The Role of Short-lived Oxygen Transients and Precursor States in the Mechanisms of Surface Reactions; a Different View of Surface Catalysis M. W. Roberts Department of Chemistry, University of Wales, Cardiff, UK CF I 3TB 1 Introduction The usual strategy adopted by the chemist to unravel the mechanism of a reaction is to identify the composition and structure of the prod- ucts and to ascertain the temperature and concentration depen- dences of the reaction rate. Two-dimensional or surface chemistry is, however, an area that did not yield easily to this approach for the very obvious reason that experimental methods which were suffi- ciently surface sensitive were not readily available until the early 1970s.Prior to the development of surface sensitive spectroscopies, and in particular Auger and photoelectron spectroscopy, progress in understanding surface phenomena had relied very much on indirect methods with kinetics playing a major role and a checker board model of the surface.The only direct method available was that of infrared spectroscopy confined largely to studies of carbon monox- ide adsorbed on metals dispersed on high surface area supports such as silica and alumina. Interest in metal oxidation had led us to explore in 1963 how studies of the energy distribution of photoelectrons might provide definitive information on the transition of chemisorbed oxygen to an oxide overlayer with a discrete band structure.’ We established that the photoelectron escape depth for the nickel-oxygen system was ca.10 A but we were limited to a maximum photon energy of 6.2 eV. Evidence was also beginning to emerge2 that ESCA, as it was known then, was surface sensitive in that core-level spectra for adsorbed iodostearic acid and chemisorbed oxygen at carbon sur- faces were reported. The VG multiphoton, multi chamber spec- trometer became available to us3 in late 1970 and we set out to explore what we could learn about chemisorption and reaction mechanisms at metal surfaces. Having established the surface sen- sitivity of XPS and UPS for studying adsorption one of the early successes was the evidence for the interplay between the molecular and dissociative states of carbon monoxide and the facile nature of ~~~~~~ Professor Roberts was born in Carmarthenshire in 1931 and edu- cated at the Amman Valley Grammar School.He graduatedfrom University College Swansea followed by postgraduate research under the supervision of Keeble Sykes. He was a Postdoctoral Fellow at Imperial College with F. C. Tompkins and then joined the ScientiJc Civil Service as Senior Scientijic Oficer at the National Chemical Luboratory Teddington. In 1959 he was appointed to a Lectureship in the Department of Chemistry at Queens’ University, Belfast, where he collaborated with Charles Kemball. He moved to the Chair of Physical Chemistry at Bradford University in I966 and in 1979 to his present post at University of Wales, Cardiff, He has published some 250 papers, is the Founder Chairman of the Surface Reactivity and Catalysis Group of the Royal Society of Chemistry, was awarded both the Tilden Medal and the Royal Society of Chemistry Award in Surface Chemistry and is a Vice-President of the Faraday Division.He is Head of the Chemistry Department at University of Wales, Cardiff, where he has also acted as Pro Vice-Chancellor bond cleavage when adsorbed at metals such as molybdenum and iron .45 The growth of surface science over the last two decades has pro- vided unprecedented insights into the atomic and electronic struc- tures of metal and oxide surfaces. They have, however, provided little insight as to whether or not transitory complexes might par- ticipate in surface reactions -the emphasis has been on static rather than dynamic studies, with EXAFS playing a dominant role.Chemical reactivity involving multicomponent reaction systems has, by comparison with structural characterization of the solid surface, been largely neglected. It was the very distinctive chemistry associated with the reactions of multicomponent gas mixtures at atomically clean metal surfaces that was first reported6 in 1986 that led us to explore whether this approach might provide experimental evidence for the transition states that controlled reaction pathways. Experimental evidence suggested that we could control the conversion of reactants to prod- ucts and therefore offer the opportunity to recognise the transition states involved in the surface reaction. It was an approach that hith- erto had not been adopted for studying chemisorption and catalysis at atomically clean metal surfaces.We also had the advantage of an extensive database, generated largely in our laboratory, of X-ray photoelectron spectra for different adsorption states, of the same element, For example C( 1s) spectra could distinguish between surface carbide, graphite, carbonate, the anionic form of adsorbed CO, and physically adsorbed CO, while N( 1s) spectra could distin- guish between the hydrogenated states of chemisorbed nitrogen; N(a), NH(a), NH2(a) and NH,(a). The concentrations of each species present could also be determined from the intensity of the relevant core spectra with, if necessary, appropriate curve fitting, illustrating the unique advantage of XPS. Although our early photoelectron studies5 had of necessity given attention to the chemisorption of inherently simple adsorbate systems (e.g.CO, CO, and NH,) our interest in nickel led us to look in some detail at this and other analogous metal oxide systems. Two points emerged with nickel which emphasised the role of surface defects in chemical reactivity -one was evidence from O( 1s) and Ni(2p) spectra for the defect states 0-and Ni3+ and the relatively unreactive nature of the ‘perfect’ nickel oxide over- Ia~er.~,~,~On the other hand the presence of 0--like surface oxygen was associated with activity in H-abstraction reactions. Oxygen activation of adsorbates had been shown to provide a mechanism by which chemisorption replacement reactions could occur at low tem- perature~~~+~-~and led to extensive studies of this phenomenon.The need to obtain atomically clean metal surfaces for fundamental studies of catalysis and the role of contaminants such as oxygen as surface poisons was an important theme in the 1960s. Electron spec- troscopy established, somewhat unexpectedly, that surface oxygen could play a dual role; it could confer on an unreactive atomically clean metal surface very specific chemical reactivity while the ther- modynamically stable perfect oxide overlayer was comparatively unreactive. This was intriguing and provided the driving force for much of our work over the last decade. Do surface oxygen states exist which have not been accommo- dated thermally, exhibit chemical reactivity which was distinct from that associated with oxygen in its final chemisorbed state and par- ticipate in reactions involving dioxygen at metal surfaces was what we set out to investigate?6 Furthermore, could an estimate be made of the surface lifetimes of the oxygen transients under the reaction conditions?In general the gas pressures used were in the range to Torr. 437 43 8 2 Atomic Oxygen Transients O~-(S) The first question to be addressed was whether the ‘oxygen’ gener- ated in the act of bond cleavage and dissociative chemisorption of dioxygen had associated with it specific chemical reactivity which distinguished it from the thermally accommodated final chemisorbed state We designated the two states as Os-(s) and 02-(a) respectively and suggested that the dynamics of the disso- ciative chemisorption of dioxygen chemisorption involved the steps shown in eqn (1) O,(g) -02(s)-0; (s) -08 (s) -O2 (a) (1) The proposition we set out to examine was whether short-lived species O$-(s) and Os-(s) had sufficiently long surface lifetimes to determine reaction pathways involving dioxygen at atomically clean metal surfaces The Mg(0001) -dioxygen system was chosen6 lo as a model system and ammonia as the probe molecule having first established that under the experimental conditions (room temperature or below) both the oxide overlayer and the atomically clean Mg(0001) surface were unreactive to NH,(g) Fig 1 shows the N( 1s) and O( 1s) spectra observed when molecularly adsorbed ammonia characterised by an 395 400 405 530 535 Binding energy / eV Figure 1 O(1s) and N(1s) spectra after physically adsorbed ammonia present at Mg(OO0 1) and characterised by an N( 1s) binding energy of 402 eV is exposed to dioxygen at 170 K Note the formation of a second N( 1s) peak at 399 eV and two O( 1s) peaks at 53 1 and 533eV These are assigned to NH,(a), chemisorbed oxygen 0’ (a) and hydroxyl OH(a), this was the first evidence for the possible role of oxygen transients in H abstraction react1 ons N( 1sj binding energy of 402 eV was exposed to dioxygen at 170 K, a new N( 1s) feature developed at 399 eV assigned to NH2(a) and two O( Is) features are present with binding energies of 530 5 and 533 eV assigned to chemisorbed oxygen, O2 (a), and hydroxyl species respectively The binding energies are accurate to ?O 15eV Although both the oxide overlayer and the clean metal are unreac- tive to NH, surface amide and hydroxyl species have been gener ated and the following mechanism involving the atomic oxygen transient Os-(s) suggested Similar chemistry is observed when an NH,-0, mixture (20 1j is exposed to the Mg(0001) surface at 295 K(Fig 2) Formation of oxygen transient R,Ofi (s) -O2 (a) ‘Oxide’ formation, the oxide route NH3 undergoing surface hopping R4NH,(s) + O8 (s) -+ NH2(a) + OH6 (a) H abstraction reaction, the amide route (2) The notation (a) refers to an adsorbed species in its final chemisorbed state while (s) refers to the surface transient Os and in the case of NH, a weakly adsorbed molecule undergoing rapid CHEMICAL SOCIETY REVIEWS, 1996 ~.‘..r-.-.’--395 400 405 525 530 535 540 Bindingenergy / eV Figure 2 (a)O(1s) and N( 1s) spectra for the formation of an oxide over layer at a Mg(0001) surface at 295 K followed by physical adsorption of ammonia at 110 K and warming the adlayer to 295 K The ammonia desorbs and the oxide overlayer is unreactive to NH, (b)O( 1s) and N( 1s) spectra when an ammonia-dioxygen mixture (20 1) is exposed to a Mg(0001) surface at 295 K Note the formation of amide (NH,) hydroxyl (OH) and chemisorbed oxygen (02) surface diffusion -the probe molecule The rate of the reaction R, is therefore given by eqn (3) R, = (No of visits of NH, to surface sites) X (fraction of sites occupied by O6 (3) assuming that the collision mechanism (R,) occurs with unit effi ciency Central to the model is the concept of an ammonia molecule with a characteristic surface lifetime T~~~~~~that is determined by its heat of adsorption AH and the surface site lifetime T,,,,determined by the kinetics of ammonia surface diffusion z e by expressions of the form eqn (4) T~~~~~~a 10 exp (AHIRT) If for example we assume EdlRca 0 (an unrealistically low value) then the ammonia molecule will visit 10” sites s Since for a AH value of 40 kJ mol I -the heat of adsorption of ammonia at a Mg(0001) surface -the value of T~~~~~~at 295 K is ca 10 s then each molecule will visit lo7 surface sites before desorbing The effective surface concentration 0 is related to the pressure or mole- ~cular impact rate Nand T~~~~~~by the expression cr = NT which ~~ in the present case gives a value for aNH, of ca 1O1O cm * which is an effective ammonia surface coverage of 10 at 295 K However, under these conditions ca 1017 surface sites will be visited, I e each site many times during the surface SOJOU~~time T~~~~~~ We have made a number of assumptions in these calcula tions in order to illustrate the principle of the probe molecule approach to search for oxygen transients for which the essential pre requisite is that R, >> R, [see eqn (2)) If the oxide route is favoured then the amide route is blocked That it was the atomic oxygen transient -rather than the molec ular species -that was active in the oxygenation reaction was OXYGEN TRANSIENTS AND PRECURSOR STATES IN SURFACE established further by coadsorbing ammonia with nitrous oxide when the dissociative chemisorption of N,O generates atomic oxygen unequivocally Similar conclusions were drawn when nitric oxide was used as coadsorbate Ion Estimates of 706 (s) under the experimental conditions used and using a steady-state model gave values of cu 10 s The assumption of the steady state model found support from a computer modelling of the reactions by solving the relevant differential equations The model predicted that the NH,(s) surface concentration was invariant throughout the reaction at a value of ca 6 X lo8molecules cm at 295 K for an assumed activation energy of diffusion Ed,, of 14 kJ mol I Secondly the O6 (s) concentration is calculated to be ca 10’ cm and decreases by about a factor of two as the coverage increases The value of da(s) which gave the best fit to the experimental data, and in par-ticular the ratio of NH,(a) to O2 (a) formed, was CCI 10 s It should be emphasised that the value of 706 (s) is with reference to the experimental conditions used in the coadsorption experiments -it is not an intrinsic characteristic of O6 species The value will vary with the metal, surface structure, temperature and pressure of reac- tants An important conclusion from these experimental data is that species may have negligible surface coverages but through surface diffusive hopping they can participate in highly efficient reaction pathways to products In the present systems both reactants 06 (s) and NH Js) are effectively transients This approach was extended to other reactions to explore the gen- erality of the concept that ‘hot’ oxygen transients were the key par ticipants in oxygenation surface chemistry The term ‘hot’ refers to oxygen atoms which are generated in the act of dioxygen bond cleavage leading to either vibrational excitation or atoms with excess translational energy and which have not been thermally accommodated at the metal surface We chose the aluminium- dioxygen system’* using carbon monoxide as the probe molecule since it was neither ‘adsorbed’ at the surface of atomically clean aluminium nor at the oxide overlayer at low temperatures However.when coadsorbed with dioxygen in a CO-rich mixture, carbonate and carbidic species were formed at 80 K (Fig 3) These species could be distinguished easily by the chemically shifted C( 1s) spectra with C( 1s) energies of 282 (C6) and 290 eV (CO,) and the following mechanism was proposed, similar chemistry was observed with magnesium [eqn (5)l 04s) -0: (s) First stage of oxygen chemisorption 0,’(s) -208 (s) Generation of hot ’0’atoms 1-.... 280 290 360 Binding energy i eV Figure 3 C( Is) spectra of an aluminium surface after exposure to a carbon monoxide-dioxygen mixture (1 9& CO) at 80 K Oxidation of CO by the transient OF (s) occurs to give a monolayer of carbonate species which are partially reduced by the aluminium to give carbidic carbon and oxide REACTIONS-M W ROBERTS 439 06 (s) + CO(s) -co; (s) Formation of reactive anionic CO; CO: (s) + Os (s)-CO,(a) Surface carbonate formation CO,(a) -C6 (a) + ‘oxide’ Reduction of carbonate at aluminium surface (5) We had clearly moved away from a checker board model for surface reactions with neither the Eley-Rideal nor the Langmuir-Hinshel wood mechanism being appropriate models for these reactions How we viewed the transition state in dioxygen dis- sociation was not clear A clue however, was that in both cases, alu- minium and magnesium, oxygen chemisorption was a highly exothermic reaction and what was envisaged was that at least one of the oxygen atoms underwent rapid translational motion during bond cleavage It was the latter that was the chemically reactive species 3 Dioxygen Transients: Oi-(s) Evidence for the participation of O6 (s) as a reactive transient in the surface chemistry of dioxygen at metal surfaces raised the possibil- ity that a molecular oxygen 0; transient could play a role in deter-mining reaction pathways We chose to investigate the Zn(0001)-dioxgyen system’ since the dissociative chemisorption of oxygen was unusually slow with a sticking probability of cu 10 3, suggesting that bond cleavage might be rate-determining Was this to be associated with a precursor transient state 0; (s)7 There was at the time at least one example we were aware of where in metal oxidation the process does not proceed at low temperatures beyond the chemisorbed molecular statei4 -the 0: (a) state at Ag(l11)The reactivity of ammonia rich dioxygen-ammonia mixtures at Zn(0001) surfaces showed13 analogous chemistry to that observed with Mg(0001) but distinctly different kinetic behaviour Fig 4 shows the temperature dependence of the formation of surface t 20 40 02 exposure/ L Figure 4 Variation of surface oxygen (chemisorbed oxygen and hydroxy species) at four temperatures, 240, 200, 160 and 120 K as a function of oxygen exposure when a Zn(0001) surface is exposed to an ammonia-dioxygen (2 1) mixture (L = Langmuir) Also shown is the data for pure oxygen Note the evidence for precursor mediated kinetics and increased efficiency for dioxygen bond cleavage in the presence of ammonia ‘oxygen’ species (hydroxyl + chemisorbed oxygen O2 -species) and XP and EEL spectra providing evidence for amide and hydroxyl species (Fig 5) Also shown for comparison is the chemisorbed oxygen concentration as a function of exposure to pure oxygen at 200 K There are two points to note (a)the inverse temperature dependence of the reaction rate and (b)the substantial increase (by a factor of ca 10,) in surface oxidation rate observed with the ammonia rich dioxygen-ammonia mixture compared with pure oxygen Both these point to the participation of an ammonia-dioxygen complex -or transition state -and the reaction scheme [(eqn (6)l was suggested This an example of how the rate of dioxygen bond cleavage is faster for the dioxygen-ammonia 0 1000 /'r 2000 3000 1 4000 Energy loss / cm-' 02-(8) 399 402 530 532 Binding energy / eV Figure 5 N( 1 s) and O( 1s) spectra for the chemisorbed adlayer at Zn(0001) surface after exposure to the ammonia-dioxygen mixture.Also shown is the corresponding electron energy loss spectra. Note the formation of NH,, OH and chemisorbed oxygen. complex than it is for dioxygen alone. We assume that the enhanced rate be associated with a lowering of the activation energy but Dioxygen accommodation O,(S)-O$-(s) Transient formation O,(g) -0;-(s) -20z-(a) Inefficient oxide route NH, undergoing surface diffusion NH,(s) + O$-(S)--* (NH,.*.O,'-)(s) Complex formation (NH,--O$-)(s) +OH(a) + NH,(a) + 02-(a) Complex decomposition; efficient pathway to dioxygen bond cleavage; amide route (6) driven by a thermodynamically favoured reaction -the amide route.We will return to a theoretical analysis of this concept later. The kinetics of the dioxygen-ammonia system conformed to a precur- sor-mediated reaction in keeping with the involvement of a charge transfer complex of the kind (O$--.NH,)(s) suggested in the above reaction scheme. Of particular note is that the rate of the formation of amide and hydroxyl species -the amide reaction pathway in the scheme -was inversely dependent not only on the temperature but also on the NH,(g):O,(g) ratio. It is the three centre precursor complex Zna+ -(0g--.NH3) that is involved in the rate-determin- ing step and clearly its concentration is directly related to the NH,(g):O, ratio, being greater the more ammonia rich is the mixture.The rate of complex formation J, is given by eqn. (7),ls where CHEMICAL SOCIETY REVIEWS, 1996 0.80.91 0.7. N 'E 0.6-0 v,? 0.5-\ b 0.4-0.3. 0.2. 0.1 ' 02 0 10 20 30 40 50 60 70 80 Dioxygen exposure Figure 6 Modelling the kinetics of the dioxygen-ammonia reaction at a Zn(0001) surface; the experimental data (-) are compared with the theoretical model (---): further evidence for the role of a dioxygen-ammonia complex (O,fi--NH,). O,,,,, is the sum of OH and chemisorbed oxygen. the terms in the square bracket are related to the surface diffusion of NH,(s) and O$-(s) involving a hopping mechanism, Oare surface coverages, vand E are frequencies and activation energies of diffu- sion respectively and AH is the activation enthalpy of complex for- mation.The experimental coadsorption kinetics were successfully modelledI5 at both 200 K and 120 K assuming an activation energy of 40kJ mol- I and a frequency factor of 10I3s-I for the decompo- sition of the precursor complex. The surface coverage of the pre- cursor complex is estimated to be ca. however, we have no firm grounds for assuming that the frequency factor for complex decomposition is loi3s-I and the calculations are illustrative only. Nevertheless the theoretical model shows how even though the equilibrium O,(g) + O,"(s) allows for only a low coverage of O:-(s), the rate of complex formation is proportional to the product of O$-(s) -a small number, and v,,, the vibrational frequency of surface diffusion of NH,(s) which is a large number, s-I.The stability of the dioxygen-ammonia complex is suggested to ariseI3J5 from electrostatic interactions between 0;-and NH, within the Zn*+.-(O:-.-NH,) complex, thus providing a route to the more energetically favourable amide pathway. Further evidence for the participation of the dioxygen O:-(s) transient in the chemistry of dioxygen at Zn(0001) surfaces was established by coadsorption with pyridine.16 Like ammonia it enhanced the rate of dioxygen bond cleavage by a factor of cu.lo2 which was in keeping with the chemistry of pyridine where dioxy- gen complex formation is well known. 4 Ammonia Oxidation at Copper Surfaces: which is the Active Species Og-(s) or 06-(s)? The activation of ammonia by preadsorbed oxygen at a Cu(ll1) surface was one of the first examples of a class of reaction~~b",~ we have referred to as 'chemisorptive replacement reactions' the chemisorbed oxygen being removed as water with the simultaneous formation of surface imide species NH(a); the reaction rate was sen- sitive to oxygen coverage and became unmeasurably slow as the preadsorbed oxygen coverage approach unity [eqn. @)I. Guided by a scanning tunnelling microscope study of the structure of the Cu( 110)-0 overlayer we carried out a Monte Carlo simula- tion of the development of the overlayer as a function of ~0verage.I~ We then compared the experimentally observed reactivity of Cu(110)-0 surfaces for imide formation as a function of oxygen coverage with the presence of structurally distinguishable oxygen states: isolated oxygen adatoms; oxygens at the end of Cu-0-Cu-0 OXYGEN TRANSIENTS AND PRECURSOR STATES IN SURFACE REACTIONS-M. W.ROBERTS expected to be relatively unreactive. We therefore associate oxygens situated at the end of Cu-O-Cu chains with 0-and anal- ogous to the oxygen atom transients Os-(s) discussed earlier. Taking a lead from our studies8J0J3 of the activation of water by oxygen at Ni(2 10)surfaces at low temperatures, the coadsorption of ammoniadioxygen mixtures at Mg(0001) and Zn(0001) surfaces -all of which provided evidence for Oa-(s) and Og-(s) transients -we studied the reactions of similar mixtures with Cu(ll1) and Cu( 110) su1faces.~~~8~~-~It was shown that for ammonia rich dioxy- gen-ammonia mixtures an efficient reaction occurred with Cu( 11 1) at 295 K leading to the formation of chemisorbed imide species (Fig.8). Even though the reaction is oxygen catalysed virtually a monolayer of NHt species is formed without any evidence for surface oxygen being present in the O( 1s) spectra. The N( 1s) binding energy was at 398 eV and the vibrational (HREEL) spectra confirmed this as the bent form of NH(a), characterised by an a,, loss feature at 1100, vNHat 3400 and vCU-NHat 700 cm-1 .Recently, Bradshaw et al.19 have shown by photoelectron diffraction that the NH species is present at the short-bridge sites of the Cu(ll0) surface. What then is the mechanism of this reaction and is it the atomic or dioxygen transient that is the active species? The kinetics did not show any characteristics associated with the participation of a precursor complex and in contrast to Zn(0001) there was no kinetic window available (see Fig. 4) in that the sticking probabil- ity of oxygen was already high for Cu( 110)at 295 K. At the Cu( 1 10) surface the oxygen dissociation rate was slower than the rate of NH, formation from an ammonia rich dioxygen-ammonia mixture. This points to a precursor dioxygen state O$-being involved in the reac- tion;lShif the reactive species was atomic oxygen Os-(s) then we would expect the rate of NH,.formation to be either less or equal to the rate of dioxygen bond cleavage. This appears to be the case for Cu( 111) (Fig. 8). Furthermore the rate of NH, formation is at least a factor of ten times higher with both Cu( 1 1 1) and Cu( 110) than the chemisorptive replacement reaction when the oxygen surface cov- erage was about 0.2. ~NH I I I 1 0 1000 2000 3000 4000 Energy loss / cm-I u [Os-,N,NH] (IO~~C~-~)5r *v1 500 1000 -02exposure / L Figure 8 Comparisons of the formation of chemisorbed oxygen when a Cu( 11 1) surface is exposed to dioxygen and the formation of NH, species when Cu(1 11) is exposed to an ammonia rich dioxygen-ammonia mixture at 295 K.Also shown is an electron energy loss spectrum of the chemisorbed layer confirming the presence of NH(a). a Occupied Site u101 B'm 0.2 0.4 0.6 0.8 Total Oxygen Surface Coverage (monolayers) Figure 7 (a) Simulated surface structure at a Cu( 110) surface after 300 Monte Carlo equilibrium steps for 0 oxygen = 0.3, E, = 2 and Eloo= 7 kJ mol-I. Note the similarity of the surface topography to the Cu-0-Cu-0 chains seen by STM. (b)The experimental data for the extent of the chemisorptive replacement of surface oxygen by NH, species (curve e) as a function of preadsorbed oxygen is compared with the surface oxygen present in four different environments (a), (b), (c) and (d). Good fit IS obtained with (d) which corresponds to oxygens at the end of Cu-0-Cu-0 chains. chains and oxygens within oxide islands.The experimental reactiv- ity data for NH formation could only be satisfactorily correlated with those oxygen adatoms that were present at the end of the Cu-0-Cu-0 chains [Fig. 7(a) and (b)j and we drew atten-tion5l7 18< to the significant role that the charge associated with the oxygen can play in determining its reactivity; 06-would best describe the former and 02-the latter. It should be recalled that although O(g) -k e 4 0-(g) is highly exothermic, the addition of a further electron is endothermic and 02-can only be formed at metal surfaces when there is a possible contribution from a Madelung term associated with oxide formation. Furthermore the O--like oxygen species would be anticipated to be highly reactive (isoelec- tronic with F) whereas 02-(isoelectronic with Ne) would be CHEMICAL SOCIETY REVIEWS, 1996 Table 1 Density function calculations for ammonia dissociation and oxidation (van Santen et af )20 Figure 9 NH species present at the short bridge site of a Cu( 110) surface after exposure to an ammonia rich ammonia-dioxygen mixture at 295 K We therefore have a hierarchy of reactivity for NHx formation coadsorption of ammonia rich NH,-0, mixtures is the most effi- cient, with the chemisorption replacement reaction the rate is always appreciably lower and tends to zero as the oxygen coverage approaches unity This led to the general proposition' I*( that the activity of oxygen in hydrogen abstraction reactions decreases as oxygen clusters (nuclei) develop and is close to zero for a 'perfect' oxide overlayer [see scheme eqn (9)l Dioxygen transient OJg) -0: (s) :O,(g) 0' (s) Rapidly diffusing oxygen adatoms -+ 0' (s) -0' (a) Isolated (reactive) oxygen adatom, end of Cu-0-Cu-0 chain O* (a) -0' (a) Growth of unreactive oxygen clusters oxidation and reconstruction (9) Although comparisons of the rate of dioxygen dissociation and imide formation suggested that the dioxygen transient participated in the dehydrogenation of ammonia at Cu( 110) it was also showni8c that isolated oxygen adatorns Os (a) at Cu( 110) were also very reactive to NH,(g) We therefore were uncertain as to whether during coadsorption of ammonia and dioxygen at Cu( 110) surfaces the reaction pathway was via the molecular or oxygen atom tran- sients Provided the oxide pathway Os (a) --* O2 (a) is not allowed to become a significant route imide formation takes place to com-plete coverage at 295 K At low temperature dehydrogenation pro- ceeds only as far as the amide speciesIs' (eqn (lo)] Oh (s) + NH,(s) -OH(a) + NH,(a) or 0; (s)NH,(s) -OH(a) + NH,(a)Ofi (s) NH,(a) -NH(a) + H(a) Reaction Activation Overall reaction energy energy /kJ rnol I /kJ rnol I NH,(g) -NH,(g) + H(g) +498 +498 NHT -NH,* + H* +344 + I76 NHT + 0"-NH,* + OH* + I32 +48 NHT + 0"-NH* + H,O(g) >200 +92 NHT + 0: -NH,* + OOH* +67 -84 NHT + 0; -NH* + O* + H,O(g) + 134 -184 As far as we were aware+ there were no similar experimental studies being pursued elsewhere and certainly no theoretical evidence for or against the role of molecular oxygen transients participating in ammonia oxidation reactions Phillip Davies, who had completed his PhD thesis at Cardiff, dealing with experimental aspects of ammonia oxidation, joined van Santen's theoretical group in Eindhoven during the summer of 1990 where there was an interest in the role of oxygen in hydrocarbon oxidation and cluster calcula- tions for ammonia adsorption on copper 20n The Eindhoven group using first principle density functional calculations extended this work to make a detailed and elegant analysis of the role of atomic and molecular oxygen precurors in the overall catalytic cycle of ammonia oxidation 2oh They used a Cu(8,3) cluster as a model of the Cu( 1 1 1) surface (Fig 10)and established that, although atomic oxygen can enhance N-H bond activation by lowering the activa tion energy for H-abstraction, it may also act as a surface poison inhibiting NH, dissociation This was in keeping with our experi- mental work Transient molecular oxygen was shown to adsorb weakly both parallel (17 kJ mol I) and perpendicular ( 10 kJ mol I) to the surface, parallel adsorption appeared to be a precursor for oxygen dissociation whereas the perpendicular form was involved in H abstraction The theoretical calculations favoured the forma- tion of OOH as an intermediate (with close to zero activation energy) in the H abstraction process rather than the simultaneous transfer of two hydrogens to form water directly Of the four reac- tion pathways analysed (Table 1) it was the nonactivated molecular oxygen transient pathway involving sequential H abstraction that was favoured -the next most energetically favoured pathway involved the 'hot' atomic oxygen transient These calculations pro- vided a theoretical basis for the oxygen transient concept in cat alytic oxidation and supported the general conclusions that had emerged from experimental coadsorption studies 5 Metastable Oxygen States at Metal Surfaces Although the presence of alkali metals in catalyst formulations is a well established approach to controlling selectivity in heteroge neous catalysis there have been comparatively few studies of the surface chemistry of alkali metals per se The investigation of the caesium-oxygen system2IN was stimulated by our earlier studies of the Cu( 1 10)-Cs-oxygen system2Ih where the oxygen was shown to be highly reactive to carbon monoxide to give Cog (a) species at 80 K and carbonate on warming to 295 K At caesium surfaces X-ray induced O( 1s) spectra have enabled three distinct oxygen states iSince this article was submitted, R J Madix has drawn attention to STM data (Surf Sci .in press) which has confirmed the model proposed leqn (9)and (lo)]for ammonia oxygenation reactions at Cu( 110) surfaces Figure 10 Abstraction of H from NH, by molecular oxygen at a Cu(8,3) cluster to form NH, and OOH In the presence of NH, oxygen prefers the perpen dicular adsorption geometry whereas at the clean cluster surface oxygen is in a configuration parallel to the surface OXYGEN TRANSIENTS AND PRECURSOR STATES IN SURFACE REACTIONS-M W ROBERTS to be assigned to isolated oxygen adatoms Os ,peroxo-type species 0; and oxide 0, The O6 state, only observed at low surface coverage, is the precursor to oxide formation in accord with ther- modynamic arguments ,lrJ A similar situation exists when oxygen is chemisorbed at Ni( 110) and Ni( lOOj surfaces2, and may indeed be a characteristic of very early stages of oxygen chemisorption and metal oxidation in general Since carbon monoxide is not adsorbed and unreactive at an atomically clean caesium surface we explored23 its catalytic oxidation and the relevance of the three oxygen states in the process We established that the Os (a) state readily gives the anionic Cog (a) species when exposed to CO(g) at 80 K, this could be oxi- dized further on exposure to dioxygen to give surface carbonate CO,(aj When CO 0, mixtures were exposed to caesium,23 the chemistry observed depended on the ratio of CO to 0, For CO-rich mixtures the reactive Os (s), via the Os (a) state, provides a facile route to Cog (a), however, for oxygen rich mixtures metal oxida- tion dominates leading to oxidation and carbonate formation It is the lifetime of the 06 (sj species, i e before it is transformed to the O2 (oxide) species that determines which reaction pathway is followed The distinction between the chemical reactivity of 0 -type and 02 type oxygen species has been one of the cornerstones of our studies of the chemistry of ‘oxide’ surfaces9 For example the ‘oxide’ overlayer at a Pb(ll0) surface is unreactive to water vapour, however the presence of Os species provides reactive sites for H-abstraction8 while in the case of Ni( 100) and Ni( 1 lo>, when the oxygen coverage is <02 the surface is highly reactive to ammonia22with the formation of NH,.species This reactivity is also observed in NH,-0, coadsorption When oxygen rich dioxy- gen-ammonia mixtures are exposed to Pt( 111) there is rapid devel- opment of N( Is) intensity at 397 eV binding energy indicative of N-adatoms -there is no intensity at any stage in the O( 1 s) spec-trum region 2J The LEED pattern indicates that the Pt(l Il)-N adlayer is well ordered, while the HREEL spectra show features at 500 cm I ( vptN) and 1185 cm (a,,,), the latter due to a small cov- erage (ca loi3cm ,) of NH,(a) -sufficient to give some asym- metry on the high binding energy side of the N(ls) peak Following our conclusions with other metal-oxygen systems we suggested24 that in the Pt( 11 l)-O system the oxygen adatoms were present either as isolated oxygens or as very small clusters Recently Ertl and his colleagues25 have reported an STM study of the Pt( 11 1)-0system and established that subsequent to dissocia- tion the oxygen atoms appear in pairs on the surface but separated by an average distance of two lattice constants This is entirely in keeping with the general conclusions suggested by the ammonia-dioxygen studies 24 6 Direct Experimental Evidence for ’Hot’ Oxygen Tr a n si e n ts: Scanning TunneI I in g Microscopy and Mass Spectrometry Experimental evidence for oxygen transients at metal surfaces relied initially on the unique chemistry observed, which could not be associated with the thermally accommodated oxygen in its final chemisorbed state and also kinetic behaviour which was indicative of the participation of precursor states There were also strong analogies with homogeneous gas phase radical chemistry (eg the NH, + 0 -NH, + OH reaction) and also the chemistry of 0 species present at bulk oxide surfaces, in that low energy reaction pathways were available only under conditions where oxygen tran- sients or metastable oxygen adatoms were present We designated the oxygen transients as Os (s), i e 0 like, and most likely to be associated with metal-oxygen systems where oxygen chemisorp- tion is highly exothermic and characterised by high oxygen sticking probabilities For metal-oxygen systems where dioxygen bond cleavage is slow, then the 026(s) transient is likely to determine the chemistry, in such systems oxidation is also less exothermic This is the case for the Zn(0001 j-dioxygen-ammonia system Earlier optical simulation studies26 of LEED patterns observed for oxygen chemisorption at Cu(2 10)surfaces had suggested that subsequent to dissociation a correlated or semi-correlated diffusion of oxygen adatoms by a hopping mechanism, occurring over large distances 1 73 A iL‘ 0 5-72L I I 1234 bb2 21 27 Number of atoms per island Figure I1 STM evidence for ‘hot’ oxygen transients at an Al( 1 11) surface, note also that for an exposure of 72L very few isolated oxygen adatoms are present At low oxygen exposures isolated oxygen adatoms predomi nate 27 (ca 10 nm), is necessary to generate the defective structures that would give rise to the observed LEED patterns However, we did not consider that this might have any influence on chemical reac- tivity In spite of these compelling observations more direct exper- imental evidence for oxygen transients being associated with oxygen chemisorption at metal surfaces was needed This became available first for the Al( 1 11j-dioxygen system through STM studies by Ertl’s group,*’ Fig I1 shows an STM image of an Al( 11 1j surface after exposure to dioxygen at 300 K Although the oxygen adatoms in the final chemisorbed state are completely immobile at this temperature, ordered patches of ‘oxygen islands’ are observed as a consequence of the generation of ‘hot’ oxygen transients The STM experiments also revealed that at low exposures (coverages) dissociative oxygen chemisorption led to the two oxygen atoms being separated from each other by at least 80 A before they became thermally accommodated with the surface through relaxation of the excess energy of several eV to phonon or electronic excitations of the substrate At low oxygen exposures (3 L+)isolated oxygen adatoms are mainly present but with increasing i.1 L (Langmuir) = 10 Torr s 444 exposure (72 L) the oxygens cluster together to give the island-type structures shown in Fig 1 1, with very few isolated oxygen adatoms Also shown is a histogram showing the size distribution of the oxygen islands for four oxygen exposures, 3 L, 13 L, 20 L and 72 L That ‘hot’ oxygen atoms are generated during dissociative chemisorption of dioxygen at aluminium surfaces was established first through the adsorption of carbon monoxide-dioxygen mixtures when even at 80 K surface carbonate is formedi2 (Fig 3) The STM observations add definitive evidence to support the mechanism involving rapidly diffusing oxygen atoms More recently Wintterlin et a1 25 have studied the dissociation of 0, at a Pt( 1 11) surface by STM The two oxygen atoms generated by the process of dissociation appear in pairs at 150 K with an average distance of two lattice constants This is an appreciably smaller distance than that observed at the Al( 1 11) surface (ca 80 A) which is compatible with the much smaller heat of oxygen chemisorption at platinum than at aluminium The authors conclude that the separation of the oxygen atoms results from transient bal- listic motion where the short range travelled is in agreement with molecular dynamics calculations The reactivity of oxygen adatoms at Pt( 11 1) to ammonia is high undergoing chemisorptive replace- ment to give NH species24 and cannot be distinguished from the similar chemistry and reactivity observed when ammonia-rich dioxygen mixtures are exposed to Pt( 111) at the same temperature This is in accord with the STM observations of Ertl -isolated oxygen adatoms being the reactive species in both cases The mode by which the oxygen atoms fly apart particularly in the AI(lll)-O system is uncertain Do they translate parallel to the surface or do they follow a parabolic, through-space type of trajec- tory? In some cases adsorbate atoms have indeed been detected in the gas phase, 0-ions from caesium surfaces28 (Fig 12) and F-ions from silicon suggesting that a ‘cannon ball’ trajec- tory involving surface hopping is the likely mechanism for such reactions [eqn (11) J F,(g) + Si +Si-F(a) + F (g) (1 1) The cleavage of the F, bond by the formation of a single F-Si bond is argued to be thermodynamically feasible on the grounds that the energy released upon adsorption at a single Si dangling bond, which does not require cleavage of a Si lattice bond, is 5-6 eV compared with 1 5 eV for the F, bond energy Some of the exothermicity is thus converted into translational energy of the scattered F atom How then do we view the transition state involved in the dissocia- tive chemisorption of dioxygen at say aluminium and magnesium d v a3-0;-v 0-Figure 12 Exposure of Cs surfacesto 0, causes ejection of 0-ions duringthe first stage of dissociative chemisorption -followed by exoelectron emission at higher exposures28 CHEMICAL SOCIETY REVIEWS, 1996 surfaces and what is the relationship between the heat of chemisorp-tion and the strengths of the metal-oxygen bonds formed? 7 Excited States generated by 0-atoms A question relevant to ‘hot’ oxygen surface chemistry is the tem- perature of the vibrational modes of the reaction products -partic-ularly if the latter might open up reaction channels not available to the ‘cold’ molecules Haller30 has recently addressed this in studies of the oxidation of carbon monoxide by ‘nascent oxygen’ -the latter being generated by a discharge It was shown that the appar- ent temperature of CO, generated by nascent oxygen atoms in the presence of a palladium catalyst is much higher than that from ther- mally accommodated 0-adatoms, i e it is rovibrationally excited A two-dimensional gas model, analogous to that put forward for oxygen transient reactions at Mg(0001) and Cu( 1 10)Io l8 sur-faces, is suggested to account for the higher temperature It is sig- nificant that Coulston and Haller3I had estimated that the CO coverage when the oxidation rate IS maximum is only of the order of 10 The mechanism proposed30 is suggested to involve a ‘weakly physically adsorbed oxygen atom with a finite lifetime’ which is similar to the arguments developed here and discussed at the Faraday Symposium,io“ where the classical Eley-Rideal mech- anism was seen to be an inappropriate mode132 for the oxygenation of ammonia at Mg(0001) The physical reason for CO, excitation arises from the less demanding energetics for the CO-nascent oxygen reaction, in that fewer surface-adsorbate bonds are broken Breaking fewer bonds will result in more energy from the oxidation reaction CO + %O,+CO,, about 280 kJ mol- I, being channelled into the product CO, In the systems studied where hot 0-atoms have been generated by dissociative chemisorption -magnesium-dioxygen, magne- sium-nitrous and nitric oxides and aluminium-dioxygen -the reac- tions are all highly exothermic The CO-O, catalytic reaction at aluminium,i2 where surface carbonate is formed at 80 K, is a par- ticularly good example of hot oxygen atom chemistry involving CO present at ‘negligible’ surface coverage The exothermicity of the reaction -oxygen chemisorption and CO oxidation being possible contributors to generating ‘hot’ CO, molecules which undergo further surface reactions via the reactive Cog-species to chemisorbed carbonate (Fig 3) Haber33 has also recently drawn attention to the role of oxygen transients in heterogeneous cataly- sis 8 Conclusions The significance of oxygen transients in providing low energy path- ways in surface oxygenation reactions was first established through using surface sensitive photoelectron and vibrational spectro- scopies in conjunction with the probe-molecule approach Both atomic 06-(s) and dioxygen O$-(s) transients have been shown to participate in a wide range of reactions at single crystal metal sur- faces including the oxidation of ammonia, carbon monoxide and hydrocarbons Which of the two transients participate in the surface chemistry depends on the metal, 0;-at the Zn(0001) surface Os-at Mg(0001) The mechanisms are non-classical in that they are analogous to two-dimensional gas phase reactions where neither of the reactants are strongly adsorbed are present at immeasurably small coverages but undergoing rapid surface diffusion to generate surface complexes (transition states) which decompose to chemisorbed products There is a clear need to consider the chem- istry of the total system rather than making deductions based on the chemistry of the individual reactants -the latter could be mislead- ing For example dioxygen bond cleavage is much faster via the transient ammonia-dioxygen complex than it is from dioxygen alone whether viewed theoretically or on the basis of experimental data Support for the transient concept has been provided by other experimental approaches, in particular the STM results from Ertl’s group, and the quantum mechanical calculations of van Santen for ammonia oxidation at copper surfaces Not only have coadsorption studies provided a different view of catalytic chemistry at single OXYGEN TRANSIENTS AND PRECURSOR STATES IN SURFACE REACTIONS- M W ROBERTS metal surfaces but also to unexpected structural assignments of surface species through the photoelectron diffraction studies of Bradshaw Acknowledgements It is a pleasure to acknowledge colleagues who have contributed to this work over the last ten years and to others who provided the platform on which this work was based They include Peter Au (Hong Kong), A Boronin and V Bukhtiyarov (Novosibirsk), A Pashuski (Weitzman Institute), S Laruelle (University of Picardie), M K Rajumon and G U Kulkarni (Bangalore), Song Yan (Xiamen) and my Cardiff colleagues Albert Carley and Phillip Davies I am grateful to Gerhard Ertl and Gary Haller for providing information on unpublished work and the EPSRC for its support I am also grateful to Harry Kroto for the invitation to write this review 9 References 1 C M Quinn and M W Roberts, Trans Faraday SOC ,1%5,61,1775 2 K Siegbahn, C Nordling, G Johansen, J Hedman, P F Heden, K Hamrin, U Gelius, T Bergmark, L 0 Wernse, R Manne and Y Baer, ESCA Applied to Free Molecules, North Holland, Amsterdam, 1969,J M Thomas.E L Evans, M Barber and P Swift, Trans Faraday Soc , 1971,67,1875 3 C R Brundle and M W Roberts, Proc R Soc London A, 1972,331, 383 4 See for example (a)K Kishi and M W Roberts, J Chem SOCFaraday Trans 1, 1975,71, 1721, (b) M W Roberts,Adv Cataf , 1980,29,55 5 M W Roberts, Surf Sc , 1994,299/300,769 6 C T Au and M W Roberts, Nature, 1986,319,206 7 A F Carley, P R Chalker and M W Roberts, Proc R SOC London A, 1986,399,167 8 A F Carley, S Rassias and M W Roberts, Surf Sci .1983.135,35 9 C T Au, J Breza and M W Roberts, Chem Phys Lett ,1979,66,340, R W Joyner,K Kishi andM W Roberts,Proc R Soc London A, 1977.358,223 10 (a)C T Au and M W Roberts, J Chem Soc ,Faraday Trans I, 1987, 83,2047, (b)M W Roberts, Chem Sac Rev, 1989,18,45I I1 P G Blake and M W Roberts, Cutaf Lett ,1989,3,379, M W Roberts, Chem Soc Rev, 1989,15,451 12 A F Carley and M W Roberts, J Chem SOC , Chem Commun , 1987. 355 13 A F Carley,M W Robertsand S Yan, J Chem Soc Chem Commun , 1988,267, J Chem SOC ,Faraday Trans , 1990,86,2701 14 A F Carley, P R Davies, M W Roberts and K K Thomas, Surf Sci , 1990,238, L467, and references therein 15 A F Carley, M W Roberts and M Tomellini, J Chem SOC Faraday Trans , 1991,87,3563 16 A F Carley, M W Roberts and S Yan, Cataf Lett 1988,1,265 17 A F Carley, P R Davies, M W Roberts and D Vincent, Top Catal , 1994,1,35 18 (a)A Boronin, A Pashusky and M W Roberts, Cataf Lett ,1992,16, 345, (6)B Afsin, P R Davies, A Pashusky and M W Roberts, Surf Sct Lett , 1991,259, L724, (c)B Afsin, P R Davies, A Pashusky, M W Roberts and D Vincent, Surf Sci .1993,284,109 19 C J Hirschmugl, K-M Schindler, 0 Schaff, V Fernandez, A Theobald, Ph Hofmann, A M Bradshaw, R Davies, N A Booth, D P Woodruff and V Fritzche, Surf Sci ,in press 20 (a) W Biemolt, G J C S van der Kerkhof, P R Davies, A P Jansen and R A van Santen, Chem Phys Lett, 1992, 188, 477, (b) M Neurock, R A van Santen, W Biemolt and A P J Jansen, J Amer Chem SOC, 1994,116,6860 21 (a) G U Kulkarni, S Laruelle and M W Roberts, Chem Commun , 1996,9,(b) A F Carley, M W Roberts and A J Strutt,J Phys Chem , 1994,98,9175 22 G U Kulkarni,C N R Raoand M W Roberts, J Phys Chem , 1995, 99,3310 23 G U Kulkarni. S Laruelle and M W Roberts, J Chem SOCFaraday Trans ,in press 24 T S Amorelli, A F Carley, M W Roberts and P B Wells, Surf Scr Lett ,1994,315, L900 25 J Wintterlin, R Schuster and G Ertl, Phys Rev Lett ,in press 26 C S McKee, L V Renny and M W Roberts, Surf Sci ,1978.75,92 27 H Brune, J Wintterlin, J Trost, G Ertl, J Wiechers and R J Behm, J Chem Phys ,1993,99,2128, H Brune,J Wintterlin,R J BehmandG Ertl, Phys Rev Lett, 1992,68,624 28 J Greber, R Grobecker, A Morgante, A Bottcher and G Ertl, Phvs Rev Lett , 1993,70, 133 1 29 Y L Li,D P Pullman,J J Yang,A A Tsekovras,D B Gosalvez,K B Laughlin,Z Zhang, M D Schulberg, D J Gladstone,M McGonigal and S T Ceyer, Phys Rev Lett , 1995,74,2603 30 G L Haller and C Wei, J Chem Phys ,to be published 3 1 G W Coulston and G L Haller, J Chem Phys , 199 1,95,6932 32 M W Roberts, ‘Contribution to General Discussion’, J Chem Soc , Faradav Trans I, 1987,83,2085 33 B Grzybowska-Swierkosz and J Haber, Ann Rep Prog Chem ,RSC, Cambridge, 1994,91, sect C, 395