colloidal deposition as method to study the influence of the support on the activity of gold...

Phys. Status Solidi B, 1–10 (2013) / DOI 10.1002/pssb.201248499 p s sb

statu

s

soli

di

www.pss-b.comph

ysi

ca

pecial Issue oneous Catalysis

Part of SMetal–Substrate Interactions in Heterogen

eature Article

asic solid state physics

Colloidal deposition as method tostudy the influence of the supporton the activity of gold catalysts inCO-oxidation

F

b

Ferdi Schuth*

MPI fur Kohlenforschung, Kaiser-Wilhelm-Platz 1, 45470 Mulheim, Germany

Received 23 October 2012, revised 12 February 2013, accepted 12 February 2013

Published online 11 March 2013

Keywords catalysis, colloidal deposition, gold catalysts, nanoparticles

* e-mail [email protected], Phone: þ49-208-3063273, Fax: þ49-208-3062995

The strong influence of the support properties on the activity of

gold catalysts has been observed in many publications. The

most studied reaction in this respect seems to be CO-oxidation,

for which gold catalysts have outstanding activity. However,

since in most studies the support properties are also important in

influencing the nature of the gold particles deposited on them

by co-precipitation or deposition–precipitation, it is difficult

to study the support effect alone. We have in a series of studies

used colloidal impregnation of preformed gold particles

approximately 3 nm in size on different supports in order to

decouple the gold particle formation from the deposition

process, in order to isolate the support effect. Even for such

similarly prepared catalysts very strong differences between

different supports were observed. The analysis of the data, also

in the light of literature data, suggests that there is no unique

factor explaining the high activity of gold catalysts, but rather a

combination of effects, which act in different proportion for

different catalysts.

� 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

1 Introduction The high activity of gold catalysts invarious reactions is a topic of tremendous current interest,both because such catalysts could be interesting for technicalapplications – Sudzucker has built a demonstration plant inwhich gold catalysts are used for the oxidation of sugars –and because their mode of action is highly intriguing andsubject of intensive scientific debate. While bulk gold is not aparticularly interesting catalytic material, small goldparticles on suitable supports are excellent catalysts for anumber of reactions, with ethylene chlorination discoveredby Hutchings [1, 2] and CO-oxidation discovered by Harutaet al. [3] being the first ones, for which the tremendouspotential of gold catalysts has been realized. The mostintensively studied reaction to date is CO-oxidation which inspite of its apparent simplicity turns out to be a very complexsystem on which no agreement has been reached, yet. Thehigh activity of gold-based catalysts for this reaction hadbeen discovered in the pioneering study of Haruta et al. [3],who had observed that gold nanoparticles supported onoxides – in this seminal publication the oxides of iron, cobaltand nickel were found to be best, later titania was the most

intensively studied support – are active for CO-oxidation attemperatures even below room temperature.

However, in spite of approximately 25 years of research onthis phenomenon, the nature of the active species and the modeof action of these catalysts are still unclear, especially sincebulk gold is not able to activate oxygen. It has, among others,been suggested, that the crucial factors are (given are onlyexemplary references, since the body of work is too large togive a complete overview): the oxidation state of the goldspecies [4–8], the size of the gold particles [9, 10], the length ofthe perimeter between gold particles and support [11, 12], thenumber of corner and edge sites on the gold particles [13, 14],the presence of bilayer gold species [15–18] and/or thereducibility of the support [19]. One should note, though, thatnot all of these effects can be clearly discriminated. Forinstance, small particles typically have a higher number of stepand kink sites compared to the same amount of gold present aslarger particles, and often different factors are not clearlydiscriminated in the literature. Moreover, for several supportmaterials the reaction is rather sensitive to the water content ofthe feed [20], which complicates the situation even more.


2 F. Schuth: Colloidal deposition as method to study the activity of gold catalystsp

hys

ica ssp st

atu

s

solid

i b

The sensitivity of the catalytic performance to the detailsof the synthesis is certainly one factor which contributes tothe inconsistencies in the literature. To give one example:normally the most active Au/TiO2 catalysts are synthesizedby deposition–precipitation or by co-precipitation. These arerather complex synthetic protocols where slight deviations inconditions can lead to strong variation in the resultingmaterial. In a systematic study on the reproducibility of suchsyntheses [21] it was found that the concentration of theNaOH used for precipitation is of strong influence, lowconcentration helps to avoid locally increased pH whichleads to inhomogeneity. Moreover, reaching always iden-tical final volume after the deposition is an importantparameter, as is the exact conditions of the drying process.Since factors like these may vary quite substantially betweenlaboratories, and are not always – in fact, rarely –communicated in detail in publications, this explains someof the inconsistencies reported in the literature. In addition,there are factors relevant during the testing, such as moisturecontent, see above.

Nevertheless, even if results are not always directlycomparable, it is clear that the nature of the support does havean influence on the performance of gold catalysts in CO-oxidation. In the following, this influence will be discussed,with special emphasis on sets of catalysts synthesized viacolloidal deposition which allows a clearcut analysis ofsupport effects.

2 The support influence Already in the very firststudy of Haruta et al. [3] on active catalysts for CO-oxidation, differences in activity for different supportmaterials were observed. However, in these early studiespossible differences in the gold particle sizes were not fullyaccounted for and thus clear-cut attribution of the differencesto the nature of the support are not possible. Numerousdifferent materials have been studied as supports for goldcatalysts which were then tested in CO-oxidation. Unfortu-nately, as stated above, syntheses for the production ofactive catalysts are often quite sensitive to conditions, and,moreover, if comparative studies are carried out, differentsupport materials do not have identical synthetic conditionsfor the synthesis of the most active catalyst. In an extensivestudy which was focussed on the understanding of thesupport effect, we had studied ten different supportmaterials, from which catalysts were perpared by depo-sition–precipitation [22]. The pH for the synthesis ofthe most active catalyst for each support differed to someextent – which is not surprising, since the isoelectric point ofthe oxide used will influence the deposition process, a resultwhich was later corroborated in the study of Moreau andBond [23]. Thus, there is an interplay between the charge ofthe gold precursor (HAuCl4 in this study) and the surfacecharge of the oxide onto which the gold is deposited at therespective deposition pH.

In spite of the existence of this complication, supportmaterials, which are suitable to induce high catalytic activityin CO-oxidation, can be identified. While categorizing


support materials in different groups in order to understandtheir behavior, such as reducible or non-reducible supports[19], initially helped in guiding research, it appears by nowclear that under the right conditions basically any supportcould lead to highly active catalysts, even if differences exist.‘‘Highly active,’’ however, is not easily defined, sincecatalysts are not always studied under identical conditions,are prepared with different gold loadings, or are analyzedin different temperature regimes. CO concentrations aremostly set at 1% with only few exceptions, but oxygenconcentration varies between 0.5% and approximately 20%,i.e., synthetic air. Space velocities used vary between around10 000 ml g�1

cat h�1 and 600 000 ml g�1cat h�1, i.e., conversions

at a certain temperature cannot be directly compared.Moreover, activities are mostly not determined underdifferential conditions, so that it is difficult to calculatemeaningful reaction rates. In the following discussion ofsupport materials which lead to active catalysts we willthus quote conversions together with the correspondingconditions, or rates, expressed as mmolCO h�1 g�1

cat at atemperature of 298 K. High activity will be defined as at least75% conversion at a space velocity of 60 000 ml g�1

cat h�1 with1% CO in the feed at a temperature of 298 K, whichcorresponds to a rate of approximately 20 mmolCO h�1 g�1

cat .It is sometimes attempted to quote activities as turnoverfrequencies, but while this is perhaps justified for singlecrystal studies, where the number of atoms on a unit surfaceis known, it is somewhat questionable for supportedcatalysts, where the number of surface atoms can only beestimated, and where it is unclear, whether all atomscontribute equally to activity. In fact, for most of thediscussed explanations of high oxidation activity, forinstance, the number of defects or the special activity ofperimeter sites, the latter assumption should definitely nothold. If all surface atoms would contribute to the activity ofthe catalyst, the rate quoted above corresponds to a TOF ofapproximately 1 s�1 for a catalyst with a gold loading of 1%and a dispersion of 10%. Figure 1 shows the elements in theperiodic table of which defined compounds have led tocatalysts with high CO-oxidation activity as defined above.

Already in Haruta’s seminal publication [3], a-Fe2O3,Co3O4, and NiO had been identified as suitable supports forhighly active catalysts which showed – at least for severalminutes to hours – full conversion of CO at a temperature of�70 8C and a space velocity of 20 000 ml g�1

cat h�1. Other ironoxides were found to be suitable supports as well, such asFe3O4 [24] or FeOx [18, 25]. Other late transition metaloxides can also lead to the formation of catalysts with highactivity. Au/a-Mn2O3, synthesized by deposition–precipi-tation, showed full CO conversion at space velocities of60 000 ml g�1

cat h�1 for 1% CO/20% O2 mixtures at approxi-mately room temperature (1% gold loading) and even down toabout�60 8C (5% gold) [26]. Also co-precipitated Au/MnOx

showed full conversion appreciably below room temperature,albeit at lower space velocity of 10 000 ml g�1

cat h�1 [27].However, this depends – as always – on the mode ofpreparation. In other studies lower activity was reported.

www.pss-b.com

Phys. Status Solidi B (2013) 3

Feature

Article

Figure 1 (online color at: www.pss-b.com) Elements which lead to high activity in CO-oxidation, if oxides (or fluorides or phosphates) areused as supports (red). Elements with somewhat lower but still appreciable activity are marked in blue. Anions leading to active catalysts inturquoise. Missing is cerium, which leads to high activity, and some of the lanthanide phosphates (see text).

The catalysts described there did not reach the thresholdactivity as defined above, with, for instance, rates of only0.041 mmolCO h�1 g�1

cat at room temperature [28]. ZnO isanother late transition metal support leading to catalysts withhigh activity, although such catalysts seem to be somewhatless active than systems based on the previously mentionedoxides. Au on ZnO nanorods showed a conversion of 85% ata space velocity of 30 000 ml g�1

cat h�1 at gas concentrations of1% CO/6% O2 [29], so this catalyst is at the border of highactivity as defined above. Rather similar maximum activitieswere reported by Carbineiro et al. [30]. In contrast to moststudies, these authors used 5% CO, and the best catalyst hadan activity of 20 mmolCO h�1 g�1

cat .The most often studied supports are probably the oxides

of the early transition metals, most notably TiO2 in all itsallotropic forms. In many reports, full conversion for typicalconditions (1% CO, space velocity of 20 000 ml g�1

cat h�1 orhigher) have been reported for temperatures below roomtemperature [11, 22, 31–33]. Also ZrO2 can induce highcatalytic activity, although for ZrO2 [22, 34] activity overallseems to be somewhat lower than for TiO2 as support. Whilein some papers relatively high activities are reported, otherpublications quote appreciable CO-conversion only close to100 8C. ZrO2 has also been used as a component of mixedmetal ceria–zirconia systems [35] which are well known inautomotive exhaust catalysis. Fifty percent conversion wasreached at approximately room temperature (conversionhad to be estimated from a figure), but space velocity was240 000 ml g�1

cat h�1 and only slightly overstoichiometricoxygen concentration was used (0.6% O2, 1% CO), so thatthis catalyst can be categorized as high activity system.Cerium oxide [36, 37] is a suitable support for highactivity gold catalysts in any case. For instance, a rateof approximately 10 mmolCO h�1 g�1

cat was reported for atemperature of 5 8C over gold on nanocrystalline ceria[37]. Especially ceria, either pure or modified by various

www.pss-b.com

additives is an intensively investigated support during thelast few years, with several ten publications per year, due toits favorable properties in preferential oxidation of CO inhydrogen rich feeds and in the water gas shift reaction [38].The other lanthanide oxides, which have been tested, seemto lead to less active catalysts. Cationic gold deposited onLa2O3 reached 10% conversion at a space velocity of30 000 ml g�1

cat h�1 at room temperature and 1.5% CO [39].Also Y2O3, while reasonably active, did not reach theactivities of the high activity systems in CO-oxidation: a rateof close to 20 mmolCO h�1 g�1

cat was only reached at 50 8C[40]. Even lower activity was observed for Pr6O11 assupport [41].

While high activity of supported gold catalysts was firstdiscovered for transition metal oxides as supports, maingroup metal oxides, including those which have very stableoxidation states and are difficult to reduce, such as MgO orAl2O3, were subsequently also found to be suitable. MgO orMg(OH)2 were reported already in the early publications ofHaruta’s group as supports leading to high activity [32], andthis high activity has been confirmed in subsequent studies[42, 43]. A 0.7% Au/Mg(OH)2 catalyst synthesized bycolloidal deposition [43] had an extremely high activity,reaching 100% conversion of a feed of 1% CO in syntheticair at a temperature of �60 8C and a space velocity of400 000 ml g�1

cat h�1. Also the lighter homolog, Be(OH)2, is asuitable support, but it appears to have been studied only inan early investigation of Haruta et al. [32], and no detailswere described. This work does not seem to have beenpicked up again, probably due to the toxicity of berylliumcompounds. Al2O3 is a support for which high activity wasreported in a number of publications. Chen and Yeh [44]observed full conversion at temperatures below 0 8C atspace velocities of 20 000 ml g�1

cat h�1 and 1% CO in air. Inour comparative study [22], alumina supported catalystswere synthesized which had 50% conversion at �44 8C at



hys

ica ssp st

atu

s

solid

i b

20 000 ml g�1cat h�1 for identical gas concentrations. Ga2O3 on

the other hand, seems to be less suitable [45]. For SiO2 assupport various different activity levels have been reported.While in most publications, the activity was found to be farbelow the values reached over catalysts prepared on titaniaor iron oxide, there are studies in which highly active goldon silica are described [45]. However, gold on silicacatalysts appear to be only active if they are synthesizedby methods deviating from the conventional pathwaysfor the synthesis of active gold catalysts (co-precipitationor deposition–precipitation), such as use of a cationicAu(en)3þ

2 precursor [46] or deposition of the gold by aCVD process [47] where full conversion of 1% CO wasreached at �15 8C at a space velocity of 20 000 ml g�1

cat h�1.Another option is the use of silica ordered mesoporousmaterials where a high dispersion of gold can be maintaineddue to the confinement in the pores, where rather similaractivities were reached [48]. One key in the synthesis ofgold on silica is the use of cationic gold species due to thenormally negative surface charge of the silica, which preventsdeposition of anionic species; alternatively, the surface has tobe functionalized to allow favorable interaction with thenegative chloroaurate precursor [49]. Tin oxide has also beenstudied, and in some reports activities just at the thresholdperformance as defined above have been reached. Moreauand Bond have optimized the synthesis of Au/SnO2, andreached 50% conversion at 26 8C at a space velocity of66 000 ml g�1

cat h�1 and a feed concentration of 0.5% CO.Similar results were reported in a subsequent publication bythe same authors [23, 50]. However, in other studies theactivities of SnO2 supported gold catalysts were appreciablylower [51]. In several publications the SnO2 supported goldcatalysts were not studied for the catalytic performance,but rather for the electrocatalytic [52] activity or for theiruse as sensor materials [53], application fields in whichalso gold on indium oxide [54] or on indium tin oxide [54]has been evaluated. However, no high activity catalystsseem to have been prepared using indium oxide as support. Ofthe other main group elements Bi2O3 has been studied [23],but its activity was very low. However, many main group –and transition metal – elements have been used as promotorsfor gold-based catalysts [45, 55, 56], but in these cases it isoften difficult to uniquely attribute possible improvements inactivity to a single cause.

In addition to doped catalysts, also defined ternaryoxides have successfully been employed as support materialsfor active catalysts, especially spinel-type oxides. MgAl2O4,the prototypical spinel, led to active gold catalysts forwhich 75% conversion of 1% CO in air at a space velocity of80 000 ml g�1

cat h�1 was reached slightly below 0 8C [57].A MgFe2O4 support lead to even higher activity (100%conversion at �40 8C at 80 000 ml g�1

cat h�1), but rapiddeactivation was observed [58]. Several chromium-spinels(MCr2O4 with M¼Co, Mn, Fe, Mg, Cu) were studied as wellas supports for gold, but while reasonable activity wasreported, the best catalyst in this series, Au/CoCr2O4,reached slightly above 80% CO conversion at 35 8C


at a rather low space velocity which is given asGHSV¼ 2400 h�1 and which cannot be converted to theunits used throughout the present contribution, since theinformation in the reference is incomplete [59]. The onlymoderate activity of Au/CuxCrYOz catalysts in CO-oxidationwas recently confirmed [60].

Oxygen availability from the support oxide has oftenbeen claimed as a decisive feature for high catalytic activityin CO-oxidation. The fact that easily reducible oxides, suchas the iron oxides, SnO2, CeO2, or ZnO as well as ratherstable oxides, such as Al2O3 and even SiO2 can lead to highactivity catalysts suggests that there are definitely additionalfactors contributing to high activity. Non-oxidic supports aretherefore especially interesting in this respect. However, oneshould keep in mind that even catalysts which nominally donot contain oxygen may have a surface layer in which highamounts of oxygen are present, and which might participatein the catalytic reaction.

BaCO3 was one of the first non-oxidic supports studiedfor CO-oxidation [61]. While the activity of this system washigh, it did not reach the levels observed for many of theoxidic supports. This catalyst reached 100% conversion atroom temperature at 0.5% CO concentration and a spacevelocity of 12 000 ml g�1

cat h�1. LaPO4, on the other hand, wassuccessfully used for the preparation of highly active goldcatalysts [62]. The best catalyst reached a rate of approxi-mately 50 mmolCO h�1 g�1

cat at�71 8C, one of the most activesystems reported, yet, although one has to mention that at8.3 wt% the gold content was relatively high. The presenceof La2O3 was carefully excluded in this study, and sincethe oxygen from the phosphate groups is tightly bound,participation of oxygen from the support appears to be notvery probable. In a subsequent publication, a number of otherphosphates were employed as supports, and for several ofthem high activity has been observed. 50% conversion wasreached for a space velocity of 44 400 ml g�1

cat h�1 at below25 8C for the phosphates of calcium, iron, praseodym,samarium, europium, and holmium. However, XRD patternsof the catalysts did not in all cases prove the presence ofdefined phosphate phases.

Carbon is another type of support for which participationof oxide from the support is not probable – although manytypes of carbon have appreciable contents of oxygen in formof different functional groups. Bulushev et al. [63] describedgold on activated carbon fibers as highly active catalysts forCO-oxidation. The best catalysts were at the high activitythreshold, as defined above. A rate of 14 mmolCO h�1 g�1

cat atroom temperature was reported. However, one has to notethat the catalysts contained appreciable amounts of ironoxide (several %) which may interact with the gold species.Moreover, rapid deactivation was observed.

We have used CaF2 as support for gold nanoparticles andstudied the performance of the resulting catalysts in CO-oxidation [64]. The best catalysts synthesized in this studywere on the borderline for high activity as defined here.Forty-five percent conversion was reached at 25 8C fora space velocity of 80 000 ml g�1

cat h�1 and a gas phase

www.pss-b.com


Feature

Article

Figure 2 CaF2 synthesized by nanocasting from CMK-3 (left) and Au/CaF2 (right) [64].

concentration of 1% CO in synthetic air, corresponding to16 mmolCO h�1 g�1

cat . The CaF2 support for this study hadbeen obtained by nanocasting from ordered mesoporouscarbon CMK-3, which resulted in high surface area CaF2

(BET surface area 144 m2 g�1). Colloidal deposition resultedin a gold loading of 1%. Figure 2 shows TEM images of theCaF2 support and the Au/CaF2 catalyst.

While it cannot be fully excluded that some oxygen ispresent in the surface of the CaF2, a Mars-van Krevelen typemechanism for the oxygen activation, which has occasion-ally been evoked to explain the high CO-oxidation activityof gold-based catalysts appears to be highly improbable forthis type of support.

3 Colloidal synthesis for elucidation of thesupport effect The previous discussion has shown thatcatalysts with high activity for CO-oxidation can besynthesized with a wide variety of different supportmaterials, ranging from supports which easily change theiroxidation state over very stable supports to supports whichdo not contain oxygen at all or only as impurities. However,within each group and for each support material, typically awide range of activities has been reported, since activity doesnot only depend on the support material, but on many otherfactors, such as synthetic protocol, resulting particle size,possible additives or impurities in the catalysts, and so on.It is therefore very difficult to clearly identify the effect ofdifferent support materials.

Colloidal deposition, i.e., the synthesis of colloidal goldnanoparticles, followed by deposition of these pre-formednanoparticles on a support material, is a method to excludemany factors which are difficult to control in co-precipitationor deposition–precipitation synthetic protocols. The methoddecouples the gold particle formation from the supportingprocess which is highly important for the identificationof the support influence. For all other synthetic protocols,the particle formation on the support is a heterogeneousnucleation process, since from solution the gold particles

www.pss-b.com

are either formed with the support or on the support. Forheterogeneous nucleation, however, the nature of thesubstrate on which the forming particle nucleates governsthe properties of the particle. Thus, different supports willlead to different types of particles. This influence can beeliminated – at least to some extent – by colloidal deposition.

The first gold-based catalysts synthesized by colloidaldeposition were independently reported by the groups ofHaruta and coworkers [65] and Baiker and coworkers [66,67]. The Haruta-group used 5 nm sized gold colloids,prepared by evaporation into a-terpineol, which weredeposited onto a titania support. For comparison, a catalystwas synthesized by deposition–precipitation with the samesupport. In this study, the colloidally synthesized catalystwas substantially less active than the reference catalyst, andonly after heat treatment at 600 8C the activities of bothcatalysts became rather similar. The authors explained thisbehavior by the weak interaction between gold and titania inthe colloidally deposited catalysts, which improved uponheat treatment. The Baiker-group deposited smaller goldcolloids (2 nm) on titania and zirconia supports and foundpronounced differences between titania and zirconia inthe as-made catalysts, although the particle sizes and thegold-particle synthesis were identical for both supports,indicating a substantial influence of the support. However,even the more active titania supported catalyst did notreach the activity of the samples synthesized by deposition–precipitation. Only one of the titania-based catalysts reached100% conversion at 27 8C, at CO and O2 concentrations of2500 ppm each, and at a space velocity of 9000 ml g�1

cat h�1.Colloidal deposition has been used for the synthesis of

high activity g-alumina based catalysts by Wen et al. [68].The resulting catalysts were very active, the best onesreaching full conversion at�20 8C with 2% CO in syntheticair at a space velocity of 30 000 ml g�1

cat h�1. Colloidaldeposition was also the method of choice for a series ofcatalysts prepared by the Dai-group [69] on SiO2, TiO2, andC. The samples were, however, treated with KMnO4 or



hys

ica ssp st

atu

s

solid

i b

Table 1 Survey of different catalysts synthesized by colloidal deposition and temperature for 50% conversion. Gold loading 1 wt%(except for Mg(OH)2 where it was 0.7 wt%), tested at 1% CO, 21% O2 and space velocity of 80 000 ml g�1

cat h�1.

support T50% (8C) refs.

TiO2 70% anatase, 30% rutile (P25, Degussa) �14 [70]ZnO (AC-45, Bruggemann) 46 [70]g-Al2O3 (Puralox SBa 200, Sasol) �11a [70]ZrO2 (self-prepared) 89 [70]MgAl2O4 (self-prepared) �9 [57]TiO2 anatase (AK 350, Tronox) 5 [71]TiO2 rutile (SP05/16, Tronox) 5 [71]a-Fe2O3, different shapes and sizes (self-prepared) 42–93 [72]MgO (self-prepared)b <�90 [43]Mg(OH)2 (self-prepared)b <�90 [43]FeOx (self-prepared) �14 [25]MgFe2O4 (self-prepared) �32 (initially higher) [58]CaF2 (self-prepared) 30 [64]

aThe alumina based catalysts showed ill reproducibility, temperature given is for the best ones; bUnusual conversion curve with decreasing conversion with

increasing temperature in a certain interval.

Fenton’s reagent in order to remove the capping ligands ofthe gold particles. This leaves manganese or iron oxidesbehind, and thus there is an additional factor which isprobably responsible for increased activity.

We have used the colloidal deposition method in anumber of studies for the synthesis of gold catalysts onvarious supports. In these studies colloidal gold particlessynthesized by the same method were used, and catalyticevaluation was carried out under identical conditions as well.There is thus a unique set of samples which can be directlycompared in order to obtain more information on the supportinfluence. Table 1 summarizes the results of the differentstudies we have performed.

All catalysts listed in Table 1 were synthesized by thesame procedure (for details, see Ref. [70]): First, a solution ofcolloidal gold particles is prepared by adding poly(vinylalcohol) (PVA) as protecting agent to a gold solution andvigorous stirring. After 10 min NaBH4 solution (0.1 N) israpidly injected, which leads to formation of an orange-brown solution. Then the support material is added, until thegold colloid is completely adsorbed, indicated by decolora-tion of the solution, resulting in a 1 wt% Au loaded catalyst.The catalyst is recovered by filtration, careful washingwith doubly distilled water, and finally drying in a vacuumdesiccator over P2O5.

All catalysts were tested under identical conditions, i.e.,50 mg catalyst (250–500 mm fraction) placed on quartz woolplug in a flow reactor with 6 mm internal diameter, a flowrate of 67 ml min�1, corresponding to a space velocity of80 000 ml g�1

cat h�1 in a gas with a composition of 1% CO insynthetic air. Reproducibility of T50% values with differentcatalyst samples from the same batch was found to bewithin �3 8C. Experiments were typically carried outunder transient conditions. This means that the reactor withcatalyst loaded was cooled down under a flow of nitrogento the base temperature (�40 8C for most systems, but lower


for the very active catalysts), then the temperature wasramped up to the final temperature, typically to 200 8C,at a heating rate of 2 8C min�1. Agreement between thesetransient experiments and steady state experiments wasrepeatedly checked, and data typically agreed within a few %conversion for identical temperatures, with some exceptions,such as ZnO or MgFe2O4. If unusual transient curves wererecorded, the catalyst performance was always checked instationary experiments.

The colloidal synthesis as used in our studies results inthe deposition of PVA-coated gold nanoparticles on thesupports. The PVA has to be removed before the activity ofthe gold catalysts can be observed. The first temperatureramp typically thus does not show high activity, fullconversion is reached only at temperatures around 150–200 8C. In subsequent runs, depending on the support muchhigher activities are then observed, conversion curves inthese runs are highly reproducible. Figure 3 shows the typicalresult of four subsequent temperature ramps for the samecatalyst. The light-off temperatures in the first vary to someextent for different supports (shifts by up to about 50 8C),so that the support properties may also influence theremoval of the PVA. However, there is no correlation withthese shifts and the final activity of the catalysts, and thusthis effect is probably only of minor importance.

Special cases are also Mg(OH)2 and MgO, since thesesupports gave catalysts with extremely high activity at verylow temperature, then falling conversion with increasingtemperature and finally again a ‘‘normal’’ increase ofconversion with temperature to full conversion (Fig. 4).The curves plotted here are obtained after the activation runto 275 8C (not shown here), so the 1st run corresponds in factto the 2nd run in Fig. 3.

The data listed in Table 1 clearly demonstrate howimportant the support indeed is. For the same type of goldparticles with identical sizes and deposited by the same

www.pss-b.com


Feature

Article

Figure 3 (online color at: www.pss-b.com) Typical conversionversus temperature curves for supported PVA-stabilized gold col-loids, in this figure supported on P25 titania. In the first run, 100%conversion is typically reached at temperatures between 150 and200 8C, since the PVA stabilizer has to be removed first.

Figure 4 (online color at: www.pss-b.com) Unusual conversionvs. temperature curves for Au/Mg(OH)2 prepared by colloidaldeposition for two different space velocities (from Ref. [43] withkind permission).

method on the support material, T50% varies by approxi-mately 200 8C (Mg(OH)2 and ZrO2 as supports)! This uniqueset of samples, for which many synthesis variables havebeen eliminated, and which was analyzed under identicalconditions, thus clearly proves that there is a substantialeffect of the support material on the catalytic properties ofgold catalysts for CO-oxidation. For a related, but differentsystem, Au@ZrO2 or the Au@C yolk-shell catalysts, thiseffect could be shown as well. Doping the surface of thegold in Au@ZrO2 by small amounts of titanium during thesynthesis led to marked increase in catalytic activity [73],and while the Au@C catalysts had very little activity, it was

www.pss-b.com

substantially higher as soon as oxide (zirconia in this case)was present in the shell [74].

As stated above, the activation process to remove thePVA stabilizer shell does not seem to be of major influence.Haruta in his early publication using colloidal depositionhad shown that activity substantially increased after heattreatment of the colloidally deposited catalysts. Thesecatalysts did not have a stabilizer shell since they had beensynthesized by evaporation of gold particles in inert gasatmosphere. After activation at 200 8C, the activity was stillvery poor, while in our studies heat treatment to 250 8C wastypically sufficient to induce high activity. Increasing thepre-treatment temperature further, as was done in the studyof different titania polymorphs as support [71], either did notlead to changes in the low temperature activity, or resulted incatalysts with decreased activity. Thus, it is most probablynot the interaction with the support which could developduring the activation, but indeed removal of the PVAstabilizer.

Interestingly, there is no clear correlation between thereducibility of the supports or their oxygen content. Activecatalysts were obtained with supports, such as MgO,Mg(OH)2, or MgAl2O4, for which high oxygen availabilityis not expected, as well as with supports which can supplyoxygen, such as iron oxide or titania. Even CaF2 whichcontains oxygen only as impurity – if at all – resulted in theformation of catalysts with reasonable activity, even if theywere not as active as the best catalysts found in our studies.

4 Possible origin of support influence While inthe preceding section it has clearly been demonstrated thatthe support has a substantial effect on the activity of thecatalysts in CO-oxidation, the mode in which the catalyticactivity is influenced is unclear. At least three differentpossibilities have been identified how the support materialmay participate in the generation of active catalysts.

4.1 Shape changes of gold particles The col-loidal particles in solution or deposited on a carbon coatedcopper grid for TEM analysis appear to be round; no distinctshapes can be identified. This changes, if the particles aresupported on the oxides. Although not all catalysts could beanalyzed by high resolution TEM (for instance, the Mg(OH)2

dehydrated at the acceleration voltage of 200 kV so that nowell resolved images could be recorded), the ones in whichsufficient resolution was achieved showed the developmentof a large contact area between gold and support, and it seemsthat the gold colloids are located on the supports mostly assubstantially facetted particles of approximately hemispheri-cal shape. Figure 5 shows this for three examples.

This facetting does not require treatment at elevatedtemperatures, the images were recorded for as preparedsamples. Obviously, the gold particles are wetting thesupport, and the interfacial energy is sufficiently high toinduce the shape transformation. The particles clearly showthe presence of edges and corner sites, and the supportinduced shape change could lead to different concentration



hys

ica ssp st

atu

s

solid

i b

Figure 5 Facetting of gold particles supported on ZrO2 (left), TiO2 (middle) and MgFe2O4 (right) (reproduced with kind permissionfrom Ref. [70] (left and middle) and Ref. [58] (right)).

Figure 6 Catalytic activity (expressed as absolute amounts ofCO converted during CO/O2 pulses) of colloidally deposited sup-ported gold catalysts versus oxygen storage capacity. Reproducedwith kind permission from Ref. [78].

of such – or other – defects, which could influence theactivity. However, the shapes do not seem to be verydifferent for the differently active support materials. InFig. 5, images are given for a relatively low activity catalyst(ZrO2) and two active catalysts (TiO2 and MgFe2O4), and theshapes which have developed appear to be rather similar.Thus, although the differences in shape could lead to activitydifferences, there are probably also other factors contribut-ing.

In a somewhat different model system, Au@ZrO2 yolk-shell catalysts, the influence of defects in the gold particleson catalytic activity could be proven, though. The sampleswere synthesized by the method introduced by Arnal et al.[75]. In this approach, gold colloids are coated with silicaspheres by a Stober-type process, then a thin layer of zirconiais coated on the silica. After thermal treatment, the silica isleached with NaOH or HF, so that a Au@ZrO2 yolk-shellcatalyst results. The gold colloids in these catalysts aresubstantially bigger than in the samples discussed so far, i.e.,16 nm. They can be leached with cyanide to sizes down toabout 5 nm, and for the size range between 5 and 16 nm theactivity per surface gold atom was shown to be basicallyindependent of particle size [76]. With different pretreatmentprotocols, it was attempted to induce different defectconcentrations in such gold-catalysts [77]. Detailed lineprofile analysis of the XRD patterns allowed to discriminatedifferent defect types, and it was found that the catalyticactivity correlated with the dislocation density of the goldparticles. The concentration of other types of defects (twinfaults and stacking faults) did not show obvious correlationswith catalytic activity. Nevertheless, the results of this studysuggest that the defect concentration can influence catalyticactivity, and thus the interaction with support materials,which leads to facetting and possibly defect formation, caninfluence catalytic performance.

4.2 Oxygen availability from the support Theavailability of oxygen from the support material has oftenbeen claimed as an important factor influencing the activityof the catalysts in CO-oxidation ([19] with further refer-ences). It thus seemed worthwhile to study the oxygenavailability for selected catalysts prepared by colloidaldeposition. The availability of oxygen from the support was


analyzed using a TAP reactor in J. Behm’s laboratory [78].For the samples studied (Au/TiO2, Au/ZrO2, Au/ZnO2, andAu/Al2O3), the catalytic activity of the samples correlatedwell with the oxygen storage capacity of the materials(Fig. 6), in line with results from earlier studies on a morelimited set of conventionally prepared samples [79, 80].Moreover, for Au/ZnO catalysts with different gold load-ings, synthesized by colloidal deposition with the identicalcluster types as used in the other investigations reported inthis chapter, the amount of oxygen vacancies, as detectedwith N2O frontal chromatography, scaled with the number ofatoms at the gold-zinc oxide perimeter sites, suggesting thatthis region in the catalysts is important in governing catalyticactivity and supplying oxygen [81]. However, for this set ofsamples, the activity in CO-oxidation has not been analyzed;only methanol synthesis was studied.

Unfortunately, the high activity observed for the aluminasupported catalyst reported earlier by us, which was alreadyin the first publication found to be ill reproducible [70], couldas yet not be reproduced for the TAP reactor study. It wouldbe very interesting to see whether a high activity aluminasupported catalyts has a high oxygen storage capacity as

www.pss-b.com


Feature

Article

well. The same holds for the Mg(OH)2 or the MgO supportedcatalysts, which have also not been analyzed with respectto their oxygen storage capacity, yet. These experimentsare ongoing and results will be published in a separatepublication.

4.3 Active oxygen species As stated, the Mg(OH)2

and MgO supported catalysts have not been analyzed, yet,in the TAP reactor setup with respect to their oxygenstorage capacity. However, for these catalysts two differentmechanisms seem to be operating, one responsible forthe very high activity at low temperatures and anothermechanism acting at higher temperatures above 100 8C.The magnesium based supports seem to be able to provide aspecific kind of oxygen – possibly molecularly adsorbedspecies, which have repeatedly been suggested as leading tohigh catalytic activity at the interface between gold andsupport, albeit for titania as support [12, 82, 83]. We haveattempted to quantify the amount of reactive oxygenspecies by titrating pre-oxidized catalysts with CO. At lowtemperatures, substantially higher amounts of reactiveoxygen were detected on the Mg(OH)2 supported catalyststhan at high temperatures, in contrast to Au/TiO2 catalysts.Estimating the amounts of oxygen detected, one has toconclude that the majority of the oxygen species are locatedon the support material. However, this still needs to beconfirmed in more detailed investigations using the TAPreactor, and in addition, studies using in situ vibrationalspectroscopy are under way to identify the nature of theactive oxygen species.

5 Conclusions The analysis of a well comparable setof gold catalysts synthesized by colloidal deposition withidentical gold colloid precursor solutions has clearlydemonstrated, that the support exerts a strong influence onthe catalytic performance of supported gold catalysts inCO-oxidation. Differences of approximately 200 8C forT50% for different supports were observed. Several possiblereasons for these differences were identified, i.e., thedifferent facetting and different defect concentrations ofthe gold colloids supported on different supports, differentamounts of support oxygen which is available for thereaction, or the presence of active molecular oxygen on thesupport which could be responsible for the very high activityin the low temperature regime of Mg(OH)2 supported goldcatalysts. For other catalyst systems, also other reasons forhigh activity have been suggested, such as the bilayer specieswhich are assumed to have especially high catalytic activity[17, 18]. However, we have carefully checked a highly activeAu/FeOx catalyst synthesized by colloidal deposition forthe presence of such bilayer structures [25], but no indicationfor bilayer species were found, which suggests that in ourcatalysts other factors are responsible for high catalyticactivity. From all evidence collected in our studies, inaddition to the evidence supplied in the literature, a uniqueexplanation for the extraordinary catalytic activity ofgold-based catalysts appears to be improbable. It is probably

www.pss-b.com

rather a combination of different factors, which simul-taneously can contribute to enhancing the catalytic proper-ties of gold nanoparticles, with different factors dominatingfor different systems, a conclusion which had also beenreached based on a theoretical study by the Norskov group[84].

Acknowledgements The work on gold-based catalystshas been funded by the DFG SFB 558, which is gratefullyacknowledged, in addition to the basic funding provided by theInstitute. I would also like to acknowledge the dedicated work of thePh.D. students and post-docs who have worked on these projects andwhose names can be found in the various references cited.

References

[1] G. J. Hutchings, J. Catal. 96, 292 (1985).[2] B. Nkosi, N. J. Coville, and G. J. Hutchings, Appl. Catal. 43,

33 (1988).[3] M. Haruta, T. Kobayashi, H. Sano, and N. Yamada, Chem.

Lett., 405 (1987), DOI: 10.1246/cl.1987.405.[4] E. D. Park and J. S. Lee, J. Catal. 186, 1 (1999).[5] M. Haruta, J. Catal. 36, 153 (1997).[6] J. C. Fierro-Gonzales and B. C. Gates, Chem. Soc. Rev. 37,

2127 (2008).[7] L. Delannoy, N. Weiher, N. Tsapatsaris, A. M. Beesley,

L. Nchari, S. L. M. Schroeder, and C. Louis, Top. Catal.44, 263 (2007).

[8] B. Yoon, H. Hakkinnen, U. Landman, A. S. Worz, J. M.Antonietti, S. Abbet, K. Judai, and U. Heiz, Science 307, 403(2005).

[9] M. Haruta, Catal. Today 36, 153 (1997).[10] N. Lopez, T. V. W. Janssens, B. S. Clausen, Y. Xu, M.

Mavrikakis, T. Bligaard, and J. K. Norskov, J. Catal. 223, 232(2004).

[11] M. Haruta and M. Date, Appl. Catal. A – Gen. 222, 427(2001).

[12] I. X. Green, W. Tang, M. Neurock, and J. T. Yates, Jr.,Science 333, 736 (2011).

[13] M. Mavrikakis, P. Stoltze, and J. K. Norskov, Catal. Lett. 64,101 (2000).

[14] N. Lopez, J. K. Norskov, T. V. W. Janssens, A. Carlsson,A. Puig-Molina, B. S. Clausen, and J.-D. Grunwaldt, J. Catal.225, 86 (2004).

[15] M. Valden, X. Lai, and D. W. Goodman, Science 281, 1647(1998).

[16] M. Valden, S. Pak, X. Lai, and D. W. Goodman, Catal. Lett.56, 7 (1998).

[17] M. S. Chen and D. W. Goodman, Science 306, 252 (2004).[18] A. A. Herzing, C. J. Kiely, A. F. Carley, P. Landon, and G. J.

Hutchings, Science 321, 1331 (2008).[19] M. M. Schubert, S. Hackenberg, A. C. van Veen, M. Muhler,

V. Plzak, and R. J. Behm, J. Catal. 197, 113 (2001).[20] M. Haruta, Faraday Discuss. 152, 11 (2011).[21] W. C. Li, M. Comotti, and F. Schuth, J. Catal. 237, 190 (2006).[22] A. Wolf and F. Schuth, Appl. Catal. A – Gen. 226, 1 (2002).[23] F. Moreau and G. C. Bond, Catal. Today 114, 362 (2006).[24] H. Yin, C. Wang, H. Zhu, S. H. Overbury, S. Sun, and S. Dai,

Chem. Commun., 4357 (2008).[25] Y. Liu, C. J. Jia, J. Yamasaki, O. Terasaki, and F. Schuth,

Angew. Chem. Int. Ed. 49, 5771 (2010).



hys

ica ssp st

atu

s

solid

i b

[26] L. C. Wang, X. S. Huang, Q. Liu, Y. M. Liu, Y. Cao, H. Y.He, K. N. Fan, and J. H. Zhuang, J. Catal. 259, 66 (2008).

[27] R. M. Torres-Sanchez, A. Ueda, K. Tanaka, and M. Haruta,J. Catal. 168, 125 (2199).

[28] S. A. C. Carabineiro, S. S. T. Bastos, J. J. M. Orfao, M. F. R.Pereira, J. J. Delgado, and J. L. Figueiredo, Catal. Lett. 134,217 (2010).

[29] X. Liu, M. H. Liu, Y. C. Luo, C. Y. Mou, S. D. Lin, H. Cheng,J. M. Chen, J. F. Lee, and T. F. Lin, J. Am. Chem. Soc. 134,10251 (2012).

[30] S. A. C. Carabineiro, B. F. Machado, R. R. Bacsa, P. Serp,G. Drasic, J. L. Faria, and J. L. Figureido, J. Catal. 273, 191(2010).

[31] G. R. Bamwenda, S. Tsubota, T. Nakamura, and M. Haruta,Catal. Lett. 44, 83 (1997).

[32] M. Haruta, S. Tsubota, T. Kobayashi, H. Kageyama, M. J.Genet, and B. Delmon, J. Catal. 144, 175 (1993).

[33] F. Moreau, G. C. Bond, and A. O. Taylor, Chem. Commun.,1642 (2004).

[34] Y. C. Hong, K. Q. Sun, K. H. Han, G. Liu, and B. Q. Xu,Catal. Today 158, 415 (2010).

[35] E. Del Rio, G. Blanco, S. Collins, M. Lopez Haro, X. Chen,J. J. Delgado, J. J. Calvino, and S. Bernal, Top. Catal. 54, 931(2011).

[36] Tana, F. Wang, H. Li, and W. Chen, Catal. Today 175, 541(2011).

[37] S. Carrettin, P. Concepcion, A. Corma, J. M. Lopez Nieto,and V. F. Puntes, Angew. Chem. Int. Ed. 42, 2538 (2004).

[38] Q. Fu, A. Weber, and M. Flytzani-Stephanopoulos, Catal.Lett. 77, 87 (2001).

[39] J. C. Fierro-Gonzales, V. A. Bhirud, and B. C. Gates, Chem.Commun., 5275 (2005).

[40] J. Guzman and A. Corma, Chem. Commun., 743 (2005).[41] P. X. Hunag, F. Wu, B. L. Zhu, G. R. Li, Y. L. Wang, X. P.

Gao, H. Y. Zhu, T. Y. Yan, W. P. Huang, S. M. Zhang, andD. Y. Song, J. Phys. Chem. B 110, 1614 (2006).

[42] J. L. Margitfalvi, M. Hegedus, A. Szegedy, and I. Sago, Appl.Catal. A – Gen. 272, 87 (2004).

[43] C. J. Jia, L. Yong, H. Bongard, and F. Schuth, J. Am. Chem.Soc. 132, 1520 (2010).

[44] Y. J. Chen and C. T. Yeh, J. Catal. 200, 59 (2001).[45] J. Vecchietti, S. Collins, J. J. Delgado, M. Malecka, E. del Rio,

X. Chen, S. Bernal, and A. Bonivardi, Top. Catal. 54, 201 (2011).[46] H. Zhu, Z. Ma, J. C. Clark, Z. Pan, S. H. Overbury, and S. Dai,

Appl. Catal. A – Gen. 326, 89 (2007).[47] M. Okumura, S. Nakamura, S. Tsubota, T. Nakamura, M. Azuma,

and M. Haruta, Catal. Lett. 51, 53 (1998).[48] H. Zhu, C. Liang, W. Yan, S. H. Overbury, and S. Dai,

J. Phys. Chem. B 110, 10842 (2006).[49] C. M. Yang, M. Kalwei, F. Schuth, and K. J. Chao, Appl.

Catal. A – Gen. 254, 289 (2003).[50] F. Moreau and G. C. Bond, Catal. Today 122, 215(2006).[51] Q. Ye, J. Wang, J. Zhao, L. Yan, S. Cheng, T. Kang, and

H. Dai, Catal. Lett. 138, 56 (2010).[52] P. Diao, D. Zhang, J. Wang, and Q. Zhang, Electrochem.

Commun. 12, 1622 (2010).[53] T. Kobayashi, M. Haruta, and H. Sano, Chem. Express 4, 217

(1989).[54] J. Y. Wang, P. Diao, D. F. Zhang, M. Yiang, and Q. Zhang,

Electrochem. Commun. 11, 1069 (2009).[55] V. Rodriguez-Gonzalez, R. Zanella, L. A. Calzada, and

R. Gomez, J. Phys. Chem. C 113, 8911 (2009).


[56] R. J. H. Grisel and B. E. Nieuwemhuys, Catal. Today 64, 69(2001).

[57] W. C. Li, M. Comotti, A. H. Lu, and F. Schuth, Chem.Commun., 1772 (2006).

[58] C. J. Jiang, Y. Liu, M. Schwickardi, C. Weidenthaler, B.Spliethoff, W. Schmidt, and F. Schuth, Appl. Catal. A – Gen.386, 94 (2010).

[59] M. Ruszel, B. Grzybowska, K. Samson, I. Gressel, andA. Klisinska, Catal. Today 112, 126 (2006).

[60] I. Sobczak, K. Szrama, R. Wojcieszak, E. M. Gaigneaux, andM. Ziolek, Catal. Today 187, 48 (2012).

[61] H. Lian, M. Jia, W. Pan, Y. Li, W. Zhang, and D. Jiang, Catal.Commun. 6, 47 (2005).

[62] W. Yan, S. Brown, Z. Pan, S. M. Mahurin, S. H. Overbury,and S. Dai, Angew. Chem. Int. Ed. 45, 361 (2006).

[63] D. A. Bulushev, I. Yuranov, E. I. Suvurova, P. A. Buffat, andL. Kiwi-Minsker, J. Catal. 224, 8 (2004).

[64] Y. Liu, Y. Meng, C. J. Jia, C. Weidenthaler, A. H. Lu,M. Comotti, B. Spliethoff, W. Schmidt, and F. Schuth,unpublished.

[65] S. Tsubota, T. Nakamura, K. Tanaka, and M. Haruta, Catal.Lett. 56, 131 (1998).

[66] J. D. Grunwaldt, C. Kiener, C. Wogerbauer, and A. Baiker,J. Catal. 181, 223 (1999).

[67] J. D. Grunwaldt, M. Maciejewski, O. S. Becker, P. Fabrizioli,and A. Baiker, J. Catal. 186, 458 (1999).

[68] L. Wen, J. K. Fu, P. Y. Gu, B. X. Gao, Z. H. Lin, and J. Z. Zhu,Appl. Catal. B – Environ. 79, 402 (2008).

[69] H. Yin, Z. Ma, M. Chi, and S. Dai, Catal. Lett. 136, 209(2010).

[70] M. Comotti, W. C. Li, B. Spliethoff, and F. Schuth, J. Am.Chem. Soc. 128, 917 (2006).

[71] M. Comotti, C. Weidenthaler, W. C. Li, and F. Schuth, Top.Catal. 44, 275 (2007).

[72] G. H. Wang, W. C. Li, K. M. Jia, B. Spliethoff, F. Schuth, andA. H. Lu, Appl. Catal. A – Gen. 364, 42 (2009).

[73] R. Guttel, M. Paul, and F. Schuth, Catal. Sci. Technol. 1, 65(2011).

[74] C. Galeano, R. Guettel, P. Michael, P. Arnal, A. H. Lu, andF. Schuth, Chem. Eur. J. 17, 8434 (2011).

[75] P. Arnal, M. Comotti, and F. Schuth, Angew. Chem. Int. Ed.45, 8224 (2006).

[76] R. Guttel, M. Paul, C. Galeano, and F. Schuth, J. Catal. 289,100 (2012).

[77] A. Dangwal Panday, R. Guttel, M. Leoni, F. Schuth,and C. Weidenthaler, J. Phys. Chem. C 114, 19386(2010).

[78] D. Widman, Y. Liu, F. Schuth, and R. J. Behm, J. Catal. 276,292 (2010).

[79] M. Kotobuki, R. Leppelt, D. Hansgen, D. Widman, andR. J. Behm, J. Catal. 264, 67 (2009).

[80] D. Widman, R. Leppelt, and R. J. Behm, J. Catal. 251, 437(2007).

[81] J. Strunk, K. Kahler, X. Xia, M. Comotti, F. Schuth,T. Reinecke, and M. Muhler, Appl. Catal. A – Gen. 359,121 (2009).

[82] Z. P. Liu, X. Q. Gong, J. Kohanoff, C. Sanchez, and P. Hu,Phys. Rev. Lett. 91, 266102 (2003).

[83] L. M. Molina, M. D. Rasmussen, and B. Hammer, J. Chem.Phys. 120, 7673 (2004).

[84] I. N. Remediakis, N. Lopez, and J. K. Norskov, Angew.Chem. Int. Ed. 44, 1824 (2005).

www.pss-b.com

colloidal deposition as method to study the influence of the support on the activity of gold...

Documents