Adsorption and Surface Reactivity of Metalsby Secondary Ion Mass SpectrometryPart 1.-Adsorption of Carbon Monoxide on Nickel and CopperBY MICHAEL BARBER, JOHN c. VICKERMAN" AND JOHN WOLSTENHOLMEDepartment of Chemistry, The University of Manchester Institute of Science andTechnology, Sackville Street, Manchester M60 1 QDReceived 10th March, 1975The adsorption of carbon monoxide on polycrystalline nickel and copper has been studied bysecondary ion mass spectrometry (SIMS), in the temperature range 77-390 K. The results indicatethat carbon monoxide is adsorbed in both a linear and a bridged form on nickel in the range 77-370 Kbut only in a bridged form at temperatures above this. For copper only a linear structure wasobserved at 77 K and only a bridged structure at 295 K.On both metals there is a build up ofcarbon on the surface at 390 K. These results are discussed in terms of those obtained from othertechniques.The adsorption of carbon monoxide on nickel and copper has been studied bymany different techniques. A review of the more recent literature reveals that areasonably consistent picture is emerging.NICKELThe picture here appears to be that carbon monoxide is adsorbed in three distinctphases which, in the terminology of thermal desorption studies, are called PI, P2and y. This agrees with the results of LEED studies.2* These indicate that threephases are formed but that these phases do not co-exist on the surface and that phasechanges occur at definite coverages. 1.r. experiments give rise to two bands whencarried out at temperatures at which the most weakly adsorbed, y, phase is expectedto be present.The high frequency band is assigned to a linear Ni-C-0 structureand the band at lower frequencies to a bridged Ni2C0 structure.There has been some argument in the literature as to whether the adsorption ismolecular or dissociative. One LEED study suggests that molecular carbonmonoxide is adsorbed on top of a layer formed by dissociative adsorption of carbonmonoxide. A recent UPS and XPS study,6 however, indicates that, at room tempera-ture, the adsorption of carbon monoxide is entirely in the molecular form and onlyprolonged heating of the metal in the presence of carbon monoxide causes dissociationto occur.COPPERThe literature is quite clear that there are two types of adsorption of carbonmonoxide on copper. The first type of adsorption causes an increase in the intensityof the single i.r.band with coverage and the second type of adsorption causes nochange in its intensity but the band shifts to higher frequencies as coverage progresses.The change-over in the i.r. spectrum coincides with a maximum in the surface potential4M. BARBER, J . C. VICKERMAN AND J . WOLSTENHOLME 41and a phase change on the surface, as observed by LEED.*'l0 The second, moreweakly bound phase can be formed only at low temperatures and is completelydesorbed at 195K. The more strongly bound species is desorbed at room tem-perature.SECONDARY ION MASS SPECTROMETRYSince SIMS, as applied to surface chemistry, is a relatively new technique we haveattempted here to investigate its scope and limitations by applying it to well-studiedsystems.Its limitations could be that the primary ions cause fragmentation of thesurface species or cause changes in the metal surface which would alter the mode ofadsorption of the gas.It is believed that neither of these phenomena occur and, further, that SIMS cannot only confirm a suggested model but also clarify and, to a certain extent, amplifysuch a model.It is the aim of this paper to show how far this is true by comparing our resultswith the overall picture of CO adsorption which emerges from the data outlined above.Exact agreement with the detail of these results would not be expected, since it is verydifiicult to reproduce precisely the conditions under which these experiments wereperformed.EXPERIMENTALThe SIMS apparatus was constructed by Vacuum Generators Ltd. and was similar indesign to that described by Benningh0ven.l' The instrument has been described in detailelsewhere.Briefly, the instrument consists of two, separately pumped chambers, a preparationchamber and an analysis chamber connected by a gate valve.The sample is cleaned in thepreparation chamber by argon ion etching using 6 kV argon ions from a V.G. AG2 ion gunat a current density of about 100pA cm-2. After cleaning, the metal is transported to theanalysis vessel by means of a u.h.v. bellows system.In the analysis chamber the primary ions are incident on the sample at an angle of 70"from the normal with an energy of 3 kV and a current density of 10-9-10-10 A ern-'.The base pressure attainable in both chambers after baking was about 2x 10-l' Torrbut the pressure in the analysis chamber rose to lo-* Torr due to argon entering the chambervia the primary ion gun when in operation.The argon for both the primary ion source and the AG2 gun was purified by passing itthrough a trap containing molecular sieve (zeolite) to remove traces of hydrocarbon. Toremove water the trap was cooled in a mixture of solid carbon dioxide and acetone.Thecarbon monoxide was treated similarly but, in this case, the coolant was liquid nitrogen.Before use each of the traps was baked in order to remove impurities from previous experi-ments.In these experiments high purity (99.99 %), polycrystalline nickel and copper foils wereused.These were supplied by Goodfellow Metals.The exposures quoted in this paper are expressed in langmuirs (L) where 1 L = Torr s(1 Torr = 133.3 Pa). The pressures used in these experiments were such that the totalexposure time was between 10 and 60 s.RESULTSNICKELThe sample was thoroughly cleaned by ion etching and spectra were run to ensurethe cleanliness of the surface. Fig. l(a) shows the mass spectrum from ~i clean surface.The lack of any peaks corfesponding to carbon or oxygen-containing species shoul42 ADSORPTION OF co ON Ni A N D CUbe noted. The peaks at mass numbers 23 and 39 are due to sodium and potassiumrespectively. Obviously, some estimate of the level of these impurities must be made.The reported value for the sputtering coefficient of K+ l 3 is found to be lo3 timeshigher than that for Ni+ and, since the potassium and nickel peaks in fig.1 areapproximately the same height then the surface concentration of potassium must beof the order of 0.1 %. The surface concentration of sodium is expected to be similarbut there is no data on its sputtering coefficient. The value of 0.1 % is only a roughestimate since Hagstrum used potassium primary ions. The value of the sputteringcoefficient when argon ions are used is not expected to be very different. The massspectrum of negative secondary ions did not indicate the presence of any impurities.The temperature of the clean sample was then reduced to 77 K.Upon the admission of 2 L of carbon monoxide at this temperature, the intensityof the peaks due to Ni+ (mass numbers 58 and 60) increased approximately 5 timesand a large pair of peaks was observed at mass numbers 86 and 88 due to NiCO+.Fairly intense peaks also appeared at 1 16, 118 and 120 due to Nii but only relativelysmall peaks at 144, 146 and 148 due to Ni2CO+, see fig.I@).Upon the admission of a further 2 L of carbon monoxide the intensity of the Ni+peaks increased by about 12 % of its former value and that of the NiCO+ peak by25 % of its original value. There was no significant change in the heights of theNi2CO+ peaks or the Niz peaks.Ni'Ni; LNa' kmle mle(4 (b)FIG. 1 .-SIMS spectrum of nickel surface (a) after argon ion etching, (b) after admission of 2 L carbonmonoxide at 77 K.Admission of a further 10 L of CO caused a similar, but far less marked, changein the appearance of the spectrum while a further 10 L made no significant change,i.e., saturation had occurred at a dosage of less than 14 L.On allowing the sample to warm to room temperature the intensities of the NiCO+and Ni2CO+ signals decreased until, at room temperature, their intensities were onlyabout 10 % of those at 77 K.Fig. 2 shows a comparison of the spectrum obtaineM. BARBER, J . C.VICKERMAN AND J . WOLSTENHOLME 43after saturation at 77 K with that obtained after heating to room temperature; notethat the spectrum at 295 K was run at a higher sensitivity.Similar spectra were obtained starting from a clean nickel surface at room tem-perature except that saturation did not occur until about 40 L of carbon monoxidehad been admitted.The value obtained by Williams et aZ.I4 using UPS and XPSwas only 22 L. There were also small peaks corresponding to Ni2C+ and Ni20+,the intensities of these were approximately equal and about 10 % of the intensityof the Ni2CO+ peaks, see fig. 3(a).After a very large number of doses of CO, small peaks appeared in the spectrumwhich could be assigned to Ni(C0): and Ni(C0);.Upon heating the sample to 390 K, after saturation of the surface at 295 K, theintensities of the NiCO+ peaks quickly diminished and could not be detected at about370K. There was no significant alteration in the intensities of the peaks due toNil, Ni2C+, NizO+ or Ni2CO+ and no new species appeared.A further dose of 40 L carbon monoxide was admitted at this temperature whichcaused the Nil, Ni2C+, Ni20+ and NizCO+ signals to double in intensity but a furtherNi:NNimle mle(a) (b)FIG.2.--Comparison of SIMS spectra (a) after saturation with carbon monoxide at 77 K and (b)after warming the saturated surface to room temperature. Note that (b) was run at 10 times greatersensitivity.20 L of carbon monoxide made no significant difference. Prolonged heating at 390 Kand several large doses of carbon monoxide (each dose being about 0.5Torr for5 min) caused a build-up of the Ni2C+ peak relative to that of the Ni,CO+ until noNi2C0 remained on the surface, fig. 3(b).This is in agreement with the resultsobtained by Joyner and Roberts.6COPPERAgain, the sample was thoroughly cleaned and the SIMS spectrum was used asan effective check on the cleanliness of the rrurfacc, fig. 4(44 ADSORPTION OF co ON Ni AND CUAdmission of 20 L doses of carbon monoxide at 77 K caused a build-up of theCuCO+ peaks (mass numbers 91 and 93) until saturation occurred at about 80 L.Apart from Cuf and CuCOf the only other peaks appearing in the spectrum wereNi,'NiCOdx 3 1 NiN i'Ni20+ Ni,C+IL.1 1 1 1 I 1 1 I I I140 120 100 00 6 0FIG. 3.-SIMS spectrum nickel surface (a) after saturation with CO at room temperature, (b) afterlarge doses of CO at 390 K.c u;c u*x 11 10CU'K'FIG. 4.-SIMS spectrum of copper foil (a) after argon ion etching, (6) at 77 K after a saturation doseof carbon monoxideM.BARBER, J . C. VICKERMAN AND J. WOLSTENHOLME 45due to Cu,'. The intensities of the Cu,' peaks are too low for them to be visible infig. 4(b) but could be seen when the sensitivity of the instrument was increased. Thisshould be compared with nickel where the mass spectrum under these conditionsindicated the presence of the species Ni,CO+ as well as NiCO+.temperature /KFIG. 5.-Variation of intensity of peak due to CuCOf as the temperature is raised from 77 to 295 K.1 1 1 1 1 1 1 1 1 1 1 1 ~190 170 150 1 3 0 . 110mleFIG. 6.-SIMS spectrum of copper foil after saturation at room temperature.When the surface had been saturated at 77 K the temperature was raised to 295 Kwhile the signal due to CuCO+ was continuously monitored.The intensity of thissignal is plotted against temperature in fig. 5. Note that there is a discontinuity onthis curve which occurs at approximately 195 K and that the CuCO+ value drops tozero at room temperature.Upon the admission of 1000 L CO at room temperature the intensities of the Cu46 ADSORPTION OF co ON Ni AND CUpeaks increased markedly and new peaks appeared at 154, 156 and 158 correspondingto Cu2CO+, see fig. 6, and saturation occurred at about 5000 L.Ton causedthe appearance of various carbide species, fig. 7. These were assigned to CuC;,CuCf , Cu2C+, Cu,Ct, Cu,Cl and Cu,C,+. There were also very small peaks whichcould be assigned to Cu,CO,+, suggesting, perhaps, that the carbon is formed bydisproportionation of the carbon monoxide followed by the desorption of carbondioxide.Allowing carbon monoxide to flow over the sample at 390 K andCU'U U ~ 2ao 210 wo 170 ' i s 0 130 110 9 0 7 0mleFIG.7.-Appearance of carbide species in SIMS spectrum after flowing CO over the surface at7.DISCUSSIONNICKELIn the temperature range 77-295 K two carbon monoxide-containing secondaryions were observed in the spectrum. One is due to carbon monoxide bonded to onenickel atom and the other to carbon monoxide bonded to two nickel atoms. Sincethe ratio of the intensities of the peaks due to these two forms is roughly constantin this temperature range, it might be argued that the singly bonded carbon monoxideis merely a fragment of the larger species, formed by the bombardment by the argonions.If this were the case then it would be expected that this ratio would remainconstant at all temperatures. This was not so. The NiCO+ peak had completelydisappeared from the spectrum at about 370K while that of Ni2CO+ remained.Thus, we assume that these two species are due to two distinct surface species andthe one does not derive from the other.We now suggest possible surface structures from which these secondary ions arederived. Obviously, it is not possible to give precise structures from a SIMS study.Since there is no evidence to say that adsorbed carbon monoxide is bonded to ametal surface via the oxygen atom, it seems likely that NiCO+ is formed from aprocess schematically illustrated in fig.8(a)M. BARBER, J . C. VICKERMAN A N D J . WOLSTENHOLME 47One can envisage a number of processes leading to the formation of Ni,CO+;these are illustrated in fig. 8(b)-(d).When considering a possible surface structure which would give rise to Ni2CO+it must be remembered that it must be different, in some way, from that which givesrise to NiCOf.The surface structure illustrated in fig. 8(b) appears to be identical with that in (a).It is, perhaps, possible that there are two types of site on the nickel surfaces and onboth of these sites carbon monoxide is bonded in a linear manner. From one ofthese sites sputtering gives rise to only NiCO+ and from the other only Ni2CO+ butthis seems unlikely.Secondary Ion(a 1 I I 1 I l l I 1p?J N N iFIG. 8.-Possible surface processes which would give rise to the appearance of NiCO+ and Ni2CO+.Process (c) suggests that the carbon monoxide is dissociatively adsorbed.It wasfound that at temperatures where carbon monoxide is known to adsorb dissocia-tively,6 we obtained peaks corresponding to Ni2C+ and Ni,O+. It is reasonable toassume, therefore, that if carbon monoxide were dissociatively adsorbed at roomtemperature then we would obtain peaks corresponding to the same species. Thusit is unlikely that process (c) gives rise to Ni2CO+.It is logical, therefore, to ascribe the appearance of Ni2CO+ to a structure of thetype illustrated in fig. 8(4. The idea of a “bridge” bonded carbon monoxidemolecule is supported by the evidence of i.r.spectro~copy,~ as mentioned in theintroduction. However, the correlation is not complete since the initial adsorptionof carbon monoxide is i.r. inactive whereas we observe “ linear ” and “ bridged ”forms even at low exposures.The decrease in intensity of the peaks due to these two species as the temperatureis raised from 77 to 295 K may be due either to a surface reaction in which carbonmonoxide is lost or to desorption.If a surface reaction were taking place then it is fair to assume that it would beone of the following :orIn either case one would expect to see extra peaks occurring in the SIMS spectrum,these were not observed. Thus, it seems likely that the reduction in the amount ofadsorbed carbon monoxide is due to desorption.toads -3- Cads+Oads * ca~ls+$~2(g)2coads -+ Cads + C02ads Cads + c02(g)48 ADSORPTION OF co ON Ni AND CUOn raising the temperature further we found that none of the linear Ni-C-0 isleft on the surface at 370 K.Although at this stage, there is a buildup of carbon andoxygen on the surface and so some dissociation must be taking place, this is small.Unfortunately, no infrared experiments have been carried out at this temperature.It would be interesting to know whether the band assigned to the linear structuredisappears at this temperature.Our observation that, on heating the adsorbed carbon monoxide at 390 K in thepresence of carbon monoxide, a surface carbide is formed, is in agreement withthe results of work by Joyner and Roberts.6 Our results would suggest alsothat the mechanism of this carbide formation is one of dissociation rather thandisproportionation since, as well as Ni2C+, we observe peaks due to Ni20+ but therewere no peaks which could be assigned to a structure containing carbon dioxide.It is possible to make a rough estimate of the relative coverages of the linear andbridged forms of adsorbed carbon monoxide but it is first necessary to make certainassumptions about sputtering coefficients.The sputtering coefficient is defined asthe number, N,, of secondary ions produced for each primary ion falling on the target,where No is the flux of primary ions.see ref. (11) and (15).when carbon monoxide is adsorbed the variation is assumed to be such thatKi = Ni/NOFor a more detailed account of sputtering coefficients and their determinationAlthough the ratio of sputtering coefficients for Ni; and Nil species may varyThus, the relative coverages of NiCO and NizCO can be estimated from theequationWhere Ii+ is the total secondary ion current for species i, summing over all oftheisotopes for a particular species.Using this approximation we find that about 30 %of the adsorbed carbon monoxide is in the bridged form and 70 % in the linear form.It must be emphasised that, as a result of the above assumption this is a fairly crudeapproximation.COPPERIn contrast with the results obtained from the adsorption of carbon monoxide onnickel at 77 K, adsorption on copper at this temperature produces only the CuCO+species and no Cu,CO+.Using a similar argument to that used for NiCO+ it isbelieved that this is derived from a surface structure analogous to that in fig. 8(a).This is in direct agreement with the results obtained from i.r. studies where thesingle, narrow band is assigned to a linearly bound carbon monoxide structure ratherthan to a bridged structure.Despite the fact that carbon monoxide is only bound in a linear form at 77 K,it is evident from our desorption curve, fig. 5, that desorption occurs from two distinctsurface species. The discontinuity on this curve at about 195 K indicates the onsetof the desorption of a second type of adsorbed species at this temperature. Thesecond species is completely desorbed at room temperature, since the signal diminishesto zero at this temperatureM.BARBER, J . C. VICKERMAN AND J . WOLSTENHOLME 49If adsorption is studied at77 K the i.r. band and the surface potential increase in proportion with coverage upto 8 = +.* This indicates that there is only one type of adsorption up to this coverage.As the coverage is increased LEED investigations suggest a compression of theadsorbed layer; the intensity of the i.r. band remains constant but the band shiftsto higher frequencies and the surface potential decreases. This indicates that asecond type of adsorption is occurring at 8 > 0.5. If, flow, the metal is warmedfrom 77 K, desorption occur's and these processes are reversed, the species formedsecond is desorbed first, complete desorption occurring at 195 K, the temperatureat which we observe the discontinuity on our desorption curve.At room temperaturethe i.r, band disappears and the surface potential returns to zero indicating completedesorption. It will be seed from fig. 8 that we, too, observed complete desorptionat room temperature.Our adsorption studies at room temperature reveal the secondary ion speciesCuzCO+ but there is no evidence for the presence of CuCO+. Arguing as before, itis probable that this is derived from a surface structure analogous to that shown infig. 8(d).Saturation of the surface with this species only occurs after a total of 5000 L.The incorporation of carbon into the metal lattice can account for the appearanceof carbon-containing secondary ions in out spectra at 390 K.The formation of thevarious secondary ions can be illustrated by the schematic representation of thesurface, fig. 9. The dotted lines in this diagram encircle groups on the surface whichThis too, is in agreement with previous studies.'*FIG. 9.-Suggested surface structure which would give rise to various carbide slructures observed inthe spectrum. 0, copper ; 0, carbon.could give rise to the secondary ions Cu,C,+ and Cu,C,+, similar boxes would showhow the other species could arise, The limitation of this approach is that it does notexplain the absence in the spectrum of species such as CuC+ and CuCl.The formation of carbon on the surface is due either to decomposition or dis-proportionation of the carbon monoxide.Disproportionation is favoured here sincethere was a weak signal in some of the spectra which could be assigned to Cu,CO;.The uncertainties both in the way the carbon arises and the structure it subsequentlyadopts indicate that it is necessary to study this system further.ASSESSMENT OF TECHNIQUEIn cases where comparable work has been done using other techniques, the resultsobtained from SIMS are consistent. Furthermore, this method will give resultsdirectly and often more conveniently. It is also capable of providing informationfrom a wide variety of samples (e.g., metal foils, thin films, single crystals, supportedmetals as well as insulators) over a wide temperature range and with good resolution.Its value for elucidating surface cleanliness has already been noted.An important point with any method of analysis is the extent to which the methodof study affects the system being analysed.As may be seen from our spectra the ion* Ref. (7) puts this at 8 = 450 ADSORPTION OF CO ON Ni AND Cubombardment does not seem to affect the surface process. For example, it is fairlyobvious that induced fragmentation does not occur since the species which might beregarded as fragments either have different thermal properties from the ‘‘ parent ”species (as in the caSe of NizCO+ and its possible fragments NiCO+, Ni,C+ or Ni20+)or do not appear in the spectrum (e.g., NiC+ or NiO+). Thus we can say that thespecies we observe faithfully reflect the situation on the surface of the sample.Furthermore, the process of bombardment does not affect the surface in such away as to alter its properties. This we may deduce both from the consistency of ourresults with those from other techniques and from the reproducibility of our resultswith the same metal sample.A further point to note is that the sensitivity of SIMS is such that we are ableto observe species which have not been reported previously in the literature (e.g.,a bridged structure for adsorbed carbon monoxide on copper).We are grateful to Vacuum Generators Ltd. for the generous loan of the equipmenton which this work was carried out and also to the S.R.C. for a Research Studentshipawarded to J. W.G. Wedler, H. Papp and G. Schroll, Surface Sci., 1974,44,463.H. H. Madden, J. Kuppers and G. Ertl, J. Chern. Phys., 1973,58,3401.J . C. Tracy, J. Chem. Phys., 1972,56,2736.A. M. Bradshaw and J. Pritchard, Surface Sci., 1969, 17, 372.T. E. Edmonds and R. C. Pitkethly, Surface Sci., 1969,15, 137.R. W. Joyner and M. W. Roberts, J.C.S. Farudzy I, 1974,70,1819.M. A. Chesters and J. Pritchard, Surface Sci., 1971, 28,460.J. C. Tracy, J. Chem. Phys., 1972,56,2748.R. W. Joyner, C. S. McKee and M. W. Roberts, Surface Sci., 1971,26,303.A. Bennhghoven, Surface Sci., 1973, 35,427.’ M. A. Chesters, J. Pritchard and M. L. Sims, Chem. Cornm., 1970, 1454.I2 M. Barber, P. K. Sharpe and J. C. Vickerman, to be published.I3 I. A. Abroyan and U. P. Laurov, Soviet Phys., Solid State, 1962,4,2382.l4 P. J. Page, D. L. Trim and P. M. Williams, J.C.S. Faraday I, 1974,70, 1769.l5 Ya. M. Fogel, Soviet Phys. Uspekhi, 1967,10,17