Electrochemical promotion by sodium of the rhodium-catalyzed NO + CO reaction

Article abstract of DOI:10.1021/jp001689x

The electropumping of sodium to a metallic rhodium film contacted with a solid electrolyte (sodium β″ alumina) under potentiostatic conditions resulted in the reversible spillover of the alkali from the electrolyte to the catalytically active surface. Sodium coverage increased linearly from 0 to approximately 0.02 monolayers while the rhodium work function decreased, also linearly, by 0.65 ev as the catalyst potential decreased over a range of 2000 mv. As sodium was pumped to the rhodium surface, the activity and N₂ selectivity of the CO + NO reaction improved by a factor of 3 and from 24 to 80%, respectively. NO promotion was caused by the enhanced NO vs. CO chemisorption and the sodium-induced dissociation of chemisorbed NO. The latter factor was mainly responsible for the sodium-induced increase in nitrogen selectivity.

Full text of DOI:10.1021/jp001689x

J. Phys. Chem. B 2000, 104, 11883-11890
11883
Electrochemical Promotion by Sodium of the Rhodium-Catalyzed NO + CO Reaction
Federico J. Williams, Alejandra Palermo, Mintcho S. Tikhov, and Richard M. Lambert*
Department of Chemistry, UniVersity of Cambridge, Cambridge CB2 1EW, U.K.
ReceiVed: May 4, 2000
XPS shows that the electropumping of Na to a metallic Rh film contacted with a solid electrolyte (Na â′′
alumina) under potentiostatic conditions results in reversible spillover of the alkali from the electrolyte to the
catalytically active surface. As the catalyst potential (VWR) is decreased over a range of 2000 mV, the sodium
coverage increases linearly from zero to ∼0.02 monolayers, while the work function (φ) decreases, also
linearly, by 0.65 eV. Over this same regime of catalyst potential, under approximately stoichiometric conditions,
both the activity and N2 selectivity of the NO + CO reaction are strongly enhanced as Na is pumped to the
Rh surface. The measured Na coverage, electrochemical behavior, and catalytic response of the system are
fully reversible with catalyst potential. Our data are understandable in terms of the Na-induced dissociation
of adsorbed NO, which is thought to be the rate-limiting step. The nature of electrochemical promotion, the
reaction mechanism, the mode of promoter action, and the inequivalence of ∆φ and ∆VWR are discussed.
1. Introduction
The present work is principally motivated by the following
questions. Can the catalytic performance of Rh be usefully
The metal-catalyzed heterogeneous reduction of CO by NO
serves as a paradigm in catalytic science. It has been investigated
over many years under a wide range of conditions and by means
of a variety of catalyst types, including single crystals, poly-
crystalline foils and wires, and oxide-supported metals. The
earliest study carried out on single-crystal surfaces under
ultrahigh vacuum conditions was that of Lambert and Comrie,1
who examined the reaction on Pt{111} and Pt{110}. More
recently, Zaera et al.2 and Belton et al.3 have investigated the
reaction on single-crystal surfaces of rhodium using molecular
beam techniques and atmospheric pressure conditions, respec-
tively. The motivation for much current work on this topic
derives, at least in part, from the important role played by the
CO + NO reaction in practical automotive catalytic converters;4
our recent contributions to this subject have involved the
application of alkali promotion using both electrochemical and
classical methods, as indicated below.
enhanced by alkali promotion? What can alkali promotion reveal
about the reaction mechanism? In addition, we address some
fundamental issues of electrochemical promotion (EP). EP
provides in situ control of promoter concentration at the surface
of a working metal catalyst.10 This is achieved by pumping
promoter species (Na in the present case) to or from a suitable
solid electrolyte (Na â′′ alumina in the present case) to a porous
metal film catalyst with which it is in contact. By varying the
catalyst potential with respect to a reference electrode (poten-
tiostatic control), we may reverse the alkali coverage, work
function, and catalytic behavior of the metal surface. The
relationship between catalyst potential, work function, and alkali
coverage is of fundamental interest, and we investigate this.
We find that alkali promotion is effective in enhancing the
catalytic performance of Rh and the data point to NO dissocia-
tion as the rate-limiting step.
We have shown that on Pt surfaces promotion by (electro-
chemically supplied) Na greatly enhances both activity and
selectivity toward N2 production in the CO + NO reaction.5 A
mechanistic model was proposed and subsequently supported
by Monte Carlo calculations.6 This work eventually led to the
development of highly active and selective alkali-promoted Pd
and Pt dispersed catalysts.7,8 Most recently, we have shown that
Pt/γ-alumina-dispersed catalysts exhibit strong promotion by
Na under simulated exhaust gas conditions.9
As we have shown, the efficacy of alkalis in promoting
activity and selectivity in NO reduction over Pt and Pd derives
from alkali-induced NO dissociation on these metal surfaces,
which otherwise are not very efficient for NO dissociation. In
this respect, clean Rh surfaces are much more efficient than
both Pt and Pd surfaces, and this indeed is why Rh is a vital
component in automotive catalytic converters. However, the
mechanism of Rh-catalyzed NO reduction by CO remains
controversial.
2. Experimental Methods
2.1. Apparatus. UPS and XPS experiments were performed
under UHV conditions (base pressure, <10-10 Torr) in a VG
ADES 400 UHV spectrometer system equipped with a reaction
cell. The EP sample was mounted on a manipulator that allowed
translation between the reaction cell and the spectrometer
chamber. Full details regarding sample mounting, manipulation,
and data acquisition are given in an earlier publication.10 Quoted
binding energies are referred to the Au4f7/2 emission at 83.8
eV from the Au wire that formed the electrical connection to
the Rh working electrode. Work function changes were deter-
mined by measuring the change in secondary electron cutoff in
the UP spectrum relative to the Fermi edge.
Reactor measurements were performed in a well-mixed
reactor operated at atmospheric pressure. The EP sample was
suspended in this reactor with all electrodes exposed to the
reactant gas mixture. Inlet and exit gas analysis was carried out
by a combination of on-line gas chromatography (Shimadzu-
14B; molecular sieve 13X and Haysep-N columns) and on-line
* Corresponding author. Tel: 44 1223 336467. Fax: 44 1223 336362.
Email: RML1@cam.ac.uk.
10.1021/jp001689x CCC: $19.00 © 2000 American Chemical Society
Published on Web 11/30/2000

11884 J. Phys. Chem. B, Vol. 104, No. 50, 2000
Williams et al.
Figure 1. XRD patterns from (a) the bare Na â′′alumina wafer, (b)
the Rh film subsequently deposited on one face, and (c) the gold film
deposited on the other face.
mass spectrometry (Balzers QMG 064). N2, N2O, CO, and CO2
were measured by gas chromatography, and NO was monitored
continuously by mass spectrometry after the necessary calibra-
tions were performed. NO (Distillers MG) and CO (Distillers
MG) were diluted in ultrapure He (99.996%) and fed to the
reactor by mass flow controllers (Brooks 5850 TR). The total
flow rate was kept constant in all experiments at 34 × 10-5
mol s-1 (500 cm3(STP)/min), with partial pressures PNO and
PCO of 0.2-1.8 and 0.3-1.1 kPa, respectively, with added
helium (PHe) giving a total pressure of 1 atm in every case.
Reactant conversion was restricted to <15% in order to avoid
mass transfer limitations. Control experiments were carried out
in which the total flow was varied by a factor of 5 in order to
verify that the observed changes in activity were indeed due to
changes in actual surface reaction rates and unaffected by mass
transfer limitations. Nitrogen and carbon mass balances always
closed to within 5%.
A galvanostat-potentiostat (Ionic Systems) was used in order
to maintain a given potential difference between the working
and reference electrodes (potentiostatic mode). All experiments
were carried out in potentiostatic mode by following the effect
of catalyst potential (VWR, measured with respect to the reference
electrode) on (i) the sodium coverage (UPS/XPS experiments)
and (ii) the reaction rates (CSTR experiments). In a separate
control experiment, the ohmic drop free potential of the working
electrode was determined using the current interruption tech-
nique. Current interruption was achieved by means of an
analogue switch (ADG201HS) with a response time of 50 ns.
A digital oscilloscope was use to record the ohmic drop. The
results showed that the ohmic drop contribution to VWR could
be neglected under all conditions. That is, the VWR values
reported below are true values. XRD data were obtained on a
Philips PW1877 automated powder diffractometer, using a Cu
KR source (λ ) 1.540 56 Å). All spectra were taken over a
range of 10° < 2θ < 70° and referenced to the JCPDS database.
2.2. Sample Preparation and Characterization. The EP
samples for spectroscopic examination and catalytic testing were
prepared by depositing a rhodium metal film on one face of a
Na-â′′ alumina wafer: this constituted the catalyst (working
electrode). The gold counter and reference electrodes were
deposited on the other face, all three electrodes being deposited
by DC sputtering of Rh or Au in argon. The wafer dimensions
were 2 cm × 2 cm × 0.1 cm (spectroscopy) and 1 cm × 0.3
cm × 0.1 cm (catalytic testing), the latter being determined by
the need to limit reactant conversion to <15%.
Figure 2. XP spectra showing the effect of oxygen pretreatments on
the amount of carbon on the catalyst. (a) Rh 3d spectra and (b) C 1s
spectra.
on one face, and (iii) the gold film deposited on the other face.
The rhodium and gold patterns show three peaks characteristic
of the relevant metal (arrowed) and three additional peaks due
to the underlying â′′alumina (filled circle).
The bottom spectra in Figure 2a,b correspond to the Rh 3d
and C 1s XP spectra of the EP sample immediately after
installation. The binding energies indicate that the film consisted
of metallic rhodium (Rh 3d5/2 at 307.3 eV) and that the only
observable impurity was elemental carbon (C 1s at 284 eV).
To remove the carbon, we treated the sample with oxygen in
the reaction cell (PO2 ) 0.2 mbar at T ) 563K and VWR
)
+1000 mV for 15 min). The spectra resulting from two
successive oxygen treatments are shown in Figure 2a,b. Cleaning
the rhodium film with oxygen resulted in a 37% increase of
the Rh signal and a 81% decrease in the carbon signal, with no
detectable oxidation of the rhodium surface.
The active metal area of the rhodium catalyst was determined
using two different methods: (i) the CO methanation technique
developed by Komai et al.11 and (ii) the electrochemical
technique developed by Vayenas et al.12 (Before measuring the
surface area, we calcined the sample in oxygen and reduced it
with hydrogen at 573 K in order to generate a clean metal
surface.) The first method makes use of a sensitive FID detector
to monitor the conversion of chemisorbed CO to methane on
metal sites. On the basis of a 1:1 CO-to-surface-metal atom
ratio, this yields a surface area equivalent to 1.5 µmol Rh (∼560
cm2 ). The second technique involves measuring galvanostatic
transients. A fixed negative current was applied between the
working and counter electrodes (i.e., sodium supply to the
catalyst) while measuring the resulting changes in catalyst
potential as a function of time. It can be shown using Faraday’s
Law and the Helmholz equation12 that
P0I
These EP samples were characterized by XRD, XPS, and
surface area measurements. Figure 1 shows the XRD patterns
from (i) the bare Na â′′alumina wafer, (ii) the Rh film deposited
A )
(1)
dφ
ꢀ
0 dt

Rhodium-Catalyzed NO + CO Reaction
J. Phys. Chem. B, Vol. 104, No. 50, 2000 11885
Figure 4. Binding energy shift of the Na 1s XPS component due to
Na in the solid electrolyte as a function of the change in VWR
.
Figure 3. Na 1s XPS spectra showing the effect of catalyst potential
on the sodium coverage of the Rh film under UHV conditions at 580
K. Invariant component due to Na on Rh; shifting component due to
Na in solid electrolyte.
where A is the catalyst surface area (m2), P0 is the initial dipole
moment of Na on Rh (1.7 × 10-29 C m13), I ) -10 × 10-6
C/s is the constant current applied between the working and
counter electrodes, ꢀ0 ) 8.85 × 10-12 C2/J m, and dφ/dt is the
initial slope of the work function versus time curve. As we
cannot determine dφ/dt experimentally, we measured edVWR/dt
and used the relationship between edVWR and dφ (discussed
below); thus, edVWR/dt ) -1.12 × 10-21 J/s, and dφ )
0.32edVWR. Inserting these values in eq 1 yields a surface area
of ∼552 cm2 (using a surface metal atom density of 1.6 × 1015
atom/cm2, which corresponds to the {111} plane of Rh). This
is equivalent to 1.4 µmol Rh, in good agreement with the value
measured by CO methanation (1.5 µmol Rh). In a similar
fashion, the metal surface area of the sample used for the reactor
measurements was found to be equivalent to 0.1 µmol Rh, or
∼37.6 cm2.
Figure 5. Integrated Na 1s XPS intensity due to sodium on the rhodium
surface and associated work function change of the rhodium film as a
function of catalyst potential.
Rh catalyst exhibits invariant binding energy (BE) and increases
in intensity with decreasing VWR. In contrast with this, the other
component exhibits constant intensity and a systematic shift in
apparent BE (∆EB). This shift is numerically equal to the change
in VWR, as shown in Figure 4, where the apparent binding energy
shift of the Na component ascribed to the Na â′′alumina has
been plotted against ∆VWR. These results confirm that this
component of the Na 1s emission arises from the solid
electrolyte whose electrostatic potential differs from that of the
3. Results
3.1. Photoelectron Spectroscopies. Figure 3 shows Na 1s
XP spectra obtained at 580 K as a function of catalyst potential
(VWR) under ultrahigh vacuum conditions. The +1000 mV
spectrum corresponds to the electrochemically cleaned sample;
increasingly negative values of VWR correspond to increasing
amounts of electropumped Na on the catalyst surface. These
observations confirm that EP works by supplying alkali promoter
species to the catalyst surface. The spectral behavior was
reversible and reproducible as a function of VWR, consistent with
the reversible and reproducible catalytic response under reaction
conditions at atmospheric pressure (see below). Note the residual
Na 1s emission from the electrochemically cleaned sample. To
understand this, recall that the rhodium film is both thin and
porous. The Na 1s emission therefore comprises two compo-
nents, one due to the Na present on the surface of the Rh film,
the other due to Na present in the underlying â′′alumina
electrolyte which was visible to XPS through the cracks and
pores in the metal film. The component due to the Na on the
Rh film by the amount ∆VWR
.
The filled circles in Figure 5 show the integrated Na 1s
emission intensity versus VWR. For VWR < ∼1000 mV, decreas-
ing the catalyst potential causes a linear increase of the sodium
coverage with potential. It is significant that over this same
regime, the work function also varies linearly with VWR (see
Figure 5 and below). A value for the sodium coverage (ϑNa)
may be estimated from the integrated Na 1s intensity of the
component associated with the Rh surface by making use of
the relationship derived by Carley et al14
YmσsNFλ cos ψ
ϑNa
)
(2)
YsσmMs
where Ym is the integrated photoelectron signal from a particular
subshell of the adatom (i.e., Na(1s)); Ys is the integrated signal

11886 J. Phys. Chem. B, Vol. 104, No. 50, 2000
Williams et al.
from the relevant subshell of the substrate (i.e., Rh(3d)), σm
and σs are the subshell photoionization cross sections for the
adatom and the substrate, respectively, Ms is the relative atomic
mass of the substrate, F is the density of the substrate, λ is the
escape depth in the substrate for the substrate photoelectrons,
ψ is the angle of detection with respect to the sample normal
of the photoelectrons, and N is Avogadro’s number. The sub-
shell photoionization cross sections and escape depth have been
taken from the calculations of Yeh et al.15 and Penn,16
respectively.
On this basis, the highest Na coverage achieved (VWR ) -900
mV, Figure 3) corresponds to 2.84 × 1015 atom/cm2, equivalent
to ∼1.7 monolayers (ML), where 1 ML is defined as one sodium
adatom per substrate atom. Such a coverage of sodium should
attenuate the Rh 3d XPS signal by ∼50% relative to that from
the sodium free surface; however, no such attenuation was
observed. The reason for this apparent discrepancy is that eq 2
is valid for an ideally flat substrate, whereas our samples are
very rough. Thus, eq 2 gives an overestimate of the actual
sodium coverage because the true metal surface area is much
greater than the geometrical area of the Rh film. We may take
this roughness factor into account in order to produce a more
realistic estimate of the sodium coverage by applying a simple
correction factor to the right-hand side of eq 2. This factor is
just the geometrical surface area divided by the true metal area.
The resulting calculated sodium coverage at VWR ) -900 mV
is then 0.013 ML, in good agreement with the observed lack of
attenuation of the Rh XPS intensity. It is also in good accord
with the linear dependence of the work function on sodium
coverage, as discussed below.
Figure 6. Effect of catalyst potential (VWR) on the CO2, N2, and N2O
formation rates and the nitrogen selectivity. Conditions: T ) 580 K,
P0NO ) P0CO ) 1 kPa.
in Figure 6 is the dependence of N2 selectivity on VWR, where
the former quantity is defined as
r
N2
S
)
(3)
N2
r
+ r
N2O
N2
These data show that the rates of CO2 and N2 production
increase with decreasing VWR, whereas the N2O rate decreases.
Thus, the system exhibits pronounced electrochemical promo-
tion: Na induces an increase in overall activity and in N2
production at the expense of N2O production, resulting in an
increase in nitrogen selectivity from 24% to 80%. The rate
enhancement ratios (F) defined as the ratio of the maximum
promoted rates to the unpromoted rates are as follows: F(CO2)
) 1.4, F(N2) ) 3.1, and F(N2O) ) 0.3. The observed changes
in turnover frequencies and nitrogen selectivity were fully
reversible; i.e., returning VWR to the initial value restored all
the reaction rates to their initial values. It is worth noting that
our TOF values for the unpromoted Rh film are very similar to
those reported for the CO + NO reaction over Rh(111) under
similar conditions of temperature and reactant pressures.18,19 This
provides additional validation for our estimate of the surface
area of the rhodium film.
3.3. Effect of Reactant Partial Pressures on Promoted and
Unpromoted Reaction Rates. The dependence of the CO2, N2,
and N2O reaction rates on PCO at fixed PNO for three different
values of catalyst potential is illustrated in Figure 7a-c. VWR
) +200 mV corresponds to the clean Rh surface (unpromoted
rate), while VWR ) - 200 mV and VWR ) - 400 mV correspond
to promoted surfaces with progressively increased sodium
loadings. Figure 7d shows the corresponding nitrogen selectivity
data: it is apparent that the highest selectivities to nitrogen
production always occur in the presence of the highest Na
loading (most negative potential). From Figure 7a,b it can be
seen that the CO2 and N2 rates exhibit Langmuir-Hinshelwood
behavior and that increased levels of Na result in a small
systematic increase in the CO partial pressure (P*CO) necessary
for inhibition. That is, Na favors the chemisorption of NO
relative to that of CO. The N2O rate also exhibits Langmuir-
Hinshelwood kinetics, but the effect of increased Na is
somewhat different: in particular, high levels of Na suppress
the N2O rate, and there is no systematic shift in P*CO. Thus
increasing the sodium coverage causes an increase in the rates
Work function data provide complementary information about
the surface processes occurring during the spillover of electro-
pumped Na. In Figure 5, the open circles show the work function
change (∆φ) when the catalyst potential is systematically
decreased from +1000 to -1000 mV at 580 K; the open
triangles show the corresponding data obtained when the process
was reversed, i.e., when the catalyst potential was increased
from -1000 to +1000 mV. Thus, the value of ∆φ corresponding
to a particular value of VWR is independent of the sample history.
This indicates that (i) the spillover process is reversible and (ii)
a given VWR corresponds to a given sodium coverage.
It is apparent from Figure 5 that ∆φ varies linearly with both
catalyst potential and ϑNa. This is a clear indication that we are
in the low alkali coverage regime, where such behavior is to be
expected.17 Specifically, ϑNa must be less than ∼0.1 ML.13 The
maximum sodium coverage achieved in the experiments whose
results are illustrated in Figure 5 caused a work function
decrease of -0.65 eV. Comparison with single-crystal data
obtained for the Rh(111)/Na system13 indicates that in the
present case, VWR ) -900 mV corresponds to ϑNa ∼0.025 ML,
i.e., about twice the value based on XPS intensity analysis. This
is a satisfactory level of agreement. For present purposes, they
key point is that two independent methods indicate that our
experimental data span the range of 0 < ϑNa < ∼0.02.
We now describe the catalytic consequences of Na pumping
to the Rh surface.
3.2. Effect of Catalyst Potential on Reaction Rates. Figure
6 shows typical steady-state (potentiostatic) rate data obtained
at 1 atm pressure and 580 K for constant inlet pressures of NO
and CO, P0 ) P0 ) 1 kPa. Turnover frequencies (TOF)
CO
NO
are expressed as molecules of product per Rh surface atom per
second. It is apparent that the CO2, N2O, and N2 turnover
frequencies are dependent on the catalyst potential, and the
observed behavior was fully reversible with VWR. Also shown

Rhodium-Catalyzed NO + CO Reaction
J. Phys. Chem. B, Vol. 104, No. 50, 2000 11887
Figure 7. Effect of the partial pressure of CO (PCO) on the rates of CO2 (a), N2 (b), and N2O (c) formation and on the selectivity of NO reduction
to nitrogen (d) at three different fixed catalyst potentials. T ) 580 K, PNO ) 0.75 kPa.
of formation of CO2 and N2 and a decrease in the rate of
formation of N2O, resulting in the increase in the nitrogen
selectivity shown in Figure 7d. Figure 7a also shows that at
high sodium loadings, high values of PCO cause an overall
decrease in the rate of formation of CO2, in agreement with the
results illustrated in Figure 8a, as discussed below. Figure 7d
shows that the nitrogen selectivity passes through a shallow
maximum as PCO increases and that this effect becomes more
pronounced as Na coverage increases.
Figure 8a-d shows corresponding results for the dependence
of the rates of production of CO2, N2, and N2O and the nitrogen
selectivity on PNO at fixed PCO for three different values of
catalyst potential. No rate maxima are observed within the
accessible PNO range. These results show that at lower partial
pressures of NO, the N2 and CO2 reaction rates exhibit maxima
as the sodium loading is increased (rate(highest Na) < rate-
(clean) < rate(moderate Na)). Figure 8d shows that the nitrogen
selectivity is relatively insensitive to PNO, other than at the
highest Na coverage; in every case, there is a very small overall
decrease in selectivity, in accord with the results of Belton et
al.20 Note also that the nitrogen selectivity of the unpromoted
Rh film is 24%sa value that is fully consistent with those
reported by Belton et al.3 for the same reactant partial pressures
and temperature, i.e., 40% over Rh(110), 26% over Rh(100),
and 24% over Rh(111). Taken together, Figures 7d and 8d
illustrate the point that over the entire range of partial pressures
explored, Na acts to increase selectivity.
between the working and reference electrodes (∆VWR) and ∆φ
deserves discussion. On one hand, Vayenas et al.21 have derived
the following expression:
e∆VWR ) ∆φ
(4)
and have validated it experimentally.22 On the other hand,
Zipprich et al.23 have observed deviations from this equality,
while Emery et al.24 and Williams et al.10 have failed to observed
this equality. Most recently, for yttria-stabilized zirconia (an
oxygen ion conductor), Vayenas et al.25 have also observed
deviations from eq 3, while Poppe et al.26 found, for their
samples, no variation of ∆φ at all with ∆VWR. Metcalfe and
co-workers have investigated these discrepancies.27,28 They
showed that for any change in catalyst potential
e∆VWR ) ∆φ + e∆Ψ
(5)
where ∆Ψ is the change in outer (Volta) potential. They argued
that eq 5 is valid when there is significant loss of spillover
species from the surface, either by desorption or reaction.
Equation 4 was taken to correspond to the case where there is
true thermodynamic equilibrium between the promoter species
in the electrolyte and those spilt over onto the metal catalyst
surface. Thus, the relative rates of spillover to the catalyst and
removal from the catalyst of the promoter species define two
limiting cases. If the kinetics of electrochemical supply are fast
whereas the rate of the loss of spillover species from the surface
is relatively slow, then e∆VWR ) ∆φ. Conversely, if the
electrochemical process is slow and the rate of the loss of
spillover surface species is fast, then e∆VWR ) ∆Ψ, since the
coverage of spillover species will be negligible and will not
4. Discussion
4.1. Relationship between∆VWR and ∆O. The relationship
between the change in the ohmic-drop free potential difference

11888 J. Phys. Chem. B, Vol. 104, No. 50, 2000
Williams et al.
Figure 8. Effect of the partial pressure of NO (PNO) on the rates of CO2 (a), N2 (b), and N2O (c) formation and on the selectivity of NO reduction
to nitrogen (d) at three different fixed catalyst potentials. T ) 580 K, PCO ) 1 kPa.
appreciably modify the metal surface work function. Thus, eq
5 may be thought to describe the behavior in the general case,
while eq 4 describes a limiting case.
{111} under UHV conditions at low alkali coverages and at
temperatures as low as 100 K. Note that although this effect
could provide a plausible explanation for our results obtained
under vacuum conditions, it is far from clear whether it is
relevant to the behavior of an EP catalyst under operating
conditions in the presence of a reactive gas atmosphere where
the alkali is present as a surface compound. The results of
Vayenas et al. and Metcalfe et al. were of course obtained at
atmospheric pressure.
4.2. Reaction Mechanism. Our results are fully consistent
with the view, advanced previously in regard to Pt-catalyzed
NO reduction,5,35 that Na acts to promote the rate-limiting step
and that this step is the dissociation of adsorbed NO. In the
specific case of the Rh-catalyzed NO + CO reaction, the identity
of the rate-limiting step has been a matter of some controversy.
The salient points are therefore considered here.
Belton et al.3,20,36-38 carried out a series of detailed investiga-
tions in which they used Rh single-crystal surfaces as model
catalysts, operated at atmospheric pressure. They concluded that
at high N coverages, nitrogen desorption is not rate-limiting,
as previously proposed.36 It was also shown that the NO
dissociation rate decreases with increasing coverage20 and that
NO and N react to form only N2O and not N2.37 At moderate
pressures, N2 formation was accompanied by N2O formation,
and there was no evidence that N2O readsorption played a
significant role.38 They showed that the reaction is structure
sensitive, with the rates increasing as the metal surface atom
density decreased in the order of (111) < (100) < (110)3.
On the other hand, studies performed under UHV conditions39
led to the conclusion that N atom recombination was the rate-
Equation 4 predicts that a plot of ∆φ against ∆VWR should
be linear with a slope of unity. Figure 5 shows that the
relationship is indeed linear, but the slope is 0.32. Equation 4
also predicts a change in φ of -2 eV over the range of VWR
explored, corresponding to a maximum Na coverage of ∼0.08
ML.13 In fact, the observed work function decrease was -0.65
eV, corresponding to ϑNa ≈ 0.025 ML.
A finite steady-state desorption rate for Na from the Rh
surface could provide an explanation for these deviations from
eq 4. However, at very low alkali coverages and under vacuum
conditions, such as those used here, Na desorption from
transition metal surfaces commences at ∼800 K (Pt{111},29 Ru-
{001}30) or ∼700 K (Cu{111},31 Ni{111}32). Although there
is no reported value for the saturation coverage of Na on Rh-
{111} at 580 K, data for closely related systems indicate that it
is highly unlikely that significant desorption occurs at 580 K
for ϑNa ≈ 0.025 ML.33 Thus, the evidence indicates that
Metcalfe’s model fails to provide an explanation for the observed
deviation from eq 4.
Partial incorporation of Na adatoms into the Rh could provide
an explanation for the deviation of the observed slope of the
∆φ versus ∆VWR plot from unity. If this were to occur, the work
function change induced by a given amount of Na would be
smaller than that induced by the same number of Na atoms on
top of the Rh. This seems a likely possibility, given the findings
of Lehmann et al.34 These authors obtained results that strongly
suggest subsurface incorporation of Na and K adatoms into Pt-

Rhodium-Catalyzed NO + CO Reaction
J. Phys. Chem. B, Vol. 104, No. 50, 2000 11889
determining step. A comparison of these UHV results with the
atmospheric pressure data of Belton et al. suggests that the
relative contribution of the different elementary steps changes
between the two pressure regimes. (Recall that no N2O is
produced under UHV conditions, whereas it is the main
nitrogen-containing product at atmospheric pressures). This may
be understood as follows. The NO dissociation rate is strongly
dependent on NO coverage (due to the effects of lateral
interactions and the availability of free sites). Under UHV
conditions, the steady-state coverage is low, NO dissociation
is extensive, the probability of the N + NO reaction is therefore
small (no N2O is observed), and the N recombination step is
rate-limiting. This situation is reversed under atmospheric
pressure conditions: now, the steady-state NO coverage is much
higher, NO dissociation therefore becomes rate-limiting, and,
as a result, N2O formation can occur.
where S represents a vacant site for adsorption. It was previously
proposed by Belton et al.3 in regard to catalysis by Rh, by
Lambert and Comrie1 in regard to the CO + NO reaction over
Pt, and by ourselves in regard to the EP of the Pt-catalyzed
reaction.5 The adsorption and dissociation of the reactants
depend on the availability of such free sites (reactions 6-8). In
this scheme, eq 8 is rate-limiting, both eq 7 and eq 8 are
enhanced by Na, and nitrogen selectivity depends on the relative
rates of reactions 10 and 11. On this basis, Na promotes activity
and nitrogen selectivity as a result of accelerating the key
reaction-initiating step (eq 8). This leads to an increase in the
coverage of N(a) and O(a) and a decrease in that of NO(a). As
a consequence, there is an increase in the rates of steps 9 and
11 and a decrease in the rate of step 10, yielding the observed
increase in overall activity and nitrogen selectivity (Figure 6).
Additionally, the kinetic data (Figure 7) demonstrate the
beneficial effect of Na in increasing the adsorption strength of
NO relative to CO such that higher CO partial pressures are
required before CO poisoning sets in.
Given that that the partial pressures and temperature condi-
tions employed in our experiments are very similar to the those
used by Belton et al., it is reasonable to propose that the same
mechanism prevails in the present case, i.e., that NO dissociation
is rate-limiting.
Figure 7d shows that the nitrogen selectivity exhibits a
maximum as a function of PCO for fixed PNO. This may be
understood as follows. Dissociation of NO(a) requires a nearest-
neighbor vacant site; therefore, nitrogen selectivity is expected
to increase with the number of such sites. When PCO is low,
the surface will be mainly covered by O atoms, inhibiting NO
dissociation; as PCO increases, the coverage of O decreases due
to formation of CO2, freeing up sites for NO dissociation. Thus,
N2 selectivity increases with PCO. When PCO is sufficiently high,
N2 selectivity eventually decreases because site blocking by CO
begins to inhibit NO dissociation. Therefore, the nitrogen
selectivity passes through a maximum. The effect is more
pronounced at higher sodium coverages as a result of the
enhanced NO dissociation induced by Na.
In the discussion that follows, the term “sodium coverage”
is used. This does not imply that the promoter phase is thought
to be present in the form of chemisorbed metallic sodium under
reaction conditions. Decreasing the value of the catalyst poten-
tial under UHV conditions does indeed pump chemisorbed
metallic sodium to the Rh surface, as we have shown (Figures
3 and 5). As would be expected and as we have demonstrated
previously,35 this electrochemically supplied sodium is in the
same chemical state as sodium deposited by vacuum evapora-
tion. However, in the presence of a reactive atmosphere, one
expects the Na to be present as a surface compound whose
identity depends on the composition of the ambient gas. This
too has been demonstrated previously. In the case of the EP of
Pt-catalyzed NO reduction by propene, XPS shows that alkali
nitrite and/or nitrate are present;35,40 in the case of propene
combustion, XANES shows that the promoter phase consists
of alkali carbonate.41
4.3. Mechanism of Promotion. All our results may be
understood in terms of the promotion of NO dissociation induced
by electropumped Na. Our XPS data demonstrate directly the
occurrence of this reversible spillover/backspillover effect, which
correlates very well with the reversible response of the catalytic
system. The dissociation of adsorbed diatomic molecules in the
electric field of adjacent alkali cations has received a detailed
theoretical analysis.42 The field lowers the energy of the NO
antibonding orbital with respect to the Fermi level. This
increases charge transfer from the metal to the NO π* orbital,
increasing the strength of the metal-N bond and decreasing
the strength of the N-O bond. In the specific case of Rh/NO,
studies on Rh(111)/K43,44 and Rh(100)/Na45 reveal the expected
strengthening the metal-N bond and weakening of the N-O
bond: both effects act to increase activity and selectivity.
The following reaction mechanism permits a rationalization
of the EP by Na of the Rh-catalyzed NO + CO reaction
As noted previously, Figure 8a,b shows that the reaction rates
exhibit a maximum as the sodium loading is increased for low
NO partial pressures at a fixed CO partial pressure. This
observation is consistent with the result displayed in Figure 7a,
namely, that at the highest CO:NO ratio, rate(highest Na) <
rate(clean) < rate(moderate Na). This may reflect the site
blocking effect of islands of Na-CO surface complexes46 that
are known to be stable under these conditions. At sufficiently
low CO:NO ratios, there is a switch in behavior: the activity
now increases with Na coverage. This may be rationalized in
terms of increased coverage by NO and its dissociation products,
especially O(a), which act to inhibit Na-CO complex formation
and, hence, concomitant poisoning.
As noted earlier, reaction rates over the unpromoted catalyst
exhibit maxima as a function of CO partial pressure (Figure 7)
and are approximately independent of NO partial pressure
(Figure 8). Such behavior was observed for the NO + CO
reaction over Pt;5 it is also in good agreement with the results
of Hendershot and Hansen,47 who studied the CO + NO reaction
over Rh(100). Belton et al.19 reported zero order in both CO
and NO under conditions similar to those used in this study.
However, close inspection of the data reveals no significant
divergence between their findings and ours. Thus, if we plot
the results presented in Figures 7 and 8 logarithmically, in the
manner employed in ref 19, it becomes apparent that our data
are in very good agreement with those of Belton et al.; the
variation of the unpromoted reaction rates with reactant partial
pressures is rather small in the range of partial pressures studied.
CO(g) + S f CO(a)
(6)
(7)
(8)
(9)
NO(g) + S f NO(a)
promoted by Na
promoted by Na
NO(a) + S f N(a) + O(a)
CO(a) + O(a) f CO2(g) + 2S
The EP by Na of the Rh-catalyzed CO + NO reaction exhibits
both similarities and differences compared with the EP of the
Pt-catalyzed reaction.5 In both cases, increasing the sodium
NO(a) + N(a) f N2O(g) +2S
N(a) + N(a) f N2(g) + 2S
inhibited by Na
(10)
(11)

11890 J. Phys. Chem. B, Vol. 104, No. 50, 2000
Williams et al.
(10) Williams, F. J.; Palermo, A.; Tikhov, M. S.; Lambert, R. M. J.
Phys. Chem B. 2000, 104, 615.
(11) Komai, S.; Hattori, T.; Murakami, Y. J. Catal. 1989, 120, 370.
(12) Ladas, S.; Bebelis, S.; Vayenas, C. G. Surf. Sci. 1991, 251/252,
1062.
(13) Mate, C. M.; Kao, C. T.; Somorjai, G. A. Surf. Sci. 1988, 206,
145.
(14) Carley, A. F.; Roberts, M. W. Proc. R. Soc. London, Ser. A 1978,
363, 403.
(15) Yeh, J. J.; Lindau, I. At. Data Nucl. Data Tables 1985, 32, 1.
(16) Penn, D. R. J. Electron. Spectrosc. Relat. Phenom. 1976, 9, 29.
(17) Verhoef, R. W.; Asscher, M. Surf. Sci. 1997, 391, 11.
(18) Permana, H.; Ng, K. Y. S.; Peden, C. H. F.; Schmieg, S. J.; Lambert,
D. K.; Belton, D. N. J. Catal. 1996, 164, 194.
(19) Permana, H.; Ng, K. Y. S.; Peden, C. H. F.; Schmieg, S. J.; Belton,
D. N. J. Phys. Chem. 1995, 99, 16344.
coverage causes an increase in the rates of formation of CO2
and N2 and results in a pronounced increase in the nitrogen
selectivity. However, in the case of Pt, increasing the sodium
coverage causes the rate of formation of N2O to go through a
maximum, whereas in the present case, under similar conditions,
the N2O rate always decreases with increasing Na loading. This
reflects the fact that clean Rh surfaces are more efficient than
clean Pt surfaces at dissociating NO. For example, NO dissoci-
ates on Rh(111) at low ϑNO (<0.25),48,49 whereas it does not
dissociate on Pt(111) under the same conditions.1,50
5. Conclusions
(20) Peden C. H. F.; Belton, D. N.; Schmieg, S. J. J. Catal. 1995, 155,
204.
(21) Vayenas, C. G.; Bebelis, S.; Lintz, H. G. Catal. Today 1992, 11,
303.
(22) Vayenas, C. G.; Bebelis, S.; Ladas, S. Nature 1990, 343, 625.
(23) Zipprich, W.; Wiemhofer, H. D.; Vohrer, U.; Gopel, W. Ber.
Bunsen-Ges. Phys. Chem. 1995, 99, 1406.
(24) Emery, D. A.; Middleton, P. H.; Metcalfe, I. S. Surf. Sci. 1998,
405, 308.
1. The coverage of electropumped Na on a thin-film Rh
catalyst contacted with a sodium-ion-conducting solid electrolyte
is determined by the catalyst potential; the effect is fully
reversible.
2. The Na coverage and the Rh work function scale linearly
with the catalyst potential. The sodium coverage varies from 0
to ∼0.02 monolayerssthe range over which the catalytic
performance also improves dramatically.
(25) Tsiplakides, D.; Neophytides, S.; Vayenas, C. G. Solid State Ionics,
in press.
3. The CO + NO reaction exhibits strong, reversible
electrochemical promotion under Na pumping to the rhodium
catalyst. This is due to the in situ control of sodium coverage
over the catalyst surface. Activity and selectivity toward N2
formation are markedly improved by a factor of 3 and from
24% to 80%, respectively.
4. NO promotion is due to (i) the enhanced NO versus CO
chemisorption and (ii) the Na-induced dissociation of chemi-
sorbed NO. The latter factor is primarily responsible for the
Na-induced increase in nitrogen selectivity.
(26) . Poppe, J.; Volkening, S.; Schaak, A.; Schutz, E.; Janek, J.; Imbihl,
R. Phys. Chem. Chem. Phys. 1999, 1, 5241.
(27) Emery, D. A.; Middleton, P. H.; Metcalfe, I. S. J. Electrochem.
Soc. 1999, 146, 2188.
(28) Emery, D. A.; Middleton, P. H.; Metcalfe, I. S. J. Electrochem.
Soc. 1999, 146, 2194.
(29) Harkness, I. R. Ph.D. Thesis, University of Cambridge, Cambridge,
U.K., 1994.
(30) Doering, D. L.; Semancik, S. Surf. Sci. 1983, 129, 177.
(31) Tang, D.; McIlroy, D.; Shi, X.; Su, C.; Heskett, D. Surf. Sci. 1991,
225, L497.
(32) Gerlach, R. L.; Rhodin, T. N. Surf. Sci. 1970, 19, 403.
(33) Diehl, R. D.; McGrath, R. Surf. Sci. Rep. 1996, 23, 43.
(34) Lehmann, J.; Roos, P.; Bertel, E. Phys. ReV. B 1996, 54, R2347.
(35) Yentekakis, I. V.; Palermo, A.; Filkin, N. C.; Tikhov, M. S.;
Lambert, R. M. J. Phys. Chem. B 1997, 101, 3759.
(36) Belton, D. N.; DiMaggio, C. L.; Ng, K. Y. S. J. Catal. 1993, 144,
273.
(37) Belton, D. N.; DiMaggio, C. L.; Shmieg, S. J.; Ng, K. Y. S. J.
Catal. 1995, 157, 559.
(38) Ng, K. Y. S.; Belton, D. N.; Shmieg, S. J.; Fisher, G. B. J. Catal.
1994, 146, 394.
Acknowledgment. We thank Norman Macleod for his
assistance with the surface area determination. F.J.W. acknowl-
edges financial support from Fundacio´n YPF, Fundacio´n An-
torchas, British Council Argentina, and King’s College Cam-
bridge. This work was supported by the U.K. Engineering and
Physical Sciences Research Council and by the European Union
under Grants GR/M76706 and BRPR-CT97-0460, respectively.
(39) Gopinath, C. S.; Zaera, F. J. Catal. 1999, 186, 387.
(40) Williams, F. J.; Palermo, A.; Tikhov, M. S.; Lambert, R. M. J.
Phys. Chem. B 1999, 103, 9960.
References and Notes
(41) Filkin, N. C.; Tikhov, M. S.; Palermo, A.; Lambert, R. M. J. Phys.
Chem. A 1999, 103, 2680.
(1) Lambert, R. M.; Comrie, C. M. Surf. Sci. 1974, 46, 61.
(2) Zaera, F.; Gopinath, C. S. J. Chem. Phys. 1999, 111, 8088.
(3) Herman, G. S.; Peden C. H. F.; Schmieg, S. J.; Belton, D. N. Catal.
Lett. 1999, 62, 131 and references therein.
(4) Taylor, K. C. Catal. ReV. Sci. Eng. 1993, 35, 457.
(5) Palermo, A.; Lambert, R. M.; Harkness, I. R.; Marina, O.; Vayenas,
C. G. J. Catal. 1996, 161, 471.
(6) Williams, F. J.; Aldao, C. M.; Palermo, A.; Lambert, R. M. Surf.
Sci. 1998, 412/413, 174.
(7) Yentekakis, I. V.; Lambert, R. M.; Tikhov, M. S.; Konsolakis, M.;
Kiousis, V. J. Catal. 1998, 176, 82.
(42) Lang, N. D.; Holloway, S.; Norskov, J. K. Surf. Sci. 1985, 150,
24.
(43) Bugyi, L.; Solymosi, F. Surf. Sci. 1987, 188, 475.
(44) Bugyi, L.; Kiss, J.; Revesz, K.; Solymosi, F. Surf. Sci. 1990, 233,
1.
(45) Hochst, H.; Colavita, E. J. Vac. Sci. Technol., A 1986, 4, 1442.
(46) Bertolini, J. C.; Delichere, P.; Massardier, J. Surf. Sci. 1985, 160,
531.
(47) Hendershot, R. E.; Hansen, R. S. J. Catal. 1986, 98, 150.
(48) Borg, H. J.; Reijerse, F. C. J. M.; van Santen, R. A.; Niemants-
verdriet, J. W. J. Chem. Phys. 1994, 101, 10052.
(49) Hopstaken, M. J. P.; van Gennip, W. J. H.; Niemantsverdriet, J.
W. Surf. Sci. 1999, 69-73, 433.
(8) Konosolakis, M.; Nalbantian, L.; MacLeod, N.; Yentekakis I. V.;
Lambert, R. M. Appl. Catal., B 1999, 22, 123.
(9) Konsolakis, M.; Macleod, N.; Isaac, J.; Yentekakis, I. V.; Lambert,
R. M. J. Catal., in press.
(50) Comrie, C. M.; Weinberg, W. H.; Lambert, R. M. Surf. Sci. 1976,
57, 619.