GENERAL DISCUSSIONDr. A. A. Holscher (Amsterdam) said : Ehrlich’s interpretation of the changesobserved in the ion images after adsorption of carbon monoxide and nitrogen ontungsten differs completely from the one presented in our paper. In his picture thedisorderly arranged bright dots are identified with individual atoms (N) or molecules(CO) on an undisturbed metal surface, whereas we concluded that the surface structureis highly disturbed by the adsorption, the bright dots being tungsten atoms. Thepresence of adsorbate is marked by a diffusiveness in some parts of the ion image.In studying the behaviour of the (1 10) face Ehrlich compared ion images obtainedafter adsorption at 20°K with those obtained after heating the imaged surface to roomtemperature or higher.By measuring the electron emission after imaging we showedthat most of the adsorbed nitrogen is removed from the Iront surface of the tip by theimaging procedure. However, the shanks of the tip and the support remain denselycovered. Heating such a locally cleaned tip assembly will most probably lead toreadsorption on the front surface, either by surface diffusion or via the gas phase.This readsorption may drastically alter the original array of bright dots, and may leadto a disappearance of the dots on the (1 10) face. An example of this effect is shown infig. a and b. Fig. a shows a clean field-evaporated tip, but by accident the zone“ decoration ’’ is also present on the (1 10) face. Fig. b shows the same surface afternitrogen adsorption at 300°K.The three tungsten atoms on the (1 10) face now havedisappeared, apparently due to field desorption of a nitrogen-tungsten complexSimilar changes can be noted in Ehrlich’s fig. 4 and 5. I think therefore that it isdoubtful whether from heating experiments as described one can decide on thenature of the bright dots on the (1 10) face of tungsten as seen after low-temperatureadsorption of nitrogen.The fact that after adsorption of nitrogen on tungsten at 300°K usually no brightdots appear on the (110) (however, we have sometimes observed them!) seems noconclusive proof that adsorption did not take place. The same conclusion would thenalso hold for the adsorption of oxygen on the (1 10) face of tungsten at 300°K. In thiscase, too, the (1 10) remains “ bare ” according to Ehrlich’s fig.3b. If a particularcrystal region does not show up in the ion image, neither before nor after adsorption,it seems difficult to draw any conclusion as to what actually happened in this region.To explain the appearance of bright dots on the (1 10) face of tungsten after adsorp-tion of CO and high coverage Ehrlich proposes a decomposition of the a-state byelectron bombardment leading to carbon formation ; the bright dots are thus identi-fied with carbon atoms. Objections are (i) It is required that after adsorption of p1and a-CO, upon imaging, ionization takes place above the (1 10) face, whereas thisplane remained “ dark ” in the clean state. (ii) In view of the low bond strength ofthe a-state it is most likely that it is completely torn away by the high field before anyconversion can take place.(iii) If any electrons are produced by field ionization abovethe (1 10) they will have no or little energy above the Fermi energy and it seems unlikelythat they should be able to achieve the proposed conversion of the a-state, even if thisspecies were still present at the surface.Dr. Gert Ehrlich (Schenectady, New Yurk) said: In our work with nitrogen thepossibility that the additional emission centres formed by adsorption were actuallydisplaced tungsten atoms was examined at length, but had to be rejected,l’ 21 Ehrlich and Hudda, J. Chem. Physics, 1962, 36, 3233.2 Ehrlich, Adu. Catalysis, 1963, 14, 255.5[To face page 54FIG. l(a).-Three tungsten atoms deposited on (1 10) of tungsten kept at 20°K.(b) Same surface after adsorption ofnitrogen. Two additional emission centres formed on (110). (c) (110) upon warming to 263°K. Tungsten atoms re-main undisturbed ; nitrogen additions have disappeared. All pictures taken at 70 % of the desorption field for W withTo facepage 55.112 sec exposureGENERAL DISCUSSION 55For the (110) plane of tungsten, our F.I.M. observations revealed that after inter-action with nitrogen, either at 300 or 190"K, this surface remains bare. At 145, 77,and 20"K, however, additional emission centres appear ; these could again be elimin-ated by heating to room temperature. The material formed at low temperaturesdiffers from tungsten atoms in being less resistant to field desorption, both in theabsence and in the presence of the image gas, as well as in its thermal stability.Moreover, in our experiments tungsten atoms interacting with nitrogen do notreproduce the behaviour of the centres observed after nitrogen adsorption.Thus, in fig.4a of our paper, three tungsten atoms are present on the (1 10) planeitself prior to nitrogen adsorption. Exposure to nitrogen at 263°K has not swept thesurface clean in fig. 4b; precisely three atoms are still present after this treatment,which has only resulted in a significant shift of one. Adsorption at 20°K is even morerevealing. In fig. a, three tungsten atoms have been deposited on the (I 10). Exposureto nitrogen at 20°K yields two additions to this plane (in b) without displacing thetungsten atoms. After warming to 263°K (c), the three tungsten atoms are still in theiroriginal locations, but the additions produced by nitrogen adsorption are gone. Thereis no indication of significant complex formation by which one might rationalize thecharacteristic properties of the emission centres found after adsorption.It is on these grounds, and not merely based on warming experiments, that weidentify these emission centres as nitrogens.Holscher's comment that it is difficultto draw conclusions about adsorption on a plane if no changes are observed on it is,however, quite appropriate. Indeed, this difficulty was already noted some time ago,in our work on nitrogen, where we state that: " although there are no changesapparent on the (110), adsorption on this plane is not necessarily ruled out.. . . Theact of forming the helium-ion image may denude this particular region of adsorbedmaterial without exerting such a strong effect elsewhere. This possibility can bereadily eliminated, however. . . . If nitrogen is allowed to interact with this sample at20"K, adsorbed material is immediately apparent on the (1 10). This clearly establishesour ability to detect adsorption on the (1 10) when it occurs, despite the high field ; thelack of change in the appearance of the (1 10) at 300°K must therefore be interpreted asindicating the absence of nitrogen adsorption." 1 Our conclusions about theoccupation of the (1 10) by nitrogen, both at 300°K and in the low-temperature range,have been verified by contact potential measurements 2 on a macroscopic (1 10) plane,as well as by independent field emission studies.3as bAs regards carbon monoxide, the intensity of an image point depends upon thelocal supply of helium, as well as on the rate constant for field ionization.An adatomwill influence both. The supply of He depends among other things upon the efficiencywith which a colliding atom transfers its kinetic energy to the lattice. Adsorbed gasincreases the accommodation coefficient of helium and should thereby raise theimage intensity. Adatoms also affect the probability of field ionization, as thisdepends upon the local electric field and the detailed shape of the barrier confrontingthe valence electron of the image gas.There is therefore nothing unusual in beingable to image adsorbed material on a plane which, at the same voltage, was quite darkprior to adsorption.The extent to which adsorbed material fragments during imaging depends upon therate of dissociation by electron impact under imaging conditions, compared with therate of field desorption. Neither the energy required for dissociation, nor the effi-ciency of this process has been measured under high fields. Without this quantitative1 Ehrlich and Hudda, J. Chem. Physics, 1962, 36, 3233.2 Delchar and Ehrlich, J. Chem. Physics, 1965, 42, 2686.3 (a) van Oostrom, Philips Res. Rep. Suppl., 1966, no. 1 ; (b) private communication56 GENERAL DISCUSSIONinformation, the interpretation of image changes observed on the (110) at high COconcentrations will remain equivocal, as already indicated in my paper.However,in flash desorption measurements from a (110) plane of tungsten, both the a and P1states have been isolated.1Apart from differences in interpretation, our experimental results, both for CO andfor nitrogen, differ significantly from those reported by Holscher and Sachtler (H & S).Although the reasons for these discrepancies are difficult to establish with certainty,the following must be kept in mind: (a) Even spectroscopically pure gases maycontain COa and other impurities at levels up to 0.2mole %. For CO, furthercontamination is introduced by disproportionation within the storage bulb duringbaking. Oxygen thus liberated is particularly dangerous, since it has been establishedthat surface rearrangement occurs during observation of oxygen covered tungsten.In our own work with CO, contaminants were therefore specifically removed justprior to the adsorption studies by selective adsorption and passage of the gas over acold finger at 63OK.2 With nitrogen, difficulties are less and purification over nickelfilms is sufficient.(b) The measurements recorded by H & S were done at 78°K.Successful operation of the F.I.M. at this temperature depends upon the use of veryfine emitters,3 for which the best image field approaches that for evaporation of thelattice atoms.4 In our studies all images were obtained with a liquid hydrogen cooledemitter, at -20°K. This by itself reduced field desorption.Judging from the workof Swanson and Gomer,s at a constant field of 4-71 V/A the rate of field desorption ofPCO diminishes by a factor of 10 just in going from 60 to 20°K. Furthermore, inour studies operation at 20°K made possible observations on larger tips, not as subjectto the deleterious influence of the field. (c) In the course of preliminary studies withCO, surface rearrangement became apparent during prolonged observations. In oursubsequent work, the first picture has therefore been taken considerably below the bestimage field, at less than 2/3 the desorption field ; the field itself was imposed for theshortest interval consistent with successful photography. In contrast, the imagefield of H & S ranges from 84 to 89 % of that for tungsten desorption, with at leasttwice the exposure to promoted field desorption.That there are severe difficulties in observing adsorbed gases, and that the partial(or sometimes complete) removal 2 , 6 of adsorbed gas is brought about by promotionof field desorption through the image gas, was clearly established some time ago.7The work function measurements by H & S provide an interesting quantitative accountof these, but do not introduce qualitatively new information requiring a re-evaluationof previous studies.Dr. D.Brennan (University of Liverpool) said : Displacement of surface atoms asconsidered by Holscher would be expected to influence the heat of adsorption.However, we underline the fact that most of the gas increments were adsorbed with acharacteristic heat, which is not as might have been expected on the basis of a largedegree of surface damage.The lower heats sometimes observed at low coverage areinteresting, but like Cernk, we believe that a more systematic study of this region mustprecede any useful discussion of the effect.Dr. J. Volter (Inst. anorg. Katalyseforschung, Berlin) said : Ehrlich has demon-strated the pronounced selectivity of individual crystal planes in chemisorption. We1 May and Germer, J. Chem. Physics, 1966, 44, 2895.2 Ehrlich, Adv. Catalysis, 1963, 14, 255.3 Drechsler and Wolf, 4th Int. Conf. Electron Microscopy, (1958), p . 835.4 Miiller and Young, J. Appl. Physics, 1961, 32, 2425.5 Swanson and Gomer, J . Chem. Physics, 1963, 39, 2813.6 Mulson and Miiller, J.Chem. Physics, 1963, 38, 2615.7 Ehrlich and Hudda, Phil. Mag., 1963, 8, 1587GENERAL DISCUSSION 51have studied various catalytic reactions on individual crystal planes of Cu and Ge.1We have found considerable selectivity, too. Now we have studied the p-H2-conver-sion of individual crystal planes of Ni and Cu single crystals in a static system. Theactivation energy on Ni varies from 3-2 kcal on the (1 11) plane to 5.2 kcal on the (1 lo),and 6.4 kcal on the (100) plane. On copper the corresponding values for the (1 11)plane are 9.1 and for the (100) plane are 13.5 kcal. The catalytic activities on thedifferent planes of Ni as well as of Cu vary by the factor of 2. These results mayshow some correlations with those of Ehrlich.(i) The accord of chemisorption pro-perties of a definite crystal plane on the small tip in the field ion microscope and on thelarge-scale single crystals give further evidence that it is possible to compare themacroscopic surface with the corresponding lattice plane. (ii) As expected fromchemisorption, in catalysis there is a specificity of the crystal planes, too. (iii) Ehrlichfinds on tungsten, that the closest-packed plane gives no or only very weak interactionwith nitrogen or CO respectively. He assigns this to an electronic rather than to a geo-metric effect. Eley 2 postulated that on the closest-packed plane the heat of hydrogenadsorption reaches a minimum, and that this plane must exhibit maximum activityin p-H2-conversion.He studied this reaction on polycrystalline metal films.We studied the reaction on single crystals and could verify Eley’s prediction. Onthe closest-packed planes of nickel and copper the activation energy of the conversionwas a minimum. This accord is not fortuitous and our results may be explainedtoo by energetic rather then geometrical factors.The importance of the energetic factor may be supported by Ehrlich’s result thatchemisorption of nitrogen on the (1 10) plane is not influenced by additional adsorptionof 3 tungsten atoms on this plane, i.e., he can vary the geometrical arrangement of thetungsten atoms to a considerable extent without altering their chemisorption properties.It might be possible to explain all specific effects of crystal planes by difference indislocation density or quality.We compared in several reactions the catalyticactivity of forms polycrystalline and monocrystalline of Cu and Ni. Considering thegrain boundaries as a special type of dislocation, we have in polycrystalline materiala much higher dislocation density. But we never found a specific positive or negativecatalytic effect of polycrystalline material compared with monocrystals. Thereforewe think the dominating factor is the lattice plane and not the dislocation.Dr. Gert Ehrlich (G.E., Schenectady, N.Y.) said: Dr. Volter’s findings thatdislocations have little effect on the catalytic reactivity of metals should hold gener-ally. A dislocation constitutes an atomic arrangement of higher energy than thatof the ideal lattice; as such, its termination at a surface can be expected to affectthe chemical reactivity of that particular crystal plane.However, after annealinga reasonably pure metal, the dislocation density is only of the order of 108/cm2or less compared with a total of - 1015 atoms/cm2. Reaction at the more numerousordinary sites should therefore overwhelm any contribution from the dislocations.Only in systems with extreme differences in reactivity, for which a reaction cannotoccur at ordinary surface atoms, are dislocations likely to play a significant role.Dr. Z. Knor (Inst. Physic. Chem., Czechoslovak Acad. Sci., Prague) said: First, Iwould like to know, if Dr. Oostrom (whose Thesis was quoted in Ehrlich’s paper)used for the study of nitrogen adsorption on the (1 10) plane of tungsten the fieldelectron or field ion emission.If the former, then was it possible to study the adsorp-tion on this plane which has such a high work function?1 Volter and Kordel-Kriiger,Z. anorg. Chem., 1964,329,261.2 EJey and Shooter, J. Catalysis, 1963, 2, 259.Volter and Schon, 2. anorg. Chem.,1963, 322, 202. Rienacker and Volter, 2. anorg. Chem., 1959, 302, 295 and 29958 GENERAL DISCUSSIONWe have used the usual type o€ F.E.M., in which the tungsten tip was treated athigh temperature in the presence of high electric field (which was the reverse of thatused for electron emission and thus the surface is protected against contamination and/or ion bombardment). The details will be published elsewhere.1 In this way,starting with the clean surface of tungsten (fig.l), we have obtained surfaces, coveredwith clusters of tungsten atoms (see fig. 2a, 3 4 , which can be converted again to theoriginal shape by mere heating to about 2300°K. During this procedure the diameterof the tip slightly increased in the ratio 1 : 1-4 (estimated by the method of Drechsler 6 ) .These clusters or microtips emit electrons at much lower voltage than the originalsurface, because of the increased intensity of the electric field in their neighbourhood(due to their small dimensions) (compare the legend to fig. 1 and 2 or 3 respectively).This behaviour of metal tips is well known, e.g., ref. (2)-(5). When hydrogen oroxygen was adsorbed on these microtips, no geometrical change was observed (fig.2 4 b and 3a, b, c). The emission current decreased in the same manner approximatelyas for the adsorption on the original surface (e.g., the change of the work functioncaused by oxygen adsorption on microtips was - 1.0 eV compared with the valueN 1.5 eV for the original surface, both estimated from the change in the slope ofFowler-Nordheim plot).From these results we conclude that probably no rearange-ment of these large clusters proceeds during these processes. This conclusion is inagreement with the results of Dr. Vernickel presented at the 3rd Int. Congr. VacuumSci. Tech. (1965 Stuttgart), who prepared the microtips by means of ion bombardment.With an adsorbed layer of hydrogen we have found a strong influence of this layeron the course of the subsequent high temperature treatment in absence of the fieldDr.Gert Ehrlich (Schenectady, New York) (communicated) : Dr. van Oostrom’smeasurements on the (1 10) plane were made in a field emission microscope equippedwith a Faraday cage, to which electrons from only a single crystal plane were admittedthrough a narrow probe hole. The work function of the clean plane was thendetermined by measuring both the slope of the Fowler-Nordheim curve and the half-width of the total energy distribution of the emitted electrons. Changes in the workfunction on adsorption were obtained as usual from changes in the Fowler-Nordheimslope.Prof. J . Oudar (University of Paris) said : During a study of reversible chemisorp-tion of sulphur on silver we have shown that this phenomenon is strongly affected bythe crystalline orientation of the metal.The heats of adsorption determined on thethree low index surfaces (1 11) (100) and (1 10) were as follows. (Values for 1 mole ofS 2 for a degree of coverage 50 % of the maximum) ; AH(ll1) = - 54 kcal, AH(100)= - 58 kcal, AH( 1 10) = - 66 kcal. These values show that the adsorbed atoms arefixed less strongly when the surface has a greater density of metallic atoms. On themost dense faces, (1 11) and (loo), the heat of adsorption is approximately constant asa function of coverage whereas it decreases for the (1 10) face. This suggests that forthe temperature range studied, 3O0-45O0C, a single absorption state exists for (1 11)and (100) while several states exist for the atomically rougher (1 10) plane. We havealso found that the structural defects associated with atomic steps are sites morereactive than the sites normalIy present on the low index planes.(fig. 24.1 Knor and Lazarov, Czech.J. Physics, 1966, 16b (in press).2 Drechsler, Z. Elektrochem., 1957, 61, 48.3 Benjamin and Jenkins, Proc. Roy. SOC. A , 1940,176, 262.4 Meclewski, Nicliborc and Wojda, Acta Physica Polonica, 1962, 22, 525.5 Vanselov, Phys. stat. SOL, 1964, 4, 697.6 Drechsler 2. Elektrochem. 1954,523, 340FIG. 1 .-Original surface of tungsten ; voltage U = 8,000 V.[To face page 58FIG. 2.-Tungsten tip treated at T = 2,300"K ; U' = 10,000 V : (a) clean surface at 78°K ; voltageU = 3,000 V ; (6) surface covered with adsorbed hydrogen at 78°K ; voltage U = 3,000 V ; (c) tipafter high-temperature treatment, T = 1,200"K; in absence of the electric field at 78°KFIG.3.-Tungsten tip treated at T = 2,300"K ; U' = 10,000 V ; (a) clean surface at 78°K ; voltageU = 4,000 V ; (b) surface covered with adsorbed oxygen at 78°K ; voltage U = 4,200 V ; (c) thesame surface as (6) ; voltage U = 5,800 VGENERAL DISCUSSION 59Finally, evidence of attractive forces between adsorbed sulphur atoms leads to theconclusion that the adsorbed layer consists not simply of sulphur atoms but of areconstructed mixed sulphur and silver layer. This hypothesis, compatible with themean separation of sulphur atoms in the layer, is in accord with recent results onother systems by low energy electron diffraction.Dr.H. Mykura (Glasgow Uniuersity) said: I would like to discuss briefly certainaspects of Ehrlich's and Gomer's papers, which they did not elaborate upon. Thisconcerns the adsorption at steps on vicinal surfaces and an experimental technique bywhich adsorption on steps can be differentiated from adsorption on singular crystalsurfaces. On the conventional terrace-and-ledge model of a surface (fig. 1) a vicinalFIG. 1.-The geometry of a vicinal surface : section perpendicular to the ledges.surface consists of" terraces " with monatomic steps (" ledges ") of height a, separatedby a distance s, so that a, the angle between the singular surface and the vicinal surface,is given by sin a = a/s. If the surface free energy of the terraces of singular surface isyo and the line free energy of the step /I, then the surface free energy of the vicinalsurface at inclination a isya = yo cos a+(B/a) sin I a I.Dividing by yo and differentiating with respect to a one obtainsPYo aa YOU1 = -sin a+- cos 1 a 1 ,which reduces to"91 = - PY o ax a=O you(3)in the limit a+O.1 a YY acthTow under suitable experimental conditions the value of - - (the " Herring torqueterm ") is directly measurable from surface-interface intersections.2,3 At present thetorque term measurements can only be done easily at high temperature and on face-centred-cubic metals by measuring the shape of ' 6 thermal etching " grooves at twin-boundary/surface intersections.In principle, grain-boundary/surface intersectionscan also be used.3 As a the step height is known from crystallographic data, the ratioP / y o , which is the relative step energy, can be evaluated.On performing the measure-ment first for a clean surface and then for a surface in equilibrium with adsorbate, thechange in the /3/ yo ratio can be measured.4 An increase in the p l y 0 ratio with increas-ing adsorption indicates preferred adsorption on the terraces, a decrease preferredadsorption on the steps.When the average surface free energy can also be measured as a function ofadsorbate concentration (using, e.g., the zero creep technique),s then the variation in1 Gjostein, Acta Met., 1963, 11, 957.2 Mykura, Acta Met., 1961, 9, 570.3 Shewnion and Robertson in Metal Surfaces, (A.S.M., Metals Park, Ohio), 1963, p.67-98.4 Robertson and Shewmon, J. Chem. Physics, 1963, 39, 2330.5 Buttner, Funk and Udin, J. Physic. Chern., 1952, 56, 65760 GENERAL DISCUSSIONadsorbate concentration with crystal orientation can be evaluated using the Gibbsadsorption equation.1The earlier experimental results obtained by this technique dealt with oxygenadsorption on Ag and Cu. For these metals on both (1 11) and (100) vicinal surfaces?oxygen adsorbs preferentially on the low index surface 2,3--the ratio increasedwith partial pressure of 0 2 in the annealing atmosphere. Other systems, however,behave differently ; sulphur on copper (1 00) vicinal surfaces showing preferred ledgeadsorption.2 Recent results on platinum heated in air at 1100°C 4 show that whileFIG.2.-Orientation dependence of surface free energy of platinum at 1100°C. Upper curves :platinum in vacuo ; middle curves : in air ; lower curves : calculated differences in surface coverageof oxygen (I?, is surface coverage at angle a from the low index orientation) ; result evaluated fromexperimentally determined torque terms.c: .- U-c,l- d, IN RA3IANS.1021- Ii 0 1 c2 0 3 0 4I1 0 2 1 1.02 11 0 0 L loo 1 0.1 0.2 0.3FIG. 3.-Torque terms and relative surface free energy for copper annealed in H2+0.12 %HzSmixture at 830°C.p / y o increases for (100) vicinal surfaces, it decreases for (1 11) surface and increasesmarkedly for (1 10) surfaces. Fig. 2 shows the surface free energies and the relativesurface coverage of oxygen derived from these measurements-while the absolutevalues of surface energy are based on debatable assumptions, the surface coverageresults are considered reasonably reliable.1 Rhead and McLean, Acta Met.1964,12,401.2 Robertson and Shewmon, J. Chem. Physics, 1963. 39, 2330.3 Buttner, Funk and Udin, J. Physic. Chem., 1952, 56, 657.7 M. McLean, Thesis (Glasgow University, 1965)GENERAL DISCUSSION 61If adsorption on the ledges is much stronger than on the terraces, then it is possiblefor the ratio to become negative. This has been observed for the system Ag/S 1and also CufS. Fig. 3 shows the torque terms and the orientation dependence ofsurface free energy obtained by graphical integration of the torque terms for the lattercase : copper heated at 830°C in a hydrogen atmosphere containing 0.12 % H2S.The negative torque terms for vicinal(100) surfaces and the surface energy maximumat (100) are very marked and imply a strong preferred adsorption of sulphur on theledges.In conclusion, relative surface energy measurements can yield much information onorientation dependence of adsorption and-particularly for vicinal surfaces-theadsorption on the different types of site on a surface of given orientation can beidentified and measured separately.Dr.Gert Ehrlich (Schenectady, New York) said: The inability to isolate discretebinding states for potassium on tungsten need not imply a qualitative difference in thedependence of bond strength on structure between metallic and covalent adsorption.The ratio of the heat of adsorption for potassium on the (110) to that on the (100)plane of tungsten ( N 1-4) is comparable to the ratio of desorption energies for the p 2and p1 states of CO, which are easily resolved.The absence of separate states is morelikely a consequence of the rapid equilibration of potassium over the surface and of itsmore continuous adaptation to a changing chemical environment as the surfacebecomes increasingly populated with other potassium atoms.At low temperatures, the adsorption of carbon monoxide itself no longer seems asvirgin as originally postulated and is now similar to the behaviour of nitrogen.1 Forthis system, in adsorption at T< 150"K, the y state successfully competes for surfacesites with the more stable p state.The population in the former as a function oftotal surface population is well described by an S-shaped curve. Adsorption at lowtemperatures, after prior exposure to nitrogen at 300"K, leads to a much lower ypopulation, with a different surface dipole from that obtained by continuous adsorp-tion at low temperatures. For nitrogen the observations are consistent with a model 2in which the strongly held state makes neighbouring sites less attractive to occupa-tion by other /3 nitrogens. Instead, a more weakly held state is formed at low tempera-tures; by its presence this interferes with strong bonding in the immediate vicinity.Inasmuch as the low temperature state now depends upon a cooperative effect (uponsite creation by adsorption itself) its concentration as a function of total coverageshould be given by an S-curve, as experimentally found. Furthermore, adsorption atlow temperatures must be irreversible-the high population of y cannot be achievedby redosing a surface saturated at higher temperatures. Without attempting aquantitative analysis of the data for CO, it appears that much the same model isapplicable to the low-temperature behaviour of this system and that a distinct virginentity may not have to be invoked.Prof. R.Gorner (Univ. of Chicago) (conmunicated) : Ehrlich is right in enumeratingsome of the reasons why discrete binding states are not observed in alkali adsorption.As my paper points out, metallic adsorption is structure-sensitive. However, there isalmost certainly a difference in the relative importance of structural and " purely ''electrostatic factors (in a sense the distinction is arbitrary since both depend on struc-ture) in covalent and metallic adsorption, because the two bond types are different, atleast in their extremes.1 Rhead and Perdereau, Acta Met., 1966, 14,448.2 Ehrlich, J . Chem.Physics, 1961, 34, 29.3 Ehrlich, Structure and Properties of Thin Films (John Wiley & Sons, N.Y., 1959), p. 42362 GENERAL DISCUSSIONWith regard to the separate existence of virgin states in CO adsorption, there is nodoubt from thermal and electron impact results that three distinct adsorption types,viz., virgin, beta and alpha exist. The virgin states may be the analogue of Dr.Ehrlich’s gamma states, but nomenclature is not the issue.The point is whether : (i)the low-temperature population is initially mostly virgin, with a relatively small fractionof beta states (possibly to be found only on certain crystal planes), with the percentageconverted to beta on heating or electron impact a function of coverage, as proposedby Menzel and Gomer,I and Bell and Gomer,2 and restated in my paper ; or whether :(ii) the amount of beta seen after heating or electron impact is on the surface ab initio,in amounts determined by the initial coverage, as suggested by Ehrlich. I favour thefirst hypothesis for the following reasons. Our results seem to show that alpha COis only formed when there is beta CO on the surface.For electron impact on a virgin(20°K) layer, it was shown by Menzel and Gomer 1 that the amount of alpha tenableby the surface went up at a rate equal to that of virgin desorption/beta conversion,while in the thermal case the results of Swanson and Gomer,3 and Bell and Gomer,2indicate that alpha is formed in appreciable amounts only after sufficient heating todrive off some virgin and (by our hypothesis) convert the rest to beta. If Ehrlich’shypothesis is correct these results can be explained only if it is further postulated thatalpha is tenable only if there is no virgin CO on the surface. On the other hand, ifthe formation of beta CO does not involve an activated rearrangement as postulatedin the desorption/conversion hypothesis, it is difficult to see why readsorption aftervirgin desorption should not again produce virgin CO.The point could be settled by electron impact experiments if the dipole moments peradmolecule in the virgin and beta states differed sufficiently ; it would then be possibleto determine from the conversion or desorption rate of virgin whether the initial depositwas virgin or beta, even if there is no desorption at all.Unfortunately the differencein dipole moments appears to be too small to make this practical. A good idea of thesituation could still be obtained by determining the amount of alpha CO tenable onpartial virgin layers before and after thermal or impact desorption ; this can be doneby electron impact desorption rate measurements.1 An increase in the amount ofalpha tenable on low 0 layers after thermal or electron impact treatment, would bestrong evidence for the formation of beta by conversion, rather than for its initial,masked presence.Dr.J. W. Geus (Staatsmijnen, Geleen, Netherlands) said: The distinction betweenmetallic and covalent adsorption made by Prof. Gomer is confirmed by the effects ofadsorption on the electrical conductance of evaporated tungsten films.4 It appearsthat caesium, which is bonded to tungsten surfaces in an analogous way to potassium,increases the conductance. Chemisorption of nitrogen, oxygen, and carbon monoxide,on the other hand, decreases the conductance. Adsorption of hydrogen brings aboutonly a small decrease of the conductance, although the amount of hydrogen adsorbedper cm2 of tungsten surface is of the same order of magnitude as that of the other gasesmentioned. The latter fact points to the chemisorption bond of hydrogen on tungstenbeing more or less of a metallic character.Dr.A. A. Holscher (Anzsterdum) said: The experimental results of Gomer on thelow-temperature adsorption suggest that part of the stable 0-state is formed in the laststage of adsorption. The model as proposed by us, invoking an activated rearrange-ment of tungsten atoms, might give a satisfactory explanation of this and fits in withmany other results on the CO/W system obtained by Gorner and co-workers.1 Menzel and Gomer, J . Chem. Physics, 1964, 41, 3329.2 Bell and Gomer, J. Chem. Physics, 1966, 44, 1065.3 Swanson and Gomer, J.Cliern. Physics, 1963, 39, 2813. 4 Geus, Surface Sci., 1964, 2, 48GENERAL DISCUSSION 63Prof. R. Gomer (Wniv. of Chicago) (communicated) : In reply to Holscher, I do notthink that it is possible to tell from our results at what stage of coverage beta CO isformed. (Probably, most of the eventual beta layer is formed only on heating.)Further, it is not yet known how the ratios for the various species vary with crystal-lographic orientation, so that a discussion of Holscher’s question seems premature.However, there is some indirect evidence from electron desorption that the beta layerobtained at low temperatures by electron impact differs from that obtainable byheating1 This evidence is summarized in fig. 38 of ref. (1) and can be interpreted tomean that heating leads to considerable surface rearrangement of the kind postulatedby Holscher and Sachtler at low temperatures.Dr.Z. Knor (Inst. Physic. Chem., Czechoslovak Acad. Sci., Prague) said: Wehave strong experimental evidence (L.E.E.D., F.I.M.) that the structure of the surfacelayer of a metal is in most cases rearranged during the chemisorption of gas particles.Therefore it seems to me improbable that the spatial distribution of the electrons onthe surface of a metal would not change during this process, even in the adsorption ofalkali metals. If the spatial distribution of the electrons on the surface changesduring the chemisorption, then we can divide the experimental value A 4 into twoparts :where 4 0 and +ads are the work functions of clean metal and a metal covered with theadsorbed layer respectively, A+e is the contribution due to the change of spatialdistribution of the electrons on the surface and A& is the part due to the dipole layer(image, permanent, induced dipoles, etc.), for which we can writewhere ,u is the dipole moment, n is the surface concentration of dipoles and a has thevalue 4n or 271, depending on the type of dipoles.Unfortunately we know nothingabout the contribution of A4e which need not be a negligible one.A4 = 40 - $ads = AA? + Ad%?A4d = a w ,Furthermore it seems to me that in the equationP = 2do4,both do (distance from image plane) and q (adsorbate charge) depends on the kind ofcrystallographic plane on which the adsorption process takes place.I therefore thinkthat the values of p calculated from the equation(as found frequently in the literature) or the values of q calculated for a special casefrom p obtained in this manner, represent only a crude approximation to p or q intheir exact physical meaning (e.g., p = dipole moment of the adsorbed particle -permanent or induced dipole, etc.).Dr. D. A. King (Imperial College) (communicated): Using two independentmethods, we have evaluated sticking coefficients s for the nitrogen + tungsten filmsystem.2 In the temperature range 78-150°K, s was high (0.9), and essentially in-dependent of temperature and also of coverage over the range 0-8 x 1014 moleculescm-2 geometric area of the film. A theoretical interpretation of this invariance wouldbe simplified if s was in fact unity, and the value 0.9 was due to some unsuspectedexperimental artefact.For this reason a cell (fig. 1) was constructed specifically to testwhether or not the sticking coefficient was unity.@o - 4ads) = a w1 Menzel and Gomer, J. Chem. Physics, 1964, 41, 3329.2 Hayward, King and Tompkins, Proc. Roy. SOC. A , to be published64 GENERAL DISCUSSIONWhen gas flows into the cell with a film deposited on the walls (but not up thegauge tubulation, a nickel disc being suspended across the mouth of the gauge duringdeposition of the film 1) it is so directed that it can only enter the gauge after collisionwith the film. Thus, the observation of a pressure increase in the gauge unambiguouslyindicates that s < 1.With the lower portion of the cell immersed in liquid nitrogen andan inflow rate of nitrogen of 1015 molecules sec-1, a pressure increase of 2 x 10-10 torrabove background was recorded in the gauge, confirming that, for N2 on W at 78"K,traps andFIG. 1.s < 1. The possibility that the observed pressure increase was due to a non-adsorbableimpurity in the gas supply was eliminated by isolating the cell from the pumps : thegauge pressure was unaltered by this procedure, whereas such an impurity would haveaccumulated in the cell, resulting in a pressure increase.The high sticking coefficients reported by Gomer for the adsorption of carbonmonoxide on a tungsten ribbon, which are completely at variance with values publishedelsewhere for tungsten filaments 2 9 3 and films,4 has prompted us to investigate theCO + W film system with the present cell.With the lower portion of the cell cooled to78"K, CO was allowed to flow into the cell at a rate of 1015 molecules cm-2; noincrease above the background pressure in the gauge was observed over a period of3 min, indicating a sticking coefficient of unity. It is estimated that the lowest pressurechange that could be detected was -2 x 10-11 torr, so that, taking into account thepressure increase for the Nz + W system where s = 0.9, we can confidently report asticking coefficient of 1.00+0-01 for CO on W at 78°K. It should be noted thats = 1 for films requires that s is also unity for an ideally smooth surface. In ourexperiment the temperature of the incoming gas was -290"K, compared with - 50°Kin Gomer's experiments ; thus, the suggestion 5 that his high values of s can beattributed to the low temperature of his molecular gas beam appears to be invalid.1 Hayward, King and Tompkins, Cltem.Comm., 1965, p. 178.2 Ehrlich, J . Chem. Physics, 1961 , 34, 39.3 Ustinov, Ageev and lonov, Soviet PhysicJ-Tech. Phys., 15165,10, 851 ; (Zh. Tekhn. Fiz., 1965,5 Bell and Gomer, J . Chem. Physics, 1966, 44, 1065.35, 1 106). 4 Ricca and Saini, Gazz. Chim. Ital., 1965, 95, 636GENERAL DISCUSSION 65Dr. L. J . Rigby (Stand. Telecomm. Lab. Ltd., Harlow) said: We have studied theadsorption and replacement of hydrogen, nitrogen and carbon monoxide on poly-crystalline tungsten wires at room temperature by means of desorption spectra.1-4It was found that two types of replacement occurred : (a) a slow replacement of thephases which were reversibly adsorbed at room temperature, e.g., the a phase ofcarbon monoxide and the phase of hydrogen were replaced by nitrogen.In thisprocess, the thermal desorption of a weakly bound phase was followed by adsorptionof the replacing gas on the vacated sites. (b) The p2 phase of hydrogen which wasirreversibly adsorbed at rooin temperature was rapidly replaced by p1 carbon monoxide.The sticking probability of carbon monoxide on a clean tungsten surface was onlyreduced by 36 % when that surface was saturated with hydrogen. This rapid replace-ment suggests that the sites for carbon monoxide adsorption are not covered by phydrogen atoms.Rapid replacement of hydrogen by carbon monoxide has now beenobserved at 78°K. The rate of surface diffusion of p hydrogen atoms at this tempera-ture must be extremely low and supports the suggestion that p2 hydrogen and p1carbon monoxide are not adsorbed on identical but on adjacent sites. This type ofreplacement is a very efficient process and may be responsible for many catalyticreactions.Prof. S . 25. Roginskii (Moscow) said : Direct electron-ionic techniques for studyingmetal surfaces, so much advanced by Muller and Gomer et al, are very efficient forinvestigation of the interaction between gases and metal surfaces in high vacuum andof the relations between the structure of surfaces and chemisorption.The paperspresented today by Ehrlich and Gomer have shown the great diversity of adsorptionspecies even for simple diatomic molecules under very pure conditions, and thepecularities of their interaction. The specificity of the techniques discussed makesthem applicable only for the simplest systems and under conditions far from thosetypical for routine catalysis on metals. True chemisorption during catalysis usuallyis considerably more complicated owing to the complexity of catalysts, to the contactwith mixtures of gases, and the greater complexity of the adsorbate molecules.Moreover, there will be considerable changes of the metal surfaces under the actionof catalytic corrosion.5 Such researches are of greater help for understanding theprocess of formation of real catalyst surfaces, rather than for establishing themechanism of the primary steps in catalysis.Also, the investigation of chemisorption by means of field emission microscopesat a coverage lower than a monolayer (8< 1) the effect of adsorption of many quitedifferent compounds on electron emission (more than forty compounds in ourexperiments) is similar.It is somewhat more specific at 1 < 8 < 2, when there appearlight spots (" molecular pictures ") of a nature that is yet insufficiently clear.6 Athigher coverage the field emission microscopes and the diffraction techniques of lowenergy are not efficient.Dr. D. W. Bassett (Imperial College) said: The way in which information aboutthe adsorption of carbon monoxide on tungsten has been deduced by Holscher andSachtler from field ion micrographs, in spite of apparently complete desorption of theadsorbate during imaging at 78"K, is encouraging.Desorption of the ad-layer is alsovirtually complete with oxygen layers, which I have attempted to examine. This had1 Robins, Trans. Amer. Vacuum SOC., 1962,9, 510.2 Redhead, Vacuum, 1962, 12,203.3 Rigby, Can. J. Physics, 1964, 42, 1256; 1965, 43, 1020.4 Holscher and Sachtler, this Discussion.5 Roginskii, Tretyakov and Shechter, Z h r . Fiz. Khim., 1955, 29, 1921.6 Shishkin and Roginskii, Dokl. Akad. Nartk S.S.S.R., 1962, 143, 37366 GENERAL DISCUSSIONforced me to conclude that using helium ion microscopy one was unable to obtaininformation about the atomic details of oxygen adsorption, and that one was restrictedto the investigation of rather more gross changes in surface structure, such as thosecaused by heating an oxygen-covered specimen.However, it may not be legitimateto extract information from ion micrographs by the methods of Holscher and Sachtlerif metal atoms are also removed from the surface during desorption of the adsorbate.Such surface damage occurs with both nitrogen, and carbon monoxide,l and is veryextensive with oxygen as the adsorbate, as is shown in fig. 1. This photograph was6 0 C>c?I2 3oc2CIHe B.I.V.I'1 I IFIG. 2.-Promoted field desorption of oxygen from tungsten at 78°K indicated by the reduction ofthe voltage required for 0.3 FA field emission, following the application for 1 min of a series ofincreasing desorption voltages (shown relative to the voltage needed for tungsten evaporation) ; (a)desorption in vacuum, (b) desorption in helium at 10-3 torr, (c) desorption in neon at 10-3 torr forthree oxygen coverages, A'p = -0.4, -0.9, and - 1.4 eV.printed from a superposition of a positive of the helium ion image of the clean surface,and a negative of the image of the same surface following the adsorption of a smallamount of oxygen.The many black spots show vacancies in the tungsten latticepresent only after the oxygen was desorbed in the act of imaging. Almost all theimage spots that appear in the second image only, white spots, can be associated withatoms of the tungsten lattice made visible by the removal of nearby atoms.The desorption of oxygen ad-layers during imaging has been examined in moredetail by using field emission measurements to characterize the amount of oxygenremaining on the surface after subjecting the layer to progressively higher appliedfields.This has been done for desorption in ultra-high vacuum, and in the imaginggases, helium and neon. The extent of desorption is greater, the higher the tiptemperature, but the general nature of the effects is indicated by some results fordesorption at 78"K, fig. 2. Vacuum field-desorption for 1 min at the best image1 Mulson and Miiller, J. Chem. Physics, 1963, 38, 2615FIG. 1 .-Vacancies, black image spots, present in a tungsten surface after the addition of a low oxygencoverage and its subsequent desorption during helium ion imaging of the surface.[To face page 66GENERAL DISCUSSION 67voltage for helium drastically reduces the oxygen coverage, and removal of the ad-layer is apparently complete if helium is present.For the purpose of chemisorptionstudies, the results for neon are more encouraging, since considerable fractions of thead-layer survive after 1 min at the best image voltage for neon. However, under theconditions of these experiments, the neon ion image was too faint to be observed orrecorded in 1 min, and increasing the neon pressure or the exposure time merelyincreased the extent of desorption. Nevertheless, the results suggest that with a high-gain image intensifier, neon imaging of the ad-layer should be possible. The tempera-ture dependence of field desorption found in these experiments, and in carbonmonoxide desorption,l suggests that imaging at 20"M, rather than 78"K, would alsobe advantageous.While it is true that field emission monitoring of the adsorbatecoverage gives little or no information about the regional variations of the extent ofdesorption or the situation in the (1 10) region, it would be desirable if field emissionevidence such as that given here for oxygen were presented in discussions of thevisibility or otherwise of adsorbates in the field ion microscope.Dr. D. Brennan (University of Liuerpool) said: I find it difficult to accept that COmolecules are able to penetrate into the subsurface region, or, alternatively, thattungsten atoms can diffuse out past adsorbed CO molecules and present new adsorp-tion sites to the gas phase.The CO molecule is large relative to the tungsten atomand carbonyl formation involves the creation of an even larger species. Qne canaccept perhaps the idea that, for example, oxygen or halogen is able to penetratebelow the surface ; in three-dimensional phases, both these species can form inter-penetrating lattices with metal atoms. But carbonyl formation, or incipient carbonylformation involves particles whose size is very different from that of the metal lattice,and surely such species are to be found only on the surface proper. Molecules of COwill be expected to surround tungsten atoms, rather than vice-versa. Is there anyinformation about the number of CO molecules adsorbed and the number of vacanciesleft after field desorption?Caution is necessary in assessing the disturbance to a surface caused by adsorption.The idea that some disturbance will occur is very reasonable having in r i n d the strengthof the bonds between adsorbate and the surface, and the large amount of energyavailable at the instant of adsorption.However, to assess the damage by the numberof layers which must first be field desorbed before a perfect surface is restored, ispossibly unreliable. The field greatly weakens the interactions between metal atomsbelow the surface, and, under these conditions diffusion of vacancies into the layerbelow might not be too difficult. That is the field itself may propagate the damage tothe subsurface regions to some extent.Prof.W. M. H. Sachtler (Amsterdam) said: Textbooks of inorganic chemistry 2state that tungsten hexacarbonyl is formed in macroscopic quantities by interaction ofcarbon monoxide and elementary tungsten at, e.g., 225°C. In the W(CO)6 moleculethe tungsten atom is octahedally surrounded by six CO ligands. Intermetallicbonds are completely broken under these circumstances. In view of the size of COmolecules and tungsten atoms, a W atom can only be lifted from the tungsten surfaceto the gas phase as a W(CO)6 molecule, if some CO groups have been attached to thisW atom from below. It follows that either CO molecules are able to penetrate intothe surface of a tungsten crystal or surface W atoms can jump above chemisorbed COmolecules. In view of this is is not surprising that similar rearrangements shouldbecome already noticeable at lower temperatures, where no macroscopic production1 Gomer and Swanson, J.Chem. Physics, 1963,39,2813.2 Remy, Lehrbuch der anotyanischen Chernie (Akademische Verlagsgesellschaft, Leipzig, 1959),11, p. 14968 GENERAL DISCUSSIONof UT(CO)6 takes place. At not too low temperatures the atomic vacancies created bythis process migrate also to deeper layers. The latter process is not influenced by anexternal electric field which, in a metal, cannot penetrate below sub-surface layers.Field effects during the act of adsorption were excluded in our work, since no field wasapplied when the gas was admitted to the tube.Dr.Gert Ehrlich (Scheizec:ady, New Yurk) said: It has often been assumed thatchemisorption of simple gases such as H2, N2, and CO on refractory metals involves aprofound disturbance of the lattice, even at room temperature. This assumption,however, appears to be only tenuously related to experimental fact. Thus in inter-preting their F.I.M. studies, Holscher and Sachtler (H & s) propose that tungstenatoms are displaced from their normal positions as a result of the interplay with CO.Even if we were to accept this, the ion images would only prove that the rearrangementand its variation with the temperature of adsorption occur for a surface under theinfluence of extremely high electric fields. During the observations by H & S atungsten atom at a kink site sits in a field ( >4 V/A) which amounts to 84-89 % ofthe field for rapid evaporation.A decrease in the binding energy of such an atom byas little as - 1.5 eV would bring about evaporation, yet in the absence of an appliedfield - 3 eV are required 1 ’ 2 just to move a tungsten atom from a kink site on to aflat, and N 8.7 eV are necessary for evaporation. The effect of the high field is furtherenhanced by promotion of field desorption through the image gas. For electro-negative materials chemisorbed on a metal this lowers the desorption field by as muchas one third the vacuum value.3 A small change in the bonding of tungsten atoms tothe lattice by an adsorbed gas can therefore be greatly magnified under observation inthe field ion microscope.Even if severe perturbations are unequivocally established,they are likely to be unique to the special conditions under which the measurementsare made and do not provide a definitive guide to the behaviour under ordinarythermal conditions.That chemisorption under ordinary thermal conditions may result in a reconstruc-tion has also been deduced from the appearance of additional intense spots in low energyelectron diffraction (L.E.E.D.) patterns of surfaces exposed to a chemisorbing gas. Intheir interpretation it has usually been assumed that scattering from the adsorbed layeritself is negligible. Additional diffraction spots therefore were rationalized as indicat-ing changes in the lattice itself. The fundamental assumption underlying this inter-pretation is not firmly based on either theory or experiment, however.Indeed, forCO on platinum, Tucker 4 has demonstrated the contrary. Probably the simplestexample of surface reconstruction deduced from L.E.E.D. has been reported for the(1 10) plane of Ni after chemisorption of hydrogen. Germer and MacRae 5 observedpatterns indicative of a lattice with twice the ordinary spacing in the [loo] direction ;this they ascribe to a reconstruction under the influence of adsorbed hydrogen, withevery second close packed row of nickel atoms missing. The interpretation of thesechanges, which require long range migration of Ni atoms, is not unique, however. Achemically more reasonable picture can be achieved on the assumption that pairs of[loo] rows are displaced towards each other (according to Tucker by -10 % theirnormal spacing).Even for hydrogen the assumption that the adatoms do not contribute directly toscattering may not be valid.In hydrogen adsorption on the (100) of tungsten,1 Sokolskaya, Soviet Physics-Tech. Physics, 1956, 1, 1147.2 Barbour, Charbonnier, Dolan, Dyke, Martin and Trolan, Physic. Rev., 1960, 117, 1452.3 Ehrlich and Hudda, Phil. Mug., 1963, 8, 1587.4 Tucker, Appl. Physics Letters, 1962, 1, 34.5 Germer and MacRae, J. Chem. Physics, 1962, 37, 1382GENERAL DISCUSSION 69Estrup and Anderson 1 find that at low coverages the diffraction patterns are typicalof a f.c.c. lattice with unit edge twice the tungsten [loo] spacing. Following Bauer,zwho pointed out that an adatom may rescatter electrons diffracted from the lattice,Estrup and Anderson interpret all their patterns as due entirely to hydrogen atoms,without any rearrangement of the underlying tungsten up to and including saturatedlayers.The same interpretation appears appropriate to L.E.E.D. studies of carbonmonoxide as well as nitrogen on the (100) of tungsten.3 For carbon monoxide uponthe (100) of nickel, Park and Farnsworth 4 also observed additional diffraction spots,corresponding to a surface mesh of twice the normal spacing. From the high intensityof these new reflections the authors concluded that either " CO produces a reconstruc-tion of the (100) nickel surface, or that diffraction intensity is not an adequate criterionfor assuming a reconstruction ".Hence, the appearance of new diffraction spots doesnot establish surface rearrangement and L.E.E.D. does not yet provide an unequivocalindication of surface reconstruction in the chemisorption of carbon monoxide,nitrogen or hydrogen on Ni and W.At the moment there is no unequivocal evidence to indicate a reorganization of thesurface, involving large-scale relocation of surface atoms, during low temperaturechemisorption ; there are, in fact, strong indications that this does not occur. Fromour F.I.M. studies we decided against any gross atomic displacements in dilute layersof nitrogen and carbon monoxide on tungsten, for which the interpretation of imagesis not yet too complicated; for CO, H & S agree with this conclusion.In principle,it should also be possible to deduce surface rearrangement from desorption measure-ments. Formation of a very strong bond between adatom and lattice, with severeweakening of lattice cohesion, should lead to desorption of an adatom together withan atom from the lattice during thermal evaporation. This would be detectable massspectrometrically, or indirectly, through differences in the amounts adsorbed deter-mined from ad- and desorption runs in flash filament experiments. At pressuresbelow 10-6 torr such an effect has not been observed for any of the systems consideredhere. 5-8Even if lattice atoms are only displaced, without forming a compound stableenough to manifest itself in the gas phase, differences should appear in the desorptionenergy, derived from kinetic studies, and from heats of adsorption, measured calori-metrically. Consider a hypothetical atom, which on chemisorption pulls a latticeatom from a kink and-on to a flat.At low temperatures the calorimetric heat ofadsorption would measure the enthalpy change in going from the gas phase to thisparticular configuration of the surface. Since it is unlikely that the extraction energyE, (liberated in restoring the displaced lattice atom to its original site) will be coupledto the desorption event, the heat of desarption would exceed the calorimetric value byEe ; if the gas is evolved in a bimolecular event, the energy discrepancy will amount to2E, per mole of gas for a one to one rearrangement. For tungsten the extractionenergy is several eV.However, the desorption energies determined from kineticmeasurements for nitrogen9 and carbon monoxide 10 on tungsten are in goodagreement with calorimetric heats for the same gases on evaporated films up toEstrup and Anderson, J. Chem. Physics, in press.2 Bauer, Physic. Reu., 1961, 123, 1206.4 Park and Farnsworth, J . Chem. Physics, 1965, 43,2351.5 Baker, Ado. Catalysis, 1955, 6, 164.6 Ehrlich, J. Physic. Chem., 1956, 60, 1388.3 Estrup, private communication.Redhead, Trans. Furaday Soc., 1961,57, 641.Ustinov, Ageev and Ionov, Soviet Physics-Tech. Physics, 1965, 10, 851.9 Ehrlich, J. Chent. Physics, 1962, 36, 1171.lo Brennan and Hayes, Phil. Trans., 1965, 258, 347.70 GENERAL DISCUSSIONmonolayer coverages, suggesting that rearrangement is not important.Finally, theagreement between the amounts of these gases adsorbed on films and their capacityfor rare gases,l gives no hint of site creation by adsorption. Although it is prudent tokeep the possibility of severe surface reconstitution in mind, at the moment theredoes not seem any necessity to invoke such an effect for the chemisorption of thesegases (Hz, N2, CO on W or Ni) under ordinary thermal conditions.Dr. A. A. Holscher and Prof. W. M . H . Sachtler (Amsterdam) (communicated) :We agree with Ehrlich’s remarks that the irr,age conditions influence the surfacestructure after adsorption. In fact, one of OUI main points concerns field effects,namely, where we have proved the occurrence of promoted field desorption ofcarbon monoxide molecules, which thus escape observation in the FIM.But! thereseems no reason for such induced effects not being equally present in Ehrlich’s ownFIM experiments on the adsorption of nitrogen and carbon monoxide on tungstenperformed in the same imaging field as in our work.While there thus are no essential differences between the experimental conditionsor the observations reported by Ehrlich and ourselves, there are differences in theinterpretation. Our conclusion of the observed disordered structures being dueto displaced tungsten atoms rather than to CO molecules is based on a number ofmutually consistent experimental results given in our paper, e.g., (a) electron emissionresults showed that under the conditions of FIM, most CO molecules are field-desorbed ; (b) field evaporation experiments revealed that the disorder had penet-rated several atom layers below the adsorbing surface ; (c) at 78°K disordered struc-tures were observed by FIM only after a high degree of CO coverage had beenobtained.The difference in temperature at which the FIM pictures were taken(Ehrlich, 20°K; our work, 78°K) seems cf little importance, as we conclude fromwork on the Nz+ W system. Here we used the same imaging temperature (20°K)as Ehrlich did, and still arrived at conclusions similar to those we obtained for theCO + W system.Once it is accepted that the disordered structures observed in the FiM are dueto displacements of tungsten atoms on a surface largely denuded of adsorbed en-tities, the question arises whether these displaceinelits are induced by the field itself.From our evidence we feel that the observed displacements predominantly takeplace already during the adsorption process; i.e., under conditions where in ourexperiments no field was applied.An important argument for this view is the ob-servation that the extent of disorder largely depends on the temperature of ad-sorption. Moreover, the disorder penetrates into deeper lying atom layers whichare shielded from the field.We realize that in the process of field desorption during imaging some atom dis-placements might be induced in addition to those already brought about by theadsorption itself. Being aware of these possible additional field effects we haverefrained from advancing too detailed an interpretation of the observed patterns,as the present knowledge of field effects and of the image formation does not justifythis.We do not see why there is so much reluctance in accepting the possibility ofsurface rearrangements at room temperature or even below it, when practice presentsso many examples of surface corrosion extending over macroscopic depths.Inany system where the bulk compound is known to be stable, chemisorption eventuallyhas to be followed by rearrangement processes. For example, for the N2+Wsystem it can be calculated from thermodynamic data 2 that at 300°K the nitrogen1 Brennan and Hayes, Phil. Tram., 1965, 258, 347.2 Kubaschewsky and Evans, Metal Physics and Physical Metalliirgy (1958)GENERAL DISCUSSION 71pressure in equilibrium with bulk W2N is about 10-11 torr.If, therefore, in theadsorption experiments at higher nitrogen pressures no bulk nitride is formed thisis due to kinetics only and not to thermodynamics.An intuitive feeling that reorganization of the surface in the presence of ad-sorbate is unlikely at room temperature because metal-metal bonds are so strongmight be misleading because the same intuitive reasoning would also fail to predicta dissociative adsorption of nitrogen at low temperatures, which requires the breakingof a bond of 226 kcal/mole. In both cases, nature apparently has at its disposala reaction path requiiing a low activation energy.Activation energies for desorption of nitrogen and carbon monoxide are deter-mined from kinetic measurements at high temperatures (TE 1000°K) where themobility of surface atoms even on a clean tungsten surface becomes appreciable,as has been shown by Muller 1 and by Ehrlich and Hudda.2 Thus, if adsorptioncauses a rearrangement of the suIface metal atoms it is most likely that upon de-sorption at high temperature the metal surface is reconstructed.The measuredvalues of the heats of adsorption and desorption then refer to the same initial andfinal states and therefore should be equal. Incidentally, we do not believe thatjumping of a tungsten atom from a surface site into a position between the adsorbedentities, where mutual co-ordination will be more favouiable, must be an endothermicprocess as Ehrlich seems to visualize.It seems to us that the inclusion of LEED and mass spectrometric results is oflittle help to the clarification of the present discussion because (a) results obtainedwith the CO + W system by either technique have not been published ; (b) the inter-pretation of LEED results still seems a controversial matter, as was pointed out byEhrlich ; ( c ) if a chemisorption complex decomposes rather than evaporates athigh temperature, mass-spectrometric analysis is of no relevance in elucidatingthe structure of the original surface complexes.As regards the question of the necessity for invoking a rearrangement of sub-strate atoms, in our view it gives a fair explanation for the irreversible conversionfrom the low-temperature into the high-temperature state in the cases of carbonmonoxide on tungsten (Gomer’s virgin states++u states) and of nitrogen ontungsten, as well as for many other experimental observations with these systems.Dr.J. M. Thomas (Univ. Coll. North Wales, Bangor) (partly commuiiicuted): Dr.Ehrlich has expressed some scepticism concerning the evidence, revealed by low-energy electron diffraction and field ion microscopy, for surface rearrangements duringor following adsorption. Gwathmey and his coworkers 3 , 4 have shown, by ordinaryoptical and replica electron microscopy, that extensive surface rearrangements takeplace when { 1001, (1 103 and certain other faces of copper single crystals function ascatalysts for the reaction of gaseous hydrogen and oxygen. From the studies carriedout by Meelheim et aZ.,5 it would appear that the surface rearrangement is initiatedand sustained primarily by the catalysis of the heterogeneous reaction.It does notappear to be identical with the phenomenon of thermal faceting (which takes placcwen in soft vacua) described by recent auth0rs.6~7 Surface rearrangement of copperatoms occurs freely during catalysis at temperatures as low as 400°C, whereas thermalMiiller, Z . Pl?vJik, 1949, 126, 642.2 Ehrlich and Hudda, J. Chern. Physics, 1961, 35, 1421,3 Gwathmey and Benton, J. Physic. Chem., 1940, 44, 3 5 .Leidheiser and Gwathmey, J. Amer. Chem. Soc., 1948, 70, 1200.Meelheim, Cunningham, Lawless, Azim, Kean and Gwathmey, Proc. 2nd Int. Cungr.Catalysis(Paris, 1960) (Technip Press, Paris, 1961), 1, 2005.6 Moore, in Metal Sitrfaces, Energetics atrd Striicture, (ASM-AIME Seminar, 1962), (A.S.M.Ohio, 1963). 7 Robertson, Acra Met., 1964, 12,24172 GENERAL DISCUSSIONfaceting, which is believed to involve 1 evaporation of the solid, requires a tempera-ture in the region of 1000°C.Ds. G. Ertl (Techn. Hochschule, Munich) said: We have studied with L.E.E.D.the interaction between N20 and a Cu( 100)-surface at 5OO0C, which leads to the forma-tion of gaseous nitrogen and adsorbed atomic oxygen. In the diffraction pattern ofthe clean (100)-surface (large circles) extra spots are visible after a short reaction timewith a twelvefold symmetry (small open circles), which correspond to the formationof two equivalent kinds of (1 1 1)-surfaces parallel to the (100)-surface, rotated by 90".These extra spots are divided into three points close together.This observation isascribed to multiple scattering between the (100)- and (1 1 1)-plane. Extra spotsresulting from multiple scattering can only be observed if there are differences in thelattice geometry. After a longer reaction time, some weak extra spots are visible inaddition (dark circles) with a nearly doubled lattice constant. These spots are attribu-ted to another surface structure with (1 1 1)-geometry.00 0 00 0 008O Q0 60004 0 0040 0 0000FIG. 1.The results can only be interpreted by the assumption that a rearrangement of thesurface copper atoms OCCUIS by the influence of atomic oxygen.Dr.T. A. Delchar and Prof. F. C. Tompkins (Imperial College) (communicated):Dr. Ertl is undoubtedly correct that rearrangement of surface copper atoms occursfollowing the chemisorption of oxygen during the N2Q decomposition. Some fiveyears ago we measured the change of work function of annealed copper (and nickel)films when oxygen was chemisorbed at 78°K and the temperature raised to roomtemperature. The negative surface potential (s.P.) due to the oxygen adatoms becamemore positive (with no desorption of oxygen) with increase of temperature; weinterpreted this as a " tunnelling " of the adatoms into the underlying copper to forman " ionic " species ; the newly exposed copper sites could chemisorb more oxygenadatoms, but the change of s.p.with number of oxygen atoms adsorbed was muchlarger showing that the electron distribution at the surface had substantially changed,and some reconstruction had probably occurred.1 Moore, in Metal Surfaces, Energetics and Structure, (ASM-AIME Seminar, 1962), ( A S M .Ohio, 1963GENERAL DISCUSSION 73Dr. A. A. Holscher (Amsterdam) said: The decrease in loglo A by a factor of 2,observed at low coverage in the carbon tetrachloride experiments by Duell et al., isascribed to a decrease in emitting area. As stated, the image (fig. 6b) does not showsuch a reduction, which was explained by assuming the emission to take place throughsmall windows with low work function. These windows, however, cannot beextremely small, because within an area with a length and width of the tunnel-barrier(10-1 5 A) a non-uniformity of the work function cannot be defined. With a resolutionof the image of about 20 A it is thus possible to have a checkerboard arrangement ofemitting and non-emitting patches which give a reduction of emitting area by a4actorof about four without this reduction being observed in the image.I think thereforethat it is more likely that another mechanism is responsible for the large drop inloglo A . It might be connected with the adsorption-induced disturbance of theperiodic potential at the surface, in the same way as we proposed for the CO/W system.Dr. D. W. Bassett (Imperial College) (partly communicated) : On the basis of fieldemission evidence given by Duell, Davis and Moss, the interaction of chlorine withtungsten surfaces is sufficiently similar to what has been observed by other workersfor oxygen on tungsten, that it should be valid to compare particular aspects of thebehaviour in these systems.However, the case for adsorbate penetration into thetungsten substrate in the initial stages of adsorption of these electronegative adsorbatesis weakened, rather than strengthened by the work function decreases claimed to occurduring oxygen chemisorption at 300°K.Considerable experimental evidence suggests that the work function does notdecrease during oxygen chemisorption at 300°K. For example, in a field ion micro-scope study of some aspects of oxygen adsorption at 300"Kl I found that work func-tions of oxygen-covered surfaces deduced from Fowler-Nordheim plots were alwayshigher than that of clean tungsten, no matter how small the coverage, or at what ratethe gas was added.The gas was added as small doses of arbitrary size, and the fieldemission data was taken under conditions of stable emission. In fact, at both 300°Kand 78°K the variation of the work function with the voltage required for a particularemission current was almost linear. Over the work function range, 4.5-5-5 eV, thepre-exponential factor of the Fowler-Nordheim equation decreased monotonicallyas the work function increased, loglo A changing from -9.55 to -10.15, and did notshow the large fluctuations given in table 2 by Duell et al. These observations areconsistent with the general features of oxygen chemisorption on tungsten as revealedin other field emission studies.2The conclusion that the work function rises steadily during oxygen adsorption ontungsten is further supported by the retarding potential measurements of workfunction made by Zingerman and Ishchuk.3 They found that when oxygen was addedto a clean tungsten surface from a molecular beam, there was initially a monotonicrise in work function at 300°K and also at 850°K.At the higher temperature, theirresults, low energy electron diffraction studies, and field ion microscopy all indicatethat extensive rearrangement of the tungsten surface accompanies adsorption. Onheating oxygen covered field-emitters, this rearrangement of the surface leads to fieldenhancement, especially over the (1 1 1) regions, that causes apparent decreases in workfunctions deduced from Fowler-Nordheim plots.Fig. Id given by Duell et al.indicates that this also happens with chlorine-covered surfaces, so that the workfunctions obtained in such heating sequences are not unambiguous evidence thatchlorine adsorption may lead to positive surface potentials. In view of the evidence1 Bassett, unpublished.2 Gomer and Hulm, J. Chem. Physics, 1957, 27, 1363.3 Zingerman and Ishchuk, Fiz. Tverd. Tela., 1964, 6, 117274 GENERAL DISCUSSIONwith oxygen adsorption, it also seems doubtful whether the interpolation procedureused by Duell et al. to deduce work functions from field emission measurements madeunder conditions of unstable emission is reliable, or whether the work functiondecreases obtained are real effects.These comments are not intended to imply that lattice penetration by electro-negative adsorbates does not occur. Oxygen certainly penetrates into tungsten at300°K in a slow activated process after completion of a saturated ad-layer, and thisuptake has been followed volumetrically,l and by field emission.2 In fact, it was thisprocess that Muller observed in the field ion microscope3: the observations of" surface corrosion " after exposure of a tungsten surface to oxygen at 20°K are notrelevant, since this was almost certainly due to damage caused in the imaging process.Dr. M. J. Duell, Mr. B. J. Davis and Dr. R. L. Moss (Warren Spring Lab.) (com-nzunicated) : In reply to Bassett, the point at issue is not whether electronegative gasescan " penetrate " the tungsten surface at 300°K but if this can occur at low coverages, inparticular with oxygen, as indicated by the present work function measurements.There are fewer field-emission studies than suggested of oxygen adsorption directlycomparable to those reported in the present paper. For example, work functions havenot been obtained from Fowler-Nordheim plots in some cases but from the emissioncurrent at constant voltage which is unreliable. Bassett reports that his results areconsistent with the general features of oxygen chemisorption on tungsten found byGomer and Hulm.4 The latter studied the spreading of oxygen deposited initiallyat very low temperatures and also work function changes on heating oxygen ad-sorbed at 300°K and " relatively high pressures '' such that A+ - 1.6-2.0 eV aftershort exposure times. In Bassett's work reported above the amounts of oxygenadmitted were not measured.5 We would emphasize the necessity for controlledadmission of oxygen or halogen compounds at very low pressures to reveal positivesurface potentials. In addition to our experiments under conditions of " unstableemission ", we used gas " doses " of chlorine or carbon tetrachloride (cf. fig. 4 and 7)and again found a reduction in work function at low coverages.The pre-exponential term of the Fowler-Nordheim equation has often beenneglected but in recent studies of nitrogen adsorption on tungsten it was found todecrease substantially from loglo A = -5.66 to -7-41.6 In this case the pre-exponential term decreased monotonically but on the (41 1) plane specifically,7 itincreased initially, followed by a decrease until A loglo A N 2. Apart from changesin emitting area, fluctuations of the pre-exponential term arise from the influenceof the applied field on the adsorbed layer depending on the polarizability of theadsorbate. It has further been suggested that a modification of the potential barrierby a corrected image-force potential should also be taken into account.71 Rideal and Trapnell, Proc. Roy. SOC. A , 1951, 204,409.2 Gomer and Hulm, J. Chem. Physics, 1957, 27, 1363.3 Miiller, in Structure and Properties of Thin Films, ed. Neugebauer, Newkirk and Vermilyea(Wiley, New York, 1959), p. 476.4 Gomer and Hulm, J . Chem. Physics, 1957,27, 1363.5 Bassett, personal communication.6 Ehrlich and Hudda, J. Chem. Physics, 1961,35, 1421.7 Van Oostrom, Philips Res. Reports Suppl., 1966, no. I