Chemisorption on Single Crystal Planes *BY GERT EHRLICHGeneral Electric Research and Development Center,P.O. Box 8, Schenectady, N.Y., 12301, U.S.A.Received 3 1st January, 1966Chemisorption on well characterized crystal planes, as well as on atomic clusters, can now beroutinely observed. Results so far obtained on tungsten are reviewed, with special emphasis on(i) bonding of metal atoms to their own lattice, (ii) the localization of distinct binding states on in-dividual crystal planes, and (iii) atomic and electronic displacements arising from the act of chemi-sorption.The relation between surface structure and adsorption has persisted as one ofthe important but unsolved problems in surface chemistry. This has again beenbrought to the forefront of attention by the realization that even in the chemisorptionof simple diatomic gases, bonding occurs in discrete states with distinct bindingenergies, surface dipoles, and chemical properties.1 Are these states adsorbed oncrystallographic features differing in atomic arrangement, and therefore chemicalreactivity, or are changes in the surface environment, brought about by the actof adsorption itself, responsible ? Experiments on polycrystalline samples suggestthat both effects may operate in one and the same system; they do not, however,permit an unequivocal assignment of a given binding state to a particular surfacefeature.Adsorption studies on well-defined single crystal planes are now feasible andcan in principle establish both the atomic configuration of the chemisorbed com-plex as well as the variations in its behaviour at different surface features.Thereare two ways of executing these : one is to make observations under high resolution,as is possible in the field ion microscope,2 to reveal the individual crystal planespresent on any specimen ; the other is the brute force approach of preparing macro-scopic single crystal planes, which can then be examined by conventional means.The former is easier and may yield information on processes at surface imperfections,as well as on the individual atomic events. Examination of large scale single crystalsurfaces, however, suffers from fewer of the uncertainties which hedge about theinterpretation of high field experiments. Moreover, it is only by learning to controlsurface structure on a large scale that we shall be able to take advantage of theinformation on structural specificities now being uncovered. Here we presentsome of the more interesting structural effects found for individual adsorbed atomsand molecules in both types of studies.SELF-ADSORPTION OF LATTICE ATOMSMetal atoms adsorbed on their own lattice constitute a simple example ofchemisorption; they are also of interest for their possible role in chemisorptionand catalysis.Despite this, experimental information on the behaviour of atomsheld on different planes has not been available. Such information can, however,*supported in part by the U.S. Air Force Office of Scientific Research, Office of AerospaceResearch.8 CHEMISORPTION O N SINGLE CRYSTAL PLANESbe derived in a straightforward fashion in the field ion microscope.A single atomis deposited on a given crystal plane at a low temperature. The mean square dis-placement (r2) at a more elevated temperature can then be determined by directobservation of the atom movements, and this in turn yields the diffusion coefficientD. The usual relationshipbetween the mean square displacement (r2) along a given direction and the period ofobservation z in a random walk is not directly applicable; boundary effects put acertain amount of correlation into the otherwise random motion of the migratingatoms. This can be simply accounted for, however, provided the separation of theboundaries a is sufficiently large so that a2> 2002.For example, in a one-dimensionalrandom walk the mean square displacement becomes 3a(r2) = 2 0 2 (1)both for boundaries that reflect as well as for those that adsorb a colliding atom.Quantitative measurements of the movement of individual tungsten adatomsat different,, temperatures have been carried out 3b on the (1 lo), (21 1) and (321)planes of the tungsten lattice and yield the activation energies for surface diffusionlisted in table 1 ; for the latter two planes, these values pertain specifically toatomic motion along close-packed atom rows. In moving over the surface thetungsten atoms serve as a convenient probe, mapping out the potential acting uponan atom at different points at the surface. Estimates of this potential have beenmade by Drechsler 4 on the assumption of pairwise additive interactions, andactivation energies derived in this way are also listed in table 1.It is apparent thatTABLE ENERGETICS OF SINGLE TUNGSTEN ATOMS ON TUNGSTEN SURFACES(kcal/mole)diffusion barrier, YDexpt. theoryplane diffusion direction(1 10) random 22 31(21 1) parallel 13 41(321) parallel 20 43desorption barrier, xexpt. theory - 50 137178- 176apart from quantitative differences, pair interactions give the wrong qualitative orderof mobilities on these three planes. Of particular interest are the differences be-tween the almost identical (21 1) and (321) planes. Although the much smaller valueon the (211) is not consistent with simple pair interactions, it is in qualitative accordwith the behaviour of conduction electrons at a surface outlined by Smoluchowski.~The electron cloud tends to concentrate in the atom troughs, leaving atoms at alattice step partly denuded.This must be expected to weaken bonding at suchsites. The (321), however, is less steeply stepped than the (211), and should there-fore be less affected by an electron redistribution, in agreement with experiment.Most important for the moment is the indication that pair interactions fail to describesurface bonding even qualitatively.Diffusion data reveal differences in the energy of an atom bound at differentsurface positions. Equally interesting but more difficult to measure are the absolutebinding energies themselves. In principle the desorption energy of a metal atomfrom a given crystal plane can be determined from measurements of the electricfield2 required for evaporation at very low temperatures (TI20"M).For aG. EHRLICH 9atom on its own lattice, this field E is related to the desorption energy x (in eV)throughHere n is the charge of the ion evolved in field desorption, In the energy for ionizinga gas-phase atom to the charge state n, and # the work function of the area fromwhich the atom if removed. The term cE2 formally resembles a polarizabilitycorrection ; in fact it also hides other effects such as field penetration into the lattice,and is best considered an empirical correction term. For tungsten atoms at a growthposition, i.e., at kinks in a lattice step, Muller and Young 6a determined an evapora-tion field of -5.4 V/A at 77°K.A value of 5.6 V/A has recently been found at20°K by a more direct method, in which the relation of applied voltage to the fieldis determined from the slope of the Fowler-Nordheim plot in field emission. Thesemeasurements suggest that the polarizability term in eqn. (3) is relatively minor,amounting to less than 15 % of the observed desorption energy of the atom.So far, only tungsten atoms held on the (1 10) plane have been studied ; theseare desorbed by fields as low as 4-4 V/A. In view of the high work function of the(1 10) plane (5.96 eV),6b it might be most favourable for tungsten atoms to evaporateas triply charged ions, and eqn. (3) then yields a desorption energy of -50 kcal/mole.These results are still preliminary; they do reveal a sizeable disparity be-tween experiment and estimates based on pair interactions only. It will be especi-ally interesting to extend these observations to atoms held on other planes, andto correlate variations in binding energy with the atomic environment.x+ I,, - n# + cE2. (3) 3,79n3/2E1/2 =SURFACE STRUCTURE AND INDIVIDUAL BINDING STATESQuantitative determinations comparable to those for tungsten atoms self-adsorbed on their own lattice are not yet available for the chemisorption of diatomicgases. However, even though essentially qualitative, observations on diatomicgases have revealed a fascinating selectivity. In the ion microscope, nitrogeninteracting with tungsten at 300°K is found to leave the (1 10) plane bare, even thoughall the surrounding areas are heavily covered by a strongly bound material.Atlow temperatures (T< 150°K) gas can be discerned on this plane. In its heat ofadsorption (9 kcal/mole) this material resembles the y state previously identified 1on polycrystalline samples. The fact that despite its low heat of adsorption theadsorbed gas withstands the field required for imaging (E24V/A) is significant.If nitrogen were held on the (110) as molecules, its field desorption should resemblethat of CO of comparable desorption energy. Carbon monoxide is removed atE13.5 V/A,8a whereas a field of 4.3 V/A is required to desorb nitrogen. Assumingdissociative adsorption, however, the binding energy of a nitrogen atom amountsto 117 kcal/mole ; at 20°K the adatoms are immobile and it is their stability thatdetermines the behaviour in the applied field.Despite the very small heat of ad-sorption from the molecular gas, nitrogen must therefore be held as atoms on the(110). These findings have been confirmed and enlarged by observations onlarge-scale single crystal planes, using a combination of contact potential and flashdesorption techniques.9 At 300°K the work function of the (1 10) remains unchangedunder continuous exposure to nitrogen at pressures as high as lO-4mm. At lowertemperatures ( T I 130"K), or higher pressures, the work function is found to dropand the material desorbed on flashing follows second-order kinetics, again indicatingthe participation of adatoms.The absence of nitrogen adsorption on the (110)at room temperature has also been verified 10 by measurements in the field emissionmicroscope10 CHEMISORPTION ON SINGLE CRYSTAL PLANESThe specificity of the (110) plane of tungsten toward nitrogen is extreme. Hereis a clean surface capable only of relatively weak interactions and therefore un-populated under ordinary conditions of temperature and pressure, under whichplanes of different orientation are saturated. Surfaces other than the (1 10) also reveala distinct though less dramatic localization of different binding states, which is indic-ated by the experimental data 9 in table 2. The a state of intermediate binding energyhas been identified only on the (1 1 1). The (100) appears unique in having two differentlow temperature y states, with dipoles of opposite sign.A split of the high tem-perature p state has so far been found only on the (1 11) plane. This highly selectivenitrogen adsorption is at least in part due to the very high dissociation energy andvalence requirements of the nitrogen molecule. To be thermodynamically profitable,dissociative chemisorption requires sites that exceed the very strong interactionsbetween nitrogen atoms held to each other in the molecular gas. This require-ment can be met at only a limited number of surface sites. However, even in themolecular chemisorption of carbon monoxide pronounced structural specificitieshave been established.TABLE 2.NITROGEN ADSORPTION ON INDIVIDUAL TUNGSTEN PLANESdipoleD plane T"K state sticking energy association(1 10) 110 Y -9 atoms > -04desorptioncoefficient kcal/mole- 10 molecules >O- -11 molecules <O*25 75 atoms *4- -9 molecules -1(100) 110 Y+300 B-(1 11) 110 Y220 a (-005 16 molecules (-1300 p1+p2 <*a -75 atoms - -02That differences in binding on different crystallographic features of a tungstensurface were responsible for the distinct states of CO was already inferred 1 frommeasurements on polycrystalline filaments, in which four distinct states have beenisolated at 300°K: an a state, forming only at high coverages, and three p sub-states. A detailed assignment of these states to a specific surface feature has notyet been accomplished.However, examination in the field ion microscope hasrevealed several interesting structural effects.As is apparent in fig. 1, adsorptionof CO at 300°K leads to a fairly uniform distribution over the higher index planes.This is in distinction to the behaviour of nitrogen,7 for which the (1 11) and its en-virons proved comparatively unreactive, and indicates that the rate of adsorptionof CO is much the same on the different atomic configurations. Again unlikenitrogen, CO does chemisorb on the densely packed (1 10). The (1 10) plane, how-ever, appears to fill up later than its surroundings, and in this respect resembles therate behaviour already established for the p1 and a states.1The nature of the material held on the (1 10) plane is clarified by experimentsat lower temperatures.In fig. 2d, CO has been adsorbed at 77"K, after which thesurface was warmed to 190°K and then recooled for another picture, shown in fig. 26.Warming removes four of the five bright spots initially present on the (110), estab-lishing that the CO is moderately labile. This is in agreement with the observationsat room temperature. Carbon monoxide impinging upon the (1 10) would migratetoward the periphery, where it can be held more strongly, leaving the (110) itselfless densely populated. Either the a or the state could conform to the generaFIG. 1.-(a) Field ion micrograph of clean tungsten, shaped by field desorption at 20"K, imageexposure to CO while at 300°K; image, taken at 15.9 kV and 20°K to minimizearrows indicate triplets.FIG.2.-(a) View of tungsten surface after adsorption of CO at 77°K ; box indicates locationlost all but one CO after warming to 190°KFIG. 3.-(a) Clean tungsten. (b) View of surface in (a) after heavy oxygen adsorption atFIG. 4.-(a) Tungsten atoms deposited on clean tungsten at 20°K. (b) Same surface after warmingarrow denotes displaced tungsten atom on (1 10) ; no nitrogen is apparenG. EHRLICH 11behaviour observed for carbon monoxide on this plane. Again the stability underthe conditions of observation can be used to good advantage. The a state has adesorption energy of -20 kcal/mole or less. For CO bound with this energy onthe high index planes of tungsten, it follows from the work of Swanson and Gomer 8athat desorption should be complete in a matter of minutes at 3-7 V/A and at 20°K.On the (110), with a work function 1.5 eV greater than the average, the ionic stateis stabilized and desorption should occur at even lower fields.The gas held on the(1 10) at moderate concentrations, as in fig. 1, must therefore be PI.Evenafter warming to 700°K this plane is not cleared of adsorbed material; desorptionis completed only at E 2 5 V/A. An unequivocal interpretation of this change isnot yet possible; it is, however, consistent with the formation of a CO on the(110). During observation of the surface the adsorbed material is subjected tocontinuous electron bombardment, which is known to fragment the a state intocarbon.8b The lability of the material on the (1 10) at low concentrations indicatesthat such fragmentation does not yet occur. The change toward an extremelystable state at high concentrations cannot just be due to filling up of the periphery-this is substantially populated even when material on the (1 10) has been found tobe labile.Rather the stability of the high concentration layer on the (1 10) pointsto the presence of decomposition products on the surface. This in turn suggeststhat a CO is formed on the (110), but only when it is heavily populated by carbonmonoxide in the j31 state.In its interactions with carbon monoxide, just as with nitrogen, the (1 10) planeof tungsten appears capable of only relatively weak interactions. Since adsorptionof CO is molecular, however, there is not the sharp distinction between planes foundfor nitrogen and it still remains to determine the surface features on which p 2 and j33states are formed.The weak interactions of the (1 10) in general must be assignedto an electronic, rather than to any strictly geometrical effect. For nitrogen thelatter cannot enter at the low concentrations at which observations have beenmade. A satisfactory explanation for the low binding energy of atoms and mole-cules on the (1 10) has not yet been offered ; qualitatively this can perhaps be cor-related with the high degree to which tungsten atoms within this plane are sur-rounded by other lattice atoms, and therefore to the saturation of their valence.Perhaps the most important feature to emerge from observations of chemi-sorption on single crystal planes is the complexity of the effects.Even the (1 10)plane appears capable of accommodating a variety of different binding states.These evidently arise as a result of the heterogeneity induced in a given plane by thepresence of adsorbed material: one state of binding changes the environment inits proximity, and therefore creates sites for an entirely different type of interaction.The nature of these induced sites differs from one plane to the next, as do the intrinsicbinding properties of the bare surfaces themselves.At higher concentrations the behaviour of CO on the (1 10) is different.STRUCTURAL CHANGES I N ADSORPTIONIt has long been recognized that the atomic arrangements and spacings at a crystalsurface may differ from those in the bulk.Chemisorption involves an energychange of much the same order as is expended in creating a surface site from the bulk ;it should therefore also be capable of altering the structure of the surface layer.This is especially true when chemisorption is a precursor to the formation of adistinct new phase, such as an oxide. The extent to which rearrangement does infact occur has not been clear in the past. Recently, however, low energy electro12 CHEMISORPTION ON SINGLE CRYSTAL PLANESdiffraction measurements have been wildly interpreted as indicating a surface re-arrangement in chemisorption,ll involving large scale movements of surface atoms.Even though the field ion microscope is not capable of detecting bond elongationsamounting to only a few per cent of the normal spacing (changes which in principlecan be measured by diffraction methods) long-range displacements should be clearlyobservable. Such observations have been made for tungsten interacting with hydro-gen.12 In this system the act of observing the ion image removes all of the adsorbedmaterial.Despite this the tungsten lattice itself appears undisturbed by the adsorp-tion: ion micrographs before and after adsorption, both at 20 and 300"K, can besuperposed, atom for atom, without showing any removal of atoms from their normalsites.Even more interesting are measurements for oxygen, recorded in fig. 3. Thesurface layer is perturbed during observation, but the adsorbed material is not com-pletely eliminated; this complicates interpretation of the pattern as a whole.Theedge of the (1 lo), however, has remained unmoved, proving that there is no corrosion,no significant reconstruction of this plane either by the adsorption of oxygen or theobservation itself. Subject only to the proviso that small changes in lattice spacingscannot as yet be detected, this is also the conclusion to be drawn from close examina-tion of the dilute layers of carbon monoxide in fig. 1, as well as for small amounts ofnitrogen adsorbed on a clean tungsten surface.7So far, only interactions with well-developed crystal planes have been considered.On real surfaces, especially on thin films and on catalysts, single metal atoms or smallatomic clusters are likely to be important. That these may react differently fromatoms embedded in the surface is revealed in fig.4. Three tungsten atoms weredeposited on the (I 10) plane of tungsten at 20°K ; this was then exposed to nitrogenwhile at 263°K. Photographs taken after this cycle give no indication that anynitrogen has adsorbed at the tungsten adatoms on the (1 10). Although further testswill be desirable, it appears that at room temperature the absence of nitrogen fromthe (110) cannot be simply related to the high co-ordination number (9 of tungstenatoms within this plane. An adsorbed tungsten atom has at most three nearestneighbours, yet does not appear capable of forming a strong enough bond withnitrogen either. We are now able to create tungsten clusters containing an arbitrarynumber of atoms on the surface.It should be possible to determine the aggregatesize required to establish strong bonding to nitrogen, and therefore also the intimatestructural dependence of bonding.Of primary interest at the moment is the fact that, as a result of the exposure tonitrogen, the tungsten atoms in fig. 4 have been displaced. On warming to 263°Kthe adatoms undergo a mean square displacement of 0.01 &/min. When the (110)is brought to the same temperature in the presence of nitrogen, the mean squaredisplacement is increased by a factor of ten. The detailed mechanism of this dis-placement has not yet been established. However, we know that at this temperaturenitrogen can be dissociated on the (110), even though re-evaporation is so rapid thatthe steady state concentration is too small for observation at pressures below 10-3 mm.If dissociation occurs in the vicinity of an adsorbed tungsten atom, part of the heatof adsorption may be imparted to the tungsten, propelling it to a new site. Whateverthe mechanism, we have here a clear example of the displacement of an isolatedlattice atom by a chemisorbing gas.Again it will be of interest to determine thesize of the atom cluster beyond which these effects are negligible.There are indications that even in a well-formed, relatively smooth plane, neigh-bouring lattice atoms may be affected by the presence of adsorbed material. Thepicture of a tungsten surface after exposure to CO shown in fig. Ib reveals the presencG . EHRLICH 13of triplets : a large bright central atom, and on either side of it, in a straight line, asmaller emitting spot brighter than the surroundings.This configuration is nbtspecific to carbon monoxide. It has been observed with adsorbed oxygen and nitro-gen, as well as with tungsten. These triplets do not represent a complex with tungstenatoms pulled out of the metal. Within the limits to which quantitative comparisonshave been made, the two satellites coincide with atom sites of the bare lattice and atnormal picture voltages are indistinguishable from their neighbours. The separationof the two satellites is quite large-ca 13 A on the (21 1) plane. This, as well as thelinear configuration, eliminates the finite size of the helium atoms used in imageformation as a possible cause of these triplets. It is known, however, that the screen-ing charge around an impurity in a metal does not decay monotonically. Instead, itoscillates, falling to zero as the inverse cube of the distance.13 At present, it is mostreasonable to attribute the appearance of triplets to such an electronic perturbationaround the adsorbed material. To what extent this could affect chemical processesremains to be established.The examination of adsorption on single crystal planes is still in its very earlystages ; however, even the few results outlined here suggest a diversity of interestingeffects that should help in coming to a better qualitative understanding of morecomplex reactions on less clearly specified surfaces.1 for a review, see Ehrlich, Ann. N.Y. A c d . Sci., 1963, 101, 722.2 Miiller, Science, 1965, 149, 591.3 (a) Ehrlich, J. Chem. Physics, 1966, 44, 1050 ;4 Drechsler, 2. Elektrochem., 1954,58, 327.5 Smoluchowski, Physic. Rev., 1941, 60, 661.6 (a) Miiller and Young, J. AppZ. Physics, 1961, 32, 2425;(b) Young and Muller, J. AppZ. Physics, 1962, 33, 91.7 Ehrlich and Hudda, J. Chem. Physics, 1962, 36, 3233.8 (a) Swanson and Gomer, J. Chem. Physics, 1963,39,2813;(b) Menzel and Gomer, J. Chem. Physics, 1964, 41, 3329.9 Delchar and Ehrlich, J. Chem. Physics, 1965, 42, 2686.10 van Oostrom, Thesis (University of Amsterdam, 1965).11 for review, see May, Ind. Eng. Chem., 1965, 57, 18.12 Ehrlich and Hudda, Phil. Mag., 1963, 8, 1587.13 Friedel, NUOVO Cim. Suppl., 1958, 7, 298.(b) Ehrlich and Hudda, J. Chem. Physics, 1966, 44, 1039