4.3.2 Removal of CO
Under fuel-lean conditions (excess O2), the oxidation of CO has been studied over a very large range of single crystals and model noble metal catalysts, one of the most intensively investigated examples being the Pd(111) surface. Although this metal is not a component of the current three-way catalyst used in the UK, it is worth considering the results in some detail for a number of reasons. The reaction on metals such as Pt is in many ways similar to that on Pd and, in any case, palladium is already being incorporated into future generations of catalytic converter, particularly for the US market. Most notable, however, is the fact that this is one of the few cases in which surface-science techniques have successfully revealed the details of a ‘real-world’ catalytic mechanism. Specifically, we will see how surface studies of the adsorption of CO and oxygen on Pd(111) – both individually and together – have led to the current understanding of the mechanism of CO oxidation.
LEED results for the adsorption of CO on Pd(111), obtained at room temperature and below, have been interpreted in terms of the structural models shown in Figure 7. One of the significant observations from this work is the readiness with which one arrangement of CO on the surface evolves into another. Thus at a surface fractional coverage of θ = 1/3 (Figure 7a), the CO occupies a hollow site where it can bind to three Pd atoms. As θ is increased to 1/2 (Figure 7b), CO moves out of the hollow to a bridging site, where it binds to two Pd atoms. Finally, at θ = 2/3 (Figure 7c), a hexagonal structure forms, in which half of the CO molecules reoccupy hollow sites, while the remainder bind to single Pd atoms at terminal sites. The readiness with which the CO molecules can reposition themselves suggests that the activation energy for surface migration in the chemisorbed state is low, and that CO is a highly mobile species under catalytic conditions.
Oxygen adsorbs dissociatively on Pd(111), and the O atoms are found to be less mobile on the surface than CO molecules. The structure of the chemisorbed layer at maximum coverage is shown in Figure 8.
Identify the adsorbate structure shown in Figure 8 in terms of the (m×n) notation, and determine the fractional surface coverage, θ, of oxygen atoms.
The unit meshes of the substrate (1 × 1) structure and of the adsorbate structure are shown in Figure 8A. Evidently, the latter is (2 × 2) in the (m × n) notation, so the full description of this structure should be Pd(111)(2 × 2)–O.
For adsorption on a single crystal surface, the fractional surface coverage is given by
θ = x/(m × n)
where x is the number of adsorbate species within the (m × n) adsorbate unit mesh. In this case (Figure 8A), O atoms occur only at the corners of the mesh, so the latter contains a total of (4 × 1/4) = 1 atom. Hence
θ = 1/(2 × 2) = 1/4
We might now assume that when CO and O2 are adsorbed together during the oxidation reaction, the properties of the system will be a simple combination of those of the two molecules adsorbed separately. The surface layer would then consist of mobile CO (maximum coverage, θ = 2/3) within a fixed lattice of O atoms (maximum coverage, θ = 1/4). The fact that this is not the case, as we shall see below, demonstrates an important point. Because of mutual interactions, the behaviour of two (or more) co-adsorbed species very often differs from their behaviour when adsorbed separately.
In the case of CO and O2, the order in which adsorption is carried out is significant. If CO is adsorbed first, to a coverage greater than one-third of a monolayer (θ = 1/3), subsequent oxygen adsorption is completely blocked. With lower coverages of CO, dissociative oxygen adsorption does occur; but the two species form separate domains on the surface (Figure 9a). Oxidation will then take place only at the boundaries between domains, and so it will be relatively slow.
When oxygen is adsorbed first to its maximum coverage, θ = 1/4, subsequent CO adsorption occurs readily and compresses the O atoms into domains in which the local coverage reaches θ = 1/3. At first, the adsorbed CO is found in separate areas (as when CO is adsorbed first), but as more is added mixed domains form, containing both CO and O, each at a local coverage of θ = 1/2 (Figure 9b).
These mixed domains bring CO(ad) and O(ad) into intimate contact, with the O atoms at twice the surface concentration possible in the absence of CO. Thus, the stoichiometry is now that required for the oxidation reaction. Moreover, an electronegative O atom will withdraw charge from the surface. In turn, the surface will withdraw charge from neighbouring Pd–CO bonds, weakening them and so making the CO more readily available for reaction. The net result is that the mixed domain is highly reactive and generates CO2 at temperatures far below room temperature.
Having thus established that, on Pd(111), rapid CO oxidation can occur by way of a Langmuir–Hinshelwood type process (a surface reaction between two adsorbed species), we are almost in a position to propose a detailed mechanism. First, however, we must consider the possibility of an alternative Eley–Rideal type mechanism, in which the rate-limiting step involves reaction between an adsorbed species and a molecule in the gas phase. In this case, there are two such possibilities:
Experiments show that exposure of a surface saturated with adsorbed CO to gas-phase O2 does not lead to CO2 formation. How does this simplify consideration of the above reactions?
This observation rules out the Eley–Rideal reaction between CO(ad) and O2(g), reaction 8.
We can make a decision about reaction 9 on the basis of the information provided in Figure 10. A Pd(111) surface presaturated with a quarter of a monolayer of O atoms was exposed to a beam of gaseous CO, and the surface coverages and the oxidation rate were monitored with time. Figure 10 shows that the rate became significant only after a population of CO had built up on the surface, and it reached a maximum when the coverages of O and CO were approximately equal. This is clear evidence for a Langmuir–Hinshelwood process. If reaction 9 had been operative, the rate would have been high initially, and would have fallen continuously as the oxygen layer was consumed by the reaction.
Given the research effort that was involved, the mechanism finally proposed for CO oxidation on Pd(111) is deceptively simple:
Although the exact nature of the surface intermediates is still not known, the depth of understanding of the catalytic mechanism is quite an accomplishment. But just how relevant are these surface studies to the practical catalysis taking place in the converter?
How do the observations outlined below lead to the conclusion that studies of CO oxidation, such as those for Pd(111) above, using surface-science techniques such as LEED, can be of direct relevance to the catalysis taking place in the converter?
During the reaction, surface contaminants such as carbon or sulfur are burnt off the Pd catalyst, so that the surface is clean, with the exception of the adsorbed reactants;
The catalytic activity does not depend on the crystallographic structure of the Pd metal surface. This is illustrated in Figure 11 where the rate of CO oxidation, over a wide temperature range, is shown to be virtually identical on five different Pd surfaces;
The rate of CO oxidation is independent of pressure, depending only on the ratio of the partial pressures of CO and O2, at least over a restricted range of conditions.
Techniques such as LEED examine adsorption at very low pressures onto clean single crystal surfaces, but the real catalytic system is more complicated than this. We are not dealing with a single crystal or a defined crystal plane, and the pressures involved are much higher. However, according to the factors listed above, CO oxidation has been shown to be independent of the Pd surface exposed, and independent of the total pressure of the reactants. In addition, the first observation suggests that under real conditions, the catalyst surface is clean; hence the effects of the adsorbed CO and O are likely to be the only relevant factors. Thus, the conclusions drawn about mechanism from surface studies may not be too far removed from those that apply in the case of the ‘real’ catalyst.
Indeed, Figure 12 shows that in the case of rhodium there is excellent agreement between the rates of CO oxidation over a Rh(111) single crystal surface and over a Rh/Al2O3 catalyst.
Although the sequence of elementary steps is quite simple, the overall kinetics of the CO oxidation reaction is not. The non-uniformity of the surface, and the segregation of the reactants in surface domains, complicates the detailed modelling of the kinetics. The exception is the special case of low surface coverages of CO and O atoms, when they are found to be randomly distributed over the surface and so satisfy one of the criteria for applicability of the Langmuir isotherm. Under these circumstances, Langmuir–Hinshelwood kinetics can be applied.
Figure 13 shows the comparative performance of single-metal catalysts for the oxidation of CO at a fixed temperature. Evidently, all three of the platinum group metals present in automotive catalysts are active for CO oxidation. In addition, results have shown that Rh may improve low-temperature activity. In the current (1996) three-way catalyst used in the UK, in which Pt constitutes 80–90% of the noble metal composition and Rh the remainder, it is the Pt that is mainly responsible for CO oxidation.
Under stoichiometric or slightly fuel-rich (reducing) conditions, where there is insufficient oxygen present to oxidise all of the CO, conversion can also occur by one of the following routes:
via the CO + NO redox reaction (reaction 6). This will be discussed in detail in section 4.4,
via the water-gas shift reaction (equation 2), because H2O is present in the exhaust gases as a product of combustion:
The water-gas shift reaction is catalysed by Pt and/or Rh, with ceria acting as an excellent promoter. Pt/CeO2–Al2O3 and Pt-Rh/CeO2–Al2O3 are particularly active combinations for the removal of CO under slightly fuel-rich conditions. The hydrogen produced in this reaction will react, in preference to CO, with any oxygen present. Hence, although the water-gas shift reaction removes CO, it also inhibits CO oxidation by producing hydrogen, which will remove any O2 present: