Reduction kinetics of surface rhodium oxide by hydrogen and carbon monoxide at ambient gas pressures as probed by transient surface-enhanced raman spectroscopy

Article abstract of DOI:10.1021/jp981287d

The transient reduction kinetics of surface rhodium oxide (Rh₂O₃) by gaseous H₂ and CO have been probed in situ by surface-enhanced Raman spectroscopy (SERS). The Rh surfaces are ultrathin films electrodeposited onto SERS-active gold, enabling surface vibrational spectroscopic information to be obtained with high temporal resolution (≈1 s) at elevated temperatures (up to 500 °C) and under ambient-pressure flow-reactor conditions. Surface Rh₂O₃ is formed by heating Rh in flowing O₂ at 1 atm and fingerprinted by an intense 530 cm^-1 V_RH-O feature. The reduction of such oxidized surfaces upon sudden exposure to either H₂ or CO over a range of partial pressures (1-760 Torr) was monitored from the decay kinetics of the v_Rh-o band intensity. Surface oxide is readily reduced by both reductants, although striking differences in the observed kinetic behavior indicate the occurrence of distinct reaction pathways. In the case of H₂, at low partial pressures (≤7.6 Torr) below 200 °C a temperature-dependent induction period is observed prior to rapid first-order oxide reduction. Such autocatalytic$ behavior is indicative of a νucleation/growth mechanism, where H2 dissociatively adsorbs to form reaction centers, followed by facile reaction between adsorbed (or lattice) oxygen and (possibly subsurface) hydrogen atoms. Immeasurably fast H₂-induced oxide reduction, however, occurs at higher temperatures (≥200 °C) and pressures (≥76 Torr). In contrast, CO-induced oxide reduction was found to be substantially (at least 10-fold) slower than with H₂ at similar pressures and temperatures, yet no induction period was detected. At lower temperatures (≤250 °C), the oxide reduction kinetics by CO exhibit fast initial removal followed by a much more sluggish zero-order response. Such autoinhibited kinetic behavior, along with the lack of an appreciable CO pressure dependence, suggests that oxide removal is hindered by the extensive formation of adsorbed CO, which is observed to develop rapidly under these conditions from the characteristic Rh-CO and C-O stretching vibrations. This mechanism is further supported by the first-order kinetic response observed throughout oxide removal at temperatures (≥300 °C) where buildup of adsorbed CO is not detected. The transient kinetics for CO-induced oxide reduction are linked to the well-known poisoning effect that Rh surface oxidation exerts on catalytic CO oxidation by O₂. Comparisons are also made between the present results and those recently obtained for methanol-induced oxide removal in order to elucidate which chemisorbed alcohol fragment(s) constitute the reactive oxygen scavenger .

Full text of DOI:10.1021/jp981287d

J. Phys. Chem. B 1998, 102, 4785-4794
4785
Reduction Kinetics of Surface Rhodium Oxide by Hydrogen and Carbon Monoxide at
Ambient Gas Pressures As Probed by Transient Surface-Enhanced Raman Spectroscopy
†
‡
§
,‡
Christopher T. Williams, Eddy K.-Y. Chen, Christos G. Takoudis, and Michael J. Weaver*
School of Chemical Engineering and Department of Chemistry, Purdue UniVersity,
West Lafayette, Indiana 47907, and Department of Chemical Engineering,
UniVersity of Illinois at Chicago, 810 S. Clinton St., Chicago, Illinois 60607
ReceiVed: February 25, 1998; In Final Form: April 20, 1998
2 3 2
The transient reduction kinetics of surface rhodium oxide (Rh O ) by gaseous H and CO have been probed
in situ by surface-enhanced Raman spectroscopy (SERS). The Rh surfaces are ultrathin films electrodeposited
onto SERS-active gold, enabling surface vibrational spectroscopic information to be obtained with high temporal
resolution (≈1 s) at elevated temperatures (up to 500 °C) and under ambient-pressure flow-reactor conditions.
-
1
2 3 2
Surface Rh O is formed by heating Rh in flowing O at 1 atm and fingerprinted by an intense 530 cm
2
νRh-O feature. The reduction of such oxidized surfaces upon sudden exposure to either H or CO over a
range of partial pressures (1-760 Torr) was monitored from the decay kinetics of the νRh-O band intensity.
Surface oxide is readily reduced by both reductants, although striking differences in the observed kinetic
behavior indicate the occurrence of distinct reaction pathways. In the case of H
e7.6 Torr) below 200 °C a temperature-dependent induction period is observed prior to rapid first-order
oxide reduction. Such “autocatalytic” behavior is indicative of a “nucleation/growth” mechanism, where H
dissociatively adsorbs to form reaction centers, followed by facile reaction between adsorbed (or lattice)
oxygen and (possibly subsurface) hydrogen atoms. Immeasurably fast H -induced oxide reduction, however,
occurs at higher temperatures (g200 °C) and pressures (g76 Torr). In contrast, CO-induced oxide reduction
was found to be substantially (at least 10-fold) slower than with H at similar pressures and temperatures, yet
2
, at low partial pressures
(
2
2
2
no induction period was detected. At lower temperatures (e250 °C), the oxide reduction kinetics by CO
exhibit fast initial removal followed by a much more sluggish zero-order response. Such “autoinhibited”
kinetic behavior, along with the lack of an appreciable CO pressure dependence, suggests that oxide removal
is hindered by the extensive formation of adsorbed CO, which is observed to develop rapidly under these
conditions from the characteristic Rh-CO and C-O stretching vibrations. This mechanism is further supported
by the first-order kinetic response observed throughout oxide removal at temperatures (g300 °C) where buildup
of adsorbed CO is not detected. The transient kinetics for CO-induced oxide reduction are linked to the
well-known poisoning effect that Rh surface oxidation exerts on catalytic CO oxidation by O . Comparisons
2
are also made between the present results and those recently obtained for methanol-induced oxide removal
in order to elucidate which chemisorbed alcohol fragment(s) constitute the reactive oxygen “scavenger”.
Introduction
been increasing success at applying and adapting some surface
spectroscopic and microscopic characterization methods to
Monitoring the chemical state of a metal surface during
heterogeneous processes involving either gas- or solution-phase
species constitutes a basic requirement in forging a molecular-
level understanding of reaction kinetics and mechanisms in
catalytic chemistry. The acquisition of such interfacial chemical
information, especially in situ under technologically relevant
conditions, such as high gas pressures, represents a substantial
challenge and hence remains a key issue in heterogeneous
catalysis. As a consequence, most molecular-level information
on the nature and reactivity of adsorbates in gas-phase hetero-
geneous catalysis, in particular, has been extracted (or implied)
from “surface-science” studies at much lower pressures, down
to ultrahigh vacuum (UHV). The need to extrapolate between
such widely differing gas-phase conditions has often been
referred to as the “pressure-gap” problem.1 Recently, there has
situations featuring even ambient-pressure gas or liquid (espe-
1
,2
cially electrochemical) systems. Important examples in this
regard include examinations of spatial atomic and molecular
surface arrangements by means of scanning tunneling micros-
3
copy (STM) and vibrational spectra for adsorbed molecules
obtained from infrared-visible sum-frequency generation (SFG).2
,3
With these objectives in mind, we have been exploring the
utility of surface-enhanced Raman spectroscopy (SERS) as an
in situ vibrational spectral probe on metals in gas-phase as well
as electrochemical environments, especially catalytic systems.
Some time ago, we found that the technique could be extended
to transition-metal surfaces by electrodeposition of thin films
4
onto SERS-active gold substrates. Significantly, several of
these films (such as Rh, Pt, Pd, and Ru) display excellent thermal
stability and SERS activity (at least up to 500 °C), enabling
their surface chemistry to be scrutinized while acting as model
catalysts in ambient-pressure gas-flow reactor systems. Such
an arrangement enables the formation and reactivity of interfacial
*
To whom correspondence should be addressed.
School of Chemical Engineering, Purdue University.
Department of Chemistry, Purdue University.
†
‡
§
Department of Chemical Engineering, University of Illlinois.
S1089-5647(98)01287-5 CCC: $15.00 © 1998 American Chemical Society
Published on Web 05/22/1998

4
786 J. Phys. Chem. B, Vol. 102, No. 24, 1998
Williams et al.
species to be examined in situ under transient as well as steady-
state (or equilibrium) conditions, involving, for example, the
acquisition of temporal spectral sequences (g1 s time scales)
following rapid alterations in gas-phase composition and/or
catalyst temperature.5 Moreover, the anticipated relationships
between surface speciation and reactivity with the overall
catalytic rates can be addressed by means of simultaneous mass
monitored in order to limit the deposition to the desired film
4b,c
thickness, usually 2-6 equivalent monolayers.
The surfaces
3
-1
were pretreated by heating to 150 °C in 100 cm min of H2
for about 5-10 min. This procedure resulted in a flat SERS
-1
background in the 200-1000 cm range, even though weak
-1
features remained in the 1300-1600 cm region due to surface
carbonaceous residues.
6
spectrometric (MS) detection of the gas-phase products.
Details of the reactor and sample holder utilized for the SERS
5
c,6d
One interesting issue that we have been exploring recently
by means of these “flow-reactor SERS” tactics involves the role
of surface oxides in catalytic oxidations. The sensitive detection
of low-frequency vibrations, such as oxide-lattice modes,
afforded by SERS makes this technique a propitious choice for
such in situ kinetic studies. These oxides, formed on transition
metals at elevated temperatures with ambient-pressure oxidants
such as O2 and NO, are typically found to inhibit catalytic
activity and/or alter reaction pathways. For example, the
differing roles of surface oxides in the methanol oxidation
reaction on rhodium and palladium catalysts, evident from
temperature-dependent kinetic data,5b-d are highlighted by a
notable dissimilarity in the reactivity of these oxides toward
methanol: rhodium oxide is capable of oxidizing methanol
above 150 °C,5 while palladium oxide is entirely unreactive
experiments have been provided previously.
The (room
temperature) gaseous reactants are introduced into the reactor
through a manifold that allows mixing of up to four inlet gases;
flow rates are measured with a bubble flow meter. Recent
reactor modifications allow for sudden introduction of such
reaction mixtures close to the sample, providing efficient mixing
characteristics required for transient spectral response experi-
ments, of central interest here. The gas-phase composition was
monitored by leaking the reactor gas into a quadrupole mass
spectrometer (Stanford Research Instruments RGA200). The
reactants utilized were 99.99% pure CO from an aluminum
cylinder (Matheson) (see footnote 18), 99.994% pure O2 (Airco),
99.9% pure H2 (Airco), and 99.9% pure Ar (Airco). Laser
+
excitation at 647.1 nm is provided by a Kr laser (Spectra
b-d
Physics), and scattered light from the sample is collected by a
three-stage spectrometer (SPEX Triplemate), equipped with a
liquid nitrogen-cooled CCD detector (Photometrics).
5
d
even at 500 °C.
We have been motivated to explore such processes on
rhodium in view of the effectiveness of Rh-based catalysts for
a variety of gas-phase reactions, together with the likely
importance of surface oxidation given the high affinity of this
metal toward oxygen. For example, the presence of rhodium
Results
The surface oxidation of Rh films by O2 at ambient pressures
7
and elevated temperatures (25-300 °C) was examined previ-
oxide inhibits CO oxidation, while dramatically increasing the
1
3b
8
ously in detail by Tolia et al. by using SERS together with
X-ray photoelectron spectroscopy (XPS). A broad Raman band
activity for alkane hydrogenolysis. However, while the reaction
of adsorbed oxygen on Rh with various gas-phase species (e.g.,
-1
7
,9
10
7,11
centered around 530 cm was evident above 150 °C along with
CO, H2, NO ) has been examined in detail, surprisingly
little is known about the reactivity of rhodium oxide, even
though this surface species readily forms at moderate O2 partial
pressures (>0.01 Torr) and elevated temperatures (g200
-
1
a second feature at 290 cm . The former band was assigned
to the metal-oxygen symmetric stretch (νRh-O) of Rh2O3 based
on comparisons with Raman spectra of bulk oxide and correla-
tions with the parallel XPS data. More recent observation of
1
2-14
°
C),
conditions which are commonly encountered in
-
1
18
the expected ca. 20 cm downshift upon substitution with
O
oxidative catalysis.
-
1
confirms that the species responsible for the 530 cm feature
is formed from O2. Assignment of the 290 cm band was
more challenging (see footnote 19). However, in the experi-
ments reported herein, this 290 cm feature was found to be
less reproducible as well as typically much weaker than the
dominant 530 cm band and is therefore not considered further
Described here is a SERS study of the reaction kinetics of
surface rhodium oxide with hydrogen and carbon monoxide
under transient flow-reactor conditions. This selection of
gaseous reductants was made mindful of comparisons with the
corresponding reaction involving methanol, since H and CO
are two likely oxide-scavenging fragments formed upon metha-
6
d
-1
-
1
-1
1
9
here.
15
nol dissociative adsorption on Rh. Striking dissimilarities are
evident in the time-dependent oxide removal kinetics with H2
and CO, which together with other vibrational spectral informa-
tion indicates the occurrence of notably different reaction
pathways. The results highlight the utility of our in situ transient
SERS approach to probe surface reaction mechanisms under
catalytically relevant conditions.
The reactivity of rhodium oxide was probed by monitoring
its removal in real time by means of SER spectral sequences
acquired following sudden exposure of the surface to CO or
H2 at various partial pressures and surface temperatures (cf. ref
5
c). Oxidized Rh surfaces were first formed by heating a sample
3
-1
to 350 °C in a 100 cm min flow of pure O2 at 1 atm, yielding
an intense broad 530 cm band associated with Rh2O3. Further
heating in pure oxygen has little effect on the intensity of the
-1
Experimental Section
-
1
5
30 cm feature, indicating that the Rh film oxidation is
The SERS-active rhodium films were prepared by utilizing
mm diameter disks cut out of 0.1 mm thick gold foil (99.95%
pure, Johnson Matthey) and polished using 0.3 µm alumina
essentially complete by 350 °C. The reactor was subsequently
flushed out with argon and the surface temperature adjusted to
the desired value. This procedure retains the oxide-covered
surface, since the temperatures investigated (50-400 °C) are
lower than those reported for thermal desorption of oxygen from
6
powder. They were then placed into an electrode holder that
2
exposed ca. 0.07 cm of the surface and subjected to oxidation-
1
6
12,16
reduction cycling (ORC) in 0.1 M KCl to produce the surface
roughness necessary for Raman enhancement.4 The surface
was rinsed thoroughly with deionized water and subsequently
transferred to another cell for electrochemical deposition of the
Rh overlayers, which utilized a 0.3 mM solution of RhCl3 in
Rh and decomposition of Rh2O3,
ca. 800 and 550 °C,
a
respectively. The reactive gas (consisting of various partial
pressures of H2 or CO in Ar at 1 atm) was then suddenly
3
-1
introduced at 2400 cm min while simultaneously monitoring
the SER spectra. Each spectrum in a sequence spanning the
desired overall time scale was acquired within periods as short
0
.1 M HClO4 at 0.1 V vs SCE. The cathodic charge was

Reduction Kinetics of Surface Rh2O3
J. Phys. Chem. B, Vol. 102, No. 24, 1998 4787
Figure 2. Plot of the relative integrated νRh-O oxide band intensity as
a function of time at 150 (circles), 100 (squares), and 50 °C (triangles)
3
-1
2
following the sudden addition of a 2400 cm min flow of 1% H in
Ar at 1 atm. Band intensities, as deduced from time-dependent SER
spectra as in Figure 1, are normalized to the maximum (i.e., initial)
values.
temperature dependence. Approximate pseudo-first-order rate
constants, k1, were extracted via semilogarithmic plots of the
relative band intensity versus (t-tind), where tind is the (induction)
time following H2 addition at which the oxide band begins to
attenuate. The temperature-dependent k1 data yield an Arrhenius
Figure 1. Real-time SER spectral sequence acquired upon sudden
exposure of an initially oxide-covered Rh surface at 100 °C to a 2400
3
-1
cm min flow of 1% H
2
in Ar at 1 atm. The surface was previously
heated in 100 cm min of pure O to 350 °C, followed by flushing of
the reactor with argon and cooling to the temperature of interest before
addition. The spectral acquisition time was 0.5 s.
3
-1
2
H
2
-1
activation energy of 3 ( 1 kcal mol and a preexponential
-
1
factor equal to 30 s . As discussed below, such intriguing
as 0.5 s. (This lower limit is constrained by our CCD camera.)
The chosen flow rate is sufficient to change entirely the
composition in the reactor chamber within ca. 3 s (i.e., 4
residence times).
The transient reactivity of surface Rh2O3 toward H2 over a
wide range of temperatures (50-400 °C) and reactant partial
pressures (1-760 Torr) is considered first. Utilizing mixtures
containing 100% (760 Torr) or 10% (76 Torr) H2 resulted in
immeasurably rapid (<0.5 s) removal of oxide, even at 50 °C.
However, lowering the H2 composition another 10-fold, to 1%
“
biphasic” Iox-t behavior is suggestive of a “nucleation-growth”
mechanism for H2 reduction of the oxide. Given the large
differences observed in the results for 10% and 1% H2 mixtures,
the effect of lowering the reactant partial pressure further was
investigated. Surprisingly, however, the kinetic responses for
2
0.1% H were observed to be nearly identical to those described
above for the 1% H case.
2
Finally, the effects of increasing the average Rh film thickness
on the oxide removal process were investigated. For these
experiments, rhodium films were deposited with 50% more
Coulombic charge, so to yield about six equivalent Rh mono-
layers. Thermal oxidation of such films in flowing oxygen to
350 °C resulted in Raman spectral features identical to those
described above (cf. Figure 1, bottom spectrum). These
“thicker” oxide films were readily reduced by hydrogen over
the range of partial pressures examined. While utilizing 100%
(7.6 Torr), resulted in measurable oxide removal rates below
2
00 °C. Interestingly, an induction period is observed after H2
addition during which no measurable (by SERS) reaction occurs,
yet followed by rapid and complete oxide removal. Figure 1
shows typical time-dependent SER spectra acquired during
reaction of oxide with a 1% H2 mixture at 100 °C. The intensity
-
1
of the 530 cm oxide band remains unchanged until about 4 s
after H2 addition, whereafter it is removed rapidly (within the
following 5 s or so). Similar kinetic characteristics were
obtained at 50 and 150 °C, the responses at higher temperatures
and 10% H again yielded kinetics that were faster than our
temporal resolution, significant differences were observed for
2
the two types of Rh films with the 1% mixture. In addition to
slightly longer induction times, the oxide removal kinetics were
significantly (2-5-fold) slower than observed for the thinner
films at the lowest temperatures. Nevertheless, the overall
“biphasic behavior” noted above was consistently observed for
both types of Rh films examined here.
Strikingly different kinetic behavior was obtained for reduc-
tion of surface rhodium oxide by gaseous CO in place of H2.
While the former reductant was able to completely remove
surface oxide at all partial pressures (8-760 Torr) and temper-
atures (150-400 °C) investigated, the process was substantially
slower than observed with H2 under similar conditions. No
induction period prior to oxide removal was discerned under
any of the conditions examined. Two distinct types of kinetic
behavior were nonetheless observed, depending on the surface
temperature.
(g200 °C) being immeasurably fast.
Such band intensity-time (Iox-t) data can be utilized to
provide at least an approximate measure of the temporal decay
of the surface oxide concentration as it reacts with the added
H2. We will assume here that the integrated band intensity is
proportional to the oxide “relative coverage,” θox. (A rough
proportionality between these quantities has been noted for
17
several metal-adsorbate SERS bands. ) Figure 2 shows the
relative oxide band intensity plotted versus time during reactive
removal by 1% H2 at 150, 100, and 50 °C. While numerous
experiments were performed at each temperature, the data sets
shown are those typical of the observed behavior. The induction
time following H2 addition is seen to decrease by at least 10-
fold from 50 to 150 °C (Figure 2). The ensuing oxide removal
kinetics are roughly first order in coverage. Surprisingly,
however, the latter Ιox-t behavior exhibits only a weak
Figure 3 shows a typical SER spectral sequence obtained upon
sudden exposure of an oxidized Rh surface held at 300 °C to 1

4788 J. Phys. Chem. B, Vol. 102, No. 24, 1998
Williams et al.
Figure 5. Plot of the relative integrated νRh-O oxide band band intensity
as a function of time at 350 °C following the sudden addition of 2400
3
-1
cm min of a 10% CO in Ar mixture at 1 atm (cf. Figure 4). The
solid trace is a “best-fit” first-order decay curve.
Figure 3. Real-time SER spectral sequence acquired upon sudden
exposure of an initially oxide-covered Rh surface at 300 °C to a 2400
cm min flow of 100% CO at 1 atm. Other details as in Figure 1
caption.
Rh2O3 removal kinetics were found to be roughly first order
under these conditions, although the rapidity of the decay
hampers a quantitative analysis. The νRh-CO band was not
discerned during oxide removal for temperatures above 300 °C,
consistent with the lack of spectrally detectable adsorbed CO
even for oxide-free surfaces under these conditions. Lowering
the CO partial pressure down to 76 Torr or below (i.e., e10%)
yielded readily measurable oxide removal kinetics even at 350
3
-1
°
C. Figure 5 is a Iox-t plot obtained from temporal spectral
data for oxide removal upon exposure to 76 Torr of CO at 350
C. The kinetic profile is seen to be roughly consistent with
°
first-order decay kinetics as shown by the “best-fit” solid curve.
Oxide removal by CO at lower temperatures (e250 °C),
however, exhibits a distinctly different kinetic fingerprint.
Figure 6 shows a representative set of time-dependent SER
spectra obtained for reduction of oxide at 200 °C by 10% CO
-
1
(
76 Torr). The intensity of the 530 cm oxide band is seen to
be attenuated by about 40% within the first 20 s and is
accompanied by the appearance of a small amount of adsorbed
-1
CO, evident as a “shoulder” at ca. 455 cm on the overlapping
stronger oxide feature. However, the subsequent diminution
and growth of the νRh-O and νRh-CO bands, respectively, are
markedly slower, taking several minutes to occur. Figure 7 is
a plot of the relative integrated intensity of the νRh-O and νRh-CO
bands (open, filled circles) versus time, extracted from data in
Figure 6. The temporal decay of the oxide band intensity
exhibits a distinctly “curvilinear” shape. The initial (e20 s)
oxide removal kinetics are very rapid before giving way to a
much more sluggish and approximately zero-order behavior (i.e.,
a θox-independent reduction rate) for the remainder of the
process (g75 s). Also noteworthy is that the νRh-CO band
intensity increases initially very rapidly, in tandem with the
oxide removal, exhibiting a similar (yet oppositely directed)
curvilinear time response (Figure 7).
Rate constants, k1, for the initial stage of oxide reduction by
CO (i.e., the first portion of the Iox-t decay) over the whole
range of conditions investigated were extracted from the kinetic
data. Figure 8 summarizes the rate data in the form of an
Arrhenius plot. The entries shown with upward-pointing arrows
are lower-bound estimates for oxide removal rates that were
faster than the temporal resolution of our instrument. At lower
temperatures (e250 °C), k1 is largely independent of CO partial
Figure 4. Plot of the relative integrated intensities of νRh-O (open
circles) and νRh-CO (closed circles) bands for oxide and adsorbed CO,
respectively, as a function of time at 300 °C following the sudden
addition of a 2400 cm min flow of 100% CO at 1 atm. Band
intensities, as deduced from the time-dependent SER spectra as in Figure
3
-1
3
, were normalized to the maximum values in the experiment.
atm of flowing CO. A severe attenuation of the 530 cm-1 oxide
band is evident even within the first spectral acquisition time
-1
(
0.5 s), along with the appearance of a band at 455 cm . While
the former feature is fully removed within the next 4 s, the latter
remains largely unchanged thereafter (see footnote 18). The
-
1
4
55 cm band is similar to that observed previously during
5c,6d
SERS studies of CO adsorption on Rh
and is assigned to
the metal-adsorbate stretch (νRh-CO) of CO adsorbed on atop
-
1
sites. An accompanying C-O stretch (νC-O) at 2030 cm is
also observed, confirming the identity of the adsorbed species.
Figure 4 shows the relative intensities of the νRh-O and νRh-CO
bands (open, filled circles), extracted from the data in Figure
3, again plotted versus time. (The integrated intensities for
νRh-CO were normalized to the final asymptotic value.) The

Reduction Kinetics of Surface Rh2O3
J. Phys. Chem. B, Vol. 102, No. 24, 1998 4789
Figure 8. Arrhenius plot of temperature-dependent rate constants, k
for the initial removal of oxide upon sudden exposure to 760 (circles),
6 (squares), and 7.6 Torr (triangles) of CO, extracted from SERS data
1
,
7
as in Figures 4 and 6. (See text for other details.)
Figure 6. Real-time SER spectral sequence acquired upon sudden
exposure of an initially oxide-covered Rh surface at 200 °C to a 2400
cm min flow of a 10% CO in argon mixture at 1 atm. Other details
as in Figure 1 caption.
has no observable effect on these zero-order kinetics. A number
of experiments were also performed involving CO-induced
reduction of oxides formed on the “thicker” Rh films described
above. However, no appreciable differences between these
kinetics and those for the thinner films could be discerned, in
contrast to the behavior with H2 (vide supra).
3
-1
Given that these data provide detailed insight into the transient
reactivity of surface oxide toward gaseous CO reactant, it is of
interest to relate this information to both the temperature-
dependent surface speciation and overall kinetics for the well-
studied catalytic oxidation of CO by O2. Experiments were
therefore performed to determine under which conditions oxide
and/or adsorbed CO forms in the presence of flowing mixtures
of CO and O2. Rhodium surfaces were heated in a stepwise
periodic fashion in various O2:CO mixtures (19:1 to 1:10) at 1
3
-1
atm with a flow rate of 200 cm min , SER spectra being
recorded at each temperature until a time-independent response
was attained (requiring ca. 3 min). Figure 9 shows typical
temperature-dependent SER spectra acquired upon heating a Rh
surface in a 19:1 O2:CO mixture. At 100 °C (bottom spectrum)
-
1
the 460 cm νRh-CO band is seen to be dominant, indicating
the extensive presence of CO adsorbed on atop sites. A weak
Figure 7. Plot of the relative integrated intensities of νRh-O (open
circles) and νRh-CO (closed circles) bands for oxide and adsorbed CO,
respectively, as a function of time at 200 °C following the sudden
addition of a 2400 cm min flow of a 10% CO in Ar mixture at 1
atm (cf. Figure 4). Data extracted from time-dependent SER spectra
as in Figure 6.
-1
broad band is also discernible at 670 cm , similar to the SERS
feature observed previously during CO oxidation on Rh.6d This
band was assigned to a bending vibration of a surface carbonate
3
-1
-1
species, based partly on the 20 cm frequency downshift
1
8
6d
observed upon O2 isotopic substitution.
Evidence for a
surface carbonate species has also been noted in a study of CO
oxidation on Rh(111), from vibrational bands at 750, 900, and
pressure, exhibiting an Arrhenius activation energy of 7.5 ( 1
-
1
-1
kcal mol with a preexponential factor of 200 s . While this
trend in k1 continues up to 350 °C for the two lowest CO partial
pressures, a distinct “knee” is observed in the case of 100%
CO (Figure 8). At 300 °C the rate constant for this reactant
composition is almost 10-fold higher than those for the two
lower pressures. This divergence increases further at 350 °C,
above which the rates even for the lower CO pressures become
immeasurably fast. For oxide removal below 300 °C, zero-
order rate constants, k0, were also obtained from the linear (i.e.,
the longer-time) portions of the Iox-t data. Arrhenius plots
extracted from these data reveal an activation energy of 11 (
-
1
1400 cm
formed during reaction under net oxidizing
7
c
conditions. While the present Raman spectra remain largely
unchanged up to 150 °C, further heating to 200 °C yields a
-
1
diminution of the 460 cm band intensity (Figure 9). Fur-
thermore, the concomitant onset of Rh surface oxidation is
signaled by the growth of the broad overlapping νRh-O feature
-
1
centered around 540 cm , assigned as before to Rh2O3. The
-
1
460 cm band is removed completely by 350 °C, as expected
-
1
from thermal CO desorption, although the 540 cm oxide
feature remains. While the temperature at which the surface
begins to oxidize under this reactant flow is essentially identical
to the case of pure O2 (vide supra), the extent of oxidation at a
-
1
-1
2
kcal mol with a preexponential factor of about 1000 s .
Varying the CO partial pressure over the range 76-760 Torr

4790 J. Phys. Chem. B, Vol. 102, No. 24, 1998
Williams et al.
yields values of the CO2 production rates (molecules site- s )
1 -1
from transient CO reduction of the surface oxide that are
-
1
numerically equal to the k1 (s ) values in Figure 8. For initial
oxide reduction at 300 °C at the same CO partial pressure (76
-
1
Torr) as above, k1 ≈ 0.3 s , corresponding on this basis to a
-
1 -1
transient CO2 production rate of 0.3 molecules site s . We
therefore deduce that the steady-state catalytic production rate
3
of CO2, at least for the above conditions, is very much (10 -
fold) higher than for transient oxide removal.
Discussion
A notable feature of the present results is the dramatic
differences in the transient kinetics of oxide removal by gaseous
H2 and CO. In addition to the much more rapid oxide reduction
rates induced by H2, the markedly dissimilar oxide coverage-
time functionalities signal that distinctly different mechanisms
govern these processes. In interpreting the results, it is
appropriate to first compare them with the (albeit sparse)
literature information on these important processes.
Surface Rh2O3 reduction by 1 Torr of CO at 145 °C has been
1
2c
investigated using imaging atom probe mass spectrometry.
-
1
An initial first-order rate constant of 0.1 s was extracted by
measuring the amount of oxygen on the surface at different
points during the reaction. This value is similar to that observed
here at 150 °C with 7.6 and 76 Torr of CO (Figure 8). Several
kinetic and UHV-based temperature-programmed desorption
Figure 9. Temperature-dependent SER spectra obtained during
3
-1
2
exposure of rhodium to 200 cm min of a 19:1 O /CO mixture at 1
atm. The spectral acquisition time was 20 s. (See text for other details.)
given temperature is apparently less, as gleaned from the
typically lower νRh-O band intensities seen here in the presence
of CO.
(
TPD) investigations have examined removal of adsorbed
9
,10,21,22
atomic oxygen by H or CO.
The rate constants for
reaction of adsorbed oxygen with hydrogen tend to be at least
00-fold higher than those with CO at a given temperature. This
Furthermore, decreasing the O2/CO ratio to 10:1 and below
yielded temperature-dependent SER spectra that are essentially
identical with those seen in the presence of pure CO. Thus,
the 530 and 670 cm-1 bands are now absent, the spectra
exhibiting only the familiar νRh-CO and νC-O vibrations at 470
1
2
is in harmony with the 10-10 -fold larger rates measured here
for H2-induced oxide removal in the “postinduction” time
regime, as compared to those found for initial reduction with
CO under similar conditions. It is nonetheless important to
recall the distinction between adsorbed oxygen atoms and a
surface oxide. The latter, but not the former, entails disruption
of the metal substrate atoms to form a distinct “surface phase”
where the oxygen atoms are located at least partly within the
metal lattice, so that oxide reduction requires displacement from
this film.
The most striking characteristic of oxide removal by hydro-
gen, yet not with CO, is the induction period observed prior to
the rapid decay in oxide coverage (cf. Figure 2). This behavior,
along with the marked effects of pressure and temperature, can
most readily be rationalized in terms of a “nucleation/growth”
type of reaction mechanism. Taking into account the findings
from numerous UHV-based investigations involving surface
reactions between oxygen and hydrogen on Rh, a plausible
mechanistic sequence for the oxide removal is
-
1
and 2030 cm , respectively. These bands are removed within
the temperature range 250-350 °C, as observed with O2-free
gaseous CO. Interestingly, then, the temperature-dependent
formation of oxide during the ambient-pressure catalytic oxida-
tion of CO on Rh requires a large O2 excess and is very sensitive
to the O2/CO pressure ratio. Adsorbed CO, on the other hand,
is formed extensively even in the presence of a large (20-fold)
oxygen excess, at least at temperatures below where thermal
desorption proceeds.
The steady-state rates of CO2 formation for a number of
reactant ratios were also obtained via mass spectrometric
monitoring of the gas-phase flow composition, in order to
compare the kinetics for the overall catalytic rates with those
for the transient CO oxidation by oxide as considered above.
The procedure utilized was identical to that outlined for the
temperature-dependent SERS experiments described above. The
steady-state CO oxidation rates were extracted from the MS
data in the conventional form of “turnover frequencies”
O2- S O(ads)
(1)
(2)
(3)
(4)
(5)
-
1 -1
(
molecules site s ), deduced by assuming a surface reaction
H2(g) + 2S S 2H(ads)
15
-2
site density of 2 × 10 sites cm . The steady-state CO2
production rates deduced on this basis, for example, at 250 and
H(ads) + O(ads) w OH(ads) + S
3
00 °C for a 9:1 O2:CO reactant ratio at 1 atm are 140 ( 50
-
1
-1
and 270 ( 50 molecules site s , respectively. The CO
oxidation kinetics are similar to those observed previously in
this laboratory on polycrystalline Rh as well as being in
OH(ads) + H(ads) w H O + 2S
2
(g)
7
approximate accordance with reports for single-crystal surfaces.
or 2OH(ads) w H O + O(ads) + S
2 (g)
A direct comparison of the CO2 formation rates with the rate
constants for transient oxide removal presented above can be
made with a knowledge of the oxide surface concentration, Γox,
involved in the latter measurements. Assuming that Γox ≈ 2
where S denotes a free site and O2 represents the lattice oxide.
Step 5 is included because it has been shown under UHV
conditions to be important when the coverage of adsorbed
-
1
5
-2
21
×
10 atoms cm , i.e., equal to the number of surface sites,
oxygen is high, as is probably the case here.

Reduction Kinetics of Surface Rh2O3
J. Phys. Chem. B, Vol. 102, No. 24, 1998 4791
Prior to introduction of hydrogen, the surface is essentially
completely oxidized to form Rh2O3 (vide supra) as well as being
saturated with atomic oxygen on available metallic sites.
the need for step 1 to occur prior to the formation of OH species.
Especially given that step 1 may well be at least partly rate
limiting under some conditions (vide infra), the occurrence of
such a pathway may account for the notably accelerated oxide
reduction rates at higher H2 pressures. The formation of Hsub
may also contribute to, or even be responsible for, the
“nucleation/growth” oxide-removal kinetics seen at lower H2
pressures and possibly also observed kinetic sensitivity to the
Rh film “thickness”. Thus, the “nucleation centers” might
conceivably be defects, etc., in the oxide film where hydrogen
atoms can more readily enter the surface lattice. Reaction
1
3b
Under these conditions step 1 can be thought of as being
essentially in equilibrium. This situation results in a very low
probability for dissociative adsorption of H2, given that two
adjacent metal sites are required for step 2. However, the
atomic hydrogen that is eventually formed will readily react
with nearby adsorbed atomic oxygen to form water, which is
1
0,21
then desorbed.
The increasing number of free metal sites
formed by this process should trigger more extensive hydrogen
adsorption, leading to a rapidly accelerated depletion of surface
oxygen. The marked temperature dependence of the induction
time (Figure 2) supports the notion that reactions along the lines
of steps 3-5 are occurring during this initial period. Oxide
removal will be initiated on this basis when the adsorbed atomic
oxygen concentration decreases to a level that causes step 1 to
be driven irreversibly, thereby decomposing the oxide lattice.
Thereafter, reactions 3-5 most likely proceed at the boundaries
2
-
between Hsub and O can then be envisaged to “spread out”
from such centers, undermining the surface lattice structure and
thereby facilitating further oxide reduction.
We now turn to a consideration of the oxide reduction
pathways with CO. One striking difference in the reduction
kinetics with CO compared with H2 is the complete absence of
a discernible induction period with the former reductant. Unlike
hydrogen, the adsorption of atop CO should require only one
free metal site. As a result, this reductant would be expected
to rapidly access the available surface sites and react with nearby
oxygen to form CO2. Indeed, both Figures 3 and 6 show that
adsorbed CO is spectroscopically detectable within the first
spectral acquisition period (0.5 s), indicating facile adsorption
under the conditions employed here. The ability of CO to
initiate the reaction promptly upon introduction to the surface
is also clear from the kinetic data. However, it should be noted
that no significant CO-induced reaction occurs on the time scales
commensurate with the induction periods observed with H2
reductant (Figure 2), when the comparisons are made at the
same temperature and reductant pressure. Consequently, then,
the observed lack of an “induction period” for CO-induced oxide
reduction does not reflect a greater intrinsic ability of CO to
initiate the reaction.
of oxide/oxygen island structures, as has been proposed by
several investigators.1
4a,b,10a,20,21
Perhaps the most direct evi-
dence supporting such a mechanism is found in recent photo-
emission electron microscopic measurements, which show that
surface oxygen removal by hydrogen during the H2 + O2
-
5
reaction on Rh(110) at ca. 10 Torr proceeds via moving
reaction “fronts”.23 Most importantly, however, such a mech-
anistic sequence can account for the striking “autocatalytic”
nature of the observed oxide-reduction kinetics.
As noted above, hydrogen reduction of the “thicker” oxide
film features longer induction periods as well as slower ensuing
oxide-removal kinetics. This may be rationalized by the notion
that thicker films are likely to be more contiguous. Framed in
the context of the “nucleation/growth” mechanism proposed
above, such oxide films would have fewer surface metal sites
available initially for dissociative H2 adsorption, resulting in
longer induction times, as are observed. A possibly analogous
effect was found by Vis et al.,1 who showed using temperature-
programmed reaction that large (>60 Å) oxidized Rh particles
were more resistant to reduction by H2 than are smaller
supported crystallites. Indeed, reactions between adsorbed
atomic oxygen and hydrogen also exhibit a significant surface-
structure dependence. For instance, an activation energy of 5.3
The second major difference between the kinetic profiles for
oxide removal by CO as compared to H2 can be characterized
in general by the apparent “autoinhibition” appearance of the
former Iox-t curves (e.g., Figure 7) at lower temperatures (e300
4b
°
C), this curvilinear response contrasting the “autocatalytic”
form of the latter data (Figure 2). The rapid buildup of
substantial CO coverages appears to play a major role in
inducing the curvilinear shape of the oxide coverage-time
profiles (Figure 7). Figure 10 shows the relative coverages of
CO and oxide (filled and open symbols, respectively), extracted
from the point at which the oxide removal kinetics switch from
an initial rapid decay to the much more sluggish zero-order
character (cf. Figure 7) plotted versus temperature, for reduction
with 760 and 76 Torr of CO (circles and squares, respectively).
The onset of zero-order removal kinetics is seen to occur
uniformly at a CO coverage (θco) of about 0.8, regardless of
the pressure or temperature employed. (Note that this relative
coverage is obtained by normalizing the νRh-CO band intensity
to the final asymptotic value achieved during a given experi-
ment.) The initial introduction of CO onto the surface is likely
to yield rapid reduction of nearby adsorbed oxygen, triggering
net removal of oxide from the metal lattice via step 1. Along
with such reactiVe adsorption, a fraction of the newly adsorbed
CO will remain unreacted and thereby rapidly build up to high
coverages, as indicated by the vibrational data. Such dense,
possibly even close-packed, CO domains will cover an increas-
ing fraction of the surface, effectively limiting the locations
where oxide removal via reaction with adsorbed CO can proceed
to island edges or other unoccupied sites in the CO adlayer.
The reaction rate is therefore depressed by an oVerabundance
-
1
10c
kcal mol was found for rate constants obtained on Rh(111),
-1
whereas a somewhat higher (7.4 kcal mol ) value was extracted
from kinetic data on the (110) face.1
0b
Given that the oxide removal kinetics are essentially identical
when employing 1% or 0.1% H2 mixtures, it is interesting as
well as surprising that the rates are accelerated so dramatically
toward higher reductant pressures (i.e., g76 Torr). One
explanation is that a new, more facile, pathway for oxide
removal is afforded by the increased H2 partial pressure. For
instance, the concentration of a surface intermediate (e.g., OH)
might be increased, thereby allowing a reaction step such as
(5) to become preferred. A more novel, although speculative,
possibility is the involvement of “subsurface” as well as
adsorbed atomic hydrogen in the reaction pathway. Such a
species, entailing diffusion of hydrogen atoms into the metal
lattice, has been invoked in a number of surface science studies
on several transition metals,24 including rhodium.
24d
The
formation of subsurface hydrogen, Hsub, will be more likely at
large H2 exposures or higher gas pressures and at surfaces
containing numerous steps, kinks, etc., as should abound on
the polycrystalline films examined here. Once formed, Hsub may
react with the oxygen within the lattice, thereby circumventing

4792 J. Phys. Chem. B, Vol. 102, No. 24, 1998
Williams et al.
fold) excess of O2. The present finding falls within the range
of reactant pressures reported to be required for oxide forma-
7
,25
tion.
The present observed effectiveness of pure CO for
oxide reduction (cf. Figures 4, 5, and 7) might appear at first
sight to be at odds with such findings. However, O2 supplies
a copious amount of atomic oxygen to the surface under reaction
conditions, pushing the equilibrium of reaction 1 toward the
left. Thus, the oxide is “pinned” under such circumstances by
the presence of large quantities of adsorbed oxygen.
Also of significance is the substantial presence of adsorbed
CO at lower temperatures even with such O2-rich mixtures
(
Figure 9). The additional presence of carbonate under these
-
1
conditions, as implied from the weak 670 cm band, is in
harmony with suggestions from kinetic data that this adsorbate
is involved centrally in the CO oxidation pathway on oxide-
rich rhodium, even that the observed overall reaction involves
7d
rate-limiting carbonate decomposition. (However, unambigu-
ous evidence that this species, rather than reaction between
coadsorbed oxygen and CO to form CO2 directly, provides the
kinetic bottleneck is apparently lacking.) While the smaller
extent of surface oxidation seen in comparison with CO-free
O2 suggests that the presence of the CO adlayer impairs oxide
formation, another factor limiting this process may well be the
reverse reaction, i.e., oxide reduction by CO. Given that
extensive CO adsorption is seen here to strongly inhibit oxide
reduction (vide supra), the observed occurrence of surface
oxidation under these reaction conditions is probably due as
much to CO-induced retardation of the forward direction of step
1. The above observation that the steady-state catalytic CO2
Figure 10. Relative intensities of νRh-O oxide (open symbols) and
Rh-CO adsorbed CO (filled symbols) bands for oxide and adsorbed
CO, respectively, during transient oxide removal by CO at 760 (circles)
and 76 Torr (squares) at the point where initial rapid oxide decay
kinetics transforms to slower zero-order decay (see text).
ν
of the adsorbed reductant, limiting severely as well as buffering
the availability of surface sites where the escaping oxide can
react (as adsorbed oxygen) with a juxtaposed CO molecule. Such
a “site-limited” mechanism can readily account for the observed
zero-order dependence in oxide, the kinetics being determined
primarily by the CO adlayer morphology.
Further inspection of Figure 10 shows that the relative oxide
coverage at which this “adsorbed CO-inhibited” pathway
emerges is significantly higher as the CO pressure is increased
from 76 to 760 Torr (open squares and circles, respectively).
This effect is consistent with the above mechanism in that the
higher CO pressure yields a markedly more rapid initial θco
increase. Given that the initial rate of oxide removal is only
slightly enhanced under these conditions (Figure 8), a greater
fraction of the oxide will become “trapped” by the enveloping
CO adlayer and will therefore be obliged to be reduced (more
sluggishly) within this adsorbate matrix.
3
formation rates are markedly (10 -fold) faster than for CO-
induced oxide removal eliminates the oxide, perhaps not
surprisingly, from contention as a bona fide reaction intermedi-
ate. However, this situation does not alter the likelihood that
the kinetics of CO-induced oxide removal play a major role in
determining the conditions for which oxide is present under
catalytic reaction conditions.
Finally, it is instructive to compare the present rates of
rhodium oxide reduction by H2 and CO with those from our
5c
recent kinetic study using methanol, in order to ascertain which
chemisorbed alcohol fragment may be responsible for oxide
removal in the latter process. Overall, the temperature-
dependent rates of rhodium oxide removal by methanol are
comparable to the initial reduction rates observed here with CO
at similar gas pressures. An interesting difference, however, is
that the methanol-induced oxide removal exhibits first-order
It remains to consider the first-order oxide removal kinetics
at temperatures (g300 °C) where the buildup of adsorbed CO
is either absent or insufficient to substantially autoinhibit the
process. Interestingly, the thermal desorption of CO coincides
with the marked divergence of the rate constants for initial oxide
removal seen for different reductant pressures under these
conditions (Figure 8). This marked upward curvature in the
Arrhenius plots suggests that an inherently more facile reaction
pathway becomes available at higher temperatures, especially
for the highest CO pressure. The reasons for this behavior,
however, are unclear. One possible explanation is that a CO
pressure-sensitive surface restructuring may be occurring which
increases the reactivity of the lattice oxide. While this notion
is speculative, even Pt(111) as well as (110) has recently been
observed to undergo surface restructuring in ambient CO
pressures at elevated temperatures.2
5c
decay kinetics even down to low oxide coverages. Also, while
adsorbed CO is seen to form on Rh under steady-state catalytic
conditions using methanol-rich CH3OH/O2 mixtures, this species
is not detected spectroscopically during transient oxide reduction
in pure methanol. Consequently, while these findings suggest
that adsorbed CO may well be the “active reductant” in the
methanol reaction, it is not formed in sufficient quantity to
autoinhibit the process, as is the case with gaseous CO.
From the observed kinetic behavior, we proposed that the
rate-determining step for oxide removal at high methanol
pressures and/or lower temperatures was the irreversible release
of oxygen from the surface lattice (step 1), the chemisorbed
methanol fragment(s) acting as an “adsorbed oxygen scavenger”
a
As already mentioned, it is of substantial interest to relate
the present observations regarding surface oxide reactivity to
CO oxidation by O2, particularly since this process is of
archetypical importance in heterogeneous catalysis. Surface
oxide formation is believed to poison catalytic CO oxidation
on Rh at elevated temperatures when O2 is in stoichiometric
5
c
and hence quenching the reverse process.
The observed
similarity in the initial oxide reduction rates with CO and
methanol lends some support to this notion. Furthermore, given
that methanol dissociative chemisorption should yield adsorbed
H as well as CO, the strikingly more facile oxide reduction
kinetics observed here with H2 strongly suggest that the former
2
5
excess. As Figure 9 shows, surface rhodium oxide forms at
around 200 °C during CO oxidation even in a large (ca 20-

Reduction Kinetics of Surface Rh2O3
J. Phys. Chem. B, Vol. 102, No. 24, 1998 4793
fragment does not act as the predominant reductant for
methanol-induced oxide removal. Most likely, this inability of
adsorbed H to react with oxide stems from a severe competition
for surface sites with other chemisorbed methanol fragments.
Nevertheless, the observation that pure H2 provides a much more
potent scavenger for surface rhodium oxide than either CO or
methanol supports the hypothesis that hydrogen can induce a
reduction pathway that circumvents the need for the oxygen to
be displaced from the metal lattice prior to reaction.
to C.G.T. and M.J.W. (CTS-9312008) from the National Science
Foundation. Some additional support was provided by the
University of Illinois at Chicago.
References and Notes
(
1) See for example: (a) Somorjai, G. A. J. Phys. Chem. 1990, 94,
1
013. (b) Goodman, D. W. J. Phys. Chem. 1996, 100, 13090.
(2) (a) Su, X.; Jensen, J.; Yang, M. X.; Salmeron, M. B.; Shen, Y. R.;
Somorjai, G. A. Discuss. Faraday Soc. 1996, 105, 263. (b) Crew, W. W.;
Madix, R. J. Surf. Sci. 1996, 349, 275. (c) Weaver, M. J. J. Phys. Chem.
1
996, 100, 13079.
Concluding Remarks
(3) (a) Cremer, P. S.; Su, X.; Shen, Y. R.; Somorjai, G. A. J. Am.
The foregoing is believed to provide concrete evidence
supporting our contention that the present SERS approach can
yield detailed kinetic as well as surface speciation information
for interfacial chemical processes for systems and experimental
conditions (elevated temperatures, ambient gas pressures, etc.)
that encompass those of direct relevance to “real-world”
heterogeneous catalysis. Admittedly, quantitative kinetic analy-
ses are hampered by incompletely known relationships between
surface morphology, adsorbate coverage, etc., and the Raman
signal intensities. The present film-deposition strategy may also
turn out to be applicable only to polycrystalline metal or related
interfaces, although encompassing numerous systems of catalytic
interest. However, the excellent temporal resolution and surface
sensitivity afforded by SERS, together with the intrinsic benefits
of Raman spectroscopy in accessing vibrational modes spanning
wide frequency ranges, conspire to grace the technique with
unusual attributes for exploring mechanistic catalytic chemistry.
Perhaps most importantly, when utilized in flow-reactor systems
as employed here, the strategy provides a versatile in situ means
of exploring the catalytic role of surface species in both transient
and steady-state circumstances. This flexibility enables direct
links to be forged between the kinetics of individual interfacial
processes and the overall catalytic rates and mechanisms. Such
a combination of temporal versatility and interfacial chemical
specificity even in high-pressure gas or liquid systems is a
unique hallmark of this approach. Thus, insight has been gained
Chem. Soc. 1996, 118, 2942. (b) Su, X.; Cremer, P. S.; Shen, Y. R.;
Somorjai, G. A. Phys. ReV. Lett. 1996, 77, 3858.
(4) (a) Gao, P.; Gosztola, D.; Leung, L.-W. H.; Weaver, M. J. J.
Electroanal. Chem. 1987, 233, 211. (b) Leung, L.-W. H.; Weaver, M. J. J.
Am. Chem. Soc. 1987, 109, 5113. (c) Leung, L.-W. H.; Weaver, M. J.
Langmuir 1988, 4, 1076.
(5) (a) Tolia, A. A.; Williams, C. T.; Weaver, M. J.; Takoudis, C. G.
Langmuir 1995, 11, 3438. (b) Williams, C. T.; Takoudis, C. G.; Weaver,
M. J. J. Catal. 1997, 170, 207. (c) Williams, C. T.; Takoudis, C. G.; Weaver,
M. J. J. Phys. Chem. B 1998, 102, 406. (d) Chan, H. Y. H.; Williams, C.
T.; Weaver, M. J.; Takoudis, C. G. J. Catal. 1998, 174, 191.
(6) (a) Tolia, A. A.; Williams, C. T.; Takoudis, C. G.; Weaver, M. J.
J. Phys. Chem. 1995, 99, 4599. (b) Williams, C. T.; Tolia, A. A.; Weaver,
M. J.; Takoudis, C. G. Chem. Eng. Sci. 1996, 51, 1673. (c) Williams, C.
T.; Tolia, A. A.; Chan, H. Y. H.; Takoudis, C. G.; Weaver, M. J. J. Catal.
1
996, 163, 63. (d) Williams, C. T.; Black, C. A.; Weaver, M. J.; Takoudis,
C. G. J. Phys. Chem. B 1997, 101, 2874.
7) (a) Peden, C. H. F.; Goodman, D. W.; Blair, D. S.; Berlowitz, P.
(
J.; Fisher, G. B.; Oh, S. H. J. Phys. Chem. 1988, 92, 1563. (b) Schwartz,
S. B.; Schmidt, L. D.; Fisher, G. B. J. Phys. Chem. 1986, 90, 6194. (c)
Peden, C. H. F.; Houston, J. E. J. Catal. 1991, 128, 405. (d) Peden, C. H.
F.; Berlowitz, P. J.; Goodman, D. W. In Proceedings of the Ninth
International Congress on Catalysis; Phillips, M. J., Ternan, M., Eds.;
Chemical Institute of Canada: Ottawa, 1988; p 1214.
(
10.
8) See for example: Gao, S.; Schmidt, L. D. J. Catal. 1988, 111,
2
(9) (a) Baraldi, A.; Gregoratti, L.; Comelli, G.; Dhanak, V. R.;
Kiskinova, M.; Rosei, R. Appl. Surf. Sci. 1996, 99, 1. (b) Campbell, C. T.;
Shi, S.-K.; White, J. M. J. Phys. Chem. 1979, 83, 2255.
(10) (a) Gregoratti, L.; Baraldi, A.; Dhanak, V. R.; Comelli, G.;
Kiskinova, M.; Rosei, R. Surf. Sci. 1995, 340, 205. (b) Comelli, G.; Dhanak,
V. R.; Kiskinova, M.; Paolucci, G.; Prince, K. C.; Rosei, R. Surf. Sci. 1992,
269/270, 360. (c) Thiel, P. A.; Yates, Jr., J. T.; Weinberg, W. H. Surf. Sci.
1979, 90, 121.
(11) See for example: Root, T. W.; Schmidt, L. D.; Fisher, G. B. Surf.
Sci. 1983, 134, 30.
in the present study regarding the kinetics of oxide reduction
by CO and H2 (and previously methanol).5c Such information
gains more broad-based importance, for example, when con-
sidered in light of commonly proposed mechanisms for oxida-
tion catalysis. The present findings suggest in this regard that
the notion of lattice oxygen (i.e., oxide) participating directly
in catalytic oxidation reactions, often known as the Mars-Van
Klevan mechanism,26 may be flawed, at least for rhodium.
Another factor contributing to our enthusiasm for the
technique as an increasingly powerful tool for in situ surface
chemical characterization in heterogeneous catalysis is the
enhanced sensitivity of Raman spectroscopy in general, due
especially to recent improvements in spectrometer and detector
technology. It is therefore now feasible not only to exploit
Raman spectroscopy for interfacial characterization even in the
absence (or greatly diminished presence) of the SERS effect
but also to explore a widening range of interesting interfacial
materials by engendering some surface Raman enhancement by
film deposition or other tactics. Especially when combined with
other means of scrutinizing surface chemical speciation and
kinetics applicable to technologically relevant interfaces, we
believe that such novel SERS approaches stand poised to make
substantial contributions to our broad-based understanding of
heterogeneous catalysis in both high-pressure gaseous and
liquid-phase environments.
(
12) (a) Kellogg, G. L. Phys. ReV. Lett. 1985, 54, 82. (b) Kellogg, G.
L. J. Catal. 1985, 92, 167. (c) Kellogg, G. L. Surf. Sci. 1986, 171, 359.
(13) (a) Logan, A. D.; Datye, A. K.; Houston, J. E. Surf. Sci. 1991,
45, 280. (b) Tolia, A. A.; Smiley, R. J.; Delgass, W. N.; Takoudis, C. G.;
Weaver, M. J. J. Catal. 1994, 150, 56. (c) Gorodetskii, V. V.; Nieuwenhuys,
B. E.; Sachtler, W. M. H.; Boreskov, G. K. Appl. Surf. Sci. 1981, 7, 355.
(14) (a) Wong, C.; McCabe, R. W. J. Catal. 1987, 107, 535. (b) Vis, J.
C.; van’T Blik, H. F. J.; Huizinga, T.; van Grondelle, J.; Prins, R. J. Catal.
985, 95, 333. (c) Martin, D.; Duprez, D. Appl. Catal. A 1995, 131, 297.
(
Sci. 1990, 240, 85. (b) Houtman, C.; Barteau, M. A. Langmuir 1990, 6,
1558.
(16) (a) Thiel, P. A.; Yates, Jr., J. T.; Weinberg, W. H. Surf. Sci. 1979,
2, 22. (b) Castner, D. G.; Somorjai, G. A. Appl. Surf. Sci. 1980, 6, 29.
(
L.-W. H. ACS Symp. Ser. 1988, 378, 303. (b) Weaver, M. J.; Hupp, J. T.;
Barz, F.; Gordon, J. G.; Philpott, M. R. J. Electroanal. Chem. 1984, 160,
21.
2
1
15) For example: (a) Parmeter, J. E.; Jiang, X.; Goodman, D. W. Surf.
8
17) (a) Weaver, M. J.; Corrigan, D. S.; Gao, P.; Gosztola, D.; Leung,
3
(
18) A series of oxide reduction experiments were also performed
utilizing CO of (ostensibly) similar purity which had been stored in a steel
cylinder. It is well-known that such gas if not purified further can contain
iron carbonyls, which may deposit on metal surfaces at elevated temper-
atures. Indeed, Rh surfaces heated in such CO in both the presence and
absence of oxygen were observed to darken above 300 °C. Parallel XPS
and SIMS surface analysis revealed that large amounts of iron, carbon, and
oxygen were present. Oxide reduction experiments with this “contaminated”
CO were therefore carried out only between 150 and 300 °C. While the
initial oxide removal kinetics were found to be very similar under these
conditions to those evaluated using Fe-free CO, the surface oxide was only
incompletely reduced using impure CO. In addition to formation of adsorbed
Acknowledgment. The XPS analysis (footnote 18) was
undertaken by Harry Chan. This work was supported by a grant
-1
-1
CO (460 cm ), a strong band at 670 cm was seen to grow in, especially
at higher temperatures.

4794 J. Phys. Chem. B, Vol. 102, No. 24, 1998
Williams et al.
(
19) In the rhodium surface oxidation experiments reported in ref 13b,
(21) Gurney, B. A.; Ho, W. J. Chem. Phys. 1987, 87, 5562.
-
1
the 290 cm band was often found to be of comparable intensity to the
main 530 cm oxide feature. The former is apparently more resistant to
CO-induced removal. This finding, together with the observed instability
or even nonappearance) of the 290 cm feature during rapid heating,
suggested that the band is due to a distinct species, possibly chemisorbed
or subsurface oxygen. The involvement of oxygen in the 290 cm band
was confirmed by an observed 15 cm downshift when formed with O2
rather than O2. However, more recent experiments undertaken using a
cleaner gas-handling/pumping system, including the present study, find that
the 290 cm feature tends to be weak or even absent. It is therefore unclear
(22) See for example: (a) Bowker, M.; Guo, Q.; Joyner, R. W. Surf.
Sci. 1993, 280, 50. (b) Campbell, C. T.; White, J. M. J. Catal. 1978, 54,
289.
(23) Mertens, F.; Imbihl, R. Chem. Phys. Lett. 1995, 242, 221.
(24) See for example: (a) Behm, R. J.; Penka, V.; Cattania, M.-G.;
Christmann, K.; Ertl, G. J. Chem. Phys. 1983, 78, 7486. (b) Kirsten, E.;
Parschau, G.; Stocker, W.; Reider, K. H. Surf. Sci. 1990, 231, L183. (c)
Chakarov, D. V.; Marinova, Ts. S. Surf. Sci. 1988, 204, 147. (d) Nichtl-
Pecher, W.; Stammler, W.; Heinz, K.; Muller, K. Phys. ReV. B 1991, 43,
6946.
(25) (a) Oh, S. H.; Carpenter, J. E. J. Catal. 1983, 80, 472. (b) Kim,
Y.; Shi, S.-K.; White, J. M. J. Catal. 1980, 61, 374.
(26) For a brief discussion, see: Masel, R. Principles of Adsorption and
Reaction on Solid Surfaces; Wiley: New York, 1995; p 500.
-
1
-
1
(
1
3b
-1
-
1
18
1
6
6d
-1
-1
at present whether this 290 cm band observed upon exposure of Rh films
to O2 is due to a species formed upon Rh oxidation as opposed to an
adventitious (albeit O-containing) impurity.
(
20) Wang, T.; Schmidt, L. D. J. Catal. 1981, 70, 187.