Heats of Adsorption of Oxygen on Evaporated Films ofMolybdenum, Tungsten, Cobalt and Nickel at 77, 90 and273'K, and Nature of Adsorbed LayersBY D. BRENNAN AND M. J. GRAHAM"Donnan Laboratories, University of LiverpoolReceived 24th January, 1966Calorimetric measurements are reported which show that oxygen adsorbed to saturation onmolybdenum and tungsten at 77°K has the same energy as oxygen adsorbed on these metals at273°K to the same coverage. The additional coverage at 273"K, either for a surface initially satur-ated at 77°K or for a surface maintained at 273°K throughout the adsorption, is effectively thesame in both cases and is associated with the same energy which is lower than that of the initialstate and falls with increasing coverage. It is argued that the final state of the adsorbed layer onthese metals at 273°K is independent of the temperature path.A model is proposed for each stageof the adsorption.For cobalt and nickel at 77"K, despite the similarity of the saturation coverages to those ob-tained at 273"K, the heats of adsorption are lower than those at 273°K. This is explained interms of the formation at 77°K of a chemisorbed layer confined to the surface proper of the metal,whereas, at 273°K incorporation of oxygen into an oxide layer occurs. Adsorption of oxygenat 273°K on a surface previously saturated at 77°K results in a further uptake amounting to 30-40 %of the initial adsorption at 77°K. It is argued that, on certain planes, the oxygen initially adsorbedat 77°K is to be found at 273°K sandwiched between the first and second layers of metal atoms,and that the metal atoms again exposed in the surface proper are able to adsorb further oxygen.Models for the adsorbed layers are proposed.Low temperatures can give rise to states of adsorption not observed at highertemperatures.In the present work, the energy of the adsorbed state and the sur-face concentration are used to establish the existence of temperature-dependentstates of adsorbed oxygen and to aid the elucidation of possible models.EXPERIMENTALMaterials, gas handling and vacuum technique have been described previously.1CALORIMETRYThe design of the calorimeter and methods of measurement are essentially the same as thepreviously described.132 Two calorimeters were used, viz., calorimeter A, which was thesame instrument as was employed for our previous measurements, and calorimeter 33.The windings of thermo-pure platinum wire (Rloo/Ro> 1.3923, supplied by Johnson,Matthey and Co.Ltd.) on the egg-shell tube of calorimeter B had a smaller pitch (0.05 in,)than hitherto.The values of (Cla)~ (see ref. (1)) for the new calorimeter at different temperatureswere determined by two methods, both of which relied on the dissipation of known amountsof electrical energy in tungsten films laid in the calorimeter, but which diflered in themethod of achieving electrical contact with the films. In one method, electrical contactwas made by brushes similar in shape to those used previously, but instead of several strandsof piano wire, four arms of spiralled steel wire, each terminating in a blob of soft solder,* present address : Division of Applied Chemistry, National Research Council, Ottawa, Canada996were used.In the other method, two bands of 43 s.w.g. steel foil, 2mm in width, wereemployed to give an even larger area of contact with the film. As pointed out previously,2the calorimeter tube is generally at a higher temperature than that of the bath. Measure-ments of (C/~)B were made for various heat inputs and rates of heating; the values ob-tained at the various temperatures are given in fig. 1. The close agreement between thedifferent measurements supports the conclusion that the quantity measured is accuratelyC/a and that there is no significant systematic error due to a contribution to the heatcapacity from the contact assembly, nor difficulty arising from the possible presence ofhigh contact resistances.The low temperature values for (C/C()A used to derive results inthis paper, have been based on the ( C l a ) ~ values given in fig. 1. This is necessary becauseADSORPTION OF OXYGEN ON METALS110100-96ao706 05031------77% 90°KI I I I35 40 45 5 0FIG. 1.-Values of (C/& as a function of the resistance R of the calorimeter. The figure givesthe results for measurements at low temperatures (0, spring contacts ; 0 , strip contacts). At195"K, (C/O~)B = 490klO cal, R = 112 D and, at 273"K, (C/CC)B = 926fl0 cal, R = 162 8;the standard deviations have been obtained from six measurements in each case.a series of (Cla)~ values is not available in the entire temperature range 77-90°K and ispermissible because the materials used in the construction of the two calorimeters wereequivalent ; calorimeter A is unfortunately no longer available for further heat capacitymeasurements to be made on it.RESULTSThe heats of adsorption at the various temperatures are presented in fig.2 and3. The parameter p is defined by the ratio, p=N(O)/N(Kr), where N(Kr) is thenumber of krypton atoms present in the monolayer at 77"K, and N ( 0 ) is the numberof oxygen atoms adsorbed by the same surface at the temperature specified. Eachpoint corresponds on the coverage scale to the mean coverage due to a gas incre-ment. The results obtained with the two calorimeters are in good accord.ADSORPTION ON CLEAN SURFACESAlthough values for the heats of adsorption of oxygen on films of these metalsat 300°K are available,l measurements were made at 273°K in order to confirmearlier work and to strengthen the characterization of calorimeter B.Satisfactoryagreement with the previously published integral heats is obtained (see fig. 2 and 3) ;correspondence between the two sets of maximum heats is not quite so close in everycaseD. BRENNAN AND M. J. GRAHAM 97The data of fig. 2 and 3 show that (a) for molybdenum and tungsten, while thesaturation coverage at 77°K is smaller than at 273"K, the heats of adsorption areeffectively the same, and (b) for cobalt and nickel, despite the similarity betweenthe saturation coverages at the high and low temperatures, the heat of adsorptionat 77°K is significantly smaller than at 273°K.(a) molybdenum0 0 .5 2.0 2.5 3.0(b) tungstenFIG. 2.-The heats of adsorption of oxygen on molybdenum and tungsten films at 273,90 and 77°K.(a) Molybdenum : a, 42.4 mg, 273"K, 192 kcal mole-1 ; A, 13-9 mg, v, 10.9 mg, 77"K, 170kcal mole-1. (b) Tungsten : 0, 28.0 mg, 6 , 28.7 mg, 273"K, 183 kcal mole-1 ; 0, 17.9 mg,90"K, 184 kcal mole-1 ; A, 23.8 mg, 77"K, 184 kcal mole-1. The open points refer to calorimeterA and the filled ones to calorimeter B ; the associated heats in each case are the average integralheats of adsorption. Previously reported 1 integral heats of adsorption at about 300% are :molybdenum, 168 kcal mole-1 and tungsten, 180 kcal mole-].ADSORPTION AT 273°K ON SURFACES SATURATED AT 77°KSurfaces were saturated with oxygen at 77°K and excess oxygen was then pumpedfrom the system.Calorimetric and coverage data for the further adsorption ofoxygen on the resulting surfaces are given in table 1 ; for comparison, the saturationcoverages obtained for the clean surfaces at 273°K are also included in the table.The values of p for cobalt and nickel are uncertain to at least 5 %.TABLE VALUES OF p FOR CLEAN SURFACES SATURATED AT 77 OR 273"K, AND FOR SUR-FACES SATURATED FIRST AT 77°K AND THEN AT 273°K; AND THE HEATS OF ADSORPTION INTHE LATTER CASEmetalp for saturation atp for saturation of the cleail 273°K of surfacespreviously saturated increments admitted77% 273°K at 77'K to surfaces previouslyheats of adsorption(kcal mole-1) atsurfaces at which had been 273°K for gassaturated at 77°Kmolybdenum 2-3tungsten 2.1cobaltnickel7.46.33.0 3.2 1 043.3 3.1 122, 120, 120,121, 114, 857-8 10-8 78, 77, 71, 666-4 8.2 798 ADSORPTION OF OXYGEN ON METALSMOLYBDENUM AND TUNGSTEN.-Table 1 shows that the saturation coveragesat 77°K are appreciably smaller than at 273"K, even though the heats of adsorptionare effectively the same. The coverage obtained on saturating at 273°K a surfacealready exposed to oxygen at 77°K is effectively the same as obtained at 273°K forsaturation in one step.Also, for tungsten, the heat of adsorption of the additionaluptake at 273°K is very similar to the heat of this portion of the coverage at 273°K.(a) cobalt100 ' * O h0 1 2 3 4 5 6 7(b) nickelFrG.3.-The heats of adsorption of oxygen on cobalt and nickel films at 273, 90 and 77°K.(a) Cobalt : e, 22.4 mg, 273°K 98 kcal mole-1 : A, 16.4 mg, V, 30-3 mg, 77"K, 84 kcal mole-1(b) Nickel: 0, 45.5 mg, 0 , 31.5 mg, 273"K, 107 kcal mole-1 ; 0, 31.4 mg, 0, 45.2 mg, 90"K,71 kcal mole-1 ; A, 29.0 mg, V, 57.3 mg, A, 35-5 mg, 77"K, 70 kcal mole-*. The open points referto calorimeter A and the filled ones to calorimeter B ; the associated heats in each case are theaverage integral heats of adsorption. Previously reported 1 integral heats of adsorption at about300% are : cobalt, 101 kcal mole-1 and nickel, 105 kcal mole-1.For molybdenum, no measurements of heat were made for adsorption at 273°Kfor coverage greater than p = 2, but it has already been found 1 that the heat fallsat these higher coverages.The heats of adsorption at 273°K on surfaces saturatedat 77°K show an analogous decrease. It is concluded, for niolybdenuin and tungsten,that two stages in the adsorption can be distinguished. The first stage is commonto the adsorption at both temperatures and terminates at about p = 2. The secondstage does not occur at 77"K, but does so rapidly at 273°K. The adsorbed layerat 273°K is the sane whether it is formed by adsorption on the clean surface at273"K, or by adsorption first at 77°K and then at 273°K.COBALT AND NICKEL.-The adsorption at 273°K on a surface saturated at 77°Kis appreciable and occurs with a heat whieh is comparable to the heat associatedwith the initial adsorption at 77°K.'Thus, the adsorbed layer obtained in the two-stage adsorption contains more oxygen, but is less stable than the layer obtainedby saturation of a surface kept at 273°K throughout the adsorption and, therefore,it is concluded that the two states are differentD. BRENNAN AND M. J. GRAHAM 99DISCUSSIONIn a consideration of the adsorbed state of oxygen, there immediately arisesthe problem of the polarity of the surface bond and with it the effective size of theadatom. The relatively small surface potential due to adsorbed oxygen was origin-ally interpreted 3 to mean that adsorbed oxygen was essentially atomic in character.However, MacRae4 has argued that the surface potential data can also be used insupport of considerable charge transfer to the adsorbed oxygen.Magnetic studiesyield conflicting evidence.5 Park and Farnsworth 6 consider obedience of photo-electric data to the Fowler curve for metals as strong evidence for adsorbed oxygenbeing mainly atomic in character. We incline to the opinion that chemisorbedoxygen is not very highly polarized and that, in considering the size of the adatom,the atomic diameter of 1.32 A is the best guide.For molybdenum and tungsten, thep values at saturation are small enough forthere to be little doubt that the adsorbed oxygen is confined to the surface proper.The values of p for cobalt and nickel are larger and raise the question whether allthe oxygen is on the surface, or whether some incorporation has occurred.Foradsorption at 273"K, the adsorbed layer must closely resemble oxide and the heatof adsorption is similar to the appropriate heat of oxidation, in keeping with thisview.1 The argument that an oxide is formed at 77"K, but that it is different fromthe oxide formed at 273°K meets with difficulties. If this were so, it would beexpected that the final state at 273°K should not depend on whether the adsorptiontakes place first at 77°K; also, a smaller value of p would be expected at 77°Kthan at 273'K and, possibly, a higher heat of adsorption, but such is not the case.The surface potential change when oxygen is adsorbed on nickel at 273°K undergoesa relatively slow increase following the initial virtually instantaneous decrease 7and this is evidence for reorganization of the adsorbed layer.No similar slowsurface potential changes were observed for nickel at 77"K, suggesting that theadsorbed layer formed initially does not undergo any subsequent change. Theseseem good arguments for supposing that oxygen adsorbed at 77°K is confined to thesurface proper, and this proposal will be examined below.MOLYBDENUM AND TUNGSTENUsing the configuration for adsorbed krypton proposed by Brennan and Graham 8and the restriction that only one oxygen atom is adsorbed per surface metal atom,the theoretical values of p for the most probable planes 1 are ~(100) = p(110) = 2 ;for the (211) plane, the metal atoms immediately below the surface are sufficientlyexposed to be counted as part of the surface proper, as far as oxygen adsorptionis concerned, and p(211) = 4.It is proposed that the adsorption up to saturationat 77°K or, at 273"K, up to the coverage at which the heat of adsorption begins tofall, viz., p = ca. 2, is confined to the surface proper. The adsorption occurringafter warming the surface saturated at 77"K, or accompanied by a falling heat at273"K, is attributed to adsorption of oxygen atoms on metal atoms immediatelybelow the surface of the (100) face (the (1 10) face does not offer sub-surface sitesof sufficient size to accommodate an oxygen atom). This description requiresadsorption on the sub-surface sites to be activated to an extent which debars occupa-tion at 77°K and this is reasonable having regard to the confined nature of thesesites.The description is also compatible with an increasing energy of activationand a decreasing energy of adsorption with increasing coverage, and with a limitingcoverage of p = 3 at 273°K. Surface potential measurements lend further suppor100 ADSORPTION OF OXYGEN ON METALSto the model. Quinn and Roberts7 have reported that there is little change in thesurface potential due to oxygen adsorbed on molybdenum when the temperatureis raised from 77 to 298°K in keeping with a common state of adsorption on thesurface proper at the two temperaturss; nor is there a significant change in surfacepotential on adsorption of oxygen at 298°K on a molybdenum surface previouslyexposed at 77"K, in keeping with the view that this extra adsorption is confined tothe sub-surface.COBALT AND NICKELTheoretical values of p similar to those measured for cobalt and nickel can beobtained for chemisorbed oxygen if it is supposed that the surface atoms of the mainfaces 1 can each account for two oxygen atoms, and, further, if the sub-surfaceatoms of the (1 10) and (100) faces can each account for one ; the (1 11) face, beingclose packed, does not offer sub-surface sites. Again using the configurations ofadsorbed krypton proposed by Brennan and Graham,s the values of p predictedfor these conditions are : p(110) = p(100) = ~(111) = 6.These values are a littlelower than the observed values, especially for cobalt, but having regard to the im-perfect nature of evaporated films, the agreement is fair.An adsorbed layer ofthis kind would undoubtedly be less stable than an oxide layer, in keeping with thesmaller heat of adsorption observed at 77°K. Additionally, such a layer wouldbe expected to result in a much larger surface potential change than an oxide layerand this accords with observation.7On raising the temperature from 77 to 273"K, it is envisaged that the adsorbedlayer undergoes a transformation in which oxygen atoms take up positions betweenthe surface metal atoms and the underlying atoms. This change is best exemplifiedby reference to the (110) plane as illustrated in fig. 4. The presence of an oxygenatom occupying a spacious sub-surface site would be expected to facilitate thereorganization.As a result of the change, metal atoms will again be exposed inthe surface plane and will be capable of adsorbing additional oxygen with a heatsimilar to that observed initially at 77°K. The state envisaged in fig. 4(b) has a vacantposition between the first two layers of metal atoms for each metal atom in thesurface proper. Such a state could have interesting catalytic properties, particularlywith respect to hydrogen. Further adsorption of oxygen on the surface shown infig. 4(b) would probably result in the state shown in fig. 4(c) and the accompanyingincrease in coverage would have Ap = 4 (cf. table 1). In the change from the staterepresented by fig. 4(a) to that by fig. 4(b), the surface potential would be expectedto increase very considerably and in fact does s0.7 Further the additional adsorp-tion resulting in the state represented in fig. 4(c) would be expected to change thesurface potential to a value nearer to that found initially (fig.4(a)), but not quiteso negative; this is again as observed.7 Similar changes in the adsorbed layercan be envisaged for the (100) plane, but are not feasible for the close-packed (1 11)plane, which is likely to remain unchanged by increase in temperature. Presumably,if the temperature were taken high enough, structures of the type shown in fig. 4(b)and (c) would collapse into the normal oxide structure, but the required activationenergy is too high for this to happen at room temperature. The act of adsorptionis accompanied by the liberation of an appreciable quantity of energy which mightbe expected to cause transformations impossible with the aid of the normal thermalvibrational energy alone.Equally, at sufficiently low temperatures, even the sumof the energy of adsorption and the thermal energy may be inadequate to bringabout a given transformation. In the present case, oxide formation is possibleat 273°K by virtue of both the energy of adsorption and the relatively great thermaD. BRENNAN AND M. J. GRAHAM 101energy of the metal; at 77"K, the difficulty of moving metal atoms prevents oxideformation and results in a genuine chemisorbed layer. Only on warming cansurface transitions occur, but not to the oxide state, since such transitions as nowoccur are due only to thermal excitation and lack the extra boost due to the energyof adsorption.FIG. 4-Proposed configurations for oxygenadsorbed of the (1 10) face of the face-centredcubic metals; (a) at 77"K, starting with theclean surface; (b) on warming state (a) to273°K; and (c) on adsorbing oxygen on state(b) at 273°K. 0, metal atoms in the plane ofthe surface ; @, metal atoms immediately belowthe plane of the surface ; 9 oxygen atoms.The authors are grateful to the former Department of Scientific and IndustrialResearch for a grant to one of them (M. J. G.).1 Brennan, Hayward and Trapnell, Roc. Roy. SOC. A, 1960,256, 81.2 Brennan and Hayes, Trans. Faraday SOC., 1964, 60, 589.3 Gundry and Tompkins, Quart. Reu., 1960, 14, 257.4 MacRae, Surface Sci,, 1964, 1, 319.5 Culver and Tompkins, Adv. Catalysis, 1959, 11, 67.6 Park and Farnsworth, Surface Sci., 1965, 3, 287.8 Brennan and Graham, Phil, Trans. A, 1965, 258, 325.Quinn and Roberts, Trans. Faraday Soc., 1964, 60, 899.