelerator Laboratory,
Pohang, 790-784, South Korea
ReceiVed: May 20, 2004; In Final Form: June 22, 2004
Temperature-programmed reduction (TPR), synchrotron-based time-resolved X-ray diffraction (TR-XRD),
extended X-ray absorption fine-structure (EXAFS), and X-ray absorption near-edge structure (XANES) were
employed in this work to systematically study the reaction of CuO and Cu2O with CO gas molecules. Both
TPR and isothermal reduction results showed that CuO was easier to reduce than Cu2O under the same
reaction conditions. In situ measurements of XRD and XANES showed a direct transformation pathway for
CuO reduction (CuO f Cu) when there was a large supply of CO, while they showed a sequential step
pathway involving one intermediate (CuO f Cu2O f Cu) with a limited supply of CO. An induction period
for CuO reduction was seen and increased with decreasing temperature. O vacancies in CuO were observed
during the induction period that could foster formation of a nonstoichiometric metastable copper-oxide species.
The metastable species can either react rapidly with CO to form metallic copper during high CO flow rates
or can relax into a Cu2O lattice when the supply of CO is limited. The Cu2O-like intermediate showed extra
disordered oxygen in the empty tetrahedral sites of Cu2O. Based on these observations, a possible mechanism
for CuO reduction is proposed. This study demonstrates that the mechanism for the reduction of an oxide by
CO can vary considerably with the experimental conditions (gas flow rate, temperature, sample size, etc.),
and its complex kinetics cannot be described by a single nth-order expression over the entire range of reaction.
The behavior of CuO-based catalysts is discussed in the light of these results.
1. Introduction
CO oxidation is of prime importance to fundamental studies
in catalysis, pollution control, and feed purification for fuel cell
applications. To meet increasingly stringent environmental
regulations in exhaust emission control, the complete oxidation
of carbon monoxide is crucial.1 Precious metals (Pd, Pt, Rh)
have traditionally been used as the most efficient catalysts with
high activity and stability for this process. However, owing to
the high cost and lack of availability of the precious metals,
considerable attention has been paid to transition metals and
their oxides.1,2 Among these systems, CuOx is one of the most
widely used catalysts because of its high activity and selectivity
as an oxidation/reduction catalyst. Furthermore, for many years
CuO-based catalysts have also been used for reactions that
involve the conversion of CO, such as alcohol synthesis (CO
+ 2H2 f CH3OH)3 and the water-gas shift reaction (CO +
H2O f CO2 + H2).4 CuO, Cu4O3 and Cu2O are well-known
copper oxides with monoclinic, tetragonal, and cubic crystal
structures, respectively (Figure 1).5-8 Thermodynamically, the
oxidation state of copper changes among CuO, Cu2O, and Cu
as a function of temperature and oxygen partial pressure.9,10
Three possible pathways for the transformation of CuO to Cu
are displayed in Figure 1, including direct reduction (CuO f
Cu, pathway 1) and reduction mechanisms that involve either
Figure 1. Possible pathways for the reduction of CuO to Cu. Notice
the significant differences in the crystal structures of CuO, Cu4O3, and
Cu2O. In Cu4O3, half of the Cu atoms have four O neighbors, while
the other half of the Cu atoms have only two O neighbors.
one intermediate (CuO f Cu2O f Cu, pathway 2) or two
intermediates (CuO f Cu4O3 f Cu2O f Cu, pathway 3). The
main objective of this work is to identify the routes for CuO
reduction with CO under different experimental conditions.
Despite the tremendous amount of work carried out to explore
the active sites and mechanism for the transformation of CO
over all kinds of copper-oxide-based catalysts (CuO/ZnO, CuO/
* Corresponding author, fax: 631-344-5815, e-mail: rodrigez@bnl.gov.
† Brookhaven National Laboratory.
‡ Yeshiva University.
§ Pohang Accelerator Laboratory.
10.1021/jp040366o CCC: $27.50 © 2004 American Chemical Society
Published on Web 08/13/2004

13668 J. Phys. Chem. B, Vol. 108, No. 36, 2004
Wang et al.
CeO2, CuO/ZrCeO4, CuAl2O4, and CuCr2O4 spinels, etc.),11-17
the behavior of copper oxides in the presence of CO still needs
to be investigated. For example, during the synthesis of alcohols
from CO and the water-gas shift reaction, CuO undergoes partial
or total reduction, and for years there has been a controversy
about the relative activity of “Cu+” and “Cu0” centers.18-20 In
the case of CO oxidation, it is believed that defects and oxygen
holes could be the reason for the high activity or selectivity of
copper-oxide catalysts.17 Nonstoichiometric metastable copper
oxide species formed during reduction or reaction could also
play a key role in CO oxidation,21 but it is not well established
how to generate them and control their concentration. This is
not a trivial task and requires the use of techniques that allow
in situ characterization under reaction conditions.
Except for the much slower flow rate (1-5 mL/min), the rest
of the experimental setup for the tape-loaded sample was the
same as the one used for normal-condition operation.
TPR experiments were carried out using the fixed-bed
sapphire microreactor27 described above, which was connected
to a 0-100 amu quadruple mass spectrometer (QMS, Stanford
Research Systems). A portion of the exit gas flow passed
through a leak valve and into the QMS vacuum chamber. QMS
signals at mass-to-charge ratios of 4(He), 18(H2O), 28(CO),
32(O2), 44(CO2), and 12(C) were monitored during the experi-
ments, and these and the temperature increase were recorded at
the same time by an online computer. Phase changes of the
sample were identified with time-resolved XRD data collected
at the same time as the reduction experiments.
To the best of our knowledge, no systematic studies have
been reported examining in situ the interaction of CO with pure
copper oxides (CuyO, y ) 1 or 2), and the concurring reactions
of CO oxidation and CuyO reduction. In general, very little is
known about the microscopic phenomena associated with the
reduction of metal oxides by CO.22 In principle, this complex
process could be controlled by one or multiple reaction steps:
creation of sites for the adsorption of CO, formation of CO2
and removal of oxygen, migration of oxygen from the bulk to
the surface, formation of suboxides, phase transformations,
growth or nucleation of the reduced metal, etc. Recently, copper
oxide reduction by H2 gas was studied in situ using time-
resolved X-ray diffraction.23,24 This powerful technique is the
result of combining the high intensity of synchrotron radiation
with rapid new parallel data-collection devices.25 Time-resolved
X-ray diffraction (XRD) was used in the present work to monitor
chemical transformations and the long-range atomic order of
CuO and Cu2O during CO oxidation. Information about the
short-range atomic order and local geometry of Cu was obtained
by collecting data of time-resolved X-ray absorption fine
structure (XAFS), which was analyzed using the principal
component analysis (PCA) method.26
2.2 Time-resolved XRD and Data Processing. The time-
resolved X-ray diffraction data were collected at the beamline
×7B (λ ) 0.9225 Å) of the National Synchrotron Light Source
(NSLS) in Brookhaven National Laboratory (BNL) using a
MAR345 area detector. An identical setup system as that in
the TPR experiments was employed for the XRD experiments
under both normal and abnormal conditions. Besides time-
resolved XRD patterns collected during TPR experiments, the
XRD patterns were also recorded during isothermal runs at
different temperatures under small or larger CO flow conditions
over Kapton-taped samples. The temperatures were calibrated
with a linear curve of the Cu cell expansion as a function of
temperature.28 The original powder rings were integrated with
the FIT2D code.29 The FIT2D parameters for the integration of
the data were obtained with a standard LaB6 crystal compound.
A two-dimensional (2D) chart document was obtained with a
Fortran CHITOUXDS code, and the time-resolved XRD pattern
files were plotted with an IDL code.30 The powder profiles from
the isothermal data at 224 °C were analyzed using the GSAS
program.31 In addition to the usual powder profile and scale
parameters, the weight fraction of each phase and selected
oxygen occupancies were refined. The ordered oxygen param-
eters were refined for copper oxides, and an “extra” oxygen
that fits in the unoccupied tetrahedral space in Cu2O was refined.
2. Experimental Section
2.3 Time-resolved X-ray Absorption Spectroscopy and
Data Processing. The time-resolved X-ray absorption spec-
troscopy (XAS) measurements were performed in beamline
×16C of the NSLS at BNL using a setup similar to that of our
previous time-resolved studies.24 First, the sample was brushed
onto an adhesive Kapton tape. Next, the strip of Kapton tape
was loaded into the same cell used for in situ XRD and heated
to the desired temperature. It was then exposed to a flow of a
5% CO and 95% He gas mixture (∼1 mL/min) while the X-ray
absorption spectra were taken repeatedly during reaction. The
isothermal reductions were repeated at several temperatures, the
lowest being 300 °C. The absorption edge steps, measured in
transmission mode, on all the samples were ca. 0.2-0.25 at
the Cu K absorption edge (8978 eV). EXAFS scans from 150
eV below the Cu K-edge to 800 eV above were measured for
the samples reduced at the lowest temperature (300 °C) for
which the duration of reaction was more than an hour; each
EXAFS scan lasted 10 min. At relatively high temperatures
(above ∼350 °C) that made EXAFS data collection impractical
due to the short (tens of minutes) reaction times, XANES
measurements were taken instead; each XANES scan took about
4 min. The start of the first scan at each temperature was
synchronized with the beginning of the CO flow, and the data
were taken until the difference between the successive spectra
became smaller than the statistical noise in the data. After
reduction, the full EXAFS measurement was taken again after
2.1 Temperature-programmed Reduction (TPR). In ad-
dition to carrying out the reaction between CO and copper oxide
powders at isothermal conditions, each copper oxide was
reduced to Cu by temperature-programmed reduction with a
linear heating rate under sufficient and insufficient gas supply
conditions.
The CuO and Cu2O powders were obtained from commercial
sources (>99.99% purity). For the normal reduction conditions
(i.e., typical flow of CO gas, standard packing of the oxide in
a flow reactor), a fresh sample (5 mg) was loaded in a clean
sapphire reactor tube27 with an 0.5-mm i.d. before each run.
Quartzwool was added on each side of the powder to keep the
sample position constant during the gas flowing. A small
resistance heater was placed below the sapphire capillary that
contained the sample. A thermocouple (Type K) was inserted
into the capillary near the powder to measure the temperature
throughout the measurements and to provide input for the
temperature controller.27 The catalyst sample was reduced by
flowing a 5% CO (99.9999% purity) and 95% He (99.9999%
purity) gas mixture (15-20 mL/min) and heating from 25 to
725 °C in 2 h. For the abnormal reduction conditions, the fine
CuO and Cu2O powders were uniformly spread with a brush
over adhesive Kapton polyamide tape. A strip of the tape was
loaded into a Kapton tube. Kapton is relatively transparent for
X-rays at the 9-10 keV energy range, does not decompose at
temperatures up to 400 °C, and thus makes a good support.

Reduction of Copper Oxides with Carbon Monoxide
J. Phys. Chem. B, Vol. 108, No. 36, 2004 13669
Figure 3. Isothermal reductions carried out in a sapphire tube at
different temperatures (gas flow rate ≈ 15-20 mL/min). (a) CuO (111)
intensity as a function of time in isothermal reduction experiments of
CuO; (b) Cu2O (111) intensity as a function of time in isothermal
reduction experiments of Cu2O.
Figure 2. Temperature-programmed reduction of CuO. (Temperature
increase rate (â) ) 350 °C/h or 5.8 °C/min; 5% CO/95% He gas flow
rate ≈ 15-20 mL/min). Panel a: TR-XRD patterns; Panel b: Intensities
for the Cu (111) diffraction line in XRD and the CO2 signal (m/z ) 44
in QMS) at the exit of the reactor.
3.2 Isothermal Reduction XRD Results. The use of linear
heating rates in TPR results in nonuniform reaction conditions
throughout the sample leading to relatively poor resolution in
the kinetic data. Thus, isothermal reduction experiments were
also carried out with the packed samples at temperatures of 180,
200, 220, and 250 °C under a large supply of CO (5% CO/
95% He gas mixture with a flow rate of 15-20 cm3/min).
The normalized CuO (111) and Cu2O (111) peak intensities
as a function of time for the reductions of CuO and Cu2O are
shown in Figures 3a and 3b, respectively. The time-resolved
X-ray diffraction patterns were similar to those obtained by TPR
methods with no intermediate phases and are not shown here.
The isothermal reduction experiments reVealed the existence
of an induction period. This induction period became shorter
with increasing temperature. For instance, the induction periods
were around 100 min, 50 min, and 25 min for the reaction
temperatures of 180 °C, 200 °C, and 220 °C, respectively
(Figure 3a). The reaction was faster at higher operating
temperatures. Figure 3b represents the change for the intensity
of the Cu2O phase during the isothermal reduction as a function
of time. Similarly, an induction period was also obserVed for
each experiment, and the effect of temperature on the induction
period and the reaction rate followed the same trends as those
for CuO reduction (Figure 3a). However, the induction time
for Cu2O reduction was longer than that for CuO reduction at
the same temperature. Moreover, unlike CuO reduction (which
could be completed during a reasonable time), a small amount
of Cu2O was always observed in the compounds at the end of
each experiment, though the amount in the system decreased
with the increase of the operating temperature. These results
indicate that Cu2O was harder to be reduced than CuO under
the same operating conditions.
the sample was cooled to room temperature, to determine how
much pure Cu was in the sample by comparing with the
spectrum from the Cu foil that was measured simultaneously
in reference mode. The sample temperatures during these XAFS
experiments were accurately calibrated by comparing the best-
fit results for the first nearest neighbor (Cu-Cu) EXAFS
Debye-Waller factor (σ2) obtained for the fully reduced Cu
sample at each temperature and their theoretical values calcu-
lated by using the correlated Einstein model.32 This procedure
can be implemented by using the software program FEFFIT.33
3. Results
3.1 Temperature-programmed Reduction. Figure 2 shows
typical results for the temperature-programmed reduction of CuO
with carbon monoxide. The time-resolved XRD patterns in
Figure 2a indicate that CuO started to reduce to metallic Cu at
a temperature of ∼200 °C, and was completely transformed to
Cu at 236 °C. No intermediate phase was seen during the
reduction process. CO was oxidized to CO2, and the QMS trace
of CO2 and the intensity change for the peak Cu(111) line in
XRD are plotted as a function of temperature in Figure 2b. The
setout point for the CO2 signal was ∼210 °C; it rapidly reached
its maximum at 236 °C, and then tailed off gradually until the
temperature reached 585 °C. The results indicate that part of
the CO2 product may have chemisorbed on the walls of the
detection system and gradually desorbed (a common problem
in CO-TPR experiments). The CO2 peak temperature observed
in this work was close to those reported for supported copper
oxide catalysts.34 Time-resolved XRD patterns for Cu2O reduc-
tion showed similar trends to those for CuO reduction with a
direct transformation to metallic Cu; the results are not shown
here.
Isothermal reductions of CuO powder on Kapton tape were
also carried out at several temperatures with a Very low supply
of CO (5% CO/95% He gas mixture with a flow rate of ∼1
mL/min). Typical time-resolved XRD patterns for this kind of

13670 J. Phys. Chem. B, Vol. 108, No. 36, 2004
Wang et al.
Figure 6. Plot of time-resolved XANES for the isothermal reduction
of CuO at 300 °C. (CuO spread evenly on a Kapton tape; 5% CO/95%
He gas flow rate ≈ 1 mL/min).
Figure 4. TR-XRD patterns for the isothermal reduction of CuO spread
evenly on a Kapton tape (Temperature: 224 °C; 5% CO/95% He gas
flow rate ≈ 1 mL/min).
In agreement with the XRD results, the XANES and EXAFS
results indicated the step reduction of CuO for all the temper-
atures studied here under a limited supply of CO. Figure 6
presents typical XANES spectra taken at 300 °C. The spectra
obtained at other temperatures showed a similar trend, and are
not shown here. The starting spectrum from CuO gradually
transformed into the corresponding spectrum for Cu. The fact
that the sample’s reduction to pure Cu was virtually 100% was
confirmed by comparing the XAS spectra in reference metal
Cu foil with the product of the reaction taken after cooling the
sample down to room temperature. The small discrepancy in
amplitude is attributed to the extra disorder in the freshly reduced
sample. The changes in the near-edge region that occurred
during the reduction are apparent. The absence of an isosbestic
point in the XANES spectra indicates the presence of more than
two species during the reaction. For comparison, in our previous
time-resolved XAS studies, we did observe isosbestic points in
those spectra for which the PCA showed conclusively no
presence of an intermediate phase at any time during the
reductions.23-24,35
Figure 5. Weight fractions of CuO, Cu2O, and Cu as a function of
time for the XRD data in Figure 4. The results were obtained with the
GSAS program.
experiment are shown in Figure 4, which displays reduction
data at 224 °C. Unlike the results shown earlier (Figures 2-3),
a Cu2O phase was clearly observed in these experiments. The
phases CuO, Cu2O, and Cu were fitted to the patterns and the
profiles refined with the GSAS program. Figure 5 shows the
weight fractions of CuxO species as a function of time. The
results indicated that, under a limited supply of CO, CuO started
to transform to Cu2O after an induction period of 35 min, and
the whole reduction process followed a sequential step reduction
mechanism
Figures 7a and 7b show the k2-weighted EXAFS and Fourier
transform magnitudes (uncorrected for photoelectron phase
shifts), respectively, of the X-ray absorption coefficient data
collected during CuO reduction at 300 °C. The data are clearly
sensitive to the changes in the Cu coordination during the
reduction. The data demonstrate a gradual decrease in the
intensity of the lower r peak corresponding to the Cu-O bond
in CuO (Figure 7b). This decrease was accompanied by the
gradual rise of the higher r peak corresponding to the Cu-Cu
first-nearest-neighbor distance in metallic Cu. However, based
on visual observation only, the presence of an intermediate Cu2O
and/or other CuxO phases in CuO reduction could neither be
confirmed nor ruled out.
2CuO + CO f + CO2
(1)
(2)
Cu2O + CO f 2Cu + CO2
that essentially involves the reverse steps for the mechanism of
Cu oxidation with O2.24 The formed intermediate phase, Cu2O,
could not be completely transformed to metallic copper, as was
observed in Figure 3b. It is also important to notice that, even
with a limited supply of CO, we neVer detected Cu4O3 as an
intermediate or suboxide in the reduction of CuO.
3.3 XAFS Results. Time-resolved XANES/EXAFS experi-
ments were carried out for the reduction of copper oxide at
different temperatures to independently verify the step reduction
mechanism and check for a possible amorphous CuOx phase as
an intermediate not detected with XRD.
PCA is not a model-dependent XAFS data analysis method
because (1) the number N of principal components is found
model-independently, and (2) the identity of individual species
composing the sample is obtained by a linear combination of
N principal components, and the “wrong” standards can often
be ruled out based on the quality of the fits.23,24 If a more
traditional (fitting) EXAFS data analysis approach were used,
their separate structural analysis would be strongly model-
dependent and less reliable. After applying the PCA to the series
of XANES and EXAFS spectra taken at different temperatures,
the PCA scree tests demonstrated conVincingly that three

Reduction of Copper Oxides with Carbon Monoxide
J. Phys. Chem. B, Vol. 108, No. 36, 2004 13671
Figure 8. (a) PCA scree test for the EXAFS data in Figure 7A. (b)
CuO, Cu2O, and Cu weight fractions during the time-resolved XANES
spectra of the isothermal reduction at 300 °C obtained from PCA. (CuO
spread evenly on a Kapton tape; 5% CO/95% He gas flow rate ≈ 1
mL/min).
Figure 7. Time-resolved EXAFS spectra in (a) k-space and (b) r-space
for the isothermal reduction of CuO at 300 °C. (CuO spread evenly on
a Kapton tape; 5% CO/95% He gas flow rate ≈ 1 mL/min). The filled
squares correspond to the initial oxides, whereas the open ones denote
the final product.
Furthermore, neither Cu4O3 nor an amorphous CuOx phase was
formed during the reduction of CuO by CO. Compared to the
XRD results obtained with the same setup (Figure 4), no
induction time and complete reduction of CuO were seen,
probably because a higher temperature (300 °C) was employed
here.
components (phases) were needed to adequately fit the spectra
(Figure 8a). Therefore, it could be expected that at least three
phases were present during the CuO reduction. It is most
reasonable to expect that these phases are the starting phase
(CuO), the resultant phase (Cu), and the intermediate phase
(Cu2O); however, this hypothesis must be confirmed by the
quantitative analysis. The three components obtained from the
PCA analysis of raw XANES (or EXAFS) spectra were first
fitted to standard compounds (e.g., metal copper and copper
oxides as the most probable species to be found in the data).
When a good fit to each standard is obtained, it points to the
presence of that respective species in the sample. After all
species are identified by fitting the principal components to the
standard spectra, these spectra are then combined to simulate
the EXAFS or XANES spectra at each time step, the mixing
fractions of these species being the only variables that differ
one time step datum from another. These mixing fractions are
then varied to best fit the experimental data collected during
reaction at all times. In our case, the different linear combina-
tions of the same three principal components were fitted
perfectly to (a) pure CuO, (b) pure Cu2O, and (c) pure Cu data
measured at the same temperatures, by only varying the
weighting factors of the components. This procedure was
performed for data taken at all temperatures. An example of
the fitting result from the data at 300 °C is shown in Figure 8b.
The figure presents the molecular fractions of three different
species as a function of time. It is eVident that the step reduction
CuO f Cu2O f Cu was confirmed with the XAFS technique.
4. Discussion
In this work, two pathways for the CuO f Cu reaction were
observed under different operating conditions. A direct trans-
formation (pathway 1 in Figure 1) was seen for samples under
a high CO/He gas flow rate (Figures 2a, 3a), while a reduction
mechanism with one intermediate (pathway 2 in Figure 1) was
obtained for samples under a low CO/He gas flow rate (Figures
5, 8). Pathway 3 in Figure 1 requires the observation of Cu4O3
as an intermediate, but this species was not seen in this work
(Figures 2a, 3a). This is somewhat surprising, because Cu4O3
is much more stable than a CuO with 25% oxygen vacancies
in DFT calculations.23 Furthermore, Cu4O3 appears as a logical
intermediate between CuO and Cu2O, with half of the Cu atoms
in a “+2” formal oxidation state and the other half in a “+1”
formal oxidation state.23 The absence of pathway 3 can be
rationalized in kinetic terms because it requires a precise
rearrangement of O atoms as shown in Figure 9. The formation
of Cu4O3 needs the specific removal of the oxygen colored
yellow in the figure from CuO without elimination of the oxygen
colored blue in the figure. However, during reduction, part of
the blue-colored oxygen could be removed faster than the
yellow-colored oxygen, making the formation of Cu4O3 very
difficult before Cu2O is seen.

13672 J. Phys. Chem. B, Vol. 108, No. 36, 2004
Wang et al.
TABLE 1
extra oxygena
occupancy in Cu2O
wt. %
Cu2O
pure Cu2O
CuO reduction
∼0.00 (2)
00
74
0.17 (1)b
intermediate (Cu2O)
a Extra oxygen position at (0.25 0.75 025) with multiplicity of 6,
while the normal lattice oxygen is at (0.25 0.25 0.25). b Occupancy
was refined when there was 74% Cu2O intermediate phase in the
system.
in stoichiometric CuO. Structures can be chosen that would have
the connectivity patterns in CuO and Cu4O3 (see Figure 9).
These structures must be highly strained because the observed
Cu2O lattice requires that all distances be 1.85Å, which is the
Cu1+-O distance. However, if there is extra oxygen in the
lattice, some of the copper must be in the +2 oxidation state,
which has the longer Cu-O distance, approaching 1.95Å. The
structural changes from CuO to Cu2O could be described as
successively distorting the observed monoclinic structure to a
hypothetical tetragonal CuO arrangement and then removing
bridging oxygen atoms between the copper atoms until the cubic
Cu2O structure is formed. The latter structures can be drawn in
a common space group I1/amd (Figure 9).7 The observation of
“extra” oxygen in the Cu2O phase can be rationalized in terms
of the requirement of this structure to rearrange from the more
random orientation in CuO. The extra oxygen occupancy in the
intermediate Cu2O is compared to that obtained from the pure
Cu2O in Table 1. These results clearly imply that the intermedi-
ate Cu2O phase formed during CuO reduction had an observable
amount of extra oxygen. However, because of the limitations
of XRD techniques for the detection of light elements, the
presence of the extra oxygen atoms needs to be confirmed in
the future with other techniques, such as neutron diffraction.
Unlike CuO reduction, Cu2O reduction has only one apparent
pathway involving two phases (Cu2O f Cu), but amorphous
suboxides could be formed. The presence of the latter can be
ruled out based on our XANES/EXAFS results. It was found
that CuO was easier to reduce than Cu2O (Figure 3). This can
be explained by the fact that the Cu(I)-O bond is much stronger
than the Cu(II)-O bond (there are fewer Cu(I)-O bonds in
Cu2O than Cu(II)-O bonds in CuO). If Cu2O is an intermediate
in the CuO reduction pathway, one would always expect to see
Cu2O formation because Cu2O would not be reduced as fast as
CuO. However, if there is another CuOx intermediate that is
less stable and more reactive than Cu2O,23 Cu2O would not be
observed. One possible intermediate is a CuO structure with
50% O vacancies that has been shown to be less stable than
Cu2O.23 This species would have a diffraction pattern similar
to CuO and would not appear as a separate phase.
Figure 9. Oxygen removal from a hypothetical tetragonal CuO in a
common space group I1/amd to form Cu4O3 (removal of oxygen in
yellow) and Cu2O (removal of oxygen in both yellow and blue).
Figure 10. Oxygen vacancies in CuO as a function of time for the
XRD results in Figure 4. The O occupancies were obtained with the
GSAS program.
Evidence for O vacancies in CuO during reduction is observed
from the profile refinement of the XRD results in Figure 4 (see
Figure 10). The maximum amount of O vacancies found is less
than 25%. The O vacancies in CuO could foster formation of a
nonstoichiometric metastable copper-oxide species.16,23 This
metastable species can either react rapidly with CO to form
metallic copper under high CO flow rates (pathway 1 in Figure
1) or relax into Cu2O when the supply of CO is limited (pathway
2 in Figure 1). At slow flow rates of CO, the O atoms have
more time for rearrangement, so a pathway to form the
thermodynamically stable Cu2O can occur.
In the X-ray structure refinements and difference electron
density maps, two different kinds of extra electron densities were
observed in addition to the densities from the lattice Cu (0 0 0)
and oxygen (0.25 0.25 0.25) in the Cu2O phase formed during
the reduction process. The less intense one at the middle of the
Cu-Cu chain (Cu-Cu distance 3.02Å) has been reported to
be the electron densities from Cu-Cu covalent bonding.36 The
intense one partially filling the positions for the yellow and blue
O atoms in Figure 9 is tentatively assigned to extra oxygen
atoms in a Cu2O lattice. The lattice of Cu2O (77.4 Å3) has a
similar volume to that of CuO (81.5 Å3), therefore, Cu2O could
embrace extra O atoms. These extra oxygens must be disordered
and have fractional occupancy less than 0.33 that would result
The coexistence of phases shown in Figure 5 is consistent
with a nonuniform reduction of the CuO samples. Based on
this and the results of refinements, one can propose the
mechanism shown in Figure 11. The initial system formed is a
modified CuO that has surface sites that are efficient for the
adsorption of CO (A f B). An induction time was always
observed and became shorter at higher temperatures (Figure 3a).
During this induction time, geometry relaxation could occur and
surface defects could be formed. A small amount of O vacancies
could be created without changing the CuO XRD pattern. Next,
CuO could evolve into a nonstoichiometric metastable species
(B f C).23 This metastable species could transform to either
metallic Cu directly (C f H), under a large supply of CO, or
to a Cu2O-like intermediate phase (C f D), under a low supply
of CO. The Cu2O-like phase generated is highly disordered with

Reduction of Copper Oxides with Carbon Monoxide
J. Phys. Chem. B, Vol. 108, No. 36, 2004 13673
be described by a single nth-order expression over the entire
range of reaction.
Acknowledgment. The research carried out at the Chemistry
Department of Brookhaven National Laboratory was financed
through contract DE-AC02-98CH10086 with the U.S. Depart-
ment of Energy (DOE, Division of Chemical Sciences). A.I.F.
acknowledges support by the U.S. Department of Energy Grant
No. DE-FG02- 03ER15477. The authors would like to thank
Avital Merl and Pesia Soloveichik for their help in XAS data
collection and processing. The National Synchrotron Light
Source is supported by the Divisions of Materials and Chemical
Sciences in the Office of Basic Energy Sciences at DOE.
References and Notes
(1) Kummer, J. T. Prog. Energy Combust. Sci. 1980, 6, 177.
(2) Huang, T. J.; Yu, T.-C. Appl. Catal. 1991, 71, 275.
(3) Klier, K. AdV. Catal. 1982, 31, 243.
(4) Newsome, D. S. Catal. ReV. 1980, 21 (2), 275.
(5) PDF #41-0254, JCPDS Powder Diffraction File; Int. Center For
Diffraction Data: Swarthmore, PA, 1989.
Figure 11. Proposed mechanism for the CuO reduction in CO gas.
(6) PDF #05-0667, JCPDS Powder Diffraction File; Int. Center For
a substantial amount of extra oxygen moving around the free
spaces of the oxide lattice. Further reduction with CO would
produce metallic copper (D f E). The metallic copper starts to
form from the surface of the Cu2O particles, and gradually
becomes thicker. As the layer of Cu becomes thicker, it is more
difficult to fully reduce the Cu2O (E f F f G). The full
reduction (G f H) would be facilitated by high temperature.
At different stages of the CuO reduction process, the rate
limiting-step can be different. Initially, it is critical to generate
sites for the adsorption of CO. Once a few of these sites are
available, the removal of O can start (i.e., the creation of more
adsorption sites for CO) and the reaction becomes autocatalytic.
In the following stages, the movement of O from the bulk to
the surface or the nucleation of a phase of metallic copper should
become extremely important. Hence, this gas-solid reaction is
typified by complex kinetics that cannot be described by a single
nth-order expression over the entire process. In fact, we attempted
to fit different portions of the kinetic data in Figures 3 and 8
with a single nth polynomial without obtaining good fittings.
Diffraction Data: Swarthmore, PA, 1989.
(7) O’Keeffe, M.; Bovin, J. O. Am. Mineral. 1978, 63, 180.
(8) PDF #083-1665, JCPDS Powder Diffraction File; Int. Center For
Diffraction Data: Swarthmore, PA, 2004.
(9) O’Keeffe, M.; Moore, W. J. J. Chem. Phys. 1962, 36, 3009.
(10) Huang, T. J.; Tsai, D.-H. Catal. Lett. 2003, 87 (4), 173.
(11) Nagase, K.; Zheng, Y.; Kodama, Y.; Kakuta, J. J. Catal. 1999,
187, 123.
(12) Wang, J. B.; Tsai, D.-H.; Huang, T. J. J. Catal. 2002, 208, 370.
(13) Dow, W.-P.; Huang, T. J. J. Catal. 1994, 147, 322.
(14) (a) Mart´ınez-Arias, A.; Ferna´ndez-Garcia, M.; Ga´lvez, O.; Coro-
nado, J. M.; Anderson, J. A.; Conesa, J. C.; Soria, J.; Munuera, G. J. Catal.
2000, 195, 207. (b) Mart´ınez-Arias, A.; Ferna´ndez-Garc´ıa, M.; Hungr´ıa,
A. B.; Igles¨ias-Juez, A.; Ga´lvez, O.; Anderson, J. A.; Conesa, J. C.; Soria,
J.; Munuera, G. J. Catal. 2003, 214, 261.
(15) Severino, F.; Brito, J. L.; Laine, J.; Fierro, J. L. G.; Lo´pez Agudo,
A. J. Catal. 1998, 177, 82.
(16) Dow, W.-P.; Huang, T. J. J. Catal. 1996, 160, 171.
(17) Liu, W.; Flytzani-Stephanopoulos, M. J. Catal. 1995, 153, 317.
(18) Nakamura, J.; Campbell, J. M.; Campbell, C. T. J. Chem. Soc.,
Faraday Trans. 1990, 86, 2725.
(19) Rodriguez, J. A.; Goodman, D. W. Surf. Sci. Reports 1991, 14, 1.
(20) Li, J.-L.; Takeguchi, T.; Inui, T. Appl, Catal. A 1996, 139, 97.
(21) Sadykov, V. A.; Tikhov, S. F. J. Catal. 1997, 165, 279.
(22) Kung, H. H. Transition Metal Oxides: Surface Chemistry and
Catalysis; Elsevier: New York, 1989.
5. Conclusion
(23) Kim, J. Y.; Rodriguez, J. A.; Hanson, J. C.; Frenkel, A. I.; Lee, P.
L. J. Am. Chem. Soc. 2003, 125, 10684.
(24) Rodriguez, J. A.; Kim, J. Y.; Hanson, J. C.; Perez, M.; Frenkel, A.
I. Catal. Lett. 2003, 85 (3-4), 247.
(25) Norby, P.; Hanson, J. C. Catal. Today 1998, 39, 301.
(26) Frenkel, A.; Kleifeld, O.; Wasserman, S. R.; Sagi, I. J. Chem. Phys.
2002, 117, 9449.
(27) Chupas, P. J.; Ciraolo, M. F.; Hanson, J. C.; Grey, C. P. J. Am.
Chem. Soc. 2001, 123, 1694.
(28) Gmelins Handbuch Der Anorganischen Chemie; Kupfer Tell A,
System-Nummer 60, 1955; Verlag Chemie, Gmbh, Weinheim, Bergastrasse.
(29) Hammersely, A. P.; Svensson, S. O.; Thompson, A. Nucl. Instrum.
Methods Phys. Res. 1994, 346, 321.
(30) IDL, Version 5.5; Research System, Inc., 2002.
(31) (a) Larson, A. C.; von Dreele, R. B. GSAS General Structure
Analysis System. Report LAUR 86-748; Los Alamos National Laboratory:
Los Alamos, 1995. (b) Reitveld, A. M. J. Appl. Cryst. 1969, 2, 65.
(32) Frenkel, A. I.; Rehr, J. J. Phys. ReV. B 1993, 48, 585.
(33) Newville, M.; Ravel, B.; Haskel, D.; Rehr, J. J.; Stern, E. A.;
Yacoby, Y.; Physica B 1995, 208 & 209, 154.
The reactions of CuO and Cu2O with CO were investigated
in detail using temperature-programmed reduction (TPR),
synchrotron-based time-resolved X-ray diffraction and X-ray
absorption fine structure techniques. The TPR results showed
that Cu2O reduced at higher temperature than CuO did under
the same operating conditions. A direct transformation pathway
was observed for the isothermal reductions of both CuO and
Cu2O under a large CO supply, while a step pathway with one
intermediate (CuO f Cu2O f Cu) was seen under a limited
CO supply. This was confirmed by both time-resolved X-ray
diffraction and X-ray absorption fine structure techniques. The
formed Cu2O was highly disordered with a substantial amount
of extra oxygen embedded in the lattice. Neither Cu4O3 nor an
amorphous CuOx phase was formed during the reduction of CuO
by CO.
(34) Tsai, D.-H.; Huang, T. J. Appl. Catal., A 2002, 223, 1.
(35) Rodriguez, J. A.; Hanson, J. C.; Frenkel, A. I.; Kim, J.-Y.; Perez,
M. J. Am. Chem. Soc. 2002, 124, 346.
(36) Zuo, J. M.; Kim, M.; O’Keeffe, M.; Spence, J. C. H. Nature 1999,
401, 49.
This study illustrates how complex the reduction of a metal
oxide by CO can be. The reduction mechanism can vary
considerably with the experimental conditions (gas flow rate,
temperature, sample size, etc.), and its complex kinetics cannot