Clusters, surfaces, and catalysisGabor A. Somorjai*, Anthony M. Contreras, Max Montano, and Robert M. RiouxDepartment of Chemistry, University of California, Berkeley, CA 94720 and Materials Sciences Division and Chemical Sciences Division,Lawrence Berkeley National Laboratory, Berkeley, CA 94720

Edited by A. Welford Castleman, Jr., Pennsylvania State University, University Park, PA, and approved November 28, 2005 (received for reviewSeptember 12, 2005)

The surface science of heterogeneous metal catalysis uses model systems ranging from single crystals to monodispersed nanopar-ticles in the 1–10 nm range. Molecular studies reveal that bond activation (C–H, H–H, C–C, CAO) occurs at 300 K or below as theactive metal sites simultaneously restructure. The strongly adsorbed molecules must be mobile to free up these sites for continuedturnover of reaction. Oxide–metal interfaces are also active for catalytic turnover. Examples using C–H and CAO activation aredescribed to demonstrate these properties. Future directions include synthesis, characterization, and reaction studies with 2D and3D monodispersed metal nanoclusters to obtain 100% selectivity in multipath reactions. Investigations of the unique structural,dynamic, and electronic properties of nanoparticles are likely to have major impact in surface technologies. The fields of heteroge-neous, enzyme, and homogeneous catalysis are likely to merge for the benefit of all three.

catalytic bond activation � high selectivity catalyst design � molecular ingredients of catalysis

Metal containing catalysts areclusters of 1–10 nm in size.These are grouped intothree types: heterogeneous

catalysts that are embedded in high sur-face area supports, usually oxides, tooptimize their surface area and thus thenumber of molecules they produce persecond and also to optimize their ther-mal and chemical stability. They areused mostly at high temperatures (400–800 K) and in the presence of vaporphase reactants and product molecules.Enzyme catalysts operate in solution,mostly in water near 300 K, and themetal-containing active sites are sur-rounded by proteins that maintain struc-tural mobility (1). Homogeneous cata-lysts function in the same solution inwhich the reactant and product mole-cules are dissolved, mostly in organicsolvents, and used in the 300–500 Ktemperature range. Because of the dif-ferent operating conditions and for rea-sons of history, the three fields of catal-ysis developed independently andbecame separate fields of science withindifferent subdisciplines of chemistry.Heterogeneous catalysis was practicedmostly by physical chemists and chemi-cal engineers, enzyme catalysis by bio-chemists, and homogeneous catalysis byinorganic and organometallic chemists.

Catalytic reactions are distinguishedfrom stoichiometric reactions by havingmany turnovers per reacting active siteproducing 102 to 106 molecules per site.Some of these catalytic reactions arevery fast. For instance, the catalytic oxi-dation of carbon monoxide (CO) pro-duces thousands of CO2 molecules permetal surface site per second above ig-nition where the exothermicity of theprocess makes it self-sustaining in tem-perature and the reaction rate is onlylimited by the speed of transport of mol-ecules to and from the catalyst surface

(2–4). Ethylene hydrogenation, anotherexothermic reaction, turns over to pro-duce ethane �10 times per metal sur-face site per second at 300 K on Pt(111)(5). The catalytic conversion of n-hexaneto benzene or to branched isomers, acomplex but important reaction thatproduces high-octane gasoline, has aturnover rate of 10�2 product moleculesper platinum surface site per second at�600 K (6). In general, the more com-plex the chemical reaction, the slowerthe turnover as the elementary reactionsteps involve more complicated molecu-lar rearrangements.

Over the past 2 decades, molecularstudies of the structure and dynamics,during catalytic turnover, of these threetypes of catalysts, heterogeneous, homo-geneous, and enzyme, have revealed fea-tures that are similar to all. The catalyststructures change under working condi-tions, and thus in situ studies are neededto verify the structures of the active,working catalyst. The active metal sitesrestructure during turnover as they bind,react, and release the adsorbed mole-cules. The reactive intermediates alsomust be mobile to free up the active siteto be able to carry out the next turnover.

Thus, dynamics of the catalyst struc-ture and that of the reactant moleculesand reaction intermediates control boththe activity (rates) and the selectivity(product distribution) of the catalysts.

In this overview, we focus on metalheterogeneous catalysts from the physi-cal chemistry perspective. This field de-veloped rapidly by use of model cata-lysts, first single-crystal surfaces (7–12),then monodispersed nanoclusters depos-ited on oxide surfaces by lithographytechniques (13–17) (Fig. 1). These two-dimensional (2D) catalyst systems couldbe characterized by a large number oftechniques on the molecular level andunder reaction conditions. In our labo-

ratory, sum frequency generation (SFG)vibrational spectroscopy was used tomonitor the structure and bonding ofthe adsorbed monolayer under reactionconditions (18–20). Scanning tunnelingmicroscopy (STM) applied at high reac-tant pressures under reaction conditionswas used to monitor adsorbate mobility(21). To observe changes in surfacestructure of the metal catalysts low-energy electron diffraction (LEED)surface crystallography was used withsingle crystal metal surfaces.

Three phenomena were discoveredthat appear to be essential features ofactive heterogeneous catalysts.

1. The adsorption and bonding of in-coming reacting molecules restruc-tures the metal surface around theadsorption site. This process occursto optimize the adsorbate-metalbonding. Bond activation (C–H,H–H, C–C, CAO) leading to bond-breaking occurs as the metal activesites restructure.

2. The strongly adsorbed molecules atthe active sites must be mobile alongthe surface to free up the active sitesfor continued turnover. This mobilityis detectable on metal single crystalsby STM. When the catalyst is poi-soned by another adsorbate, mobilitystops, and the turnover is inhibited.

3. Oxide–metal interfaces are activesites for catalytic turnover. There isevidence from diverse experiments

Conflict of interest statement: No conflicts declared.

This paper was submitted directly (Track II) to the PNASoffice.

Abbreviations: LEED, low-energy electron diffraction; SFG,sum frequency generation; STM, scanning tunnelingmicroscopy.

*To whom correspondence should be addressed. E-mail:somorjai@berkeley.edu.

© 2006 by The National Academy of Sciences of the USA

www.pnas.org�cgi�doi�10.1073�pnas.0507691103 PNAS � July 11, 2006 � vol. 103 � no. 28 � 10577–10583

SP

EC

IAL

FE

AT

UR

E:

PE

RS

PE

CT

IVE

for their unique activity even afterthe metal sites are deactivated.

Below we give examples that describethese phenomena followed by sugges-tions of some of the directions of cataly-sis science for the future.

Model catalysts are used less fre-quently in the fields of enzyme and ho-mogeneous catalysis. Nevertheless, asthe experimental conditions used for thethree types of catalysts become similarby creative design, it is likely that themolecular characteristics, which controlthe activity and selectivity of the differ-ent types of catalysts (surface, enzyme,and homogeneous), also become similar.It is our hope that these three fields ofcatalysis will merge to become one fieldin the foreseeable future.

C–H Bond Activation of Ethylene overthe Platinum (111) Crystal SurfaceThe activation of C–H bonds is at theheart of the catalytic transformation oforganic molecules and the production ofhydrogen from light alkanes.

At 52 K, ethylene adsorbs molecularlyon the Pt(111) surface, forming a�-bonded ethylene surface species (22).At 200 K on the Pt(111) crystal faceand at low pressures of ethylene (�10�6

torr; 1 torr � 133 Pa), the SFG vibra-tional spectrum is interpreted as owingto an ethylene molecule that forms di-�bonds with the platinum surface (23).LEED surface crystallography studiesindicate that ethylene occupies a three-fold surface site, and the C–C bond is ata 23° angle with respect to the metal

surface plane (Fig. 2a) (24). Upon heat-ing, the SFG spectrum changes, indi-cating the formation of ethylidene(HC–CH3) at 250 K by intramolecularhydrogen transfer (25). Slightly abovethis temperature, ethylidyne, C–CH3,forms as another C–H bond is activatedand dissociates, and there is a transferof a hydrogen atom to the metal sur-face. SFG spectra show this transforma-tion, and a mechanism is presented inFig. 3. Ethylidyne remains stable up to

430 K when further loss of hydrogen, aswell as C–C bond breaking, occurs withthe formation of C2H (acetylide) andCH (methylidyne) species that turn intographite at �800 K (26). Ethylidyne onPt(111) occupies the same threefold fccmetal site as di-�-ethylene, as shown byLEED crystallography (Fig. 2b) (27). ItsC–C bond is normal to the metal sur-face and the nearest- and next-nearest-neighbor metal atoms change their loca-tions as compared with their positions

Fig. 2. The structure of adsorbed ethylene and the C–H activation induced restructuring of the platinumsurface. (a) The best fit structure of di-�-bonded ethylene on Pt(111) is a mixture of 60% at fcc sites and40% at hcp sites. The symbols b1 and b2 represent Pt–C bond lengths, and bu represents buckling in thetop layer. These data were obtained by LEED surface structure analysis. (b) The restructuring of the Pt(111)crystal face, when ethylidyne forms from ethylene by C–H bond breaking, is obtained by LEED surfacestructure analysis. Shading distinguishes buckled metal atoms in each layer.

Fig. 1. Model surfaces for surface science and catalysis studies. (Upper) Three single crystal surface diagrams representing the (111), (100), and stepped (557)surfaces of a face-centered cubic crystal lattice. (Lower) A 2 � 2-�m atomic force microscopy image of a platinum nanoparticle array supported on a thin filmof alumina. This array was fabricated with electron beam lithography.

10578 � www.pnas.org�cgi�doi�10.1073�pnas.0507691103 Somorjai et al.

on the metal surface before C–H bonddissociation occurs. Adsorbate-inducedrestructuring of the metal surface is,therefore, an important part of C–Hactivation, even on the (111) crystal faceof platinum, which is the closest-packedand lowest surface free-energy plane ofthis face-centered cubic metal. The ther-mal activation of the C–H bond in ethyl-ene on the (111) crystal faces of rhod-ium and palladium is similar to thatfound for the (111) crystal face of plati-num (28, 29). On Rh(111), ethylidyneoccupies a hexagonal close-packed hol-low site; that is, there is a rhodium atomfrom the second metal layer directly

under the carbon that is bound to themetal surface (30). This different siteoccupancy of ethylidyne on the Rh(111)surface changes the nature of adsorbate-induced restructuring of the metal sur-face around the chemisorption bond, asshown by LEED crystallography. On thePd(111) surface, ethylene dehydroge-nates to ethylidyne at 300 K, which re-acts to form methylidyne �400 K (29).

When hydrogen is introduced in ex-cess, in the 1–100 torr range, along withethylene at 295 K, the dehydrogenationof the molecule slows down such thatdi-�-ethylene and ethylidyne coexist onthe Pt(111) surface (18, 25, 31, 32). The

SFG spectrum demonstrating this pointis shown in Fig. 4a. Under these reac-tion conditions, ethylene hydrogenationoccurs with a turnover rate of �10ethane molecules being produced perplatinum surface site per second. Thesame rate of catalytic hydrogenation isfound for the Pt(100) crystal face be-cause this reaction is a structure-insensi-tive one (32). Only the di-�-ethylene toethylidyne ratio is somewhat differenton the two crystal faces. Hence, al-though C–H dissociation of ethylene issurface-structure sensitive, its catalytichydrogenation to ethylene is structure-insensitive. Detailed reaction studies ofhigh pressures of ethylene in the pres-ence of excess hydrogen have revealedthat ethylidyne remains strongly ad-sorbed on the platinum surfaces for �1million turnovers (molecules of ethaneformed per platinum site per second)(33). Hydrogenation occurs throughweakly adsorbed �-bonded species thatoccupy atop sites (18). This bonding siteis available on all crystal planes, whichexplains the lack of structure sensitivityof the catalytic hydrogenation of ethyl-ene. �-bonded ethylene hydrogenatessequentially to an ethyl intermediateand then to ethane, as both of theseweakly adsorbed species have been de-tected on the Pt surface under reactionconditions by SFG in surface concentra-tions of 4% of a monolayer (Fig. 4b) (31).

In Tables 1 and 2, we list the temper-

Table 1. Surface structure sensitivity ofC-H activation

Crystalface

Ethylene Cyclohexene

Temp., K Ref. Temp., K Ref.

Pt(111) 250 25 200 34Pt(100) 275 (1 � 1) 35 220 (1 � 1) 36

255 (5 � 20) 35 200 (5 � 20) 36Pt(110) 280 37 – –

Table 2. Temperatures of observedC–H activation

Reactant species Temp., K Ref.

Alkenes*Ethylene 250 25Propylene 230 38Isobutene 270 391-Hexene 250 40Cyclohexene 200 34

Alkanes†

Methane 250 41Ethane 275 41n-hexane 296 402-methylpentane 296 403-methylpentane 296 40

*Pressure �10�6 torr.†Pressure 1–1.5 torr.

Fig. 3. Temperature-dependent rearrangement of adsorbed ethylene as monitored by SFG on Pt(111)surfaces. (Left) SFG spectra showing conversion from di-�-bonded ethylene to ethylidyne with increasingtemperature stepwise from 243 to 352 K. These spectra are taken after a 4-langmuir dose of ethylene ontoPt(111) single crystal. (Right) Proposed mechanism for this surface transformation.

Fig. 4. The H2 pressure dependence of ethylene surface species as monitored by SFG on Pt(111) surfaces.(a) SFG spectrum of the Pt(111) surface during ethylene hydrogenation with 100 torr of H2, 35 torr ofethylene, and 615 torr of He at 295 K. (b) SFG spectrum of the Pt(111) surface during ethylene hydroge-nation with 727 torr of H2 and 60 torr of ethylene.

Somorjai et al. PNAS � July 11, 2006 � vol. 103 � no. 28 � 10579

atures and pressures at which C–H bonddissociation was first observed in ourstudies and in those of others. It shouldbe noted that bond dissociation occursat a certain rate under the experimentalconditions used. Thus, its detection bySFG spectroscopy or by the formationof surface carbon is somewhat uncertainand also subjected to detection limits.Nevertheless, it is clear from the inspec-tion of the tables that C–H activation isfacile at �300 K for all of the moleculeslisted.

CO Bond Activation over Platinum SingleCrystal SurfacesBy using SFG surface vibrational spec-troscopy, the interaction of CO withplatinum single crystals was investigatedat high pressure and high temperatures(42). Under 40 torr of CO, the moleculewas found to dissociate on Pt(111),Pt(557), and Pt(100) at 673, 548, and500 K, respectively. The CO top sitefrequency was observed to shift to lowerfrequencies as a function of temperature(Fig. 5). At a particular temperature,dependent on the surface structure, theSFG spectra evolved with time, indicat-ing the surface was being modified. Theobserved frequency shift before CO dis-sociation is attributed to a harmoniccoupling to the frustrated translationalmode. At the dissociation temperature,the frequency shift is attributed to sur-face roughening. The surface roughen-ing is believed to result from the forma-tion of platinum carbonyl species, whichwould be a driving force to extract plati-num atoms from the surface lattice. Forboth the (111) and (100) surfaces of

platinum, the crystal must be heated toa temperature at which platinum car-bonyls are formed to produce step andkink sites, which are needed for dissoci-ation, which deposits carbon on themetal surfaces. The Pt(111) surface ex-hibits a much higher CO dissociationtemperature as compared with Pt(100)because it is the most stable surface forplatinum. Pt(557) is essentially a (111)surface with steps already present in thestructure (Fig. 1), so the crystal doesnot need to be heated to a high temper-ature to produce step sites necessary forCO dissociation and carbon depositionthrough the production of platinumcarbonyls.

Mobility of Adsorbed Molecules DuringCatalytic Turnover: Poisoning ofReactions and Adsorbate Mobility by COSTM studies indicate that ethylidyne ismobile on the Pt(111) crystal face at300 K, both at high and low pressures,and with or without hydrogen, during

the catalytic reaction (21, 26). Ethyli-dyne could only be slowed to the timescale of the STM scanning rate and ob-served upon cooling to180 K or lower.Extended Huckel calculations indicatethat the activation energy barrier to eth-ylidyne surface diffusion is low (i.e., 0.1eV; 1 eV � 1.602 � 10�19 J) (43). Thismobility is essential for maintaining thecatalyst activity. For the reaction to oc-cur under high pressures where the sur-face is at nearly saturated coverage, sta-tistical f luctuation of adsorbate densitymust be maintained to open up activesites crucial for the catalytic reaction.Adsorbate mobility is necessary so thatfavorable surface metal sites can be ac-cessed by reactants for adsorption anddiffusion to sustain the hydrogenationreaction. We tested this concept by add-ing CO to poison the reaction. By usinghigh-pressure STM and mass spectrome-try, we find that the catalytic activity ofthe Pt(111) and Rh(111) catalysts stopssuddenly when CO is coadsorbed (21).

Fig. 5. CO top-site frequency as a function oftemperature for Pt(111), Pt(557), and Pt(100) under40 torr of CO. The observed frequency redshiftbefore CO dissociation is attributed to a harmoniccoupling to the frustrated translational mode.

Fig. 6. Shown are 100 � 100-Å STM images of a Pt(111) single crystal after the sequential addition of 20mtorr of H2 (Left Upper), 20 mtorr of C2H4 (Center Upper), and 5.6 mtorr of CO (Right Upper and Lower).The catalytically active mobile adsorbate layer becomes immobile upon catalyst deactivation caused bycoadsorption of CO.

Fig. 7. STM images of active and CO-poisoned Pt(111) catalyst surfaces during cyclohexene hydroge-nation/dehydrogenation. (a) A 75 � 75-Å image of Pt(111) in the presence of 200 mtorr of H2 and 20 mtorrof cyclohexene at 300 K. No discernable order is present. (b) A 90 � 90-Å STM image of Pt(111) in thepresence of 200 mtorr of H2, 20 mtorr of cyclohexene, and 5 mtorr of CO at 300 K. The surface forms anordered CO structure, and the catalyst is deactivated.

10580 � www.pnas.org�cgi�doi�10.1073�pnas.0507691103 Somorjai et al.

STM results suggest that in the presenceof CO, the adsorbed species becomelocked into static-ordered structures.Surface mobility is suppressed to inhibitthe surface density fluctuation necessaryfor the availability of reactive sites, andthe catalytic reaction is poisoned. Fig. 6displays STM images showing this be-havior on the Rh(111) surface.

High-pressure STM studies suggestthat mobility within the adsorbed layeris also key for hydrogenation and dehy-drogenation of cyclohexene to cyclohex-ane and benzene, respectively, on thePt(111) surface (M.M., M. Salmeron,and G.A.S., unpublished results). At 300K and above and with high pressures ofcyclohexene (�20 mtorr) and hydrogen(�200 mtorr), surface species presenton Pt(111) are disordered and cannot beimaged on the STM time scale (Fig. 7a).SFG studies indicate �-allyl (C6H9) spe-cies are present on the surface (32, 44).The addition of CO causes all catalyticactivity to cease and orders the surface(Fig. 7b). These results indicate that theC6 reaction intermediates that includecyclohexadienes and �-allyl are mobileduring the dehydrogenation and hydro-genation reaction. There is evidence

from both SFG and STM studies thatformation of the �-allyl (C6H9) surfaceintermediate may be critical for the hy-drogenation of cyclohexene to cyclohex-ane to proceed. Just as for ethylenehydrogenation, introduction of CO stopsthe mobility and results in the formationof ordered surface structures and thetotal poisoning of catalytic activity.

Reactivity of Oxide–Metal InterfacesEffect of Oxide Monolayers on Rhodium-Catalyzed CO Hydrogenation Reactions. COand CO2 hydrogenation to methane overa Rh foil decorated with submonolayerquantities of TiOx, VOx, ZrOx, NbOx,TaOx, and WOx were carried out (45).The rate of methane formation wasmeasured at 1 atm (1 atm � 101.3 kPa),and the state of the working catalyst wascharacterized by x-ray photoelectronspectroscopy immediately after reaction.Each of the oxides was found to en-hance the rate of CO methanation rela-tive to that observed over the Rh metalalone in the absence of any of the ox-ides. The maximum degree of rate en-hancement occurs at an oxide coverageof approximately half a monolayer,

which corresponds to an optimum ox-ide–metal interface area. Fig. 8 showsthat niobium oxide, titanium oxide, andtantalum oxide show the largest increasein turnover rates, on the order of a 12-fold increase with respect to the cleanrhodium foil. Clearly the oxide–metalinterface is implicated in increasing theactivity of rhodium of these hydrogena-tion reactions. It also should be notedthat the oxides alone are not active forthese reactions.

Platinum Nanoclusters on Silica and Alu-mina: CO Does Not Poison Ethylene Hydro-genation Activity at Oxide–Metal Interfaces.When ethylene hydrogenation is carriedout on platinum single crystal surfacesand platinum nanoclusters deposited byelectron beam lithography on silica oralumina supports (Fig. 1), the two typesof systems yield roughly equal turnoverrates at 300 K (17, 46). However, majordifferences in catalytic behavior emergewhen the platinum catalysts are poi-soned by the addition of CO. CO poi-soning of the platinum single crystalincreases the activation energy to 20kcal�mol from �10 kcal�mol and de-creases the turnover rate at 300 K by 7orders of magnitude to �10�6 s�1.

Upon CO adsorption, the platinumnanoparticle arrays show dramaticallydifferent behavior than the Pt(111) sin-gle crystal. The CO-poisoned activationenergy for ethylene hydrogenation onalumina and silica is 11.4 and 15.6 kcal�mol, respectively (17). These values aremuch lower than for the Pt single crys-tal. The turnover rates remain in therange of 5 � 10�2 s�1, which are ordersof magnitude greater than for the singlecrystal surface. It appears that on theseplatinum nanoparticle arrays depositedon the oxides there are reaction sites

Fig. 8. Model catalyst system to study the effectof the oxide–metal interface on CO2 hydrogena-tion. (a) Diagram of submonolayer metal oxideislands formed on Rh foil. (b) Effect of differentmetal oxides, as a function of coverage, on the rateof methane formation from CO2 and H2 over Rhfoil.

Fig. 9. The oxide–metal interface (highlighted bythe red and black arcs) is catalytically active.

Fig. 10. Transmission electron microscopy of platinum nanoparticles synthesized by two differenttechniques to obtain size and shape control of the particles. (a) Monodispersed platinum nanoparticles inthe 1–8 nm range are synthesized in solution. They are capped with a polymer coating (polyvinylpyrro-lidone in this case) that prevents their aggregation. (b) In the presence of silver ion, the shape of theplatinum nanoparticles is altered because of preferential adsorption on one of the crystal surfaces.

Somorjai et al. PNAS � July 11, 2006 � vol. 103 � no. 28 � 10581

that do not deactivate for ethylene hy-drogenation in the presence of coad-sorbed CO. The unpoisoned turnoverfrequencies for the Pt nanoparticle sam-ples on alumina and silica are 7.3 and5.3 s�1, respectively, assuming that allavailable platinum surface atoms areactive for the reactions. By using thesame assumption, the CO-poisonedturnover frequencies for the aluminaand silica supported samples are 0.071and 0.041 s�1, respectively. However, ifthe oxide–platinum interface sites areconsidered to be the only active sites forreaction during CO poisoning (Fig. 9),and the turnover frequency is calculatedfrom just these sites alone, the alumina-and silica-supported samples have turn-over frequencies of 7.1 and 4.2 s�1, re-spectively. These turnover frequencies

are almost identical to those of the un-poisoned samples. Although not conclu-sive, these results indicate that oxide–metal interface sites remain active underpoisoning conditions.

Future Directions of Catalysis ScienceWe need chemical processes that pro-duce only the desired molecules withoutany by-products that need disposal aswaste. We call this green chemistry, andit requires catalysts that exhibit 100%selectivity toward needed products formultipath reactions where each reactionchannel is thermodynamically feasible.We need catalysts that will help us toachieve this goal. However, our under-standing of the molecular ingredients ofa catalyst system that control selectivityis not as well developed as our knowl-

edge of the molecular properties thatcontrol activity (turnover rates). Thesize, surface structure, and shape of themetal cluster catalyst are known to in-fluence reaction selectivity. For this rea-son, there are investigations in manylaboratories, including our own, to syn-thesize 2D and 3D catalyst systems withmonodispersed metal clusters with con-trolled size, surface structure, andshape.

Synthesis and Characterization of 2Dand 3D Metal NanoclustersWe synthesize platinum and rhodiumnanoparticles in the 1–10 nm range incolloidal solutions in the presence ofpolymers (47, 48). As the metal clustersform, they are capped with the polymerthat inhibits their aggregation but stillpermits their growth and to maintaintheir structure and chemical stability.They are characterized by transmissionelectron microscopy, small-angle x-rayscattering, and x-ray diffraction (Fig.10) (47).

To deposit 2D arrays of nanoparticlesthe polymer-coated metal clusters aredispersed on a liquid surface in a Lang-muir–Blodgett trough and are capturedat a given surface pressure on an oxidesurface (49). The surface density can bealtered by changing the applied surfacepressure as shown in Fig. 11.

These particles are subjected to low-temperature oxidation and reductiontreatment to remove the polymer andcovalently bind the nanoparticles on theoxide surfaces. These particles canreadily be used for catalysis and moni-tored by the characterization techniquesused in surface science such as Augerelectron spectroscopy, x-ray photoelec-tron spectroscopy, and atomic force mi-croscopy (47, 50).

The 3D monodispersed platinumnanoparticles can be produced by en-capsulating polymer-coated platinumor rhodium nanoparticles in meso-porous silica. The mesoporous silica isactually synthesized around the nano-particle. In this circumstance, usuallyno more than one platinum nanopar-ticle is embedded in a mesoporouschannel (Fig. 12). After oxidation andreduction treatments to remove thepolymer that coats the metal nanopar-ticles, catalytic reactions can be carriedout where the variables are the sizeand shape of the cluster. We havestudied ethylene hydrogenation, ethanehydrogenolysis, cyclohexene hydrogena-tion, and other multipath reactions tomonitor changes of reaction selectivity(R.M.R., H. Song, S. Habas, M. Grass,K. Nietz, J. Hoefelmeyer, P. Yang, andG.A.S., unpublished results).

Fig. 12. Monodispersed platinum nanoparticles are encapsulated in mesoporous silica with a channelstructure (SBA-15) to form a 3D model catalyst system. The particle size is varied while keeping theplatinum loading at 1%.

Fig. 11. The polymer-capped monodispersed platinum nanoparticles are compressed using a Langmuir–Blodgett trough and captured on an oxide surface to form 2D arrays of different density. The surfacedensity of nanoparticles is controlled by the surface pressure.

10582 � www.pnas.org�cgi�doi�10.1073�pnas.0507691103 Somorjai et al.

Our model studies of heterogeneouscatalysis that started with the use ofmetal single crystal surfaces are beingcontinued using monodispersed metalnanoparticles.

Studies of Structural and ElectronicProperties of Metal NanoclustersAs nanoscience becomes more devel-oped, many interesting physical-chemicalsize-dependent properties are being un-covered that are important for under-standing surfaces and catalysts. Themelting temperatures of metal clusters issize dependent (51). Pressure-inducedstructural transformations are more fac-ile for smaller clusters and have loweractivation energy (52). The nanocrystalsare more perfect, because they cannotsupport dislocations because of theirsmall size (51). The 2D phase diagramthat is applicable to surface systemspermits miscibility of metals to formsolid solutions that are immiscible in 3

dimensions (53). There are severalstructural arrangements for a givensize cluster, all of them thermodynami-cally equally stable, and there is possi-bility of facile rearrangement amongthese structures (54).

The electronic properties of nano-clusters are equally interesting. Be-cause the electron mean free path inmetals is in the 5–15 nm range, thesmaller nanoclusters permit unscat-tered electron f low (55). The plasmafrequency of electrons can shift withsize, giving rise to color change ingold, for example (56). The roughedges of clusters exhibit large electricfields leading to resonance enhance-ment observed in Raman spectroscopy(57). The band gap of insulator andsemiconductor nanoparticles increaseswith decreasing cluster size (58). Theseand other properties yet to be discov-ered impart unique opportunities forapplications in surface technologies

ranging from catalysis and microelec-tronics to information storage andsensors.

Let us close with the expectationwith which we started this work. Heter-ogeneous, enzyme, and homogeneouscatalysis share so many characteristicson the molecular level that we hopethey become one field of science thatpermits deeper understanding of themolecular ingredients of structure anddynamics that makes them function toobtain high reaction selectivity. Thelearning across these broad fields thatwould follow would benefit our capa-bility to generate and convert energy,to maintain and enhance environmen-tal quality, and increase the qualityand length of human life.

This work was supported by the Director,Office of Energy Research, Office of BasicEnergy Sciences, Materials and ChemicalSciences Divisions, of the U.S. Departmentof Energy under Contract DE-AC02-05CH11231.

1. Somorjai, G. A. (2004) Nature 430, 730.2. Fassihi, M., Zhdanov, V. P., Rinnemo, M., Keck,

K.-E. & Kasemo, B. (1993) J. Catal. 191, 438–452.3. Rinnemo, M., Fassihi, M. & Kasemo, B. (1993)

Chem. Phys. Lett. 211, 60–64.4. Kulginov, D., Rinnemo, M. & Kasemo, B. (1999)

J. Phys. Chem. B 103, 3170–3178.5. Zaera, F. & Somorjai, G. A. (1984) J. Am. Chem.

Soc. 106, 2288–2293.6. Davis, S. M., Zaera, F. & Somorjai, G. A. (1984)

J. Catal. 85, 206–223.7. Salmeron, M. & Somorjai, G. A. (1981) J. Phys.

Chem. 85, 3835–3840.8. Salmeron, M. & Somorjai, G. A. (1982) J. Phys.

Chem. 86, 341–350.9. Davis, S. M. & Somorjai, G. A. (1983) J. Phys.

Chem. 87, 1545–1552.10. Imbihl, R. & Ertl, G. (1995) Chem. Rev. 95, 697–733.11. Imbihl, R. (1993) Prog. Surf. Sci. 44, 185–343.12. Backman, A. L. & Masel, R. I. (1991) J. Vac. Sci.

Technol. A 9, 1789–1792.13. Jacobs, P. W., Wind, S. J., Ribeiro, F. H. &

Somorjai, G. A. (1997) Surf. Sci. 372, L249–L253.14. Eppler, A. S., Rupprechter, G., Guczi, L. & So-

morjai, G. A. (1997) 101, 9973–9977.15. Eppler, A. S., Rupprechter, G., Anderson, E. A. &

Somorjai, G. A. (2000) J. Phys. Chem. B 104,7286–7292.

16. Grunes, J., Zhu, J., Anderson, E. A. & Somorjai,G. A. (2002) J. Phys. Chem. B 106, 11463–11468.

17. Contreras, A. M., Grunes, J., Yan, X.-M., Liddle,A. & Somorjai, G. A. (2005) Catal. Lett. 100,115–124.

18. Cremer, P. S., Su, X., Shen, Y. R. & Somorjai,G. A. (1996) J. Am. Chem. Soc. 118, 2942–2949.

19. McCrea, K. R., Parker, J. S. & Somorjai, G. A.(2002) J. Phys. Chem. B 106, 10854–10863.

20. Su, X., Cremer, P. S., Shen, Y. R. & Somorjai,G. A. (1997) J. Am. Chem. Soc. 119, 3994–4000.

21. Tang, D. C., Hwang, K. S., Salmeron, M. &Somorjai, G. A. (2004) J. Phys. Chem. B 108,13300–13306.

22. Cassuto, A., Kiss, J. & White, J. M. (1991) Surf.Sci. 255, 289–294.

23. Cremer, P. S. & Somorjai, G. A. (1995) J. Chem.Soc. Faraday Trans. 1995, 3671–3677.

24. Doll, R., Gerken, C. A., Van Hove, M. A. &Somorjai, G. A. (1997) Surf. Sci. 374, 240–246.

25. Cremer, P., Stanners, C., Niemantsverdriet, J. W.,Shen, Y. R. & Somorjai, G. A. (1995) Surf. Sci.328, 111–118.

26. Land, T. A., Michely, T., Behm, R. J., Hemminger,J. C. & Comsa, G. (1992) J. Chem. Phys. 97, 6774–6783.

27. Starke, U., Barbieri, A., Materer, N., Van Hove,M. A. & Somorjai, G. A. (1993) Surf. Sci. 152�153,532–542.

28. Dubois, L. H., Castner, D. G. & Somorjai, G. A.(1980) J. Chem. Phys. 72, 5234–5240.

29. Gates, J. A. & Kesmodel, L. L. (1983) Surf. Sci.124, 68–86.

30. Wander, A., Van Hove, M. A. & Somorjai, G. A.(1991) Phys. Rev. Lett. 67, 626–628.

31. Cremer, P. S., Su, X., Shen, Y. R. & Somorjai,G. A. (1996) Catal. Lett. 40, 143–145.

32. McCrea, K. & Somorjai, G. A. (2000) J. Mol. Catal.A 163, 43–53.

33. Davis, S. M., Zaera, F., Gordon, B. E. & Somorjai,G. A. (1985) J. Catal. 92, 240–246.

34. Henn, F. C., Diaz, A. L., Bussell, M. E., Hugen-schmidt, M. B., Domagala, M. E. & Campbell,C. T. (1992) J. Phys. Chem. 96, 5965–5974.

35. Hatzikos, G. H. & Masel, R. I. (1987) Surf. Sci.185, 479–494.

36. Lamont, C. L. A., Borbach, M., Martin, R., Gard-ner, P., Jones, T. S., Conrand, H. & Bradshaw,A. M. (1997) Surf. Sci. 374, 215–228.

37. Yagasaki, E., Backman, A. L. & Masel, R. I.(1990) J. Phys. Chem. 94, 1066–1072.

38. Cremer, P. S., Su, X., Shen, Y. R. & Somorjai,G. A. (1996) J. Phys. Chem. 100, 16302–16309.

39. Cremer, P. S., Su, X., Shen, Y. R. & Somorjai,G. A. (1996) J. Chem. Soc. Faraday Trans. 92,4717–4722.

40. Yang, M. & Somorjai, G. A. (2004) J. Am. Chem.Soc. 126, 7698–7708.

41. Somorjai, G. A. & Marsh, A. L. (2005) Philos.Trans. R. Soc. London A 363, 879–900.

42. McCrea, K., Parker, J. S., Chen, P. & Somorjai,G. A. (2001) Surf. Sci. 494, 238–250.

43. Nomikou, Z., Van Hove, M. A. & Somorjai, G. A.(1996) Langmuir 12, 1251–1256.

44. Yang, M., Chou, K. C. & Somorjai, G. A. (2003)J. Phys. Chem. B 107, 5267–5272.

45. Boffa A., Lin, C., Bell, A. T. & Somorjai, G. A.(1994) J. Catal. 149, 149–158.

46. Grunes, J., Zhu, J., Yang, M. & Somorjai, G. A.(2003) Catal. Lett. 86, 157–161.

47. Riuox, R. M., Song, H., Hoefelmeyer, J. D., Yang,P. & Somorjai, G. A. (2005) J. Phys. Chem. B 109,2192–2202.

48. Hoefelmeyer, J. D., Nienz, K., Somorjai, G. A. &Tilley, T. D. (2005) Nano Lett. 5, 435–438.

49. Song, H., Kim, F., Connor, S., Somorjai, G. A. &Yang, P. (2005) J. Phys. Chem. B 109, 188–193.

50. Zhu, J., Konya, Z., Puntes, V. F., Kiricsi, I., Miao,C. X., Ager, J. W., Alivisatos, A. P. & Somorjai,G. A. (2003) Langmuir 19, 4396–4401.

51. Tolbert, S. H. & Alivisatos, A. P. (1995) Annu.Rev. Phys. Chem. 46, 595–626.

52. Chen, C. C., Herhold, A. B., Johnson, C. S. &Alivisatos, A. P. (1997) Science 276, 398–401.

53. Moran-Lopez, J. L. & Falicov, L. M. (1979) Surf.Sci. 79, 109–116.

54. Mann, B. E. (1997) J. Chem. Soc. Dalton Trans. 9,1457–1472.

55. Hohlfeld, J., Wellershoff, S. S., Gudde, J., Conrad,U., Jahnke, V. & Matthias, E. (2000) Chem. Phys.251, 237–258.

56. Doremus, R. H. (1964) J. Chem. Phys. 40, 2389–2396.

57. Baldelli, S., Eppler, A. S., Anderson, E., Shen, Y.& Somorjai, G. A. (2000) J. Chem. Phys. 113,5432–5438.

58. Alivisatos, A. P. (1996) J. Phys. Chem. 100, 13226–13239.

Somorjai et al. PNAS � July 11, 2006 � vol. 103 � no. 28 � 10583