Density functional theory in surface chemistryand catalysisJens K. Nørskova,b,c,1, Frank Abild-Pedersena,c, Felix Studta,c, and Thomas Bligaardc

aSUNCAT - Center for Interface Science and Catalysis, SLAC National Accelerator Laboratory, Menlo Park, CA 94025; bDepartment ofChemical Engineering, Stanford University, Stanford, CA 94305; and cCenter for Atomic-Scale Materials Design, Department of Physics,Technical University of Denmark, DK-2800 Lyngby, Denmark

Edited by Charles T. Campbell, University of Washington, Seattle, WA, and accepted by the Editorial Board November 18, 2010 (received for review July13, 2010)

Recent advances in the understanding of reactivity trends for chemistry at transition-metal surfaces have enabled in silico design ofheterogeneous catalysts in a few cases. The current status of the field is discussed with an emphasis on the role of coupling theory andexperiment and future challenges.

Surface chemistry is interesting andchallenging for several reasons. Ittakes place at the border betweenthe solid state and the liquid or gas

phase and can be viewed as a meeting placebetween condensed-matter physics andchemistry. The phenomena at the solid–gasor solid–liquid interface are complicatedfor this reason. A chemical reaction ata metal surface, for instance, has all ofthe complexity of ordinary gas-phase re-actions, but in addition the usual electronconservation rules do not apply becausethe metal provides a semi-infinite sourceof electrons at the Fermi level. A new setof concepts therefore needs to be de-veloped to describe surface chemistry.An understanding of surface chemical

reactions is necessary to describe a largenumber of surface phenomena includingsemiconductor processing, corrosion,electrochemistry, and heterogeneous ca-talysis. Heterogeneous catalysis alone hasbeen estimated to be a prerequisite formore than 20% of all production in theindustrial world (1), and it will most likelygain further importance in the years tocome. The development of sustainableenergy solutions represents one of themost important scientific and technicalchallenges of our time, and heterogeneouscatalysis is at the heart of the problem.Most sustainable energy systems rely onthe energy influx from the sun. Sunlightis diffuse and intermittent, and it is there-fore essential to be able to store the en-ergy, for example chemically as a fuel or ina battery. Such storage can then provideenergy for the transportation sector and indecentralized applications, as it evens outtemporal variations (2, 3). The key toproviding an efficient transformation ofenergy to a chemical form or from onechemical form into another is the avail-ability of suitable catalysts. In essentiallyall possible sustainable energy technolo-gies, the lack of efficient and economicallyviable catalysts is a primary factor limitingtheir use (3, 4).

In the present paper, we will discussthe status of the development of an un-derstanding of surface chemistry. We willdo this from a theoretical perspective, butit is important to stress that it is a closecoupling between theory and experimentwhich has enabled the developments thusfar. Surface science experiments have beeninvaluable in providing a quantitative de-scription of a range of surface phenomena(5–14). This has been essential in bench-marking computational surface sciencebased on density functional theory (DFT)calculations and in providing experimentalguidance and verification of the conceptsdeveloped. This forms a good backgroundfor the development of an understandingof heterogeneous catalysis, which is theother part of this paper. We will show howthe concepts developed allow the under-standing of trends for certain classes ofcatalysts and reactions.The ambition is to develop concepts that

are useful in understanding which prop-erties determine the activity and selectivityof a catalyst and to be able to use calcu-lations to search for new catalyst leads. Wewill discuss how far we are in this respect

and point out some of the many chal-lenges ahead before this becomes a regularway of designing new catalysts.

Density Functional Theory Calculationsof Surface ChemistryThe description of the chemical bond be-tween a surface and a molecule is thefundamental basis for understanding sur-face chemical reactivity and catalysis. Aconsiderable amount of understanding hasbeen developed for the adsorption ofsimple atoms and molecules onto transi-tion-metal surfaces (15–19).The d-band model (20, 21) has proven

particularly useful in understanding bondformation and trends in reactivity amongthe transition metals. The d-band model isan approximate description of the bond

Fig. 1. Bond formation at a transition-metal surface. Schematic illustration of the formation ofa chemical bond between an adsorbate valence level and the s and d states of a transition-metal surface.The bond is characterized by the degree to which the antibonding state between the adsorbate stateand the metal d states is occupied. The higher the d states are in energy relative to the Fermi level, themore empty the antibonding states and the stronger the adsorption bond. Adapted from ref. 19. DOS,density of states.

Author contributions: J.K.N., F.A.-P., F.S., and T.B. designedresearch, performed research, analyzed data, and wrotethe paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission. C.T.C. is a guesteditor invited by the Editorial Board.1To whom correspondence should be addressed. E-mail:norskov@stanford.edu.

www.pnas.org/cgi/doi/10.1073/pnas.1006652108 PNAS | January 18, 2011 | vol. 108 | no. 3 | 937–943

SPECIA

LFEATURE:PERSP

ECTIV

ED

ownl

oade

d by

gue

st o

n Ju

ne 9

, 202

0

formation at a transition-metal surface, asillustrated in Fig. 1. It describes the inter-action between adsorbate valence statesand the s and d states of a transition-metalsurface. Coupling to the itinerant s statesgives rise to a shift and broadening of theadsorbate states, but this contributes toa first approximation equal to the bondenergy for all transition metals (they allhave half-filled, very broad s bands). Thedifferences between transition metals aredue to the formation of bonding and anti-bonding states between the (renormalized)valence states and the metal d states. Thestrength of the bond is given by the fillingof the antibonding states but, unlike gas-phase chemistry, where this is determinedby the number of electrons in the system, ata metal surface the filling is given by theenergy of the antibonding states relative tothe Fermi level. Because the antibondingstates are always above the d states, theenergy of the d states (the center of thed states) relative to the Fermi level is a goodfirst indicator of the bond strength. Thehigher the d states are in energy relative tothe Fermi level, the higher in energy theantibonding states are and the strongerthe bond.The details of the bonding picture have

been confirmed in a series of X-ray spec-troscopy experiments (Fig. 2). It explainstrends in reactivity from one transitionmetal to the next, and the effects of al-loying, structure, strain, defects, and soforth. A large number of calculated andexperimental results have been accounted

for in this way (22–36). An example ofexperimental results explained by thed-band model is included in Fig. 3.It is important to stress that variations in

the width of the d band will lead to var-iations in the position of the d-band cen-ter. The reason is that d filling does notchange for any transition metal, and thiswill fix the d band to the Fermi level.Hence, each metal system will have tocompensate for variations in the width byshifting the d states up or down in energydepending on the nature of the change.The width of the d band varies significantly

with metal coordination number and thed band will shift accordingly; however, thebinding energies shift with an equalamount, as seen in Fig. 4 (37).As one sweeps through the transition

metals from right to left, the metals be-come more and more reactive and bindinggeometries of adsorbates might change.Geometrical changes in the adsorbate willaffect the position and number of adsor-bate levels that couple to the metal states,and one must address such cases morecautiously.The correlation between interaction

energy and d-band center has been foundboth for adsorption energies and transi-tion-state energies. Because different ad-sorption energies and transition-stateenergies are correlated with the same un-derlying electronic structure properties ofa surface, it is not surprising that correla-tions betweendifferent adsorption energies[so-called scaling relations (38)] and be-tween adsorption energies and transition-state energies (so-called Brønsted–Evans–Polanyi relations) are found through-out transition-metal surface chemistry(e.g., Fig. 5).

From Surface Science toHeterogeneous Catalysis: CatalystDesign PrinciplesThe kinetics of a catalytic reaction is thecentral propertyof acatalyst.Thekinetics—rate and selectivity for different products—is the starting point (together with massand heat transport) for the design of chem-ical reactors. The kinetics also provides thelink between the microscopic propertiesof the catalyst, adsorption energies, and ac-tivation barriers of elementary steps, andexperiments (39–48). We will focus in thefollowing on the trends in reactivity, thatis, the variations in kinetics from one surface

Fig. 2. Experimental verification of the d-band model. (Upper) Experimental X-ray emission (XES) andX-ray adsorption (XAS) spectra for N adsorbed onto Ni and Cu surfaces. The bonding and antibondingstates originating from the adsorbate pxy and pz states and the metal d states are clearly seen. For Cuwith the lowest-lying d states, the antibonding states are filled and the adsorption bond is weak.Adapted from ref. 19.

Fig. 3. Illustration of the use of the d-band model. Electrochemically measured changes in the hydrogenadsorption energy (ESCE) for Pd overlayers on a number of metals are shown to scale well with thecalculated shift of the d-band center (δεd). Adapted from ref. 32.

938 | www.pnas.org/cgi/doi/10.1073/pnas.1006652108 Nørskov et al.

Dow

nloa

ded

by g

uest

on

June

9, 2

020

to the next.An understanding of these trendsis the basis for the design of new catalysts.In principle, kinetics can be obtained for

a given catalytic reaction if one calculatesall reaction (free) energies and activation(free) energy barriers (as a function ofcoverage and surface structure). There arenow several examples where the macro-scopic kinetics of a catalytic reaction hasbeen modeled successfully on the basis ofelectronic structure calculations (38, 43–46). These are important benchmarksshowing that the theoretical methodswork, but the procedure for calculating allenergy parameters for all possible ele-mentary steps under all possible con-ditions is extremely demanding.The same procedure can, in principle, be

used to search for new catalysts: For eachnew catalyst structure and composition,a new microkinetic model is developed andthe catalytic properties are evaluated. The

effort to repeat this procedure thousandsof times in search for new catalysts is,however, prohibitive with present-daycomputers. An alternative procedure is todevelopmodels of surface reactivity andusethese to pinpoint the most important mi-croscopic materials properties of the cata-lyst which determine the macroscopic cat-alytic performance. These “descriptors”can then be scanned much more readily fora given class of materials. An additional—and perhaps even more important—resultof such an approach is that it typicallyestablishes some design concepts thatexperimentalists and catalyst developerssubsequently can use in their daily lab-oratory work.

Deriving Descriptors of CatalyticActivityModels of reactivity trends that single outthe important parameters describing cat-

alytic activity or selectivity are the essentialprerequisites for the tailoring of surfaceswith specified catalytic properties. Inprinciple, there are many parameters thatenter into the kinetics of a given reaction:The energy of all intermediates and ofthe transition states separating them isneeded to specify the full reaction ener-getics. Even for simple reactions, this easilygives 10–20 energy variables for a specificcatalytic reaction. The complexity of theproblem is greatly reduced by the energycorrelations discussed above. Because ofthe high level of correlation between var-ious adsorbate and transition-state ener-gies, the number of independent descri-ptors can often be reduced to only afew, for example one or two, without in-troducing errors larger than the averagesimulation errors that would be expectedbased on the quality of the modern gra-dient-corrected exchange-correlationfunctionals used in electronic structurecalculations. This limited number of de-scriptors can then be screened muchmore readily. Fig. 6 illustrates this workingprinciple. In the following, we will showhow this principle can be applied.Take the methanation reaction (CO +

3H2 → CH4 + H2O) as an example. Asseen in Fig. 5, the scaling relations of theCHx species mean that if the C adsorptionenergy is known, the CH, CH2, and CH3adsorption energies are also known fora given surface. Likewise, the OH ad-sorption energy is given by the O adsorp-tion energy and, as shown in Fig. 5, thetransition-state energy for CO dissociationscales with the dissociative adsorption en-ergy of CO (hence the C and O adsorptionenergies). Finally, it turns out that the CHxand OH formation transition-state ener-gies are also functions of the C and Oadsorption energies. The CO adsorption

Fig. 4. Illustration of the extent of the d-band model. Calculated CO and O adsorption energies fora range of different Au (A) and Pt (B) surfaces including 12 atom clusters are seen to correlate with thecalculated d-band center (εd). Adapted from ref. 37.

Fig. 5. Illustration of scaling relations. The plot on the left shows how all calculated CHx adsorption energies scale with the adsorption energy of C fora number of metal surfaces (close-packed, black; stepped, red). The slope is given entirely by bond counting. Adapted from ref. 38. The plot on the right showshow the activation energy for CO dissociation over stepped surfaces scales with the dissociative chemisorption energy of the reaction.

Nørskov et al. PNAS | January 18, 2011 | vol. 108 | no. 3 | 939

Dow

nloa

ded

by g

uest

on

June

9, 2

020

energy also scales with the C adsorptionenergy, and the same is roughly true of theH adsorption energy (although here thevariation from one metal to the next isvery slight). The result is that in a firstapproximation the activity of a given metalis given by only two descriptors: the C andthe O adsorption energies.Fig. 7 shows the calculated rate of

methanation under industrial conditions asa function of these two parameters. Theinput into the model is the set of scaling

relationships for reaction energies andtransition energies. The result is a volcano-shaped relationship between the catalyticrate and the adsorption energies. Themean absolute errors introduced by thescaling relations are on the order of 0.2–0.3eV, errors that are within the limit of DFT.These errors are very small comparedwith the energy scale on which the volcanois defined. Hence, trends will be well-described even based on approximatescaling relations. The volcano relationshipis a very fundamental concept in hetero-geneous catalysis (49, 50) because it allowsthe identification of the best catalytic ma-terials for a given reaction as a function ofthe chosen descriptors (51). A search forgood catalysts therefore amounts to findingmaterials with descriptors that are close tothe maximum of the volcano.Themost critical issue is to determine the

active site for a specific reaction. For themethanation reaction, it is the step or kinksites on the metal particles that providethe proper structure for the breaking of thestrong CO bond. Each structure definesa new set of scaling and Brønsted–Evans–Polanyi relations. Hence, identifying whichsites are present during the reaction andidentifying which of the resulting relationslead to the highest rate are the essenceof determining the proper catalytic mate-rial for the process.For the methanation reactions, the ori-

gin of the volcano is simple. We find thatthe breaking of the CO bond is the rate-

determining step for relatively unreactivecatalysts such as Ni and that this reactionstep proceeds via bond breaking of the C-OH intermediate yielding adsorbed C andOH on the surface (52–55). On the morereactive metals, such as Fe, this reactionstep becomes fast, and the surface be-comes poisoned by adsorbed carbon andoxygen. A good methanation catalysttherefore represents a reasonable com-promise between a low CO dissociationbarrier and a high carbon and oxygen de-sorption rate (in terms of the CH4 andH2O desorption, respectively) (56). As canbe seen in the volcano relation in Fig. 7,the theoretical results describe experi-mental findings (49) very well: Ru andCo lie at the top of the volcano, whereasRh and Ni on the right and Fe on theleft are calculated to have somewhatlower activities.Industrially, the catalyst that is com-

monly used is based on Ni. Even though Ruand Co have been found to be the mosteffective (39), Ni is still preferred due toits lower price. Recently, DFT calculationsidentified Ni-Fe catalysts as a cheaper al-ternative with a higher activity than Ni(see also Fig. 7). These findings have beenverified experimentally (56).

Descriptor-Based Search for NewCatalystsThe identification of Ni-Fe as a possiblecatalyst for methanation is a very simple,early example of descriptor-based searchesfor new heterogeneous catalysts. There areseveral others in the recent literature (7,25, 57–62). In the following, we will illus-trate the approach with another example,which is a little more complex because itinvolves not only the rate but also the se-lectivity of the process.We will take the selective hydrogenation

of acetylene in the presence of an excess ofethylene as an example. The reaction isimportant industrially, because it is used toremove trace amounts of acetylene fromthe ethylene used to make polymers. Fig. 8shows the calculated pathway for the hy-drogenation of acetylene to ethylene andthe further hydrogenation of ethylene toethane on the Pd(111) surface. The pro-cess starts with the adsorption and suc-cessive hydrogenation of acetylene toethylene. Once ethylene is formed, it candesorb or react further to produce ethane,which is unwanted in this process (if thiswere fast, the ethylene would all disap-pear). A selective catalyst should have anactivation barrier for the hydrogenation ofethylene that is higher than its desorptionenergy to favor desorption of the productrather than its further hydrogenation.Whereas these two energies are compa-rable for Pd(111), the desorption barrier issmaller than the activation barrier for thePdAg(111) surface (Fig. 8), making this

Meta-data

Catalytic properties

(Selectivity, turnover, ...)

Data(Energies, ...)

Experiment

Models

Descriptors

Quantumcalculations

Fig. 6. Illustration of the link between the mi-croscopic surface properties and the macroscopiccatalytic properties as measured in a catalyticconverter. They are directly linked through thekinetics, but it is far more instructive to developmodels that map the large number of microscopicproperties characterizing the catalyst onto a fewdescriptors.

Fig. 7. Theoretical volcano for the production of methane from syngas, CO, and H2. The turnoverfrequency (TOF) is plotted as a function of carbon and oxygen binding energies. The carbon and oxygenbinding energies for the stepped 211 surfaces of selected transition metals are depicted. Reactionconditions are 573 K, 40 bar H2, 40 bar CO.

940 | www.pnas.org/cgi/doi/10.1073/pnas.1006652108 Nørskov et al.

Dow

nloa

ded

by g

uest

on

June

9, 2

020

surface more selective and explaining theaddition of Ag to industrial Pd catalysts(63–66). Importantly, the addition of Agto the Pd catalyst not only enhances itsselectivity, it also decreases the activityby lowering the adsorption energyfor acetylene.One can now use the acetylene and

ethylene adsorption energies, ΔEC2H2 andΔEC2H4 , as the descriptors determiningturnover and selectivity. These two de-scriptors are interrelated by a scaling re-lation as shown in the top of Fig. 9 (59). Itwas found that both acetylene and ethyl-ene adsorption are well-described by themethyl adsorption energy. The slope ofthe scaling relations is 4 for acetylene and2 for ethylene, a result that can be viewedas a manifestation of bond-order conser-vation (67, 68). Importantly, the resultsshow that ΔEC2H2 and ΔEC2H4 are corre-lated. The optimal catalyst is hence a com-promise between selectivity and activity.This observation allows for the determina-tion of awindowwhere good hydrogenationcatalysts can be found. The window is lim-ited by selectivity on the left side and ac-tivity on the right side as schematicallyshown in the bottom of Fig. 9. Screening ofa number of different bimetallic catalystswith respect to the descriptor, the methyladsorption energy, and the metal price perkilogram alloy led to the discovery of theNiZn andNiGa catalysts. These alloys werepredicted to have an inherent selectivitythat is comparable to the industrially usedPdAg catalysts, while at the same timebeingconsiderably cheaper in price. Experi-mental testing of both NiZn and NiGa (59)has indeed shown that both catalysts havea high selectivity comparable to that ob-served for PdAg.

Challenges in Theoretical SurfaceReactivity and HeterogeneousCatalysisFinding a new catalyst lead for a givenreaction is only the first of a number ofnecessary steps toward making a new

technical catalyst. Whereas a high activ-ity or a good selectivity are necessaryrequirements, and a high cost of theconstituents can become prohibitive forindustrial-scale use of the catalyst, thereare commonly a multitude of other im-portant requirements for a new technicalcatalyst. The ability of the catalyst to workin the presence of potential poisons, itsability to avoid Ostwald ripening and other

sintering mechanisms, and its stability interms of, for example, avoiding slowevaporation are typically crucial catalystproperties for achieving an economicallyacceptable lifetime. The production costs,the ease with which the catalyst is reduced,its capacity to be improved by secondarypromoters, the role of defects, and how thecatalyst interacts with different supportmaterials can also be factors that need to betaken into account. The above-mentionedfactors can all be simulated to some ex-tent, but it would be reasonable to say thatsubstantial improvements are neededwith respect to the simulation of all thesematerials properties before they will pro-vide an alternative to carrying out thecorresponding experiments. Naturally,however, the push in the theoretical com-munity is toward carrying out as much aspossible of the design process in silicorather than by experiment. In the end it isonly an experiment that can confirm thediscovery of a good catalyst.The couple of examples mentioned so

far refer to relatively simple reactions, andthe catalysts had rather well defined activesites on simple vicinal surface structures

Fig. 8. Potential energy diagram obtained from DFT calculations for the hydrogenation of acetylene toethane on close-packed surfaces of Pd (black) and PdAg (red) surfaces. Adapted from ref. 59.

Fig. 9. (Upper) Adsorption energy for acetylene and ethylene plotted against the adsorption energy ofa methyl group. The solid lines show the predicted acetylene (red) and ethylene (blue) adsorption en-ergies from scaling. The dotted lines define the region of interest, where the ethylene adsorption energyis less than the barrier for further hydrogenation (blue) and where the reactivity of the acetylene hy-drogenation step is estimated to be 1 s−1 per site (red). (Lower) Cost (in 2006 metal prices) of 70 binaryintermetallic compounds plotted against the calculated methyl adsorption energies. The smooth tran-sition between regions of low and high selectivity (blue) and high and low reactivity (red) is indicated.Adapted from ref. 59.

Nørskov et al. PNAS | January 18, 2011 | vol. 108 | no. 3 | 941

Dow

nloa

ded

by g

uest

on

June

9, 2

020

of pure and alloyed transition metals. In-organic compounds such as zeolites, oxidesin general, carbides, sulfides, and so forthare playing an increasingly importantrole in industrial catalytic conversion. Ex-tending systematic computer-based designbeyond transition-metal catalysts providesa considerable challenge from a theoret-ical point of view. We know from detailedcomparisons between theory and experi-ment that density functional theory worksquite well for the transition metals, but wealso know that it often works somewhatworse for other classes of potential cata-lysts, such as, for example, strongly corre-lated oxides (69–73).As ambition grows toward treating the

catalysis of larger organic molecules, forexample, the inclusion of van der Waalsinteractions becomes important (74, 75).Although steps have already been taken inthat direction (76–78), it appears thatfurther improvements are needed. Meth-odological improvements are also neces-sary to address the fundamental challengerelated to finding the ground-state struc-tures of unknown materials (79–81) and todo this systematically and self-consistentwith in situ reaction kinetics. When a cata-lyst operates under conditions where thereaction is occurring without one singleelementary step being the rate-determiningstep but with the reaction rate being theresult of several competing rates, onecannot define a unique chemical potentialof all of the reactive and product species.The oxidative or reducing power of thedifferent species may be said to be de-termined by kinetics and not thermody-namics alone. Establishing this kind of self-consistent phase-diagram analysis underkinetic conditions poses a major challenge.Even though some progress has been

made toward understanding electro-catalytic processes and photocatalyticprocesses from DFT, methodologicalimprovements are needed to systematicallyinvestigate the adiabatic assumption underelectron transfer processes at interfacesand better deal with the limitations of DFTin treating the electronic bandgap andexcited electronic states. Current state ofthe art is to use oversimplified treatments

of entropic contributions such as, for ex-ample, harmonic transition-state theory forthe different free-energy states of thereactants. The harmonic transition-statetheory approach works reasonably forsmaller well-defined systems, but as thelevel of ambition grows toward includingmore of the environment and looking atlarger reactant molecules, an improvedsampling of the entropic part of free-en-ergy contributions will become increasinglyrelevant. This in itself will require morecomputational power, more efficientcomputer codes, and codes that scale wellon massively parallel computer systems.It is therefore clear that, in a number of

cases, we need higher accuracy than cur-rent density functional theory methods canprovide to have predictive power. In ad-dition, we need methods and associatedcomputer codes that are more efficientcomputationally to be able to treat thecomplexity that characterizes most tech-nically interesting systems and their envi-ronments—real materials includingdefects and two- or three-phase systems.On the conceptual level, a number of

improvements are equally desirable. Fortransition-metal alloys there are knownexceptions to the d-band model, and thiscan result in outliers on the linear energyrelations and suggests that improved de-scriptors may exist. Development of sys-tematic approaches for determining theunderlying electronic structure-materialsfunctionality descriptors without as muchinput of external knowledge could be de-sirable. Scaling relations carrying over toinorganic compounds suggests that theideas of the d-band model have wider ap-plicability than to only metallic surfaces,but a counterpart to the d-band modelworking for oxides so far remains elusive.As there is a push toward single-active-

site control and high selectivity, we canperhaps find inspiration from the fact thatenzymes developed in nature tend to haveexactly these attributes. Although manyenzyme processes are not able to competewith modern industrial processes in termsof energy management, much can still belearned about low-temperature/highly se-lective catalysis by studying the active sites

in enzymes, and a further integration of theconcepts in homogeneous, biological, andheterogeneous catalysis is highly desirable.To address some of the challenges out-

lined above, it seems that the catalysisscience community as a whole could helpitself by establishing a generally accessiblestructured database of simulated materialsproperties. If most groups systematicallyused this database it would make it mucheasier to look up what systems have alreadybeen looked at by other groups. It wouldalso improve reproducibility of results, andsystems that were originally calculated forone purpose by one group could readilybe reused for different purposes by anothergroup. Establishing such a database couldpotentially have the same enormous im-pact on the computational materials designcommunity as the Brookhaven database ofbiological structures has had on thebioinformatics community.There are, as outlined above, a number

of challenges on the road ahead. Thechallenges above are stated primarily aschallenges for theory. They should, how-ever, also to a large extent be viewed aschallenges to experimentation. In the sameway that well-defined experiments in ultra-high vacuum on single-crystal surfaces wereof central importance in establishing thecurrent reliable level of theoretical treat-ment of adsorption of small molecules ontotransition-metal surfaces, so will new de-tailed experiments be of central impor-tance in extending the reliability ofelectronic structure calculations to newareas. The fact that it has been possible ina few simple cases to tailor surfaces withimproved catalytic properties on the basisof insight and DFT calculations providessome hope that this may with time developinto a general and versatile design strategy.The rapid developments that we are seeingnow are, we hope, only the beginning.

ACKNOWLEDGMENTS. The authors acknowledgesupport from the US Department of Energy, Officeof Basic Energy Sciences, the Center for Atomic-scale Materials Design funded by the LundbeckFoundation, and the Catalysis for Sustainable En-ergy (CASE) initiative, which is funded by theDanish Ministry of Science, Technology, andInnovation.

1. Maxwell IE (1996) Driving forces for studies in catalysis.Stud Surf Sci Catal 101:1–9.

2. Lewis NS, Nocera DG (2006) Powering the planet:Chemical challenges in solar energy utilization. ProcNatl Acad Sci USA 103:15729–15735.

3. Office of Basic Energy Sciences, US Department of En-ergy (2005) Basic Energy Needs for Solar Energy Utili-zation, Workshop Report. Available at http://www.er.doe.gov/bes/reports. Accessed June 1, 2010.

4. Office of Basic Energy Sciences, US Department of En-ergy (2007) Basic Energy Needs Catalysis for Energy,Workshop Report. Available at http://www.er.doe.gov/bes/reports. Accessed June 1, 2010.

5. Valden M, Lai X, Goodman DW (1998) Onset of catalyticactivity of gold clusters on titania with the appearanceof nonmetallic properties. Science 281:1647–1650.

6. Schlögl R (2003) Catalytic synthesis of ammonia—A“never-ending story”? Angew Chem Int Ed Engl 42:2004–2008.

7. Linic S, Jankowiak J, Barteau MA (2004) Selectivity drivendesign of bimetallic ethylene epoxidation catalysts fromfirst principles. J Catal 224:489–493.

8. Campbell CT, Parker SC, Starr DE (2002) The effect ofsize-dependent nanoparticle energetics on catalystsintering. Science 298:811–814.

9. Tsirlin T, Zhu J, Grunes J, Somorjai GA (2002) AFM andTEM studies of Pt nanoparticle arrays supported onalumina model catalyst prepared by electron beamlithography. Top Catal 19:165–170.

10. Ertl G (2008) Reactions at surfaces: From atoms tocomplexity (Nobel Lecture). Angew Chem Int Ed Engl47:3524–3535.

11. Yeo YY, Vattuone L, King DA (1997) Calorimetric heatsfor CO and oxygen adsorption and for the catalytic COoxidation reactiononPt{111}. J ChemPhys106:392–401.

12. Goodman DW, Kelley RD, Madey TE, Yates JT, Jr.(1980) Kinetics of the hydrogenation of CO over singlecrystal nickel catalyst. J Catal 63:226–234.

13. Lytken O, et al. (2008) Energetics of cyclohexeneadsorption and reaction on Pt(111) by low-temperaturemicrocalorimetry. J Am Chem Soc 130:10247–10257.

14. Xu B, Zhou L, Madix RJ, Friend CM (2010) Highlyselective acylation of dimethylamine mediated byoxygen atoms on metallic gold surfaces. Angew ChemInt Ed Engl 49:394–398.

15. Bligaard T, Nørskov JK (2008) In Chemical Bonding atSurfaces and Interfaces, eds Nilsson A, Petterson LGM,Nørskov JK (Elsevier, Amsterdam), 1st Ed.

942 | www.pnas.org/cgi/doi/10.1073/pnas.1006652108 Nørskov et al.

Dow

nloa

ded

by g

uest

on

June

9, 2

020

16. van Santen RA, Neurock M (2006) MolecularHeterogeneous Catalysis (Wiley-VCH, Weinheim,Germany).

17. Ertl G (2009) Reactions at Solid Surfaces (John Wiley &Sons, Hoboken, NJ).

18. Nilsson A, Pettersson LGM (2004) Chemical bonding onsurfaces probed by X-ray emission and densityfunctional theory. Surf Sci Rep 55:49–167.

19. Nilsson A, et al. (2005) The electronic structure effect inheterogeneous catalysis. Catal Lett 100:111–114.

20. Holloway S, Lundqvist BI, Nørskov JK (1984) Electronicfactors in catalysis. Proceedings of the EighthConference on Catalysis (Springer, Berlin), Vol IV, p 85.

21. Hammer B, Nørskov JK (1995) Why gold is the noblestof all metals. Nature 376:2238–2240.

22. Gajdos M, Eichler A, Hafner J (2004) CO adsorption onclose-packed transition and noble metal surfaces: Trendsfrom ab initio calculations. J Phys CondMat 16:1141–1164.

23. Hammer B (2006) Special sites at noble and latetransition metal catalysts. Top Catal 37:3–16.

24. Roudgar A, Gross A (2003) Local reactivity of thin Pdoverlayers on Au single crystals. J Electroanal Chem548:121–130.

25. Greeley J, Mavrikakis M (2004) Alloy catalysts designedfrom first principles. Nat Mater 3:810–815.

26. Kitchin JR, Nørskov JK, Barteau MA, Chen JG (2004)Modification of the surface electronic and chemicalproperties of Pt(111) by subsurface 3d transitionmetals. J Chem Phys 120:10240–10246.

27. Nikolla E, Schwank J, Linic S (2009) Measuring andrelating the electronic structures of nonmodelsupported catalytic materials to their performance. JAm Chem Soc 131:2747–2754.

28. Markovic NM, Ross PN (2002) Surface science studies ofmodel fuel cell electrocatalysts. Surf Sci Rep 45:121–229.

29. Tripa CE, Zubkov TS, Yates JT, Jr., Mavrikakis M,Nørskov JK (1999) Molecular N2 chemisorption-specificadsorption on step defect sites on Pt surfaces. J ChemPhys 111:8651–8658.

30. Mills G, Gordon MS, Metiu H (2003) Oxygen adsorptionon Au clusters and a rough Au(111) surface: The roleof surface flatness, electron confinement, excesselectrons, and band gap. J Chem Phys 118:4198–4205.

31. Zhang JL, et al. (2005) Mixed-metal Pt monolayerelectrocatalysts for enhanced oxygen reductionkinetics. J Am Chem Soc 127:12480–12481.

32. Kibler LA, El-Aziz AM, Hoyer R, Kolb DM (2005) Tuningreaction rates by lateral strain in a palladiummonolayer. Angew Chem Int Ed Engl 44:2080–2084.

33. Behm RJ (1998) Spatially resolved chemistry onbimetallic surfaces. Acta Physiol Pol 93:259–272.

34. Schnur S, Gross A (2010) Strain and coordinationeffects in the adsorption properties of early transitionmetals: A density functional theory study. Phys Rev B81:033402.

35. Menning CA, Chen JG (2009) General trend foradsorbate-induced segregation of subsurface metalatoms in bimetallic surfaces. J Chem Phys 130:174709.

36. Gross A (2008) Adsorption on nanostructured surfacesfrom first principles. J Comput Theor Nanosci 5:894–922.

37. Jiang T, et al. (2009) Trends in CO oxidation rates formetal nanoparticles and close-packed, stepped, andkinked surfaces. J Phys Chem C 113:10548–10553.

38. Abild-Pedersen F, et al. (2007) Scaling properties ofadsorption energies for hydrogen-containing moleculeson transition-metal surfaces. Phys Rev Lett 99:016105.

39. Boudart M, Djega-Mariadassou G, eds (1984) Kineticsof Heterogeneous Catalytic Reactions (Princeton UnivPress, Princeton, NJ).

40. Stoltze P, Nørskov JK (1985) Bridging the “pressuregap” between ultrahigh-vacuum surface physics andhigh-pressure catalysis. Phys Rev Lett 55:2502–2505.

41. Dumesic JA, Rudd DF, Aparicio LM, Rekoske JE,Treviño AA (1993) The Microkinetics of HeterogeneousCatalysis (Am Chem Soc, Washington, DC).

42. Stegelmann C, Andreasen A, Campbell CT (2009)Degree of rate control: How much the energies ofintermediates and transition states control rates. J AmChem Soc 131:8077–8082.

43. Hansen EW, Neurock M (2000) First-principles-basedMonte Carlo simulation of ethylene hydrogenationkinetics on Pd. J Catal 196:241–252.

44. Linic S, Barteau MA (2003) Construction of a reactioncoordinate and a microkinetic model for ethyleneepoxidation on silver from DFT calculations andsurface science experiments. J Catal 214:200–212.

45. Reuter K, Frenkel D, SchefflerM (2004) The steady stateof heterogeneous catalysis, studied by first-principlesstatistical mechanics. Phys Rev Lett 93:116105.

46. Honkala K, et al. (2005) Ammonia synthesis from first-principles calculations. Science 307:555–558.

47. Kandoi S, et al. (2006) Prediction of experimentalmethanol decomposition rates on platinum from firstprinciples. Top Catal 37:17–28.

48. Temel B, Meskine H, Reuter K, Scheffler M, Metiu H(2007) Does phenomenological kinetics provide anadequate description of heterogeneous catalyticreactions? J Chem Phys 126:204711.

49. Boudart M (1997) In Handbook of HeterogeneousCatalysis, eds Ertl G, Knozinger H, Weitkamp J (Wiley-VCH, Weinheim, Germany).

50. Bligaard T, et al. (2004) The Bronsted-Evans-Polanyirelation and the volcano curve in heterogeneouscatalysis. J Catal 224:206–217.

51. Nørskov JK, Bligaard T, Kleis J (2009) Rate control andreaction engineering. Science 324:1655–1656.

52. Andersson MP, et al. (2008) Structure sensitivity of themethanation reaction: H2-induced CO dissociation onnickel surfaces. J Catal 255:6–19.

53. Bengaard HS, et al. (2002) Steam reforming andgraphite formation on Ni catalysts. J Catal 209:365–384.

54. Coenen JWE, van Nisselrooy PFMT, de Croon MHJM,van Dooren PFHA, van Meerten RZC (1986) Thedynamics of methanation of carbon-monoxide onnickel-catalysts. Appl Catal 25:1–8.

55. Van Ho S, Harriott P (1980) The kinetics ofmethanation on nickel-catalysts. J Catal 64:272–283.

56. Andersson MP, et al. (2006) Toward computationalscreening in heterogeneous catalysis: Pareto-optimalmethanation catalysts. J Catal 239:501–506.

57. Besenbacher F, et al. (1998) Design of a surface alloycatalyst for steam reforming. Science 279:1913–1915.

58. Toulhoat H, Raybaud P (2003) Kinetic interpretation ofcatalytic activity patterns based on theoreticalchemical descriptors. J Catal 216:63–72.

59. Studt F, et al. (2008) Identification of non-preciousmetal alloy catalysts for selective hydrogenation ofacetylene. Science 320:1320–1322.

60. Greeley J, et al. (2009) Alloys of platinum and earlytransition metals as oxygen reduction electrocatalysts.Nature Chem 1:552–556.

61. Greeley J, Jaramillo TF, Bonde J, Chorkendorff IB,Nørskov JK (2006) Computational high-throughput

screening of electrocatalytic materials for hydrogenevolution. Nat Mater 5:909–913.

62. Strasser P, et al. (2003) High throughput experimentaland theoretical predictive screening of materials—Acomparative study of search strategies for new fuel cellanode catalysts. J Phys Chem B 107:11013–11021.

63. Thanh CN, Didillon B, Sarrazin P, Cameron C (2000) USPatent 6, 054,409.

64. Sheth PA, Neurock M, Smith CM (2003) A first-principles analysis of acetylene hydrogenation over Pd(111). J Phys Chem B 107:2009–2017.

65. Sheth PA, Neurock M, Smith CM (2005) First-principlesanalysis of the effects of alloying Pd with Ag for thecatalytic hydrogenation of acetylene-ethylenemixtures. J Phys Chem B 109:12449–12466.

66. Mei D, Sheth PA, Neurock M, Smith CM (2006) First-principles-based kinetic Monte Carlo simulation of theselective hydrogenation of acetylene over Pd(111). JCatal 242:1–15.

67. Bond GC (1962) Catalysis by Metals (Academic, NewYork).

68. Shustorovich E, Bell AT (1988) The thermochemistry of C2

hydrocarbons on transition metal surfaces—The bond-order-conservation approach. Surf Sci 205:492–512.

69. Kohan AF, Ceder G, Morgan D, van de Walle CG (2000)First-principles study of native point defects in ZnO.Phys Rev B 61:15019–15027.

70. Solans-Monfort X, Branchadell V, Sodupe M, Sierka M,Sauer J (2004) Electron hole formation in acidic zeolitecatalysts. J Chem Phys 121:6034–6041.

71. Pacchioni G (2008) Modeling doped and defectiveoxides in catalysis with density functional theorymethods: Room for improvements. J Chem Phys 128:182505.

72. Chrétien S, Metiu H (2008) O2 evolution on a cleanpartially reduced rutile TiO2(110) surface and on thesame surface precovered with Au1 and Au2: Theimportance of spin conservation. J Chem Phys 129:074705.

73. Huang P, Carter EA (2008) Advances in correlatedelectronic structure methods for solids, surfaces, andnanostructures. Annu Rev Phys Chem 59:261–290.

74. Eder F, Lercher JA (1997) Alkane sorption in molecularsieves: The contribution of ordering, intermolecularinteractions, and sorption on Brønsted acid sites.Zeolites 18:75–81.

75. Swisher JA, et al. (2010) Theoretical simulation of n-alkanecracking on zeolites. J Phys Chem C 114:10229–10239.

76. Dion M, Rydberg H, Schröder E, Langreth DC,Lundqvist BI (2004) van der Waals density functionalfor general geometries. Phys Rev Lett 92:246401.

77. Chakarova-Käck SD, Schröder E, Lundqvist BI,Langreth DC (2006) Application of van der Waalsdensity functional to an extended system: Adsorptionof benzene and naphthalene on graphite. Phys RevLett 96:146107.

78. Zhao Y, Truhlar DG (2008) Density functionals withbroad applicability in chemistry. Acc Chem Res 41:157–167.

79. Jóhannesson GH, et al. (2002) Combined electronicstructure and evolutionary search approach to materialsdesign. Phys Rev Lett 88:255506.

80. Curtarolo S, Morgan D, Persson K, Rodgers J, Ceder G(2003) Predicting crystal structures with data mining ofquantum calculations. Phys Rev Lett 91:135503.

81. Oganov AR, Glass CW (2006) Crystal structureprediction using ab initio evolutionary techniques:Principles and applications. J Chem Phys 124:244704.

Nørskov et al. PNAS | January 18, 2011 | vol. 108 | no. 3 | 943

Dow

nloa

ded

by g

uest

on

June

9, 2

020