Bond Et Al. (Gold Catalysed CO)

8/3/2019 Bond Et Al. (Gold Catalysed CO)

1/10

Gold-Catalysed Oxidation ofCarbon Monoxide

Geoffrey CBondBrnnelUniversity, Uxbridge, Middlesex, UB83PH, UK

andDavid T Thompson'Newlands'The Village, Whitchurch Hill, Reading. RG87PN, UK

Of the many important reactions that are known to be catalysed by gold particles supported on transition metaloxides, the oxidation of carbon monoxide is of outstanding significance: the current state of understanding of itsmechanism is reviewed, and discordant observations reported in the literature are noted and analysed. A generalmechanism involving reaction at the edge of a particle containing both gold atoms and ions, and involving thesupport, is suggested, although not all features of it are necessarily always operative. Possible practical uses for goldin catalysing this reaction include pollution control, fuel cells, and gas sensing.

The last decade has witnessed a rapid growth of interest inthe catalytic properties of gold; when it is suitablyemployed, and much skill is needed to achieve this,supported gold is capable of catalysing a wide variety ofreactions under comparatively mild conditions, includingoxidation, oxidative dehydrogenation, hydrogenation,and many others. We have recently reviewed this sector ofcatalysis in depth (1), and other review articles have beenpublished (2, 3); the earlier Gold Bulletin review alsoprovides a general survey (4). One of the mainconclusions from our previous review (1) was that the wayin which gold acts catalytically is substantially differentfrom that adopted by metals of the Platinum Group.

Probably the greatest surprise has been the ability ofappropriately prepared gold catalysts to effect theoxidation of carbon monoxide at very moderatetemperatures; the best have even been reported tofunction below ambient temperature, and to be moreactive than conventionally prepared palladium andplatinum catalysts. Because of the potential for practicalapplication, as well as intrinsic scientific interest, the flowof publicat ions continues unabated. We have recordedsome 60 publicat ions in 1999, and at least half of theseconcern carbon monoxide oxidation. However, a readerof the recent literature, as well as of the earlierpublications already reviewed (1), cannot fail toexperience a sense of confusion at the many methods ofpreparation and characterization that have been used, themany different supports, and in particular at the various

@ GoldBulletin2000, 33(2)

and often contradictory reports on the importance of theexperimental variables. It is little wonder, therefore, thatalmost all discussions of reaction mechanisms are of anelementary and somewhat speculative nature, because ofthe lack of a firm foundation of reaction kinetics andspectroscopic data analysis.

The main reason for this review is to try tounderstand the reasons for the many discrepancies andcontradictions which the literature reveals. These reflectthe very critical and multiple criteria which must be metbefore high catalytic act ivity can be obtained, and theextreme sensitivity of the final catalyst to all stages of itspreparation and treatment before use. Although all is asyet far from clear, there are enough indicators towards apossible model that will accommodate many of theobservations, and a mechanist ic concept is tentat ivelyadvanced later in this paper.

CONSENSUAL OBSERVATIONSNotwithstanding the seemingly confused state of theliterature, there is a large measure of agreement as to theprincipal factors responsible for the activity (or inactivity)of gold catalysts evaluated for carbon monoxideoxidation, and it will be helpful to start by listing these.1 Unsupported gold in the form of sponge, wire,

powder, etc, and large supported gold particles areat best only weakly active for carbon monoxide

41


2/10

oxidation (5 - 7), although fine gold powder (meanparticle size 76 nm) shows activity between 249and 294 K (8).

2 Much higher activities are shown by oxidesupported smallgold particles less than about 5 nmin size, with the highest rates being shown at sizes ofabout 2 to 3 nm (1, 9 - 11). This is true whateverthe method of preparation employed; they mayoriginate as a colloidal dispersion with uniform sizedgold particles (12 - 14) or be formed by vapourdeposition (2, 15- 18), or by controlled aggregationof atoms or ions in simple gold compounds (1, 19-22, etc).

3 Choice of support is very important: successdepends on using an oxide of the first row of th eTransition Series elements in Groups 4 - 12 (2),compounds which have no t previously foundmuch use as supports for the Pla tinum GroupMetals. Of these, the oxides of t itan ium and ironhave recently commanded the most interest,although those of manganese, cobalt and nickel arealso effective (1). Many other oxides, includingmagnesia (23 - 26), and the conventional aluminaand silica (3, 27, 28) have also been studied but aregenerally less efficient.

4 The method of prepara tion is critical. Th etraditional methods, such as ion exchange andimpregnation (10, 20, 29, 30), are no t verysuccessful although successive reduction andcalcination can generate activity in the impregnatedmaterial (10). Three methods in par ti cu la r havebeen widely used: (i) coprecipi ta tion, in which thesupport and gold precursors are brought out ofsolut ion, perhaps as hydroxides, by adding a basesuch as sodium carbonate; (ii) depositionprecipi ta tion, in which the gold precursor isprecipitated onto a suspension of the pre-formedsupport by raising the pl-l, and (iii) Iwasawasmethod in which a gold-phosphine complex (eg[Au(PPh;0JN03) is made to react with a feshlyprecipitated support precursor. These methods havebeen descr ibed in more detail in our earlier review(1), and the chemistry of lwasawas method has beendescribed in detail (2). The use of urea as ahydrolysing agent has now been introduced (31).Other procedures such as the use of colloids(12 - 14), grafting (3) and vapour deposition (2, 15- 18) meet with varying degrees of success.

5 The mode of pretreatment before use is alsoimportant: calcination is frequently used withgood results, while reductive procedures are usuallynot helpful.

42

DISPUTED OBSERVATIONSAll would agree that the foregoing paragraphs correspondto the general experience of workers in the field; it is,however, when one tries to elucidate the quantitativeimportance of these factors that disagreements start toemerge. For example:1 How effectively does gold itself catalyse oxidation

of carbon monoxide (8, 1O)?2 Does the activity of a catalyst decrease with time on

stream (20), or does it increase (21), beforeattaining a steady state?

3 Inwhat circumstances is calcination either desirableor essential, and what are the best conditions touse? Some investigators report highest activitieswith uncalcined catalysts (1, 21, 32).

4 Are the catalytically active gold species zero-valentatoms (Au") or oxidized species (Au I or Au III)?

5 Kinetic parameters (orders of reaction, activationenergy) seem to vary widely and irrationally; whatare the factors that are responsible?

UNSATISFACTORYASPECTS OF THEEXPERIMENTALWORKWhen a new field of research potentially as large as 'catalysisby gold' opens up, it is inevitable that each piece of researchand each publication will only be able to tackle a very smallpart of the whole, so that opportunities for cross-checkingwill be minimal. Moreover 'catalysisby gold' exhibits manynovel and unexpected features no t shown by othercatalytically active metals, and for which research workershave been unprepared: these features, and the tightconstraints within which activity appears, largely explainthe late development of the field and add to the problemsfaced when trying to review the topic. The unsatisfactorynature of the published work therefore stems from thenecessarily limited and very specific nature of eachcontribution, and thus from the sheer scope of the field,and its numerous relevant variables. It is nevertheless worthlisting a number of points which, were they to receivesufficient attention, would probably remove many of theambiguities and uncertainties which surround past work:1 Physical characterizat ion of catalysts is of ten very

sketchy; for example, BET surface area measurements(surely one of the simplest things to do) are often no treported, and estimates of mean gold particle size (egby TEM) are often lacking. There are differentopinions about the value of EXAFS for characterizingsupported gold (eg3, 22, 33).

@ ' GoldBulletin2000, 33(2)


3/10

2 Some papers focus predominantly on the use ofone method for characterization (eg XPS (22) orMossbauer spectroscopy (eg 3, 21, 26, 32). Themore the merrier!

3 More importantly, there is little appreciation of thechemical and physical changes that can occurduring instrumental characterization, t hrough theeffects of UHV and of radiation (33). There is aparticular risk that oxidized gold species may bereduced to Au" during examination: a kind ofHeisenberg's Uncertainty Principle applies (ielooking at the catalyst changes it). It might be quitedangerous to infer what species are catalyt ical lyactive on the basis of what is revealed by XPS, XRDetc because the irradiation used may change theelectronic structure of the catalyst being examined.

4 Post-reaction characterization is more useful thanpre-reaction studies, but it is no t often done.There have been few in situ studies duringreaction, egby FTIR (10).

5 Activity measurements are usually based on a singleset of conditions (temperature, 02/CO ratio, flowrate or contact time): possibly important changes ofrate with time on stream are no t always reported, andwhen they are the s ta tements are usually onlyqualitative. Th e times at which rates are measured arenot always recorded, an d observations on reactionkinetics are few and far between (seeTable 1).

6 Comparisons are made between rates and otherkinetic parameters based on the use of very differentreactant concentrations (see Table 1), contact timesand other variables: in par ticular there seems to be

Table 1 Kinetics ofCarbon Monoxide Oxidationon Supponed GoldCatarysts. Rateex Pco x PofSupport Method* [Au](%) d(nm) T(range) x y E O zlCO Ref

(K) (kJmar l)aFe20 ) COP 0.66 4 304 0 0.05 35. 1 20 34

(273 - 333)FeOx COP 3.15 6.5 353 0.55 0.27 31 19(313- 373)FeOx IW A 3 2.9 2 17 0.03 0.35 28 37(-)

CO)04 DP 1.2 6 -7 273 0.05 0.27 16.3 20 34(253 - 323)

Ce(Ca)Ox CO P 5 8 283 - 327 0.30 0.18 53.7** 8 35(283 - 327)

T i02 DP 3.3 2 273 0.05 0.24 34.3 20 34(263 - 313)Ti0 2 IMP 2.3 25 - 30 3 13 0.24 0.4 -3 8. 6(-) -0.6 -1 0***T i02 COP 1.0 33 293 0.43 0.03 30.5 10(-)Ti0 2 CO P 1.0 33 3 13 0.56 0. 13 10Ti0 2 IW A 3 263 0.25 0.4 1 24 38(- )

NotesI The first temperature isthat at which orders of reaction were measured:the range in brackets isthat usedfor activation energy.2 A dash meansthat information was not provided.3 The quoted 02/CO ratio isthat usedforthe activation energy.* COP = coprecipitation; DP = deposition precipitation; IMP= impregnation; IWA = lwasawa's method in which gold-phosphinecomplexes react with freshly precipitated support precursor.** This activation energy is true: the others are apparent***The lower value was found above 360 K

@ ' GoldBulletin 2000, 33(2) 43


4/10

little appreCiatlOn of how for example apparentactivation energy may depend on reactant pressures,or how orders of reaction may depend ontemperature.

7 Statements of reaction mechanism, when given, areoften of th e flimsiest nature, leaning heavily on th eideas from quite different systems and ignoringrelevant literature.

8 There have been no systematic studies of th eimportance of calcination conditions (temperature,time, gas composition, flow-rate, etc).We stress that the reasons for these implied

criticisms are understandable and forgivable, bearing inmind the mul tipl ic ity of variables and techniquesavailable, and the t remendous scope of the field. Whatis advocated above is no doubt a counsel of perfectionbut there's nothing wrong with a little perfection!

REACTION KINETICSThe determination of orders of reaction an d apparentactivation energy ought to supply valuable informationon reaction mechanism, as more than any othertechnique they speak directly of what is occurring atthe active centres; and differences in their values oughtto reveal changes in mechanism. Unfortunately, this isonly true if measurements are made under the same orvery similar conditions, but the data collected in Table1 show that this is far from bein g th e case. Thus,activation energies are measured either with a greatexcess of oxygen or wi th eq uimolar amounts: orders aremeasured at 217 - 353 K depending on the activity ofthe catalyst. Overarching generalizations will clearly behard to find, but some attempt must be made.

Orders of reaction are invariably reported in aPower Rate Law form, ie as exponents of th e reactantpressures. There has only been one a ttemp t (6) tointerpret resul ts in terms of Langmuir-Hinshelwoodkinetics (it was concluded that the reactan ts wereadsorbed non-competitively), and only in three (6, 31 ,35) were orders measured at more than on etemperature. There is one disturbing report (3) onreaction orders depending on the type of apparatusused. Orders in carbon monoxide are very variable;many are close to zero, indicating near saturation of theadsorption sites, but quite a few are in the range 0.25 0.6, suggesting weaker adsorption. Orders in oxygen arein general g rea te r than fo r ca rbon monox id e (,-'0.2 0.4), bu t are sometimes low (10, 34) and on occasionlower t han th at for the reductant (10). Because of thevariety of conditions and catalys t types, it is impossible

44

to detect any clear connection between orders ofreaction an d catalyst structure.In considering activation energies we may consult ahelpful compilation (2) as well as Table 1: their values fallin three fairly dist inc t brackets. Several catalysts showvalues between 50 and 60 k] moP (2, 35), and theseinclude two of very low activity. Some, no t all very active,show low values (10 - 18 k] mol."), while many includingthe most active have values between 24 and 41 k] mol",averaging about 30 k] mol:'. It would be nice if each rangecould be associated with a specific mechanism bu t thereare indications (36) that activation energy varies withreactant pressures, and indeed with the temperature rangeused (10), so such an inference would be premature .Incidentally, attempts (2) to express activities in terms ofturnover frequencies based on a presumed number ofexposed gold atoms are likely to be misguided if thepopular concept of reaction at the metal-support interfaceturns ou t to be correct.

ANAPPROACHTO UNDERSTANDINGTHE REACTION MECHANISMThe practical application of an y gold catalyst woulddemand (i) a totally reliable method for itsmanufacture, (ii) a knowledge of its long-termmechanical and catalytic stability, and (iii) informationon its response to variations in opera ting conditions(temperature, 02/CO ratio etc). A fuller and betterappreciation of the factors determining catalyticactivity an d th e means of achieving it, an d of thereaction kinetics an d mechanism, are thereforeessential for progress towards practical use.

What follows now is an attempt to rationalize andharmonize the body of information relating to goldcatalysed oxidation of carbon monoxide. It must beunderstood, however, that we are offering speculative bu treasoned interpretations of the perplexing observations inthe literature. One objective is to account for thedivergent opinions as to whether the active species is Au"or Aux+; a second is to understand the role played by thesupport, especially the claims (21) that disorderedstructures perform better; and a third is to considerwhether the mechanism should be the same on all typesof catalyst.The InterfacialHypothesisMany authors have concluded on th e basis of soundevidence that the site of very effective reaction is to befound at the junction between a small gold particle and



5/10

a Transition Series metal oxide support (2, 3, 10, 19,29, 37, 38). This idea receives confirmation from thefact that (i) large (unsupported) gold particles areeither inactive (39) or only poorly active (6) undercomparable conditions; (ii) that large supported metalparticles on alumina, silica and other oxides formed byimpregnation with HAuC14 solution and reduced inhydrogen are also of very low activity, (iii) that mostsupport oxides by themselves have little if any activity,although Fe203 has been reported to have some (40);and (iv) that active catalysts can be made by depositingtitania onto gold powder (39), or gold onto a titaniasingle crystal (41). Clearly, where peripheral goldspecies are required, rate will increase as particle size isdecreased, and it has been claimed that the expectedquantitative relation can be observed (34). At thetemperature needed to reduce the AuC14- ion,aggregation of the Au" atoms will occur, and largeparticles will result. Furthermore, especially onalumina and titania, chloride ions will be hard toremove (except by washing or steam treatment), andthey are likely to act in some way as a catalyst poison.Sodium ions on the other hand are said to promote thereaction and are usually present during depositionprecipitation (3).The Role ofthe SupportA feature of the oxides of the elements from titaniumto copper and zinc is their comparative ease ofreduction by hydrogen or carbon monoxide.Numerous studies have described significantchemisorption of both oxygen and carbon monoxideon oxides used as supports (1): the general conclusionis that carbon monoxide on a support is unreactive orof low reactivity, while the behaviour of oxygen is morecomplex and perhaps more important (8, 37, 38),depending much on the ease with which anionvacancies (Os) can be created. With alumina and silica(42) it is mos t improbable tha t these can occur undernormal conditions of pretreatment or reaction, so thatoxygen chemisorption will be negligible andinvolvement of lattice oxide ions unlikely. Inconsequence the mechanism on large gold particleswhere there is no support participation and where thetotal length of the periphery is comparatively small willbe of the Langmuir-Hinshelwood type, ie oxygen(weakly) and carbon monoxide (more strongly)adsorbed adjacently onto gold (42).

At least two of the preparative methods generatesupports which initially are non-crystalline (they may infact be hydroxides or hydrated oxides), and even afterdrying or calcination (if performed) may be highly


disordered or of complex structure. This applies tolwasawas method and to coprecipitation: it isnot clear towhat extent deposition-precipitation may cause damageto the pre-exisiting support structure, although it isclaimed that the presence of gold species (and even oftriphenylphosphine) can cause erosion of the surface (2,38). Good catalytic activity is, however, very oftenassociatedwith disorder in the support, either generallyorjust in the neighbourhood of the gold particle. Suchdisorder may in part be caused by the heat released bychemisorption of the reactants (40).

There are several possible reasons for this. Use offreshly precipitated hydroxide undoubtedly encouragesadsorption of the gold-phosphine precursor in a highlydispersed form, and therefore helps to achieve a small sizeof the ultimate gold particles. With coprecipitation,effective dispersion of the gold precursor (Au(OHh,AU203.xH20 or AuO(OH)) amongst the supportparticles is also guaranteed, although there is a risk of theactiveparticles being buried within the support: therefore,we suggest that this is why calcination is needed - tocreate porosity by loss of water, and hence access to thegold particles. In disordered zones of the support it mayalso be easier to create anion vacancies into which oxygenmolecules can be chemisorbed.

Structural complexity in the support is wellexemplified by ferric oxide. Coprecipitation using sodiumcarbonate leads to the formation of 'ferrihydrite', astructurally disordered material having the compositionFe5HOs.4H20 and to FeO(OH) (goethite), which aftercalcination transforms to a-Fe203 (haernatite) (21, 32).In Iwasawasmethod, the initially formed Fe(OHh* uponcalcination gives first a mixture of Fe304 (magnetite) anda-Fe203 and at a higher temperature the Fe304 isoxidized to ),-Fe203 (2). Veryactive catalysts are obtainedin both cases, but activity is not simply a function of BETarea (2).The Oxidation State ofthe GoldA major bone of contention is the oxidation state ofthe gold in the active centre. As noted above, this isoften inferred by methods such as XPS, XRD, andEXAFS, which employ conditions of vacuum and/orirradiation quite unlike those experienced duringcatalysis. Several different kinds of gold species havebeen identified using FTIR on chemisorbed carbonmonoxide (42); these include Au", in high or lowcoordination, "positively polarized" gold atoms at theperiphery (13, 42), and gold species on which bothcarbon monoxide and oxygen atoms are adsorbed (1).*Asterisk (*) denotes freshly precipitated

45


6/10

Oxidation state is also infer red f rom the effect ofpretreatment on actlvity: calcination (in air) issometimes essential, sometimes beneficial, sometimesunnecessary an d somet imes harmful . Oxidized goldspecies are unstable even in air, an d decompose to Au "as temperature is increased (22) : this change leads toloss of chemisorption capacity an d catalytic activity.Change of activity during use has also b een used toidentify the oxidation state: deactivation is oftenobserved an d has been ascribed (20) to reduction ofions by carbon monoxide to Au", as for example:

although activities sometimes increase to begin with,before reaching a stable high level, egwhen dried (ie non-calcined) materials are used (21). Most authors are quitestrongly polarized in their views, believing that only Au"(3, 10) or only AuX+ (22, 42) species are responsible foractivity. The mode of action of the latter, an d the reasonsfor their superiority over Au", is no t explained by thosewho believe in them, and no one seems to consider thatthe presence of both may be necessary. Electrocatalyticoxidation studies using gold electrodes (1, 43, 44)strongly suggest that gold atoms having low coordinationnumber (CN) are more electropositive and are thereforemore easily oxidized than those of high CN, and thissuggests that peripheral gold atoms in small supportedparticles may readily transform into ions. Thesimultaneous presence of both atoms and ions is thereforeno t ruled out.Chemisorption ofReactants and ofU7teterIt is, however, generally agreed that the carbonmonoxide which is oxidizable is reversibly adsorbed onthe gold (ei ther Au" or Au'") (19, 37, 38): if it is onthe support it is as a strongly held spectator (37, 38) oras a st abl e carbonate ion (3, 10,38,41). The route foroxygen appears to involve its adsorption as a moleculeor superoxide molecular ion ( 2-) on an ion vacanciesin the surface of the supports, if these are easy to form(37, 38). Although several authors suggest that latticeoxide ions (02-) are the oxidant (40), there is littledirect evidence for this, and very clear indications thatthe active species is an oxygen molecule (2, 10, 37, 38)or 2-, No dissociation of oxygen occurs excep t byreaction with carbon monoxide (37).

Various results have been reported for the effect ofwater on the reaction (10, 22, 25, 37, 38). It sometimesincreases the rate, perhaps by ensuring that some or all ofthe gold remains in an oxidized state (3), bu t it has noeffect on Au/Fe(OH)3* (37) and acts as a poison on

46

Au/Ti(OH)4* (38) and Au/Ti02 (10), perhaps byoccupying anion vacancies (ie hydrating the surface), andpreventing oxygen adsorption. A positive role for surfacehydroxyls has been suggested with magnesia (25), andindeed many of the oxide supports are l ikely to remaincoated by at least a par tial layer of hydroxyl ions duringuse. Water promotes the reaction on Au/MgO at hightemperature, bu t poisons it at low temperature (25). It hasrecently been noticed t ha t w ith Au/Fe203 the reactiondoes no t proceed if the reactants are completely dry, bu tthat 100% conversion occurs at 273 K when water ispresent even in traces?".Two Further FactorsBefore attempting harmonization of all these very diverseobservations, two other factors need to be considered:1 The reactivity of gold atoms may depend on their

coordination number (CN), strongerchemisorption being shown theoretically for lowervalues of CN (45). The concept of varying electrondensity across s teps on model surfaces is wellestablished (46), and the fract ion of edge atomswill obviously rise as particle size decreases (1, 45).

2 The stability of noble metal (Pt, Rh) particles onalumina and titania is attributed to a 'chemicalglue' of metal cations at the interface between thesupport and the particle (see, for example,reference 47). At small particle sizes th e fraction ofthe metal in ionic form can be quite high.Structures of t hi s i nt er faci al zone, and changeswith ambient atmosphere, have been elucidated byEXAFS (47), bu t similar studies have no t ye t beenperformed with gold on reactive supports.

Classes ofSupportIt remains now to draw these threads together an d in avery speculative way to imagine what structures mayexist in gold ca talys ts that are active for carbonmonoxide oxidation, and by what mechanisms thereaction may occur. It has been very sensibly suggestedthat it may be necessary to conside r each type ofcatalyst by itself, an d although for example Au/FeOxcatalysts made an d pretreated in different ways mayexhibit some different characteristics these must beconsidered i n the framework of a single model . We will,therefore, adopt the following classification of supports:1 FeOx (and perhaps other easily reducible oxides suchas C0304, NiO, CuO, etc, for which detailedinformation is lacking)

** G. J. Hutchings, personal communication



7/10

AuOparticle

ox. atron

Active particle

calcinatio n,redu ction

...

X Je H eM e Aum O Auo Od I

Figure 1 Pictorialrepresentation o/supportedgold catalyst indicatingpossible changes underconditions giving oxidationor reduction0/theactivegoldparticles. Thegold ions provide a 'chemical glue'which binds theparticle to thesupport.2 1102 (and perhaps other less easily reducible oxides

such as zr02)3 MgO (and perhaps other strongly basic supports)4 Ah03, Si02.We have already said something about alumina andsilica as supports and there is only a little informationavailable about magnesia (23, 25): we will, therefore,focus on the more commonly studied iron oxide andtitania supports. Iwasawas Au/FeOx are more active thanhis Au/1102 catalysts (2), probably because FeOx is themore easily reducible support and because the disorderedstructure noted above facilitates the formation of anionvacancies near gold particles: the reverse is, however, truefor coprecipitation catalysts (3), which may suggest thatanion vacancies are oflesser importance in this case.Possible Reaction MechanismsWe suggest tha t many of the reported observations canbe understood on the basis of a model in which theactive catalyst contains both gold atoms and ions (whichwe will take to be Au IlI) , and tha t the l att er form the'chemical glue' which binds the particle to the support(see Figure 1). Each particle is, therefore, bounded by aring of Au l l ions. The structure is not, however, fixed:the AuIIlIjAuo ratio may change (i) during calcination,when depending on temperature a certain fract ion ofthe Au l l will autoreduce as oxygen is lost (20, 22), (ii)during reduction by hydrogen if performed (40), and(iii) during reaction, where initial changes in reactivitymay reflect changes in this rat io (10, 20 - 22). Thus

@ ' GoldBulletin 2000, 33(2)

complete reduction, due to excessive calcination or useof hydrogen, is harmful because the glue is lost fromthe interface and sintering can then occur easily(Figure 1): bu t equally the total absence of Au'' isundes irable , as it is needed to provide a locus for thechemisorption of the carbon monoxide, and a partiallyreduced support surface may conta in anion vacancieswhich assist oxygen adsorption.

Thus, the sense of initial activity changes will dependon the Au l l / Au? ratio at the outset and on the 02/COratio used. We believe t ha t not only reduction but alsooxidation of the gold phase can occur during reaction,and we have noted the very widely varying reaction ratiosthat are used to measure activity (Table 1). This modelcan be refined to assign various tendencies for carbonmonoxide to chemisorb on Au" atoms of different CN(45). Large gold particles may possess only relatively smallinterfacial areas and few atoms having low CN and thusshow only diminished activity.It has been suggested in the case of magnesia (25) thatthe mechanism may start by a support hydroxyl ionattacking a carbon monoxide molecule on the gold(Au"...CO). We think this may occur more generally, andpropose a 'periphery' mechanism as follows. TheAu?...CO is attacked by an hydroxyl group either on asupport cat ion or on a peripheral Au IlI ion, forming acarboxylate group attached to the latter. This is in turnattacked by a superoxide ion, which must be responsiblefor oxidizing two carboxylate ions: the hydroxyl groupreturns whence it came and is ready to re-engage in the

47


8/10

catalytic cycle. Consequently, steps 1n the mechanismcould include:AuO+ CO -7 Auo CO (1)Au ill + OH,' -7 Aull OH (2)Au?...CO + Aull...OH -7 Aull..COOH +Auo (3)Oz + 0,' -7 0 z".o , (4)Aull COOH + 0 z' ...Il,-7 AuII + COz + HOz'.o, (5)Aull COOH + I-10 z'...O, -7 Aull+ COz +2 OHs' +0, (6)Aull+ 0, -7 Auill + 0,' (7)

The nett reaction, ie 2CO + O2 -7 2C02, isobtained by doubling the processes represented byEquations (1) - (3) and then adding all the processes inthe set.

There is infrared evidence for the presence of bothAu"...CO and AuX+ ...CO species with the relevantabsorptions occurring at 2112 and 2151 cm",respectively (42). It should be noted that AuO(OH) isdetected by Mossbauer spectroscopy in pre-calcinedferrihydrite-supported catalyst (21) and Au+ is present at

I e M"

the interface between gold and magnesium hydroxidesupport (3, 26). XPS results are also consistent with thepresence of gold hydroxyl species on FeZ03, 110z, andAl2 0 3 (22).A graphic representation of the initial stages of thismechanism is given in Figure 2. There the anion vacancyused in Equation (4) is created in Equation (2), and byimplication is removed when the hydroxyl ion (which actscatalytically) is restored to the support surface in theprocess represented by Equation (6).

We noted above a suggestion th at the mechanismmight not be the same on all catalysts. Thus, forexample, one possible variant of the above scheme mightapply with those oxides or under those conditions whereanion vacancies are not formed: the oxygen moleculemight simply be weakly bonded to a support cation,instead of as in Equat ion (4). Similarly, where mobi lesupport hydroxyls do not exist, reactions (2) and (3) willnot take place, and the mechanism may entail directreaction of a chemisorbed carbon monoxide with anoxygen molecule or a superoxide ion. Much of thedetailed kinetic information currently available (10, 37,38) is confined to specific catalysts, and generalizationsare hard to make.

Figure 2 A representation ofthe earlystages ofthe oxidationofcarbon monoxideat theperipheryofan activegoldparticle. At the left,a carbonmonoxidemolecule is chemisorbed on a low eN goldatom, and an hydroxylion hasmovedfrom the supportto anAuD! ion, creatingan anion vacancy. At the right theyhave reacted to[arm a carboxylate group, and an oxygen moleculeoccupies theanion vacancy asO2 ' , This thenoxidizes thecarboxylate groupby abstracting a hydrogen atom, fOrmingcarbondioxide, and the resulting hydroperoxide ionH02' then oxidizesa[uriher carboxylate species fOrminganother carbondioxideand restoring twohydroxylions to thesupportsuiface. This completes thecatalyticcycle, No attempt ismade tosuggest the charges carriedby the reacting species.

48 @ ' GoldBulletin2000, 33(2)


9/10

However, in this manner we can understand (as in aglass darkly!) how contradictory conclusions about theoccurrence and importance of Au'' and AuX+ species canarise both from physical methods of characterization(which can alter their ratio) and from chemical changestaking place during calcination and reaction. Subtlerdifferences between the various types of oxide remain tobe explored.

Whilst it is dangerous to general ize and drawconclusions on the basis of results recorded from awide variety of catalysts operating under such diverseconditions, the mechanistic concepts advanced aboveprovide a working hypothesis which may fit thecurrent state of knowledge on gold catalysed carbonmonoxide oxidation. If this is correct , then Au", Au'";and the metal oxide support all have a role to play andthe interaction of these species at the interface betweenthe gold and the support is particularly important.Testing the MechanismAny mechanistic concept has to withstand experimentaltest: of the several cr iteria to be applied to ou r proposal,perhaps the most significant would be detecting thesimultaneous presence ofAuO and Au3+ (or perhaps Au")in the working catalyst. Possibly EXAFS measurementsmight show if this is indeed the case, and find where theions (i f found) are located. The ratio ofAuX+:Au" shouldincrease as the O2 : CO ratio is increased. Carefulspectroscopic studies might indicate the presence ofcarboxylate or similar intermediate groups. Theoreticalwork using density functional theory should be directedtowards disclosing the energy barriers involved in the CO+ OH- reaction and to the oxidation of carboxylate ion byoxygen molecular ions (02}

POTENTIAL APPLICATIONS FORCATALYSIS BY GOLDPract ical applications for gold catalysts wil l naturallyrequire reliable methods of manufacture and a secureknowledge of their long-term mechanical and catalyticstability as well as their performance criteria, includingactivity and durability under practical operatingconditions. Nevertheless, there is no reason why suchinformation should not become available in due courseand the po tent ia l of these catalysts for pollutioncontrol, fuel cell, sensor, and chemical processingapplications could then be realized.

It is possible that gold could be usefully incorporatedinto automobile catalysts: the price of pal ladium is


current ly rising rapidly due to the demands for low lightoff 'start-up' catalysts in these systems and supported goldhas the potential of meeting at least some of therequirements for such systems. Some patents (48, 49)have been filed claiming gold catalysts for use inautomobile pollution control (see Gold Bulletin, 1999,32, 30, 102). Their use in this application will, however,require demonstration of adequate stability at operatingtemperatures.

Preferential oxidation of carbon monox ide inhydrogen-rich reformer gas is effected by a goldcatalyst (Au/a-Fe203) which is s ignificant ly moreactive than the commercial PROX catalyst, ie Ptl-yAl203 (19). Gold catalysts are therefore likely to beconsidered for carbon monoxide removal from thistype of gas stream, a process which is required for fuelcell applications. The hydrogen / oxygen reactionwhich is also used in fuel cells is catalysed by gold (1).Patents have been filed for the use of gold catalysts infuel cells (50, 51). The removal of carbon monoxidefrom air to prov ide high purity nitrogen and oxygenhas also been envisaged (52).

Haruta et aL (3, 53, 54) have studied theabsorbance of thin films of transition metal oxides (egnickel, cobalt and copper oxides) containing gold andshown that by monitoring the absorbance change attwo different wavelengths (eg 600 and 900 nrn) therecognition of hydrogen and carbon monoxide becamepossible when these gases are present at 1vol% levels.Consequently, composite films of cobalt oxide crystals(5 - 10 nrn) and small gold particles (20 - 40 nm) canform a basis for optical gas sensors to measure thelevels of these two gases simultaneously.

ACKNOWLEDGEMENTSWe t hank Donka Andreeva, Alfons Baiker, MasatakeHaruta, Graham Hutch ings , Yasuhiro Iwasawa, BenNieuwenhuis and Walter Vogel for providing reprintsand preprints of their papers.

ABOUT THEAUTHORSGeoffrey Bond held academic posts at Leeds and HullUniversities before joining Johnson Matthey plc in 1962as Head ofCatalysis Research. In 1970 he was appointedProfessor in Brunel University's Chemistry Department,and is now an Emeritus Professor at that University.

David Thompson has a special interest in catalysis bygold, having previously held catalysis research posts in

49


10/10

leI and Johnson Matthey plc where he was involved inprecious metals catalysis projects. In his current freelancework as a chemical consultant he advises on catalystdesign for chemical processing, pollution control and gasdetection. He isTechnical Editor of Gold Bulletin.

REFERENCES*Thereferences denoted byanasterisk were also included inourprevious review (1)1 G.c. Bond andD.T. Thompson, Cat, Rev, Sci, Eng., 1999,41, 3192 AI. Kozlov,AP Kozlova, H, Liu andY Iwasawa, AppL CtuaL A, Gen, 1999, 182,93 The Abilities andPotential of Gold as a Catalyst', Report of theNational Research

Institute,Osaka.japan, 19994 D,T. Thompson, GoldBull. , 1998, 31,111; 1999,32,125* NW.CantandW.K Hall,J Plys, Chern., 1971, 75,29146" S,D, Lin,M, Bollinger andMA.Vannice, G:ttaL Lett, 1993, 17,2457* WA.Bone andG,w.Andrew, Proc. JWy. Soc. (London), 1925, 109A, 459

Y Iizuka, T.Tode,T.Takao, K,Yatsu,T.Takeuchi, S, Tsubota andM,Haruta,J C:ttal.,1999,187,50

9* M Vaiden,X.Lai andD,W. Goodman, Science, 1998, 281, 164710* MA Bollinger andMA.Vannice,AppL C:ttal. BiEmironmental;1996,8,41711 M Vaiden, S,Pak, X.Lai and D,W. Goodman, Cual. Letters, 1998, 56,712* ]..0, Grunwaldt, C.Kiener, C.Wogerbauer andA Baiker,J Cuai, 1999, 181,22313 J,.D, Grunwaldt, M Maciejewski, O,S, Becker, P. Fabrizioli andABaiker, I C:ttal.,

1999,186,45814 J,.D,GrunwaldtandA Baiker,JPhys, Cbem.B, 1999, 102, 100215 K.Sayo, S, Deki, and S,Hayashi.j, Co&idandInteiface Science, 1999,212,59716 K.Sayo, S, Deki andS,HayashiJMater, Cbem; 1999,9,93717 Y.A Bondzie, s,c. Parker and c.T.Campbell, G:ttaL Leu, 1999,63, 14318 M Arai,M Mitsui,p, Ozaki and Y, NishiyamaJ Co/hidandInterface Science, 1994,

118,47319 MJ,Kahlich,H.A Gasteiger andRJ,Behm,JCasal, 1999, 182,43020 AM Visco, E Neri, G,Neri, A Donato, C.Milone and S, Galvagno, Phy" Chem,

Chern. Phys" 1999, 1,286921 RM Finch, NA. Hodge, G.]. Hutchings, A Meagher, QA. Pankhurst, MRH,

Siddiqui, EE. Wagner and R Whyman, Phys, Chem, Chern. Plys" 1999, 1,48522 E.D,Park and LS, Lee, I C:ttal., 1999, 186, 123 A Sanchez, S,Abbet, U. Heiz, W.D, Schneider, H,Haekkinen, RN, Barnett and U,

LandmanJ Phys, Chern. A, 1999, 103(48),957324 U. Heinz,A Sanchez, S,Abbet andW.D, Schneider, Eur, Plys.]., D,1999,9, 3525 DA.H,Cunningham,W. \bgelandM,Haruta, Caial. Lett,1999,63,4326 Y, Kobayashi, S,Neu, M Haruta and T. Tsubota, /(yow Daigaku Genshiro Jikkensho

GakujulJuKoenkai Hobunshu, 2000, 34, 14527 c.x Chang, Y]. Chen andC.T, Yeh,App! CtuaLA, 1998, 174,1328 Y, Yuan, K,Asakura,AP. Kozlova,H Wan, K Tsai andY Iwasawa, CtuaL Today, 1998,

44,33329 N,].Osipoffand NW.Cant, Topics inC:ua!ysis, 1999, 8,16130 K,.HChoi, B..Y Koi andH,LLee, CtuaL Today, 1998,44, 20531 BE Niewenhuys, MA.P. Dekkers and Y.]. Mergler, Polish Journal ofEnvironmental

Sw.dies, 1997,6, 5332* FE.Wagner, S, Galvagno, C.Milone, AM, Visco, L Srievano and S.D. Calogero,

50

I Clem. Soc, Faraday Trans,,1997,93, 340333 D.AH Cunningham, W. Vogel, RM Torres Sanchez, K,Tanaka andM, Haruta,

I C:ua!, 1999, 183,2434* M, Haruta, S, 'Isubota, T.Kobayashi, H Kageyama, M]. Genet and B. Delmon,

I C:ua!, 1993, 144, 17535* W. LinandM Eyrzani-Srephanopoulos.j.Catal, 1995, 153,304,317,36 E Bocuzzi,A Chiorino, M,Manzoli, n Andreeva and T.Tabakova, I Catal, 1999,

188,17637 (a) H Liu,AI. Kozlov,AP. Kozlova,T.Shido andY Iwasawa, Phys, Chern. Chern. Ply"

1999,1,2851; (b)"APKozlova,AlKozlov, S,Sugiyama, YMatsui, K.Asakura andY Iwasawa,J CtuaL, 1999, 181,37

38 H,Liu,AI. Kozlov,AP. Kozlova, T.Shido, K.Asakura andY Iwasawa,J Catal, 1999,185,252

39* 2,M,Liu andMA.Vannice, G:ttal, Lets,1997,43, 5140 (a) AX, 'Iriparhi, Y.S, Kamble and NM, Gupta, I Catal, 1999, 187,332; (b) HM

Gupta andAK.Tripathi,J G:ttal, 1999, 187,34341 (a) VA. Bondzie, sc Parker and cr Campbell, I Sci, Vac, TechnoL A, 1999, 17,

1717; (b) s,c. Parker,AW. Grant, Y.A Bondzie and c.T. Campbell, Suif. Sci, 1999,441,10

42 (a) MA.P. Dekkers,M,J, Lippits and BE Nieuwenhuys,Catal Letz, 1998, 56,195 (b)MA.P. Dekkers,M.]. Lippits and B.E. Nieuwenhuys, CataL Today, 1999, 54,381

43* LD, Burke and P.E Nugent, GoldBull., 1998,31, 3944 LD, Burke and P.E Nugent, GoldBull., 1997,30,14345 M,Mavrikakis, P. Stolze and].K, Norskov, CtuaL Lett,2000, 64,10146 LM,Falikov andG.ASomorjai, Proc. Am,Sci, USA, 1985, 82,220747 EW.HKampers and nc. Konigsberger, Faraday Disc, Clem. Soc, 1990, 89,13748 LA.Petrov, Laman Consultancy Ltd, Bulgaria, W.O, Patent 9851401 Al 1998111949 P. Marecot andR Emmanuel, Rhone Poulenc Chimie, French Patent 2771310 Al

99052850 Gold Catalyst rorFuel Cells,Y.MTatchev, Laman COnsultancy, Bulgaria,W.O, Patent

9916546A199040851 B. Grigorova, ]. Mellor andJ.AJ, Tumilty, Anglo American Research Laboratories,

SouthAfrica,WOPatent 2000013791 Al (16Mar 2000)52* M,Haruta, CtuaL Today, 1997,36, 15353* M,Ando,T.Kobayashi andM Haruta, G:ttal, Today, 1997,36,13554* MAndo, T.Kobayashi, S, Iijima andM,HarutaJ Mater Chem, 1997,7, 177955 L Prati and G,Martra, GoldBull., 1999,32,9656 TA.NiJhuis, B.]. Huizinga, M Makkee andJ.AMoulijn, lnd. Eng, Chern. &s., 1999,

38,88457 YA. Kalvachev, T.Hayashi, S,Tsubora andM HarutaJ Catal,1999, 186,22858 B.S,Uphade andM Haruta, App! C:ttal., A (GeneraiJ, 1999, 190,4359 0, Kaikawa, X.5,Wang, T,Tabata and0, Okada, Organohahgen Compounds, 1999,

40,58160 H,u YY Luk andM,Mrkisch, Langmuir, 1999, 15 (15),495761 n Andreevaand T. Tabakova, Proc Europa CJt.4, Rimini.Italy, September, 1999, P 69962 ].E Bailie and G,J, Hutchings, Chem, Commun. 1999,215163 H,.G,Ahn andH,Niiyama, Kongop Hwahak, 1999, 10 (7),100364 E. Seker, J Cavataio, E. Gulari, P. Lorpongpaiboon and S, Osuwan,AppL Catal

A, 1999, 183, 12165 W. Cwikla and0, Naimek,Adsorpt. SciTecbnol; 1998, 16,77366 A Ueda andM.Haruta, GoldBull.,1999,32,367 T.Wang and JA.Broussard, Celanese International Corporation,WOPatent9962632

@ ' GoldBulletin 2000, 33(2)

Bond Et Al. (Gold Catalysed CO)

Documents

Transcript of Bond Et Al. (Gold Catalysed CO)