Chemisorption and Surface Corrosion in the Tungsten +CarbonMonoxide System, as Studied by Field Emission and FieldIon MicroscopyBY A. A. HOLSCHER AND W. M. H. SACHTLERKoninklijke/Shell-Laboratorium, Amsterdam, The Netherlands(Shell Research N.V.)Received 5th January, 1966A new interpretation of the occurrence of different adsorption states of carbon monoxide ontungsten has been achieved by combining field emission and field ion microscopy techniques. Itis shown that the adsorption of CO on a field-evaporated tungsten surface causes a re-arrangementof the surface atoms. The activation energy required decreases with increasing coverage and issmallest for the less densely packed crystal faces. At 78°K re-arrangement takes place to a verylittle extent only.Two adsorption layers can be distinguished : a chemi-adsorption layer and aphysisorbed layer on top. At 300°K re-arrangement leads to complete surface corrosion, creatingvacancies in the layers beneath the corrosive chemisorption layer. On top of the latter layer, addi-tional CO can be weakly adsorbed.Hydrogen does not cause any re-arrangement of tungsten. However, CO adsorption on tungstenprecovered with hydrogen, again gives rise to re-arrangements with removal of hydrogen. Theresults, together with the evidence obtained with other systems, justify the conclusion that surfacere-arrangement can be an important factor in adsorption, even at low temperatures and on metalswith a high cohesive energy.Recent experimental work 1, 3 on single crystal faces has revealed that crystallo-graphic specificity in adsorption is an important factor in understanding the simultan-eous occurrence of different adsorption complexes on one and the same metal.Buteven when allowance is made for this finding, the complexity of many compositeadsorption layers is still not completely resolved, because even on one single facevarious complexes can often be distinguished.In the discussions on the nature of such different complexes it has in most casesbeen assumed that the positions of the surface metal atoms remain unchanged, thephenomena being ascribed to either different types of bonding or different adsorptionsites on the surface. However, it has been shown4 that chemisorption causes ade-metallization of the surface atoms.In those gas-metal systems where the corres-ponding bulk compounds are stable, it is logical to expect that eventually a re-arrangement of substrate atoms will take place, as it is the first stage to bulk compoundformation.Experiments with the low-energy electron diffraction (LEED) technique haveshown 5 9 6 that in several cases the formation of chemisorption complexes involves adrastic re-ordering of substrate atoms. These experiments were confined to adsorp-tion temperatures of about 300°K and above. A major restriction of LEED is thatsurface re-arrangements can only be detected if they give rise to new surface structureswhich exhibit long-range order.The question arises whether any re-arrangement can be observed at temperaturesbelow 300°K even on metals with a high cohesive energy, like tungsten, and beforetwo-dimensionally ordered structures appear.The most promising experimentaltool for studying these phenomena is the field ion microscope,7 because it allows230 F.E.M. AND F.I.M. OF co ON wobservation of metal surfaces in atomic detail. The effect of gas adsorption canbe investigated by comparing the ion images of the same surface before and afteradsorption. By applying field evaporation, ideally ordered clean surfaces can beobtained to start with.So far, gas adsorption on ideally ordered surfaces has received little attention.George and Stier 8 studied the adsorption of oxygen on an ordered tungsten surfacewith the field electron microscope. They used the ion microscope mainly to character-ize the structure of the initially clean surface (either highly ordered or thermallyannealed) and concluded that the field emission results are independent of the degreeof initial surface order.After adsorption at 20°K on a field-evaporated surface andremoving the oxygen by field evaporation they obtained an ion image of a surfacelittered with out-of-place atoms. Ehrlich and Hudda have used the ion microscopefor studying the adsorption of nitrogen 9 and carbon monoxide 10, 11 on tungsten.They conclude that the metal surface is only negligibly disturbed by the adsorption.Ln the present work we have combined the ion microscope and field emissiontechniques in a more comprehensive way to study the adsorption of carbon monoxideon clean ordered tungsten surfaces and on tungsten precovered with hydrogen.Inthis way we obtained different kinds of information on one and the same state ofadsorption. First, the change in work function caused by adsorption is derived frommeasurements of the change of electron emission and applying the Fowler-Nordheimrelation. Although the exact meaning of changes in average work function is not asclear-cut as those obtained from measurements of individual crystal faces on accountof the complicated averaging inherent in the method, the average work function canstill be used as an important parameter to describe the adsorbed state.A second useful parameter is provided by the change in the pre-exponential factorof the Fowler-Nordheim relation.It is often found that adsorbed states characterizedby the same work function differ markedly with respect to the change in the pre-exponential factor.The field emission results can be fitted into a more detailed picture of events atthe surface by making use of the observations with the ion microscope. Theseessentially fall in two categories. First, we may compare the ion image of the cleansurface with the one obtained after adsorption, the adsorption state being characterizedby the Fowler-Nordheim data. Secondly, the procedure may be reversed by measur-ing the electron emission current after taking ion pictures of successively advancedstages of field desorption and field evaporation following adsorption. The lattertype of experiments provides information on the effect of adsorption on atom layersbeneath the outermost surface layer and constitutes a unique feature of the ionmicroscope.Observations made with the ion microscope are purely visual ; conclusions areobtained by comparing photographs of ion images.The reproduction of the photo-graphs in the present article has been restricted to those necessary for illustratingthe most important points.EXPERIMENTALFig. 1 shows the microscope tube. An outer Dewar, normally filled with liquid nitrogen,shields an inner Dewar, which can be filIed with a suitable coolant to adjust the tempera-ture of the tip. The tip assembly is mounted on tungsten rods, which are ledthrough the inner Dewar. The distance from the tip to the flat conductive fluorescentscreen (1 1 cm dim.) is about 9 cm.A polished stainless steel cylinder, closely fitting the lower end of the Dewar and keptat the same potential as the screen, serves two purposes. First, it prevents spurious electroA.A. HOLSCHER AND W. M. H. SACHTLER 31emission from the conductive coating during field desorption and helium-ion experiments.Secondly, it effectively helps in accommodating the helium atoms before they are ionizedat the tip surface, thus improving the intensity as well as the resolution of the ion image.External electric contact with the conductive coating is made via a Kovar tube, sealed witha glass bead.The vacuum system is schematically shown in Fig. 2. It consists of an ultra-high vacuumpart with an ion pump and a fore-vacuum line with an oil-diffusion pump backed by a two-stage rotary pump.The ultra-high vacuum part can be baked out to 400°C, all valvesexcept A1 being in open position. After the normal bake-out procedure and closing valveA2, the pressure was of the order of 10-10 torr.I/ f 1, I,L - - -- - - - - - - - - -FIG. 1 .-Field electronand ion microscope.FIG. 2.-Schematic representation of vacuum apparatus.T, working table ; V, 75 l./sec VacIon pump; 0, 10 l./sec oildiffusion pump ; R, 2-stage rotary pump ; M, microscope tube ;C , cold trap ; P, Penning gauge ; I, ionization gauge ; A 1, 2, 3,4, 5 , 6, bakeable u.h.v. valves Granville Phillips type C ; A 7,ditto type L; B, magnetically operated ball joint valve; G, gassupply ampoules or diffusion thimbles.‘The gases used in our studies were of an ultra-pure grade and stored in glass bulbs Gprovided with a break seal, except hydrogen, which was purified by diffusion through anickel thimble.In the field electron emission experiments, use was made of a highly stabilized 0-5 kVpower supply.The output voltage could be regulated in steps of 10 V, while a fine regulationprovided for intermediate values. As the accuracy of the dial setting was better than 0.2 %,no separate voltage measurements were carried out. Emission currents were determinedwith a vibrating reed electrometer with an accuracy of about 2 %. The high voltagesrequired for field desorption, field evaporation and ion microscopy came from a stabilized0-50 kV power supply and were measured with a precision potentiometer (accuracy aboutPictures of the electron and ion images were taken with a Nikon camera (1.4 opening)on Gevaert spectroscopic G film.To increase the speed and lower the contrast of thisfilm, we pre-exposed it just to the point such that after developing a slight fog became visible.Unless otherwise stated, we started all adsorption experiments with a clean tungstensurface, field-evaporated at 78°K. The process of field evaporation was checked by visualobservation of the electron image and measurement of the emission characteristic afterreversing the polarity of the tip with respect to the metal cylinder and the screen. Thesewere always kept at earth potential. When after prolonged field evaporation successivelyhigher emission voltages were required to give a constant emission current of 0.5 PA, it was1 %>32 F.E.M.A N D F.I.M. OF co ON wassumed that the field-evaporated end form had been obtained. Ion images of the surfacesprepared in this way invariably showed perfect order, which allowed us to omit imaging ofthe clean surface when this was less convenient.A desired amount of gas was dosed on to the metal tip by manipulating valves A4, A5and A6, respectively, the ball joint shutter and valve A7. Valve A3 was normally kept inopen position and closed only when the main system had to be exposed to air for replacementof the metal tip. Thus it could be arranged that A4, A5 and A6 never experienced atmo-spheric pressure once the break-seals of the gas storage bulbs were broken.Before introducing helium into the microscope tube, valve A7 was closed and the requiredimage-voltage applied to the tip to prevent any contaminant in the helium from beingadsorbed. All ion pictures were taken at a helium pressure of about 2 x 10-3 torr and mostlyat 78°K.Though lower tip temperatures, down to 20"K, would improve the intensity aswell as the resolution of the images, the image quality at 78°K was sufficient for our purposes.After taking pictures the main part of the helium was removed via the fore-vacuum line byopening valves A1 and A2. When a pressure of 10-5 torr was reached (in about 5 min),valve A1 was closed. Upon opening valve A7 the ion pump quickly restored a pressure ofless than 10-9 torr.In those experiments where the electron emission was to be measured after taking anion picture, the tip was kept at the high voltage required for image formation when pumpingoff the helium.Under these conditions a clean surface remains clean, adsorption beingprevented by the presence of the high field, which ionizes all the reactive gas moleculesapproaching the surface.When the pressure was again in the 10-9 torr range we quickly changed the polarity anddetermined the electron emission characteristic. The constancy of the emission current at agiven voltage was found to be a criterion that no re-adsorption took place either from thegas phase or by diffusion from the shanks of the tip.Field electron emission is described by the Fowler-Nordheim relation,' which in asimple approximate form can be written as I = A Wexp (- B#%/ V).( A and B being constants.)By comparing the slopes of the linear plots of loglo (IIV2) against l/V for the clean andcovered surfaces, the relative work functions of covered surfaces may be calculated. SuchFowler-Nordheim plots were determined in the current range of 10-8-10-~ A by varying thevoltage stepwise. The experimental data were processed on a computer to calculate slopesand values of loglo A . The accuracy of the results, although differing slightly in the variousexperiments, was always better than 0.4 % for the derived work functions and better than1 % for the values of loglo A. For the average work function of the clean surface we assumeda value of 4.50V. All the Fowler-Nordheim plots of gas-covered tips were taken afterreducing the gas pressure to 5 x 10-9 torr.RESULTS AND DISCUSSIONCARBON MONOXIDE ON TUNGSTENELECTRON MICROSCOPYWe have studied the adsorption states at 300 and 78°K.The results obtaindewith saturated surfaces are summarized in the following scheme (see top of page 33).The data are averaged from several experiments.(a) The adsorption state at78°K depends on the way it is obtained. Direct adsorption at 78°K (step 1) leads toa large increase in #, while adsorption at 300°K and subsequent saturation at 78°K(steps 3 and 4) give a much smaller increase in #. The same result can be obtainedby temporary heating of the 78"K(I) state to 300°K (steps 2 and 4).Thus, theattainment of the stable 78°K (10 state involves an activated process, which has theeffect of decreasing #,(b) The same work function and electron pattern result for the 300°K state, itbeing immaterial whether this state is reached directly (step 3) or via the 78°K (I) stateSome of these results need additional comment.Step 4 was found to be completely reversible (step 5)FIG. 3.-(a) Clean tungsten (field-evapor ated at 9.6 kV) ; image at 78°K and 8.05 kV ; (b) samesurface after CO adsorption at 78°K (partial coverage) ; image at 78°K and 8.05 kV.[To face page 32FIG. 4.-Tungsten surface after saturated CO adsorption at 78°K; image at 78°K and 7.20 kV.FIG. 5.-Tungsten surface after saturated CO adsorption at 300°K; image at 78°K and 9.6 kVFIG.6.-Tungsten surface after CO adsorption at 300°K (partial coverage) ; image at 78*K and9.2 kV.FIG. 7.-Tungsten surface after CO adsorption at 78°K (partial coverage) and heating to 280°Kfor 2 min ; image at 78°K and 9.7 kVFIG. %--Field desorption and field evaporation of surface layers after saturated CO adsorptionat 300°K ; for experimental data and conditions, see table 2A. A. HOLSCHER AND W. M. H. SACHTLERA 4 = 0.73 V3 Aloglo A == -1.4 1A# = -0.31 VA loglo A = -0.3 A 4 = -0.47 V4 v A loglo A = 0.033A 4 = 1-20VA loglo A = - 1.55(steps 1 and 2). This also holds for the stable 78°K (XI) state, which can be obtainedvia the 300°K state. Although there is a slight difference in the decrease in loglo Afor both routes, it seems that the 300°K state is not greatly influenced by pre-adsorp-tion at lower temperatures.(c) Adsorption at 300°K as well as at 78°K causes a large decrease in the pre-exponential factor A .In contrast to this, relatively small if any changes in A occurupon heating the 78°K (I) state to 300°K and redosing at 78°K.Although the 300°K state is independent of the way it was reached, it is not atrue equilibrium. We observed that the emission current at 300°K rose slowly withtime, even in the absence of gaseous CO. No current rise was noticed at 78°K.The rise could be accelerated by temporarily heating the tip to 330°K. Afterheating for 5 min, the work function (measured at 300°K) was dropped by 0.03 V,while loglo A had increased by 0.3.After redosing such a tip with CO at 78°K thework function returned to the 78°K (11) value, but loglo A was lower by 0.3 comparedwith the 78°K (11) value. From the temperature dependence of the current rise anactivation energy of 8 kcal was derived for this process.ION MICROSCOPY(a) ADSORPTION AT 78"K.-We have compared the ion images of a clean, ideallyordered surface with those obtained after adsorbing various amounts of CO at 78°K.In each of these experiments, ultra-high vacuum conditions were restored and thesurface subjected to field evaporation at 78°K before the desired amount of CO wasadmitted. The adsorption states were characterized by their F.-N. plot, the para-meters of which are given in table 1. Up to a surface coverage characterized byA 4 = 0.35 V not the slightest difference was observed between the ion images takenbefore and after adsorption of CO.This striking result proves that adsorbed CO molecules cannot be observed in theion microscope. Apparently, they have been removed from the surface by the highfield without disturbing the metal surface.This field desorption is probably pro-moted 12, 13 by the impact of helium atoms, which, due to polarization, acquire akinetic energy of about 0.2 eV.When with increasing coverage A# becomes equal to about 0-5 V, the first smallchanges of the ion picture occur. This is seen from fig. 3, where 3(a) is the image ofthe clean surface and 3(b) the one obtained after adsorption. The disordering of theoriginal array of bright spots is most pronounced in the regions between the (21 1) and(100) faces.The four atoms visible on the (100) face of the clean surface are absen34 F.E.M. AND F.I.M. OF co ON wfrom the ion image taken after adsorption. Closer inspection reveals that also inother regions atoms have disappeared and " holes " are left behind. At a coveragewhere A$ = 0.65 we observed a more severe disordering, still concentrated in theareas between (21 1) and (100).TABLE 1 .-FOWLER-NORDHEIM DATA AND OBSERVED CHANGES IN ION IMAGES AFTERADSORPTION OF INCREASING AMOUNTS OF co ON A FIELD-EVAPORATEDTUNGSTEN SURFACEchange in ion image comparedwith image of clean surface4-59 - 0-40 no change at all4.61 - 0.24 no change at all4.85 - 0.43 no change at all5.03 - 0.55 very slight disordering between (21 1) and(100) see fig.35.15 - 1.27 increased disordering, also around (1 1 1)5.70 - 1.52 generally spread disordering see fig. 4Upon saturation (A$ = 1.2 V) the disordering is more generally spread over thewhole surface, as shown in fig. 4. Similar images obtained from tips less sharp thanthe one of fig. 4 showed slightly greater effects of disorder, but the symmetry of thepattern always remained clearly visible.How are these changes to be interpreted? The new dots are not due to CQ mole-cules adsorbed on an undisturbed surface as such molecules would have been tornoff by the combined action of the high field and impinging helium atoms. There isno reason either to assume that field desorption is incomplete at high coverage.On the contrary, it is well known that the heat of adsorption decreases with coverage.We therefore propose identifying the changes with a disordering of the tungstensurface itself.The following model can then be set up to fit the experimentalobservations. With increasing coverage, metal atoms move out of their originallattice sites into more favourable positions surrounding adsorbed CO molecules.Imaging of such a re-arranged surface in the ion microscope will result in a completeor partial field desorption of the adsorbed complexes (including both CQ moleculesand tungsten atoms), leaving behind the remnants of a disordered surface structure.The re-arrangement, and thus the observed changes depending on the surface cover-age can be understood for two reasons.First, it is reasonable that the re-arrangedstructure is thermodynamically more stable only above a certain minimum coverage.Secondly, re-arrangement of surface atoms is an activated process. It is conceivablethat the activation energy decreases with increasing coverage owing to the weakeningof the intermetalIic bonds upon formation of the chemisorption bonds.4 Moreover,it is to be expected that the activation energy required for re-arrangement is differentfor the various crystal faces and smallest for the atomically rough areas, as e.g., (41 1)and (61 1). These are in fact the regions where the first changes are observed.A further verification of this interpretation is provided by experiments reportedbelow, (b) and (c).(b) ADSORPTION AT ca.30O0K.-Fig. 5 shows a surface which had been saturatedat 300°K ( A 4 x 0.73 V, A loglo A = - 1.4). The symmetric pattern of the cleansurface (not shown) has completely changed into a random array of dots all over thesurface. The disorder is far more intense than was observed after direct adsorptionat 78°K (fig. 4). An exactly similar pattern is obtained after adsorption at 78°Kfollowed by heating to 300°K and redosing at 78°K ( A 4 = 0.42 V ; A loglo A = - 1.7)A. A. HOLSCHER AND W. M. H. SACHTLER 35These experiments prove that the disordering as observed in the ion picture of fig. 5involves an irreversible activated process. This may be expected when metal atomsparticipating in a chemisorption layer have to re-arrange in order to achieve a morestable configuration.The occurrence of an activated process in going from the low-temperature state 78°K (I) to the 300°K state was already concluded from fieldemission experiments alone. The ion microscope studies have now revealed thenature of this process.Dosing experiments, similar to those reported under (a) showed that the disorderincreased with increasing coverage. Although a completely unchanged pattern wasnever obtained after adsorption at 300"K, the changes at low coverage were small.A more intense disordering, but still much less than at saturation, is shown in fig. 6,for a surface for which after adsorption at 300°K A 4 = 0.39 V and A log10 A = - 0.52.There is a marked parallelism between the degree of disorder and the decreasein loglo A .To illustrate this, a tungsten tip was dosed at 78°K up to a coveragewhich, at this temperature, involved little or no re-arrangement, but would havecaused severe disorder had dosing taken place at 300°K. The tip was then heated toabout 280°K for 2 min and a F.-N. plot recorded. The characteristic parameterswere A 4 = 0.51 V and A loglo A = -0.42 with respect to clean surface. Aftercooling to 78"K, the ion image showed that only very small changes had occurred(fig. 7). Apparently, the rate of the activated re-arrangement on a partially coveredsurface is low at 280°K.A comparison of the F.-N. data and the ion images of fig. 6 and 7 shows that themore pronounced changes of fig. 7 are accompanied by a larger decrease in loglo A ,but a smaller increase in work function.A surface saturated at 300"K, giving ahighly disordered ion-image, is characterized by a work function of 5.23 V, whichis only 0.2 V more than the value obtained in the 280°K case. However, the valueof A loglo A decreases from -0.4 to - 1.4. These results suggest that at a givencoverage re-arrangement is accompanied by lowering of both the work function andof log10 A .(C) FIELD DESORPTION AND FIELD EVAPORATION OF SURFACE LAYERS.-The resultspresented under (a) and (b) suggest that, under the conditions of image formation,the greater part of the adsorbed CO is removed from the surface by field desorption.To investigate this point in more detail and to clarify the nature of the surface layersleft behind, we carried out experiments in which ion pictures were taken of successivelyadvanced stages of field desorption and field evaporation after a saturated COadsorption at 300°K. After taking an ion picture we reversed the polarity of themetal tip and determined a Fowler-Nordheim plot.Table 2 summarizes the experi-mental data and conditions. The corresponding ion images of a representative partof the surface are shown in fig. 8. All the ion pictures were taken under exactly thesame conditions. All A 4 and A loglo A values are referred to the clean surface afterfield evaporation at 78°K in the presence of helium. This circumstance might beresponsible for the relatively low work function increment after CO adsorption.Examination of the F.-N.data and ion pictures reveals many remarkable facts.In the successive stages of field desorption the work function passes first through aminimum value of 4.50 V and then through a maximum (see later).It is clear that the mere application of the required image voltage of 7-8 kV removesa major part of the chemisorbed layer and that the field desorption effect is markedlyenhanced by the presence of helium. After field desorption at 8.4 kV the surface isregarded as clean, as from this stage the values of 4 and loglo A are no longer signifi-cantly different from those of the initially clean surface36 F.E.M. AND F.I.M. OF co ON wThe ion image of fig. 8(b) represents the characteristic disordered pattern whichis always obtained after adsorption at 300°K.Relatively small changes have occurredafter field desorption at 8-2 kV (8(c)). Field desorption at 8.4 kV (8(d)) involvesmore pronounced changes, particularly visible in the (1 1 1) region. Although accord-ing to the F.-N. data the surface is now completely clean, the arrangement of brightspots, showing strong differences in intensity, is still disordered. We also observerelatively large dark areas which are not imaged at all, in contrast with the same areasTABLE 2.-EXPERIMENTAL DATA OF FIELD DESORPTION AND FIELD EVAPORATIONOF SURFACE LAYERS AFTER SATURATED CO ADSORPTION AT 300°Kpicture A loglo A = conditions after whichnumber * ' log10 -4co,e,d-h310 dclean F-N plot and/or picture8a8b8c8dSe8f8s8h8i4.504.954.504-614.624.534.564.5 14.550- 1-52- 0.66+0*17 + 0.21- 0.03 + 0.090.00+ 0.07field evap.at 8-8 kV, 78°K with HeCO adsorption at 300°Ktip 5 min at 7.8 kV without Heafter taking 8btip 1 min at 8-2 kV with Hetip 1 min at 8.4 kV with Hetip 1 rnin at 8.6 kV with Hetip 1 rnin at 8-7 kV with Hetip 1 min at 8.8 kV with Hetip 1 rnin at 8.9 kV with Hetip 1 min at 9.0 kV with He* Ion pictures taken at 2 x 10-3 torr He, 78"K, 7.8 kV, and 15 min exposure time.on photograph 8(a) of the clean surface. An interesting phenomenon can be noted bycomparing 8(d) with 8(b). The latter photograph shows in some regions a slightfuzziness, which is absent from 8(d). As the experimental conditions for all photo-graphs were the same, the observed effect is significant ; it may be attributed to smallamounts of chemisorbed CO, which in view of the F.-N.data were still present on thesurface depicted in 8(b), but not on the surface of 8(d).Photographs S(e), (f), (g), (h) and (i) show the effect of field evaporation byapplying successively higher voltages : surface atom layers are torn off, the disordergradually decreases and atoms " fill up " the dark areas. After about four layers havebeen removed, a completely ordered surface is again exposed. The number of layerstorn off can most easily be determined by comparing the structure around the (211)and (1 11) faces on the various photographs.That the removal of surface layers can occur already below the voltage initiallyrequired to give the field evaporated end state shown in fig.8(u) can be explained asfollows. When, during the formation of the chemisorption layer, metal atoms moveout of their original lattice positions, vacancies are left behind. By diffusion processes,these vacancies are partly distributed among deeper lying atom layers until anequilibrium is established. As a consequence, the surface which becomes exposedafter the chemisorption " crust " is field-desorbed, shows a " hill and valley " structureon an atomic scale, the valleys being either single vacancies or vacancy clusters. Anion image of such a surface reveals only (part of) the atoms on the hills, because abovethe valleys the ionization probability for the helium atoms is comparatively small.Field desorption from such a disordered structure proceeds more easily, i.e., at alower applied voltage, than from a completely smooth surface.The attack commencesat the relatively protruding atoms and edges where the field strength is highest andcontinues until the increase in the (local) radius of curvature causes the field strengthto drop below the value required for desorptionA. A. HOLSCHER AND W. M. H. SACHTLER 37Eventually atomlayers become exposed which have no more vacancies than are normally observedand the surface is again ideally ordered.From these experiments we conclude that : (i) the disorder observed in the ionimages obtained after CO adsorption cannot be attributed merely to adsorbed COmolecules. Even after CO has been removed from the surface the disorder persistsand should therefore be identified with disurbances in the initially perfect arrangementof the metals atoms; (ii) there is evidence that chemisorbed CO molecules, only avery small part of which can escape field desorption during image formation, lowerthe image contrast instead of showing up as bright spots ; (iii) the adsorption-inducedre-arrangement of metal atoms strongly influences deeper lying atom layers by in-jecting vacancies into these layers.We think that these conclusions and those obtained from the experiments under (b)confirm the correctness of the interpretation put forward at an earlier stage to explainthe phenomenon observed in the ion microscope after adsorption of CO on field-evaporated tungsten surfaces.At consecutively higher voltages the whole process is repeated.DISCUSSIONIn the foregoing section we have given a description of the adsorption of CO ontungsten which satisfactorily explained the results of both ion microscope and fieldemission experiments.It implied a re-arrangement of surface metal atoms, whichwas dependent on coverage and, being an activated process, was far more pronouncedat 300°K than at 78°K.We now complete this description with further details, including a discussion ofthe changes in 6 and loglo A , and arrive at the following picture of the adsorption ofCO on tungsten. Adsorption at 78°K up to a fairly high coverage leads to a chemi-sorption layer uniformly spread over the surface, while no re-arrangement of substrateatoms occurs.The dipole moments of the surface bonds have their negative endpointing outwards, which results in a high work function increment. Upon formationof the chemisorption bonds the intermetallic bonds are weakened and this eventuallyallows a restricted surface re-arrangement to take place, particularly on the lessdensely packed crystal faces. The re-arrangement causes a tighter bonding of a partof the adsorbed CO molecules by embedding them in a two-dimensional tungstencarbonyl structure. New adsorption sites are thus created, which we visualize asbeing located on top of tungsten atoms that have penetrated through the adsorbedCO layer. CO molecules adsorbed on these sites are less strongly bonded thanthose in the carbonyl layer, the latter interacting with more than one tungsten atom.Upon saturation, an additional very weakly bonded layer may be formed on top of thechemisorbed layer.In assessing the effect of this complex adsorption process upon the work function,it is conceivable that any re-arrangement in the chemisorbed layer has the effect ofdecreasing the effective dipole moment vector perpendicular to the surface by intro-ducing components which point with the negative pole parallel to the surface or eveninwards.However, the rearrangement at 78°K takes place to a very limited extentonly. Therefore the work function is decreased only slightly, as compared to thehypothetical case when no re-arrangement takes place. The physisorbed layer mostprobably lowers the work function.Upon heating the 78°K state to 300°K drastic changes occur.First, the physi-sorbed layer is desorbed. If no other processes took place the work function wouldincrease. However, instead, there is a considerable decrease in work function o38 F.E.M. AND F.1.M. OF co ON wabout 0.5 eV, which now, from the ion microscope results, can be correlated with astrongly increased re-arrangement in the surface layer. The degree of this re-arrangement and the ultimate surface structure is probably different for the variouscrystal faces. In some areas it presumably takes place to an extent which justifiesthe term “ surface corrosion ”, i.e., CQ molecules are completely incorporated in thesurface structure.By analogy to what is observed in the formation of oxide layers, we assume thatthe corrosion layer is mainly formed by metal atoms diffusing outwards. Such asupposition is strongly sustained by the field evaporation experiments which provethe existence of vacancies underneath the chemisorption layer. As CO moleculesin this layer are more strongly bonded than those at the surface, different statesbecome distinguishable with respect to the energy required for desorption.Moreover,the states become discernible with respect to the surface potentials, as re-arrangementschange the component normal to the surface of the bond dipoles.After peeling off one or more surface layers, the work function drops to thevalue of the clean surface, although there is still some CO present.This can beroughly explained by visualizing that the surface now contains tungsten atoms withCO molecules underneath. Further field desorption reverses the situation again ;this results in an increase in work function.Upon re-dosing the 300°K state at 78°K a weakly bonded adsorption layer is formedon top of the corroded surface, which is accompanied by a further decrease in workfunction.The critical coverage where re-arrangement sets in, is much lower at 300°K thanat 78°K. This result proves that the critical coverage is determined by the pre-requisite that metal-metal bonds must be sufficiently weakened by chemisorptionsuch that the activation energy for re-arrangement becomes sufficiently low.Adsorption both at 78°K and at 300°K greatly decreases the pre-exponentialfactor A of the Fowler-Nordheim relation.Such a decrease has also been foundwith other adsorption systems; it is certainly not simply caused by a reduction ofthe emitting surface. Two other explanations have been put forward in the literature,both being concerned with the large decrease in loglo A at low temperature. In thefirst model, adopted by Menzel and Gomer 14 for CO on tungsten, the high electro-static field polarizes the adlayer, inducing an extra work function increment, whichto a first approximation does not alter the slope of the F.-N. plot, but decreasesloglo A. In the second model proposed by van Oostrom 15 for nitrogen on tungsten,the highly polarizable adlayer modifies the shape of the potential barrier throughwhich the electrons tunnel.One of the models, or a combination of both, might explain the observed changesin loglo A upon direct CO adsorption at 78°K.Neither model, however, accounts forthe decrease observed at temperatures where no weakly bonded polarizable layer canbe present. The relevant mechanism is in this case related to the rearrangement ofthe surface at these temperatures.We tentatively propose the following qualitative picture. The corroded layer andthe induced vacancies are expected strongly to disturb the periodic potential at thesurface. As a consequence, the number of electrons arriving at the potential barrier,i.e. the supply function, may be drastically decreased by an increased scattering inthe surface layers.16 If, in the small range of electron energies relevant in fieldemission, the scattering is assumed to be independent of the electron energy, it hasthe same effect as a decrease in emitting area, i.e., in loglo A.A weakly bonded layer on top of a corroded chemisorbed layer, formed at lowtemperature, gives rise to a relatively small change in loglo A .In terms of Corner’A. A. HOLSCHER AND W. M. H. SACHTLER 39model, CO molecules in such a layer are apparently less polarized by the field thanthose on top of what may be called a chemi-adsorption layer.Our view on the different types of adsorption and the specific changes in Cp andloglo A they induce, can be summarized in the following scheme.weak adsorption physisorptionA#<O 1 AlogloA$Ocorrosive - chemi-adsorptionchemisorp tion4A loglo A < 0 1 A#>OA loglo A $0clean cleanThe slow rise in emission current subsequent to CO chemisorption at 300”K,which rise has an apparent activation energy of 8 kcal, might tentatively be ascribed toan ordering of either the corrosion layer or of the disturbed tungsten layers beneath it.We may now review the bearing of our results on those obtained previously byothers.CO on tungsten was studied by many authors with a variety of techniques.For a survey of this work we refer to articles by Swanson and Gomer,l7 and byEhrlich.18 In general our experimental observations agree well with previous work.However, the preser,t combination of field electron and ion microscope studies hasresulted in a new interpretation of the phenomena observed.The new picture is atvariance with one of the conclusions by Ehrlich,los 11 who interpreted the disorderlyarranged bright spots in the ion images after CO adsorption as individual moleculeswhich had remained unaffected by the severe conditions of ion image formation.He concluded that the tungsten surface itself had been changed only negligibly.The same difference in interpretation holds for the adsorption of nitrogen on tungstenwhich we have also investigated.Our experimental results, although obtained on field-evaporated tungsten tips,are in fzir agreement with those of Gomer 7 9 1 7 s 19 et al., if we identify virgin layers,/?-states and cc-states of their terminology with what we have called the chemi-adsorptionlayer, CO molecules incorporated in corrosive chemisorption layers, and the moleculesmore weakly adsorbed on sites created on top of the corroded layers.The rearrangement Gomer proposes is “ a flipping of CO into a more tightlybound configuration locally ”.However, it seems unlikely that the potential barrierbetween the unstable and stable sites (in this model at a distance of about 2 A fromeach other) is so high that a molecule hitting the surface on an “ unstable” sitecannot immediately “ flip over ”, upon making an initially highly activated surfacebond.From their electron impact desorption experiments Menzel and Gomer 20 con-cluded that a thermal conversion of virgin states to p- and cr-states (the latter occurin their and in our model only when there is p-CO) starts at about 270”K, in completeagreement with our findings. A careful examination of other results derived withthe electron impact technique reveals that most conclusions fit well in the modelproposed in the present work40 F.E.M.AND F.I.M. OF co ON wH2CARBON MONOXIDE ON TUNGSTEN PRECOVERED WITH HYDROGENAfter saturation with hydrogen both at 78 and 300°K on a field-evaporated tung-sten tip, the ion image had not changed compared to that of the clean surface. Thus,hydrogen does not induce any re-arrangement of tungsten atoms. If, however, ahydrogen-covered tip was exposed to 10-6 torr CO for about 5 min both at 78 and300°K we obtained the same type of pattern as was found for CO adsorption onclean tungsten at these respective temperatures (see fig.4 and 5). Apparently COcan displace adsorbed hydrogen, and subsequently induces a re-arrangement oftungsten atoms. The field emission results are summarized in the following scheme.A# = 0.48VA loglo A = -0.7field-evaporated at 78°K# = 4.50, loglo A x -5.5clean clean300°K 78°K4300°K 4 = 4-98 V€32 A$ = 0.80VA loglo A = -0.4 ! 78°K 4 = 5-30VAt 78°K CO adsorption lowers the work function of a hydrogen-covered tip.This can be understood by assuming that CO is mainly present in a physisorbedlayer, while the major part of the strongly bonded chemi-adsorbed hydrogen layeris not removed. In complete agreement with this assumption is the large drop inloglo A , characteristic of a higher polarizable layer.Upon heating to 300°K the workfunction and loglo A further decrease. Presumably, the greater part of the hydrogenis now replaced by CO and a corrosive chemisorption layer is formed. The samestate can be achieved by direct CO adsorption at 300°K on a precovered hydrogen tip.Displacement of hydrogen by CO is accompanied by a relatively small increase inwork function and a large decrease in loglo A , this being again in agreement withthe conversion of a chemi-adsorbed hydrogen layer into a carbonylic corrosion layer.The 300°K state of the mixed adsorption is characterized by a slightly lower workfunction than that of the pure CO adsorption state at 300°K. This is most probablydue to hydrogen still present in the corrosion layer.After adsorption at 300°K orheating the 78°K state to 300°K the same steady increase in current was observed asafter CO adsorption on clean tungsten. After intermediate heating to 330"K, thework function had changed little (A# = - 0.03 V), but loglo A had increased markedly(A loglo A = 0.35). The activation energy of the process was found to be about12 kcalA. A. HOLSCHER AND W. M. H. SACHTLER 41Hydrogen admitted to a tip which was precovered with CO and temporarilyheated to 330"K, caused no further change in work function and loglo A. However,when intermediate heating to 330°K was omitted, a decrease in both work function( - 0.3 V) and loglo A (- 0-3) was observed. Heating to 330°K restored the originalvalues of 4 and loglo A characteristic of the pure CO chemisorption layer.This behaviour qualitatively resembles the results found by Siddiqi and Tompkins,21who studied the adsorption of hydrogen on nickel precovered with CO at 78°K bymeans of a contact-potential technique.They observed a large reversible decreasein work function caused by the adsorption of hydrogen on the CO-covered surface.CONCLUSIONSThe field emission and field ion experiments reported in the present work haveallowed us to set up a detailed picture of the adsorption of CO on clean tungstensurfaces and on tungsten surfaces precovered with hydrogen. The most remarkablefeature of the proposed model is a re-arrangement of substrate atoms, which evenoccurs, although to a slight extent, at very low temperatures.We now formulatetentatively some conclusions which may be of more general importance, in thatthey have implications for other adsorption systems as well. Also, the possibilitiesof the electron emission and ion microscope techniques deserve some further comment.(i) The formation of chemisorption bonds involves a weakening of the intermetallicbonds in the metal surface. The degree of weakening increases with the coverageand with the strength of the chemisorption bonds formed.(ii) Owing to weakening of intermetallic bonds the activation energy required formetal atom re-arrangement of a covered surface can be much less than for the cleansurface. The stronger the chemisorption bond and the smaller the cohesive energyof the metal, the more readily surface re-arrangement takes place.Since the strengthof chemisorption bonds goes parallel with the bond strength in the corresponding bulkcompounds,4 thermodynamic data of bulk phases can give a rough indication whetherand, if so, how easily surface re-arrangement will occur.(iii) The possibility of surface re-arrangement even at low temperatures and onmetals with a high melting point, should be taken into account in any interpretationof chemisorption phenomena and of composite layers in particular. It can no longerbe assumed d priuri that the surface consists of an assembly of adsorption sites ofwhich the configuration is not altered upon chemisorption. Chemi-adsorption maybe followed by the formation of corrosive chemisorption layers already at low cover-ages. Molecules or atoms incorporated in the corrosion layers will be more stronglybonded than those on top of these layers. Also, it is to be expected that the newsurface structures will be different for the various crystal faces, this giving a furtherdifferentiation in the " adsorbed states " on one and the same metal.(iv) From an experimental point of view the present work justifies the conclusionthat the ion-microscope technique offers many possibilities for studying chemisorption,particularly when applied in combination with field electron emission techniques.Ion microscopy can provide, at least in part, the atomic detail which is needed for abetter understanding of chemisorption phenomena. Apart from the fact that only" frozen-in " states can be observed, this work clearly shows that a severe limitationis set by the phenomenon of promoted field-desorption, which prohibits theobservation of the chemisorbed species themselves. Possibly the application ofother image gases, such as neon or argon, which give images at lower electric fieldsthan helium does, might prevent field desorption during imaging42 F.E.M. AND F.I.M. OF co ON w1 Ogurj, J . Physic. Soc. Japan, 1964, 19, 83.2 I-Iolscher, J. Chem Physics, 1964, 41, 579.5 Delchar and Ehrlich, J. Chem. Physics, 1965, 42, 2686.5 Lander, Surface Sci., 1964, I, 125.6 Farnsworth, Adv. Catalysis. 1964, 15, 31.Slachtler and Van Reyen, J. Res. Inst. Cat., 1962, 10, 87.Gomer, Field Emission and FieM Ionization (Harvard University Press, Cambridge, Massa-chusetts, 1961).8 George and Steer, J. Chem. Physics, 1962, 37, 1935.9 Ehrlich and Hudda, J. Chem. Physics, 1962, 36, 3233.10 Ehrlich, Trans. N. Y. Acad. Sci., 1963, 101, 722.11 Ehrlich, Adv. Catalysis, 1961, 14, 255.12 Nishikawa and Miiller, J. Appl. Physics, 1964, 35, 2806.13 Ehrlich and Hudda, Phil. Mag., 1963, 8, 1587.14 Menzel and Gonier, J. Chem. Physics, 1964, 41, 33 1 1 .15 van Oostrom, Thesis (Amsterdam, 1965).16Toya, J. Res. Inst. Cat., 1962, 10, 236.17 Swanson and Gomer, J. Chem. Physics, 1963,39, 2813.18 Ehrlich, Proc. 3rd Znt. Congr. Catalysis (North-Holland Publ. Co., Amsterdam, 1964), vol. 1,19 Gorner, J. Chem. Physics, 1958, 28, 168.20 Menzel and Gomer, J. Chem. Physics, 1964,41, 3329.21 Siddiqi and Tompkins, Actes du 2 i h e conqrG.7 international de Catalyse (Editions Technip,p. 113.Paris, 1961), part 2, p. 1767