Gold has fascinated mankind for millennia, it being generally regarded as the most beautiful and noble of precious metals. And yet, this beauty is more than just skin deep. When particles of gold are synthesised in the nanoscale, they exhibit physical and chemical properties which are in near complete contrast to the material in bulk (r3-5 of 133), but are just as fascinating, if not more so.
Extended gold surfaces are well established as being chemically inert (1,2 OF 2b), and are therefore poor heterogenous catalysts. This lack of activity is mainly attributed to the low adsorption energies of gases and high dissociation barriers and/or lack of activation of the weakly adsorbed molecules, which are essential for most catalytic reactions(1 of 45, 3 of 133).
Generally, as we depart from the bulk state of a typical catalyst and tend towards progressively finer particle sizes, an increase in surface area would usually result in an increase in catalytic activity. Gold however, in keeping with its nobility, doggedly resists this trend, but when at least one of the material's dimensions is reduced to the nanoscale (â‰ˆ5nm (2b), as ultra-thin films, nanowires or clusters(133)), catalytic behaviour ensues.
Get your grade
or your money back
using our Essay Writing Service!
Indeed, at these dimensions, gold nanoparticles are very effective catalysts for a number of oxidation reactions (3-17 of2b), the most notable of which is the oxidation of CO. It has been demonstrated that for this reaction, gold-based nanocatalysts are the most active (21 of 3) and, at low temperatures (ambient to sub-ambient), exceed the activity of Pt or Pd based catalysts by a factor of five (9 of 3). On the other hand, the latter remain the most active for extended surfaces at high temperatures (2b).
This report focuses on the catalysis of low-temperature CO oxidation by gold nanoparticles. This reaction has been the subject of intense study due to environmental and health implications, principally though its potential applications in hydrogen purification in Proton Exchange Membrane fuel cells (22 of 132), and air purification (3). Moreover, the relative simplicity of the reaction per se, coupled with the lack of side reactions and products, makes it an ideal probe reaction for surface chemistry studies. (9 of 76)
In order to present a comprehensive picture of the most pertinent aspects of low-temperature CO oxidation by gold nanoparticles, this report will seek to cover the following objectives, but in a level of detail commensurate with the academic background of advanced under-graduate chemistry students:
To explain very briefly the importance this reaction in science and industry.
To outline the intrinsic properties which make Au nanoparticles so valuable for low-temperature gold-catalysed CO oxidation.
3. To explain in brief how the nature and properties of oxide support materials and subsequent metal-support interactions affect catalytic activity.
4. To provide an in-depth review of (published) plausible reaction pathways/mechanisms for low-temperature CO oxidation using supported and unsupported Au nanoparticles, comparing and contrasting these processes and assessing their merits based on recent published findings.
5. To give an overview of other factors (positive/negative) which have been found to affect the catalytic activity/stability of supported Au nanoparticles in low-temperature CO oxidation, and their influence on the adoption of Au nanocatalysts in industry.
To briefly discuss current/future research and goals with regards to objectives 1 through 5.
The project draws reliable information from varied sources; specialist textbooks, articles from peer reviewed journals, and material garnered from reputable websites such as www.gold.org. Due to the ongoing and rapid progress in the research of the subject matter contained herein, and the sheer volume of relevant material being published, a continuous effort was expended in keeping abreast with such developments throughout the execution of this review. Consequently, at the time of submission, the information contained in this project is deemed to be up to date.
2 Catalytic properties of Gold nanoparticles
2.1 Factors affecting catalytic activity
While it is generally agreed that catalytic activity depends on gold particle size, (1 of 133, 134) have stressed that many of the structural, energetic and electronic properties of gold nanoparticles cannot be deduced or extrapolated from those of coarser aggregates, i.e. they are non-scalable (xxx). There is still considerable debate over which of these properties is/are primarily responsible for gold's nanocatalytic activity (15 of 2b) and no overarching consensus has been reached thus far. The situation would be relatively simple were catalytic activity limited to pure (non-supported) gold clusters (42,43 of 115), but in real catalytic systems, gold nanoparticles are typically supported on metal oxides such as TiO2, Al2O3, Fe2O3 or SiO2, and gold-support interactions must also be taken into consideration(11).
Always on Time
Marked to Standard
Catalytic activity in gold-catalysed CO oxidation is considered to be directly related to the ability of the catalyst surface to adsorb O2 and CO simultaneously (120). A wide variety of factors have been presented as contributing towards this activity. Lopez et al. (140) proposed that the most significant of these is the increase in concentration of low-coordinated gold atoms with decreasing particle size. This view has been corroborated by several groups (84,115,14 of 115,4,5), although quantum size effects (84), the nature of the support (124), Au-support interface properties (27 of 84), and strain effects (32 of 84) have also been suggested to play significant roles too.
2.2 Chemisorption of CO and oxygen on gold surfaces
The adsorption of CO to gold nanoparticles is straightforward; bonding occurs exclusively via the C atom(120) and has been found to be stronger at low-coordinated corner, edge, step and kink sites than at facet Au surface locations(24, 49, 25,28, 29 of 84). This relationship is discussed in greater detail in section 2.x.
Except at very high temperatures, molecular oxygen does not adsorb either dissociatively or intact on extended gold surfaces (488 of 11). O2 dissociation is a prohibitively endothermic process (197 of 11). However, this barrier decreases along with particle size (196 of 11) and has been calculated to become surmountable but only at very small particle sizes of 10 atoms (<2nm) or less, which corresponds to a high degree of coordinative unsaturation of surface Au atoms(221 of 11). For larger particle sizes (2-5nm), Hakkinen, Landman and co-workers (220, 474 of 11) have argued that since CO oxidation (on gold nanoparticles) has been experimentally found to take place at low temperatures with very low activation energies, this implies that the reaction cannot proceed via a highly energetic dissociation pathway. Indeed, many groups (226, 247, 375, 377 of 11) have proposed mechanisms involving molecular oxygen. Fig xyz illustrates the PE profile of dissociative vs. molecular O2 adsorption.
< diagram b4 pg130, without balls . Footer to include abbrev only>
Curve A describes an oxygen molecule as it approaches the gold surface. After a weak physical adsorption (-âˆ†Hp), repulsion sets in due to lack of adsorbate/substrate orbital interactions. In order to proceed, O2 will have to surmount a very high PE barrier (activation energy, Eca) before it can meet Curve B (at I3), which represents the energy barrier of dissociated O2. An alternative route is presented if the physisorbed O2 acquires an electron (as, for example, by transfer from small Aun- clusters) in order to become an adsorbed O2-, for which activation energy is very small, thereby allowing access to Curve C and possible O2 dissociation (B4 129).
2.3 Coordinatively unsaturated Au cluster atoms and catalytic activity
Terrace sites on Au surfaces are well established as being catalytically inactive. This is not the case for low-coordinated surface, edge or corner sites (3, 14 of 76, 10-13 of 115), the presence of which have been found to strengthen the adsorption of both carbon monoxide and oxygen (15 34 of 125). As can be seen in Fig 1, using model octahedral clusters, 15,32 of 2b show that the fraction of these coordinatively unsaturated atoms relative to the total number of surface atoms increases with reduction in particle size (15,32 of 2b).
FIG. 1. Percentage of surface atoms on perfect octahedra in corner (O), edge (âˆ†), and face (â-¡) positions, as a function of the number of atoms along each side. Also shown are the corresponding dispersions (â-) and sizes (---). (31 of BT)
Using DFT based calculations, Lopez et al(140) determined the adsorption energy of CO and oxygen on a number of different gold surfaces (figure 5 (top)); a decrease in coordination number results in a stronger Au-adsorbate bond. The poor interaction between the very low energy d states of Au(111) surface atoms and oxygen 2p valence states results in a bond so weak that O2 activation (dissociation or noticeable O-O stretching) does not occur (34,35 of 140). The d states of Au atoms at lower coordination sites (Fig. 5 (bottom) are closer to the Fermi level, giving a stronger bond and therefore lower adsorption energies for both CO and oxygen. These in turn translate into lower surface reaction barriers at these sites.
Fig. 5. (Top) The correlation between the binding energies, for CO molecules and O atoms, with respect to the coordination number of Au atoms in a series of environments. Binding energies, in eV, reported are referred to gas-phase CO and O2, for O2 the energies are given per O atom. (Bottom) The dependence of the binding energy of O atoms with the position of the d band of Au atoms in different environments.
This Essay is
a Student's Work
This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.Examples of our work
Phala and Steen (134) extend the relationship between catalytic activity and Au d states by correlating the onset of catalytic activity to the cluster size where the contribution of the d-band hybridization energy to the total CO chemisorption energy switches from being positive (in the bulk state) to negative (for gold nanoparticles in the 5-6nm range) (Fig x). This in turn corresponds to an upward shift of the d-band centre with respect to the Fermi level.
Fig. 3 Size dependency of the contribution of the d-band hybridization energy to the total energy of CO chemisorption onto gold particles.
2.4 Quantum size effects in free gold clusters
There is considerable debate regarding whether the catalytic activity of free Au clusters or Au-oxide systems derives from anionic or cationic gold atoms (2,29,30,58,-65 of 84). Gold particle anions show a strong even-odd alternation with particle size (of 20 atoms or less) in their reactivity toward O2(160 ofNC, 61 of 84). This variation in anion reactivity is thought to arise from quantum-size effects (goodmanvalden), i.e. energy levels in the bulk state are considered to be continuous, but for materials in the nanoscale, these energy levels split up into electronic levels that can no longer be approached as a continuum. These quantum-size effects are particularly distinct for clusters of the noble metals with singly occupied s-orbitals (NCpg 95). The resultant quantized electronic particle structure leads to a pronounced even-odd alternation in the electron binding energy (160 of NC). These alternating open and closed shell valence electron configurations are depicted in fig.4 in terms of measured vertical electron detachment energies (VDEs), along with the chemical reactivity towards molecular oxygen involving the corresponding electronic levels.
Fig. 4. Electron binding energies measured as VDEs of gas phase Aun- ions in comparison to the reactivity of the same particle sizes toward O2 (NCpg 95)
Fig 4 shows that binding between molecular oxygen and Au clusters is favoured where n is even, i.e. where charge transfer between the low electron-binding energy Au s-orbitals and the O2 Ï€* antibonding orbitals is facilitated (resulting in an activation of the O-O bond through bond stretching (8-10 of 120)). Experimental results derived from various sources support this pattern (Fig.5 (a)). In contrast, no electron transfer occurs between cationic Au clusters and O2. However, neutral Aun clusters for which n<4 show moderate O2 adsorption(62 of 84), but this is weaker than for anionic clusters and is thought to follow a different mechanism(60a of 84). Interestingly, DFT studies carried out by Laursen and Linic show that even though anionic gold is needed for the adsorption and activation of O2, the progressive increase in oxygen's chemical potential results in the formation of predominantly cationic gold sites. These sites then take part in a favourable interaction with CO in the Au-O2 bond formation process. (60a of 84).
Au-CO interactions are stronger than for O2, regardless of the charge on Au (84). These reactions have also been extensively covered by several groups [352, 372, 375-378 of NC] and are illustrated in Fig. 2, (b). Although reactivity is experimentally shown to depend on cluster size, no alternating odd-even relationship has been observed.
Fig. 5. Literature derived compilation of experimental results (normalized) on the relative reactivity of gold particle anions in the adsorption reaction of one O2 or one CO molecule, respectively, as a function of the particle size n. (a) Reactions of Aunâˆ’ with O2: (filled squares) , (open triangles) , (open circles) . (b) Reactions of Auâˆ’ n with CO: (filled squares) , (open circles) .
Whereas quantum size effects are shown to have a significant impact on adsorption energies in clusters of less than 10 atoms (â‰¤1nm size), they level out as the cluster size increases and remain considerably less than coordination number effects (in terms of binding energies) in the 2-5nm region, where most catalytic activity is observed for supported systems. (140)
2.5 Role of the support
Essentially, a support must present a high surface area over which gold nanoparticles are well dispersed. Ideally, it should also serve to stabilise said particles against sintering and degradation during the reaction (68, 6, 7 of 131). An added bonus would be if the support somehow played an active part in the promotion of the catalytic process, and indeed this has been observed for selected supports and is discussed in chapter xxx.
2.5.1 Nature of the support
For the purposes of this review, only metal oxide supports are covered. Whereas other forms of support have been explored and papers published for low-temperature gold-catalysed CO oxidation, the volume of said literature is relatively sparse and mechanistic studies even more so.
Metal oxide supports for Au cat CO oxid have been extensively studied(58,84,85,3,40) and are usually classified according to their reducibility(21,22of124).; Al2O3, MgO and SiO2 are considered 'inert' (irreducible), whereas transition metal oxides such as TiO2, Fe2O3, and CeO2 are reducible or 'active'. This classification is derived from the ability of the support to provide oxygen to the catalytic system and thereby enhance catalytic activity(124), although ultimately, tight control of supported Au preparation and pre-treatment techniques have been recorded to render Au nanoparticles active even on traditionally 'inert' supports(23,24of124,5of03).
At this point, it should be noted that until recently, it was thought that an oxide support was an absolute prerequisite for gold nanoparticles to exhibit any appreciable catalytic activity(B4pg161), but evidence has emerged which indicates that non-supported gold, when taking specific atomic configurations which expose adjacent under-coordinated Au atoms, can also be highly active(42,43 of 115,115).
The above findings infer that while certain supports may help to promote catalytic activity, they are not fundamental to nanogold's catalytic properties(xxx).
2.5.2 Au-support interactions
"The structure and stability of small gold particles is a function of the chemical and physical nature of the support on which they reside"(b4pg59) However, each Au cluster/oxide support combination presents its own distinctive 'mix' of properties, and this has made it very difficult to clearly distinguish between intrinsic Au particle effects and those derived from Au-support interactions(150ofb4). Despite this analytical hurdle, the following interactions are generally accepted as typically having the most impact on catalytic activity vis-à-vis CO oxidation.
- Various groups(2,11,3 others) have presented arguments in favour of the gold-support interface periphery being the principal active site of reaction between CO and oxygen. Indeed, Haruta and co-workers(2in45) have identified this 'contact structure' as the main contributor defining the performance of supported gold nanoparticles, followed by the nature of the support and particle size. In turn, all these factors are largely dependant on the preparation method, and pre-treatment of the supported systems prior to use(11of45). The best catalytic performance has been obtained for preparation methods yielding systems in which the particle perimeter has been maximised(2in45). For a given gold loading, this equates to small particles which retain a hemispherical shape on the support(b4194). Experimental work has shown that coordinatively unsaturated gold atoms are prevalent at the particle periphery, a situation which is favourable towards both O2 and CO adsorption(16,22-26of45,121of43).
- Au clusters tend to be stabilized above anion vacancies located on the support surface. Charge transfer to the Au clusters from these vacancies leads to the formation of catalytically active AuÎ´- species(116ofb4,143,241of11). The ease of charge transfer is a measure of the activity of the catalyst as a whole. Reducible oxide supports are characterised by a small band gap which promotes charge transfer. In contrast, irreducible oxides have a large band gap which has the opposite effect(84) With semi-conducting oxide supports, AuÎ´- particles have also been said to induce further vacancy formation in their vicinity due to the Schottky junction effect at the metal-support interface(181ofB4,28of135).
- Various groups(32,67,69of84) have shown that lattice mismatch at the Au-support interface gives rise to a tensile strain due the effort expended in Au-substrate atom alignment. This strain results in a simultaneous up-shift in Au d-band centres which is calculated to enhance O2 adsorption.
2.5.3 The effect of different preparative methods on catalytic performance
Table 1 below provides a basic yet informative outline of Au/oxide catalyst preparation techniques most relevant to use in low-temperature CO oxidation.
An aqueous solution of HAuCl4 and water-soluble metal salts, such as a nitrate, is poured into an aqueous alkaline solution (Na2CO3 and/or NH4OH) and agitated for a few minutes. The two hydroxides (or hydrated oxides) are precipitated simultaneously. The coprecipitates are washed, dried, and calcined in air to obtain the supported metallic Au particle.
High gold particle dispersion
May contain a significant concentration of sodium ion and chloride ion, if a metal chloride is used as precursor. Both can act as a catalyst poison
Only applicable to Î±-Fe2O3, Co3O4, NiO and ZnO
Some of the gold particles could be embedded in the bulk of the support
The conventional method usually involves suspending the support in a larger volume of gold precursor solution, from which the solvent is then removed. After drying, the precursor has to be calcined at temperatures as high as 1073K and reduced.
Applicable to most supports
Relatively poor dispersion (much larger gold clusters) with large particle size distribution
Yields spherical particles
The precursor to the active species (HAuCl4) is brought out of solution in the presence of a suspension of the support, usually by raising the pH in order to precipitate a hydroxide. Product is washed. Reduction or calcination is then applied to remove the chloride.
Applied to the widest range of different support materials
All of the active component remains on the surface of the support and none is buried within it
Na+ and Cl- ions are removed during the wash step, thereby eliminating major source of poisoning.
High dispersion and low size distribution
Yields hemispherical particles with their flat planes strongly attached to the oxide support
Requires a great deal of meticulous control of preparation conditions
Photochemical deposition (PD)
Metal cations with appropriate redox potentials are reduced by photoelectrons created by band gap illumination of semiconducting oxides (e.g. those of Zn, W, Ti)
Extent of gold deposition is almost quantitative.
Thermal treatment is unnecessary because gold is already reduced by the UV irradiation.
Efficiency is very sensitive to both bulk and surface structural
Yields spherical particles
Table 1: The four main types of Au/oxide catalyst preparation techniques most relevant to systems for low-temperature CO oxidation
Table 2 below presents the turnover frequencies (TOFs) and the reaction rates (per single surface Au atom per second) of CO oxidation over a Au/TiO2 catalyst prepared by DP, PD and IMP methods(30of123).
Table 2: CO oxidation over Au/TiO2 at 300K and prepared by different methods(30of123)
As one can see, catalysts prepared by DP methods give the smallest particle sizes, and result in the lowest activation energies and highest reaction rates for CO oxidation. Of special interest, however, are the TOF values, which differ by as much as four orders of magnitude. This remarkable difference points to the contact structure as being the most critical factor in supported Au/oxide catalysts(123)
3 Reaction mechanisms for low-temperature carbon monoxide oxidation
Currently, published mechanisms addressing low-temperature gold catalysed CO oxidation can be split into two broad categories in which:
- the reaction proceeds only on the metallic gold particle,
- the support is assigned an active role.
It is generally accepted that, in all likelihood, more than one mechanism may come into play, depending on the type of support(226, 413, 487 of 11), particle size(xx), and dispersion(xxx). Moreover, different processes may predominate with changes in temperature(6.2.5 b4) and moisture levels(22.214.171.124 of b4), etc.
3.1 Mechanisms on gold particle only
Much of the work put forward to explain the 'metal-only' reaction pathway is based on a variety of analytical techniques applied to size-selected anionic gaseous gold clusters in the 2 to 20 atom scale(1,52,124,125 of b4).
Hagen et al.(34of ncp106) performed temperature dependent rf-ion trap experiments with anionic gold trimers, as tabulated below:
Trimer exposed toâ€¦
100 to 300K range
CO and O2 (simultaneously)
Absorption of up to two CO molecules
CO and O2 (simultaneously)
Only one CO molecule adsorbed onto the trimer
cooling down to
Up to two additional O2 molecules were adsorbed. (Au3(CO)O2- and Au3(CO)(O2)2- co-adsorption products were detected).
Table 1: Tabulated summary of observations recorded for rf-ion trap experiments performed on gaseous gold trimers by Hagen and co-workers (34, 185 of ncp107)
The phenomenon presented in the last entry of Table 1 has been attributed to a 'conditioning' of the gold cluster by the pre-adsorption of CO, thereby permitting subsequent O2 co-adsorption(NC 107), i.e. a cooperative process whereby Ïƒ donation from CO to the gold timer facilitates oxygen adsorption as O2- at a neighbouring site(1,8 of NC 107). Similar results were obtained using gold anion dimers, but in this case on exposure to the reactant mixture, the dimer was found to react about 10 times faster with O2 than with CO resulting in the initial formation of Au2O2- only. On cooling, a major peak corresponding to Au2(CO)O2- was also detected(34of nc107).
Experimental evidence of cooperative co-adsorption behaviour on gold clusters has been recorded by various groups, and has also been confirmed for clusters of up to 10 atoms thus far(15-20 of 120).
The above mentioned experimental studies are in good agreement with DFT calculations for CO oxidation on unsupported gold nanoparticles(126-129 b4). Indeed, data from both experiment and theory have been used to formulate viable reaction mechanisms(33, of NC, 87 of zz), but any such proposals cannot truly hold water unless they stand up to rigorous kinetic analysis too(BT) and many published papers fail to give adequate detail.
Socaciu et al.(nc110) have carried out ion-trap reactor experiments on Au2- in order to study the relevant reaction kinetics. In the presence of oxygen only and at T=300K, Au2- was found to react rapidly and completely to form Au2O2-. With a gradual increase in p(CO) such that p(O2) = p(CO), experimental data yielded a kinetic reaction profile (Fig Xa) which gave a 'best-fit' when expressed via the equilibrium reaction (i):
Au2- + O2 -> <- Au2O2- (i)
That is, an increase in the concentration of CO was found to favour the backwards reaction (Au2- formation), and this could only be explained by the presence of intermediate reaction steps involving CO in the reaction mechanism.
Fig 6 = fig pg 110of NC
The full mechanism was elucidated by applying variations in reaction temperatures and reactant partial pressures to the system; the results of some of these experiments are depicted in Fig Xb and Xc. By fitting this data to the simplest possible reaction mechanism, and by applying complex DFT based calculations, the following pathways were resolved:
Au2- + O2 -> <- Au2O2- (i)
Au2O2- + CO -> Au2CO3- (ii) Scheme 1
Au2CO3- + CO -> Au2- + 2CO2 (iii)
Present as cycles instead of schemes
Au2- + O2 -> <- Au2O2- (i)
Au2O2- + CO -> <- Au2(CO)O2- (iv)
Au2(CO)O2- + CO -> Au2CO2- + CO2 (v) Scheme 2
Au2CO2- -> Au2- + CO2 (vi)
Theoretical simulations performed by xxxxxx(33 of NC) were employed to characterize/confirm the structures of co-adsorption complex intermediates formed in steps (ii) and (iv). The results led to the exclusion of a Langmuir-Hinshelwood (LH)-type mechanism in favour of a Eley-Rideal (ER)-type process. That is, in step (ii), gas phase CO inserts via an ER-type process into the O-O bond in Au2O2- to form a carbonate-like anion. A second gaseous CO then binds to a free CO3- oxygen in (iii) to give two molecules of CO2 and this completes this particular cycle (Scheme 1). Alternatively to (ii) and (iii), CO could insert into the Au-O bond in Au2O2- (iv) followed by an ER, terminal addition of another CO to the resultant peroxy complex, again leading to the formation of 2CO2 but in a stepwise fashion as presented in steps (v) and (vi) (Scheme 2)(ZZ) The first pathway has been has been found to be experimentally predominant in conditions of very high p(CO) and low temperatures(NC) but energetically, there is very little to differentiate between the proposed mechanisms (Fig.zz), i.e. both exhibit very low barriers to intermediate formation and are thus equally viable(33 of NC).
The essential mechanistic and energetic aspects of the work presented by Socaciu et al.(nc110) have been seconded by Liu and co-workers(23 of 120), who also calculated the existence of a meta-stable four-centre intermediate (CO-OO) via an ER pathway. In contrast, adsorption energetics work carried out by Davran-Candan et al.(120) on anionic gold planar hexamers arrived at the same intermediate in the first of a two-step LH mechanism. In this step, CO and molecular O2 are co-adsorbed in a Î·1 end-on fashion at the lateral and apical sites, respectively, or vice-versa. The lateral species then migrates towards and couples directly with the molecule adsorbed at the apex via the CO-OO intermediate leading to the formation a O-Au6-complex and the first CO2 molecule.
Whether or not this mechanistic difference may be directly related to the quantum size effects for Au6- in particular (as expressed in Fig 5.) is unclear, but its unique structural and electronic properties (symmetry, stability) as observed by Zhai and co-workers(18of120) are indicative.
Along similar lines, Lopez and Norskov(87 ofZZ) applied DFT based calculations to a neutral, two-layer Au(7,3) cluster, comparing results obtained from two plausible LH-type reaction paths, one involving the facile dissociative adsorption of O2 (Scheme 1) and the other where adsorbed, molecular oxygen reacts directly with adsorbed CO (Scheme 2). These schemes are summarised below (Schemes 3 and 4):
In contrast to O2 adsorption on the previously discussed Au6-, fig zzz shows that on Au10, O2 is initially adsorbed in a Î¼1,1 fashion at the base of the cluster's (100) face.
Figure zzz. Profiles for the formation of the first CO2 molecule on Au10 particles. All energies are given with respect to CO and O2 in the gas phase. Black color, direct path; blue, indirect path. Thicker lines represent stable states, while thinner lines correspond to transition states. Yellow spheres represent Au atoms, red spheres represent O atoms, and gray spheres represent C atoms (LOPNOR)
The plausibility of the subsequent dissociative pathway (marked in blue) has been attributed not just to the inherent properties of low-coordinated surface Au atoms, but also to their spatial arrangement. To be more specific, three of these atoms are positioned in such a way as to permit simultaneous coordination to the transition state (TS) oxygen atoms (Structure 4 in Figzzz)(LOPNOR). This stabilization of the TS radically reduces the barrier to O2 dissociation, after which one of the nascent O atoms can easily react to form the first CO2.
As for the pathway involving molecular oxygen, this involves just a single TS arising from the simultaneous O2 cleavage and O...CO bond formation. Of the two schemes presented, related studies performed by various groups(88,68ofZZ) have indicated that the second is most likely initial oxidation route.
3.2 Mechanisms involving the oxide support
As for unsupported systems, most published mechanisms for oxide supported systems are mainly concerned with oxygen adsorption and its activation, a process which, in this case, is taken to involve the participation of the support. As for the adsorption of CO, this is generally accepted as occurring exclusively on the metallic gold particle(b4pg193).
Oxygen activation via the support occurs mainly as a function of the number of anion vacancies (F centres) on the support surface(143). These vacancies, coupled with the reducibility of the oxide support itself, will determine which mechanistic pathways are viable for CO oxidation(135).
3.2.1 Mechanisms involving non-reducible ('inert') oxide supports
MgO(100) is a typical example of a non-reducible oxide support surface which has found frequent use in various Au-support studies(x,x,x,x). Arenz et al.(45) investigated CO oxidation on monodispersed Au nanoparticles (<1nm) on defect-rich (â‰ˆ5%) and defect-poor MgO(100) films. By combining experimental results with first-principle quantum mechanical calculations, they were able to identify two competing LH-type mechanisms.
With Au8 clusters deposited on the defect-poor support, adsorption and subsequent reaction between CO and a superoxo (O2-) species (of unspecified origin) were deduced as occurring only below 250K and only on the top facet (second layer) of the Au8 cluster(45) (Fig xyz, LHt mechanism).
Fig xyz, LHt mechanism
Similar work performed on (defect-free, MgO (100) supported) Au34 rods by Molina and Hammer(121of43) also determined that the low-coordinated second layer corner sites are most favourable for CO adsorption, but they argued that the theoretical binding energy values for O2 dissociation at these sites were not sufficiently low enough to be reconciled with the reaction barriers to CO oxidation obtained through experiment. The closest match could only be obtained via an ER process whereby pre-adsorbed CO captures a gas-phase 02 molecule to form a O2â€¦CO dimer intermediate.
Above 250K, Arenz et al.(45) found that CO on Au8/defect-poor MgO desorbs before any reaction can take place, but on a defect-rich MgO support, a stronger CO binding energy keeps it in position. This is a direct result of charge transfer from oxygen vacancies in the support to the overlying Au8 cluster, increasing the binding energies and thereby activating the reactants (quantum size effect)(14,15of45,143). Theoretical calculations have shown that only one oxygen species will adsorb per cluster, and on defect-rich MgO, this occurs at the periphery of Au8 cluster-support interface (first Au layer) (Fig xyz, LHp mechanism).
Fig xyz, LHp mechanism
In this system, the reacting peroxy (O22-) species is inferred as deriving mainly from oxygen molecules adsorbing directly onto the MgO anion vacancies and then migrating to the Au8 cluster periphery (reverse spillover)(45,148ofb4), although an ER process involving adsorbed O2 and gas-phase CO was also calculated as being viable, especially at very low temperatures (c. 90K) Fig xyz, ER mechanism)(45).
3.2.2 Mechanisms involving reducible oxide supports
TiO2 rutile(110) is the most extensively studied reducible oxide support(11). DFT studies performed by Liu et al.(138of143) with a Au/stoichiometric TiO2 support have shown that oxygen (as O22-) adsorbs simultaneously with a surface Ti cation and peripheral Au cluster atoms. Apart from the charge transfer to the adsorbed oxygen via these Au atoms, the local electric field of the Ti cation is suggested as lowering the energy of the O2 Ï€* antibonding orbitals with respect to the Au Fermi level, thus permitting O2 dissociation at these interfacial sites and subsequent (straightforward) reaction with adsorbed CO.
As regards defect-rich Au/TiO2 systems, Remediakis et al.(140of143) assert that the above situation is still applicable, but there is an increase in Au/support binding energies which, according to Pacchioni(238OF11), is brought about by the redistribution of charge over the Ti ions surrounding the anion vacancies. Such charge distribution was not observed on defect-rich MgO surfaces.(239of11) This rationale is given to explain the overall reduction in the reaction barriers at peripheral and other low-coordinated Au atom sites of the cluster Fig 25 of 143.
Fig 25 of 143
This lowering of barriers, in conjunction with activated oxygen (O2-) being readily provided by vacancies surrounding the Au particles, effectively remove dissociation as the rate limiting step, resulting in the high CO oxidation rates as recorded in experiment(b4 33).
It should be noted that experiments with isotopically labelled O2 have shown that lattice oxygen does not participate in any of the above reactions(B4). Such involvement is prescribed for the more reducible oxide supports such as Fe2O3 and CeO2(33ofb4).
Hutchings and co-workers(14,15ofb433) employed a variety of spectroscopic techniques to study low-temperature CO oxidation with Au/Fe2O3. Gold was revealed to be present mainly as Au3+ in the form of interfacial AuOOH.xH2O, and to a much lesser extent, Au0, with the support being hydrated (i.e. Fe5HO8.xH2O). They invoked a mechanism originally put forward by Bond and Thompson(refBTorig)(Fig ABZpg194ofb4). CO was taken to adsorb at a low-coordinated Au0 atom and subsequently attacked by a hydroxyl group bonded to an interfacial Aux+ ion, in order to form a carboxylate group. This group is in turn attacked by O2- adsorbed at an adjacent anion vacancy forming the CO2 plus HO2-. The cycle is completed by HO2- attacking yet another carboxylate to produce a second CO2 molecule, regenerating the surface hydroxyl in the process, as given below.
(Scheme from 11 pg113)
By applying sophisticated reactor techniques, Daniells et al. delved deeper into the Au/Fe2O3 system and proposed the mechanism presented overleaf (FigB4Pg196). Of particular interest is the participation of lattice oxide ions adjacent to the Au cluster, and the involvement of surface hydroxyls and H20, resulting in a bicarbonate intermediate which decomposes to give the product. The abstracted oxygen leaves vacancies which are then occupied by externally introduced O2-, leading to the formation of another carboxylate and so forth.
As one can discern from the above schemes, the requirement for O2 dissociation (and associated barriers) is no longer an issue since the O-OH is cleaved instead(11)
Of particular note are the results of DFT studies carried out by Camellone and Fabris(113), which have yielded an innovative CO oxidation mechanism which centres upon the dispersion of Au3+ ions within a CeO2 lattice as substitutional point defects. The proposed cycle (as depicted in Fig113-8) involves three main steps, these being:
(A) the production of the first CO2 molecule by interaction of gaseous CO with an apical O atom coordinated to the lattice Aud+ of the stoichiometric AuxCe1-xO2 surface.
(B) the adsorption of molecular O2 on the resultant anion vacancy induces a 'local rearrangement of the AuO4 unit and the formation of O adspecies'.
(C) a gaseous CO molecule reacts with the O adspecies, recovering the AuxCe1-xO2 catalyst.