Molecular Beam Investigations of Surface Chemical Reactions and Dynamics Citation Mullins, Charles Buddie (1990) Molecular Beam Investigations of Surface Chemical Reactions and Dynamics. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/64s4-0182. https://resolver.caltech.edu/CaltechETD:etd-03122007-135158 Abstract Experimental results from molecular beam investigations of trapping and dissociative chemisorption phenomena for several gas-surface systems are presented. The dissociative chemisorption of oxygen on Ir(110)-(1x2) in the limit of zero coverage S o was studied as a function of incident kinetic energy E i , incident angle θ i and surface temperature T s . Results from this investigation indicate that two mechanisms account for the initial chemisorption. At low incident kinetic energy (less than 4 kcal/mol) chemisorption mediated by trapping is primarily responsible for the dissociative adsorption while at high energies a direct mechanism can account for the results. In both energy ranges the initial dissociative chemisorption probability is insensitive to incident angle. The trapping of molecular ethane as well as the dissociative chemisorption of ethane on the clean Ir(110)-(1x2) surface has also been investigated. The initial trapping probability ζ o is found to decrease with incident kinetic energy from a value of ~0.98 at 1 kcal/mol to ~0.1 at 16 kcal/mol. These data scale with E i cos 0.5 θ i . The initial dissociative chemisorption of ethane on Ir(110)-(1x2) occurs via a trapping-mediated mechanism at low E i and a direct mechanism at high kinetic energies. In the trapping-mediated regime S o decreases rapidly with increasing T s . These data quantitatively support a kinetic model consistent with a trapping-mediated chemisorption mechanism. The difference in the activation energies for desorption and chemisorption from the physically adsorbed, trapped state E d -E c is 2.2±0.2 kcal/mol. Chemisorption at high kinetic energies, in the direct regime, is independent of surface temperature. Additionally, the trapping probability of Ar on Pt(111) has been measured as a function of incident kinetic energy, and angle for T s = 80, 190 and 273 K. The trapping probability decreases with increasing E i in a manner that depends on both θ i and T s . The angular scaling law governing the trapping is a function of T s such that ζ o scales with E i cos 1.5 θ i at 80 K, E i cos 1.0 θ i at 190 K and E i cos 0.5 θ i at 273 K. These results suggest that parallel momentum dissipation becomes increasingly more important to the trapping dynamics as the surface temperature is increased. Item Type: Thesis (Dissertation (Ph.D.)) Subject Keywords: Molecular beams, surface chemistry Degree Grantor: California Institute of Technology Division: Chemistry and Chemical Engineering Major Option: Chemical Engineering Thesis Availability: Public (worldwide access) Research Advisor(s): Weinberg, William Henry Thesis Committee: Weinberg, William Henry (chair) McKoy, Basil Vincent Beauchamp, Jesse L. Gavalas, George R. Defense Date: 29 September 1989 Non-Caltech Author Email: mullins (AT) che.utexas.edu Funders: Funding Agency Grant Number IBM UNSPECIFIED Link Foundation UNSPECIFIED Grace Chemical Company UNSPECIFIED Caltech UNSPECIFIED NSF UNSPECIFIED Record Number: CaltechETD:etd-03122007-135158 Persistent URL: https://resolver.caltech.edu/CaltechETD:etd-03122007-135158 DOI: 10.7907/64s4-0182 Related URLs: URL URL Type Description https://doi.org/10.1116/1.575985 DOI Article adapted for Chapter III. https://doi.org/10.1063/1.457808 DOI Article adapted for Chapter IV. https://doi.org/10.1063/1.457762 DOI Article adapted for Chapter V. https://doi.org/10.1016/0009-2614(89)80020-x DOI Article adapted for Chapter VI. https://doi.org/10.1063/1.458151 DOI Article adapted for Appendix A. https://doi.org/10.1063/1.455838 DOI Article adapted for Appendix B. https://doi.org/10.1063/1.454597 DOI Article adapted for Appendix C. https://doi.org/10.1021/ja00273a002 DOI Article adapted for Appendix D. ORCID: Author ORCID Mullins, Charles Buddie 0000-0003-1030-4801 Default Usage Policy: No commercial reproduction, distribution, display or performance rights in this work are provided. ID Code: 938 Collection: CaltechTHESIS Deposited By: Imported from ETD-db Deposited On: 20 Mar 2007 Last Modified: 16 Apr 2020 18:00 Thesis Files Preview PDF - Final Version See Usage Policy. 12MB Repository Staff Only: item control page

MOLECULAR BEAM iNVESTiGATtONS OF SURFACE CHEMtCAL REACT)ONS AND DYNAMtCS Thesis by Charιes Buddie Muιiins tn Partiaι Futfi)ιment of the Requirements for the Degree of Doctor of Phi)osophy Catifornia institute of Technoiogy Pasadena, Catifornia 1990 (Submitted September 29, 1989) i i ACKNOWLEDGEMENTS ι am sincere!y gratefuι to the organizations which heιped fund my graduate education. )BM generousty provided me with a feιiowship for two years. The Link Foundation awarded me a feιιowship for one year and the Grace Chemicat Company provided funds to support me during one summer. The Caιtech Chemicat Engineering Department arranged a feitowship for me during my initia! year. The Natio∩a) Science Foundation has provided the butk of the funding for the research reported in this thesis. During my tenure at Cattech ) have had the great fortune of making many friends, some of which hetped in this work a great deat. t hope at! of you know how gratefui t am for your assistance and most of at! for your friendship. My advisor, Henry Weinberg, has attowed me co∩siderabte freedom throughout the course of my research coupted with very strong support, even when he disagreed with the approach t pursued. Henry was especiatty supportive during the "tea∩ years" when t had considerabte trouble making the apparatus operabte. We have become very good friends over the past six years and it is this aspect of our retationship that t appreciate the most. very much Henry! Thanks so tII Without the support of the excellent staff at Ca!tech this work wouid have been impossibte. i owe speciai thanks to Kathy, Floyd, Chic, John and Heien in Chemicai Engineering and Tony, Guy, Ray and Tom of the Division Eiectricai and Mechanicai Shops. The Weinberg group has had severai members that have devoted significant amounts of effort to this work and t thank aii of them, especiaiiy Maiina Hiiis, Jenna Zinck, Jim Engstrom, Brad Anton, Eric Hood, Steve George, Chuan Kang and Randy Verhoef. John Parmeter, Yong-Kui Sun and Udo Schwaike have been good friends as weit as research coιieagues and ι appreciate a!) their heip and friendship over the years. Wang. ! owe a particulariy iarge debt of gratitude to Youqi Youqi has been an inspiration with his boundiess energy, enthusiasm and integrity, it is truiy a p!easure and honor to be his friend. The Caltech environment is such that it is virtua)iy impossibie to study and work without also making friends outside of your research area. Howard Stone has been a close friend and sounding board for several years. interact with Howard. It has always been fun and educational to Greg Voth has shared significant amounts of his time, energy and advice with me over the past 5 years, in addition to being my regular tennis practice and doubles partner Greg has been a friendly and enthusiastic tutor regarding theoretical chemistry and other esoteric topics, i miss the "good oid days" when i could talk with Greg and Howard virtually everyday. iV ) spent 5 months at the )BM Aιmaden Research Center in San Jose in 1988 working with Chariie Rettner on a project inctuded in this thesis. This experience was everything ι had hoped it woutd be and more thanks to Chartie's support and enthusiasm. Chariie is one of a kind and i am extremeiy gratefuι to have had the chance to work with him and to be his friend. During my stay at Aιmaden t found many heipfui and friendiy peopie and i thank them at). However, Joe Schiaegei, Erhard Schweizer and Dan Auerbach deserve speciai thanks for at) their efforts. Finaiiy, words cannot express the gratitude and overwetming feeiing of warmth and cioseness i have toward my famiiy for their undying support over the past six years. maintain my sanity and enthusiasm My mom and dad he!ped me during those previousiy mentioned "iean years" and shared in my excitement when "good things" started to happen. Thanks mom and dad for aii your iove. wife Meiissa shouid be given the Nobei Prize in patience. My She has put up with more baιoney during this period than any one person deserves in a ιifetime. ) wiιι try very hard to make it up to her. gives my )ife meaning and purpose and t )ove her very much. She V ABSTRACT Expérimenta) resutts from mo)ecu)ar beam investigations of trapping and dissociative chemisorption phenomena for severa) gas- surface systems are presented. The dissociative chemisorption of oxygen on ιr(110)-(1x2) in the )imit of zero coverage So was studied as a function of incident kinetic energy Ej, incident ang)e Θ; and surface temperature Tg. Resuιts from this investigation indicate that two mechanisms account for the initiai chemisorption. incident kinetic energy ()ess than 4 kca)∕moι) At tow chemisorption mediated by trapping is primariιy responsibιe for the dissociative adsorption white at high energies a direct mechanism can account for the resutts. t∩ both energy ranges the initiât dissociative chemisorption probabitity is insensitive to incident angte. The trapping of motecutar ethane as wet) as the dissociative chemisorption of ethane on the ctean tr(110)-(1x2) surface has atso been investigated. The initia) trapping probabitity ζo is found to decrease with incident kinetic energy from a vatue of -0.98 at 1 kca)∕mo) to -0.1 at 16 kcat/mot. These data sca)e with Ejcos° 5θj. The initia) dissociative chemisorption of ethane on )r(110)-(1x2) occurs via a trapping-mediated mechanism at tow Ej and a direct mechanism at high kinetic energies. regime So decreases rapidty with )n the trapping-mediated increasing Tg. These data Vi qua∩titativeιy support a kinetic mode) consistent with a trapping- mechanism. mediated chemisorption activation energies for desorption physicatty adsorbed, trapped state The difference in the and chemisorption from the Ed-Ec is 2.2+0.2 kcaι∕moι. Chemisorption at high kinetic energies, in the direct regime, is independent of surface temperature. Additional, the trapping probabi)ity of Ar on Pt(111) has been measured as a function of incident kinetic energy, and a∩gιe for Ts=80, 190 and 273 K. The trapping probabitity decreases with increasing Ej in a manner that depends on both θj and Tg. The angutar seating taw governing the trapping is a function of Tg such that ζo scates with Ejcosi $θ{ at 80 K, Ejcosi Oθj at 190 K and E{cos°5θj at 273 K. These results suggest that parattet momentum dissipation becomes increasingty more important to the trapping dynamics as the surface temperature is increased. Vii TABLE OF CONTENTS ACKNOWLEDGEMENTS ii ABSTRACT V L )NTRODUCT!ON )-1 ) !. DYNAMιCS OF THE DιSSOC)AT)VE CHEMtSORPT)ON OF OXYGEN ON AN tr(11O)-(1x2) SURFACE ! 11. A MOLECULAR BEAM STUDY OF THE DιSSOC!ATtVE CHEMιSORPT)ON OF Op ON ιr(110)-(1x2) ί V. ÎV-1 TRAPP<NG-MED)ATED DιSSOCιAT)VE CHEMιSORPT)ON OF ETHANE ON tr(11O)-(1x2) V). ίίί-1 TRAPP)NG OF MOLECULAR ETHANE ON THE ιr(110)-(1x2) SURFACE V. ι)-1 V-1 VAR<AT)ON OF THE TRAPP)NG PROBABtUTY OF Ar ON Pt(111 ) W)TH K)NET)C ENERGY AND ANGLE OF ÎNCÎDENCE: THE CHANGÎNG ROLE OF PARALLEL MOMENTUM WiTH SURFACE TEMPERATURE Vι-1 V)ι) A. MOLECULAR ADSORPTtON OF ETHANE ON THE )r(110)-(1x2) SURFACE: MONTE-CARLO StMULATιONS AND MOLECULAR BEAM REFLECTιV)TY MEASUREMENTS B. A-1 DESORPTιON AND TRAPPιNG OF ARGON AT A 2H-W(100) SURFACE AND A TEST OF THE APPL)CAB!UTY OF DETA!LED BALANCE TO A NONEQUιUBRιUM SYSTEM C. D. B-1 ELECTRON ENERGY LOSS SPECTROSCOPY OF AMMONιA ON Ru(001) C-1 )NTERACT)ON OF ETHYLENE W)TH THE Ru(001 ) SURFACE D-1 ί-1 CHAPTER ί iNTRODUCTiON The study of gas-surface dynamics and reactions is important At a basic ievet, the for both fundamentai and practicaι reasons. chemicat and physicai processes that occur when a partic)e coiiides with a surface are oniy pooriy understood. For exampte, the energy transfer mechanisms that account for the physicai adsorption of a rare gas atom at a soiid surface are just beginning to be recognized. The dynamics and energetics of the dissociative chemisorption of gas phase mo!ecu)es at soiid surfaces are even more compiicated. Thus, this type of chaiienging probiems. phenomena provides many inteiiectuaiiy The resuits of such studies are frequent)y usefui in understanding and enhancing industriai processes aiso. For exampie, quantifying and understanding dissociative chemisorption is usefui in improving the performance of heterogeneousiy cata)yzed commerciai reactors. Fundamentai studies over the past 30 years have carried surface science and in particular the dynamics of gas-surface interactions to a new ievei of maturity^. increasingiy more detaiied questions are being asked and answered. The past five years have witnessed a virtual explosion in activity with regard to studies of the dynamics and reactions of gas-surface systems by moiecular beam techniques^Much of the recent progress in surface science ί-2 is due to these studies. Thus, this thesis reports findings from severat gas-surface investigations of dynamics empioying the mo)ecuιar beam technique. Chapters ti and ίίί contain the resutts of research concerning the chemisorption dynamics of the interaction of oxygen with the )r(110)-(1×2) surface. resutts of the dissociative Expérimenta) adsorption as a function of incident kinetic energy, incident ang)e Chapter H a)so contains a and surface temperature are reported. detai)ed description of the apparatus used in obtaining the expérimenta) data reported in Chapters ))-V and Appendix A. Chapters )V and V and Appendix A report resutts from an investigation of the interaction of ethane with the )r(110)-(1×2) )n particu)ar, Chapter )V is an account of the initia) surface. probabi)ity of physica) adsorption on the surface at 77 K. The incident kinetic energy and ang)e dependence of the trapping probabi)ity are reported and discussed. resu)ts of experiments concerning Chapter V describes the the initia) dissociative chemisorption of ethane as a function of incident kinetic energy, incident ang)e and surface temperature. Appendix A reports findings from a comparison of Monte-Car)o simu)ations and mo)ecu)ar beam ref)ectivity measurements of the physica) adsorption probabi)ity of ethane on )r(110)-(1x2) as a function of time. Chapter V) and Appendix B contain descriptions of resutts from experiments probing the interaction of argon at re)ative)y )ow i-3 kinetic energy with the Pt(111) surface and the 2H-W(100) surface. These studies were motivated by a desire to better understand the trapping step in trapping-mediated chemisorption. Chapter Vi reports studies of the trapping probabi)ity as a function of incident energy and angie as does Appendix B. in addition, Chapter Vi reports the surface temperature dependence of the trapping probabiiity. Appendices C and D dispiay and discuss expérimentai resuits from eiectron energy ioss measurements of ammonia and ethyiene on Ru(001). t-4 REFERENCES 1. A. ZangwiH, Pbys/cs at Surfaces, Cambridge University Press, Cambridge 1988. 2. D.J. Auerbach, H.E. Pf∩ur, C.T. Rett∩er, J.E. Sch)aegei, J. Lee and R.J. Madix, J. Chem. Phys. 81, 2515 (1984); H.E. Pfnur, C.T. Rettner, D.J. Auerbach, R.J. Madix and J. Lee, ∕b∕d. 85, 7452 (1986). 3. C.T. Rettner, H.E. Pfnur and D.J. Auerbach, Phys. Rev. Lett. 54, 2716 (1985). 4. S.T. Ceyer, Ann. Rev. Phys. Chem. 39, 479 (1988); and references therein. 5. M.B. Lee, Q.Y. Yang and S.T. Ceyer, J. Chem. Phys. 87, 2724 (1987). 6. M.P. D'Evetyn, A.V. Hamza, G.E. Gdowski and R.J. Madix, Surf. Sei. 167, 451 (1986). 7. A.C. Luntz and D.S. Bethune, J. Chem. Phys. 90, 1274 (1989). 8. C.T. Rettner and H. Stein, Phys. Rev. Lett. 59, 2768 (1987); J. Chem. Phys. 87, 770 (1987). 9. G.R. Schoofs, C.R. Arumainayagam, M.C. McMaster and R.J. Madix, Surf. Set. 215, 1 (1989). 10. C.B. Muι)i∩s and W.H. Weinberg, J. Chem. Phys, (submitted). 11. A.V. Hamza, H.-P. Steinruck and R.J. Madix, J. Chem. Phys. 85, 7494 (1986). 12. A.C. Luntz, M.D. Williams and D.S. Bethune, J. Chem. Phys. 89, 4381 (1988). 13. C.T. Rettner, L.A. DeLouise and D.J. Auerbach, J. Vac. Sei. Techno!. A 4, 1491 (1986); J. Chem. Phys. 85, 1131 (1986). ί-5 14. C.T. Rettner, Η. Stein and E.K. Schweizer, J. Chem. Phys. 89, 3337 (1988). t ί-1 CHAPTER ίί DYNAMιCS OF THE DιSSOC)AT)VE CHEM)SORPTιON OF OXYGEN ON AN ιr(110)-(1x2) SURFACE [The text of Chapter H consists of an articιe coauthored with W.H Weinberg, to be submitted.] ίί-2 ABSTRACT The dissociative chemisorption of oxygen on )r(110)-(1x2) has been investigated empιoying moιecu)ar beam techniques. The probabi!ity of dissociative chemisorption as a function of incident kinetic energy between 1 and 23 kcat/mo), incident angιe between 0 and 45°, and surface temperature between 150 and 600 K is reported in the !imit of zero-coverage. approximate^ 4 kcaι∕moι the chemisorption are mechanism. data For kinetic energies ιess than initiai expiained probabiiity of dissociative via a trapping-mediated For kinetic energies greater than 4 kca!∕moi a direct mechanism is primariιy responsibie for dissociative chemisorption. The data are insensitive to surface temperature except at greater than 500 K in the trapping-mediated chemisorption regime where the temperature effect is manifested in a decrease in the trapping probabiiity. The height of the activation barriers for desorption and chemisorption from the trapped state are equivaient to within 100 cai/moi. Both the direct and trapping-mediated chemisorption regimes are very insensitive to incident angie scaiing with totai energy. tt-3 ). )NTRODUCT)ON Quantifying and understanding the dynamics and energetics of the dissociative chemisorption of a gas-phase diatomic motecuιe on a transition meta) surface is a prob)em attracting a great dea) of interest currency, Comparative)y )ndeed, the prob)em is muιtifaceted and comp)ex. titt)e is known, for examp)e, regarding the mechanism with which the impinging mo)ecu!e transfers its kinetic energy to substrate and interna) mo)ecu)ar excitations. The surface site at which dissociation occurs for a given incident kinetic energy remains a pertinent question. The rote d e)ectrons p)ay in the dissociation mechanism is on)y beginning to be understood. There are far more questions than answers concerning this c)ass of chemica) reactions. tt is c)ear this prob)em is important for fundamenta) reasons but answers to the above questions are a)so of practica) significance. Heterogeneous)y cata)yzed chemica) reactions are an important part of industria) chemica) processing and the dissociation of a diatomic at a surface is among the simp)est of the profusion of reactions occurring in an industria) reactor. Obvious)y the dissociation of a diatomic mo)ecu)e at a surface wit) have to be understood before significant progress on other more complex reactions can be made. Perhaps more re)evant to this discussion, the dissociative chemisorption of oxygen is an important step in the epoxidation of ethy)ene and the oxidation of carbon monoxide. ίί-4 Recentιy severa! groups^ dissociative chemisorption have investigated the dynamics of uti!izing mo)ecutar beam techniques. These studies have great)y enhanced our understanding of the gross features of the mo!ecu!e-surface interaction potential· Thus we emp)oy a motecu)ar beam apparatus in this investigation of the dissociative chemisorption of oxygen on )r(110)-(1×2) to probe and determine the genera! features of the potentia! hypersurface for this system. The resu!ts from a targe body of work inctuding the previous!y mentioned mo!ecu!ar beam investigations indicate that there are two basic mechanisms that account for most dissociative surface reactions in the !imit of zero surface coverage. There is a direct mechanism whereby the mo!ecu!e dissociates upon impact with the surface at re)ative)y high incident kinetic energy and with )itt!e or no surface temperature dependence. There is a!so a trapping- mediated mechanism in which the mo!ecu!e transfers sufficient trans!ationa! energy to substrate and interna! excitations during the co!!ision to trap in a mo!ecu!ar!y adsorbed state. The probabi!ity of trapping at the surface increases with decreasing incident kinetic energy. Once it is trapped the mo!ecu!e equi!ibrates with the surface and either desorbs or reorientates such that dissociation can occur. Thus, trapping-mediated chemisorption typica)!y strong functions of temperature. rates are However, this is not a!ways true and the system discussed in this paper (θ2∕ιr(110)(1x2)) is one of the exceptions. !ndeed, increasing surface temperature Tg can serve to both decrease and increase reaction ίί-5 rates or have no effect at aifs. The chemisorption rate decreases with increasing Tg if the activation barrier to dissociation is iess than the barrier to desorb from the moιecuiar!y adsorbed (trapped) On the other hand, the chemisorption rate increases with state. Tg if the reverse is true. increasing characteristics are reiativeiy faciιe Systems with the former so that chemisorption probabi)ities can be quite high whiie the tatter characteristics are re)evant to activated chemisorption systems since the barrier to dissociation is higher than the vacuum zero energy tevet. tt is our betief that both cha∩nets exist for most motecute/surface systems. However the probabitity of dissociative chemisorption can be quite tow in the activated systems. Obviousty if the probabitity of dissociation in a trapping-mediated chemisorption mechanism is surface temperature independent the barrier to desorb is the same height as the barrier to dissociate. Some of the eartier motecutar beam studies of chemisorption dynamics focussed on the direct mechanism of dissociation for activated energy systems^. can be very These studies reveated that trans)ationai effective in surmounting the barrier to dissociative chemisorption and that this barrier is often very nearty one-dimensionaL However, this work genera))y has not reveaιed any detaiis regarding the trapping-mediated mechanism for activated systems. That this channeι exists for these activated systems has been demonstrated with experiments at reiativeiy high pressure in batch reactors'* s,ι 7 where the avaiiabie kinetic energies are iimited to iow vaiues according to a Ma×weii-Boitzma∩n distribution. Thus H-6 indicating the importance of channe) this in industriai heterogeneous!y cataιyzed chemicat reactions. Some of the more recent accounts of motecu)ar beam studies of surface chemica! reaction dynamics have described re)ative)y faciie chemisorption systems which have been particu)ariy usefuι in uncovering the mechanismio*i4. re)evant ]∩ dynamics in a trapping-mediated study of N2 on W(100) Rettner, et ai/4 convincingty demonstrated the existence of a trapping-mediated channei at tow incident kinetic energies, same system Rettner, tn a separate paper on the et at J 8 atso showed that the trapping probabitity was onty weakty inftue∩ced by surface temperature, a criticat assumption in severat analyses prior to these measurements. Rettner and coworkers have atso investigated the initiât and coverage-dependent dissociative chemisorption of oxygen on W(110). This study has not supported assignment of a trapping-mediated mechanism for dissociative chemisorption for Tg between 280 and 800 K. tn fact, the initiât probabitity of dissociative chemisorption So increases rapidty with increasing E⅛ from 2.3 to 11.5 kcat/mot which fits better with a direct dissociation mechanism. Onty weak indirect evidence for trapping-mediated dissociative chemisorption was found for incident kinetic energies of tess than 1.1 kcat/mot. However a study of O2 chemisorption on W(110) using an effusive beam source by Wang and Gomeri9 indicates that a trapping- ίί-7 mediated chemisorption channet may be significant for surface temperatures tess than 200 K. Luntz and coworkers have investigated the dissociative chemisorption of oxygen on Pt(111) recentiy using moιecutar beam This study unambiguousty shows both trapping- techniques^. mediated and direct channeis to dissociative chemisorption. The high incident kinetic energy direct channet for this system atso showed a weak but suprising temperature dependence, !n a separate paper Luntz, et ah have atso demonstrated the existence of a physicatty adsorbed state of C⅛ on Pt(111) which is betieved to be the precursor to the motecutar chemisorbed state^o. Of particutar interest with regard to the present study is the work of Taytor, et at.21 on the dissociative chemisorption of O2 on tndeed this work provided some motivation for this 1r(110)-(1x2). study. tn their investigation Taytor, et at. emptoyed thermat desorption mass spectrometry, tow-energy etectro∩ diffraction and work function measurements to determine initiât and coverage- dependent chemisorption probabitities. Their resutts indicate that the initiât probabitity of chemisorption is independent of surface temperature. They atso conctuded for Tg between 300 and 700 K that the adsorption kinetics coutd be described by a second-order precursor model. This paper is a presentation and discussion of expérimentât data concerning the interaction of C⅛ with the tr(110)-(1x2) surface H-8 empιoying a mo!ecu!ar beam apparatus, ιn particutar, measurements of the initia! probabi)ity of dissociative chemisorption as a function of incident kinetic energy, incident ang!e and surface temperature are presented and discussed in tight of the mechanisms described above. We find evidence at !ow kinetic energies for a trapping- mediated chemisorption channe! and a direct channe! at high Ej. We a!so find that the initia! probabi!ity of chemisorption is re!ative!y insensitive to incident ang!e and surface temperature. A detai!ed description of the apparatus and procedures used to obtain the data is presented next. A brief, pre!iminary account of this work has appeared previous!y22. H. EXPER!MENTAL The expérimenta! apparatus is shown schematica!!y in an overhead view in Fig. 1 and consists of an u!tra-high vacuum (UHV) scattering chamber and a thrice differential pumped mo!ecu!ar beam source. The apparatus is particu!ar!y we!! suited for studies of the dynamics of adsorption because of its re!ative!y sma!! vo!ume and high pumping speed. These attributes simp!ify and increase the of the ana!ysis probabi!ities of adsorption accuracy is necessary from uptake that to determine measurements. Determination of the probability of adsorption by coverage versus desorption mass spectrometry or Auger ana!ysis can a!so easi!y be performed. Ease exposure measurements either by therma! of operation and maintenance are a!so a consideration in the design of a mo!ecu!ar beam apparatus. The machine has a particu!ar!y H-9 simpιe yet effective samp)e maniputator which aιtows e×tremety rapid cooting via ιiquid nitrogen. The nozz)e, nozzιe heater and nozzιe temperature contro!ιer were a!so very easy to construct and have provided simp)e operation and stabte performance. Experiments are typicatty conducted by directing a supersonic beam of oxygen at an tr singte crystattine sampte and measuring the scattered ftux. Detaits of the expérimentât procedure and the retevant portions of the apparatus are provided betow. The scattering chamber is the targest vacuum chamber in the system and is atso the focus of the apparatus since it houses the crystat sampte and att the spectroscopic instrumentation. The scattering chamber has a votume of -100 titers and is pumped by a turbo-motecutar pump (Batzers TPU2000) with an effective pumping speed of approximate^ 1000 titers/s. A typicat base pressure of -1x10*1° τorr is achieved within a few days of a 36 hour bakeout at 150°C. The third beam-source chamber (C3 in Fig. 1) is atso pumped by a turbo-motecutar pump (Batzers TPU270) and after bakeout has a pressure of -1x10^° Torr. The other beam-source chambers (C1 and C2) are pumped by diffusion pumps and do not need to be baked to attain adequate vacuum. The function and contents of the beam- source chambers is exptained betow. The iridium crystal was cut from a boute that was oriented with a Laue diffractometer and then techniques. potished using standard The surface was cteaned routinety by argon sputtering and annealing the surface in an oxygen background. ion This ίί-1 ο treatment reduced impurity )eve)s on the surface (mainty carbon) to beιow the detection ιevei of our Auger spectrometer (Physicai Etectronics industries, 15-255G). sharp (1×2) The cιea∩ surface exhibited a eιectron ιow-energy diffraction indicative of the reconstructed ιr(110) surface. (LEED) pattern LEED measurements are made using a 4-grid screen (Varian) which can be viewed and photographed through a standard 6 in. Hanged window. The azimuthat orientation of the sampte as determined by LEED is such that an [001] vector tying in the ptane of the sampte surface is approximate^ 15° from the axis of rotation of the maniputator (i.e., the potar angιe). The single crystattine sampte is mounted on a maniputator which can be rotated to vary the angte of incidence of the motecutar beam on the surface. The crystat is mounted on the maniputator by smatt copper btocks that damp on two short Ta wires spot wetdedto the backside of the sampte. The crystat is heated by passing a current through the Ta supports and the sampte itsetf. With tiquid nitrogen flowing through the manipulator the sampte can be cooted from 1600 K to 80 K in less than 90 seconds. temperature is determined by a 0.003 in. The sampte W-5%Re∕W-26%∕Re thermocouple spot-welded to the backside of the crystat. The copper mounting blocks are attached to the copper rods from a miniconflat combination power and thermocouple feedthrough. The upper end of the feedthrough copper rods, tocated inside the stainless steel manipulator shaft, are immersed in tiquid nitrogen when cooling is desired. The liquid nitrogen flows down a tube H-11 inside and coaxiai to the manipu)ator shaft directiy to the attached feedthrough whereupon the fiuid turns 180° and fiows back over the inner tube to exit the manipu)ator shaft. This aspect of the design together with the sma)ι therma) resistance from the copper rod to the crysta) is )argeiy responsibie for the rapid coo)ing of the crystal· The moiecu!ar beam is horizonta) and therefore the surface norma) of the crysta! is a)igned in a horizonta) p)ane during mounting using a HeNe )aser beam. The ang)e of incidence of the mo)ecu)ar beam can be read and repositioned to -0.1° and its abso)ute va)ue has been determined to -1° using a HeNe )aser beam coaxia) with the mo)ecu)ar beam apertures. A supersonic beam of mo)ecu)ar oxygen is generated using a heatab)e nozz)e source with a 100 μm diameter orifice and a tota) of three differentia) pumping stages. The vacuum chamber housing the nozz)e is shown in Fig. 1 and denoted C1. A 0.48 mm skimmer (Beam Dynamics, Mode) 1) is mounted in this chamber and the nozz)e is positioned from 2-8 mm from the knife-edge orifice of this skimmer, depending on the gas, nozz)e temperature and pressure. The second beam chamber, C2, has a 0.98 mm skimmer (Beam Dynamics, Mode) 1) mounted inside, as we)) as a beam stopping f)ag (Newport). The f)ag has a response time of -0.001 seconds and can be used to accurately dose the crysta) with the mo)ecu)ar beam (as short as 0.01 seconds). The flag is also used in making ref)ectivity or uptake measurements, as is discussed )ater. Fina))y, the third beam chamber, C3, has a 2.0mm plate aperture which is the )ast the beam must pass through before entering the scattering chamber. The H-1 2 third beam chamber aιso contains a chopper so that moduιation of the beam is possib)e. The two skimming apertures and the fina) p)ate aperture were a)igned concentricaι)y visuatiy with a teiescope and atso by diffraction of tight from the HeNe taser. an optica) fiber attached which The taser has directs tight coaxiat with the motecutar beam to aid in atigning the crystat in the beam. The nozzte has a window at the end outside the vacuum chamber which can atso be used with the taser for atignment of the nozzte with the apertures. As mentioned above the third beam chamber contains a high speed chopper which attows beam energies to be determined via time-of-ftight measurements using a quadrupote mass spectrometer (E×traNuctear) as a detector. This operation is conducted in putse counting mode using a mutticha∩net scater board (Ortek, ACE-MCS) inside a persona) computer. The chopper is composed of an atumi∩um disc with 2 narrow stits machined into it radiatty 180° apart, near the disc edge. The disc is mounted on and driven by a synchronous motor (Gtobe-TRW, 75A6000). As the disc rotates and the upper stit passes through the motecutar beam, tight from an incandescent butb passes through the tower stit ittuminating a photodiode. This creates a vottage which putse muttichannet scater. is used as a trigger signât for the t∩ practice there is a smatt tag (or tead) time due to misatignme∩t of the diode and tight butb with respect to the motecutar beam. Thus the reat zero time is estabtished by running the chopper both ctockwise and counter-ctockwise and averaging the tt-1 3 resutts of the time-of-ftight measurement to canceι the tead and tag effects. To accuratety determine beam kinetic energies it is atso necessary to measure the ion ftight time to the etectron muttiptier ton ftight times have been catibrated using a fottowing ionization, beam motecutar of SFß and comparing the time-of-ftight measurements of the various mass fragments from the cracking pattern. These data are then fit to an equation of the form tion"(CjonxMion)°5 where Mid is the ion mass and C}o∏ is a constant determined by the fit. ton ftight times are between 5 and 10% of the totat ftight time. Beam kinetic energies between approximate^ 1 and 23 kcat/mot can be generated by co∩trotting the nozzte temperature and seeding in Ar and He. Nozzte heating is accomptished by passing an AC current through a ceramic insutated inconet ctad nichrome wire which is tightty wrapped around the finat 9 inches of tength of the nozzte. Even with this tong heated tength there is a maximum 15% discrepancy between the measured nozzte temperature and time-of- ftight determined gas temperature (for rare gases) for nozzte temperatures above 300 K. The nozzte temperature is measured with a Chromet-Atumet thermocoupte spot wetded to the end of the nozzte and the temperature is controtted and stabitized with the aid of a temperature controtter (Omega). is typical of ∆EFWHM∕Epeak"O.1 5. most The energy spread of the beam superso∩icatty generated beams with H-14 Dissociative chemisorption probabitities are measured primari)y using an uptake method commoniy known as the beam ref)ectivity method of King and Weiis^s. ]∩ this transient measurement the surface acts as a getter and the experimental determined partiai pressure of oxygen is used as a measure of the motecuιes that ref)ect from the surface without adsorbing. )n principie this technique can yieid the vaιue of the chemisorption probabi)ity as measurement. (e.g., a function of surface coverage in a singie For a system with negtigibie vacuum time constant Fig. 2) the initiai probabiiity of chemisorption wouid be determined by a comparison of the initiai oxygen partiai pressure, foiiowing the opening of the high speed shutter (at tι), to the oxygen partiai pressure from the beam scattering from the saturated surface (at t3). in this case the initiai probabiiity of dissociative chemisorption is simpiy P1∕(P1 + P2)- Data is acquired during a refiectivity measurement in the puise counting mode using the same muitichannei quadrupoie mass spectrometer mentioned eariier. scaier and The data can be anaiyzed using the same persona) computer that is used for acquisition. This analysis is reiativeiy straightforward for an ideai system as demonstrated above. However, for a reai system with finite vacuum time constant and increases in the observed partiai pressure which are not totaiiy due to the supersonic beam impinging on the surface, extra measures must be taken. Fortunateiy another complicating factor, unstable pumping speed, has a negiigibie effect ii-15 on our measurements. The high pumping speed of the turbomotecuiar pump (effectiveιy 1000 t∕s) and the smaιi chamber decreases the effect of pumping by chamber waits to an immeasureabιe effect. in cases probabiiity is with reactive near unity such gases where the chemisorption as the present study, oxygen dissociation on )r(110)-(1x2), the initia) rate of change of the surface coverage is too rapid to negiect the vaccum time constant. We have accounted for this compiication with two measures. Firstiy, we diiuteiy seed ai! beams to reduce the f)ux of reactive gases to the surface. Secondiy, we directiy measure the vacuum time constant, ignore the initiai transient and iineariy extrapoiate back to the beginning of the transient using data that is at ieast 5 time constants (< 0.1 s) removed from the time zero. As pointed out by Rett∩er et aι∕3 there is a component in the partiai pressure rise in the scattering chamber im addition to the component due to the supersonic beam impinging on the crystai surface. This additionai nonnegiigibie partia) pressure rise is due to an effusive beam of gas emanating from the finai differential pumped chamber, C3. This gas scatters out of the supersonic beam and eventuaiiy effuses into the main chamber contributing to the oxygen fiu×. To measure this component we note the pressure rise in the third beam chamber with the beam on and the fiag open and then measure the partiai pressure of oxygen increase in the scattering chamber. Then, with the fiag shut a ieak vaive mounted on the third chamber is opened and adjusted to give the same pressure rise in H-1 6 this chamber as with the beam on and f)ag open. The partiat pressure of oxygen in the scattering chamber is then measured once again and attributed entirety to the effusive beam component. effusive beam component has been The measured for severat gas mixtures (oxygen and carrier gas) and in at) cases yte!ds a vaιue that is 1.75% of the supersonic component. radiates into Since the effusive component a hatf space with a cosine distribution that is azimuthaι)y symmetric the number of partic)es that interact with the crystai surface is negiigibιe. This means that the entire effusive component adds to the partiat pressure recorded during a reftectivity measurement at each instant in time after the initiât 0.5 seconds, where time constant effects are present. Hence, the effusive component is subtracted from the reftectivity measurement from tι+0.5 s through t3. The initiât probabitity computed as fottows. of dissociative chemisorption is The vatues of the partiat pressures from to to tι are averaged and this average is tabeted Pt.-^ and an uncertainty σ0-1 is computed for this average. A simitar computation is performed for partiat pressures from t2 to ⅛ and the average is tabeted P^.^. The uncertainty in Pt,-t,, is tabeted σ2-3. Finatty, as mentioned eartier a tinear teast squares fit is performed on the initiât transient partiat pressure data suitabty removed from tι (>5 vacuum time constants) and the tine generated by this fit is extrapotated to t2. This vatue is tabeted Pj and the associated uncertainty is tabetted oj. The initiât probabitity of dissociative chemisorption is then computed as: ii-1 7 S.=(Pt,-t,-Pi)∕{(Pt,-t, -P^,)0.9825}. The factor of 0.9825 in the equation accounts for the effusive component in the partia) pressure. Uncertainties for each measurement are estimated from the individuat uncertainties that make up the computation of the initiai dissociative chemisorption probabiiity. The uncertainty in the effusive component is not inciuded in the propagation of uncertainty since the contribution to the totai uncertainty is much iess than that of the other components (aiways < 5% of the totai uncertainty). The propagation of uncertainties is straightforward and ieads to the foiiowing simpie formuta for computation: (as.)2[Pt^-Pt..tj2=[1-S.]2(o2-3)2+S.2(oo-l)2+(Oi)2. Uncertainties in So are aiways iess than 0.05 and typicaiiy -0.025. Finaiiy, the surface coverage during a refiectivity measurement is represented by the cross-hatched area in Fig. 2 so that probabiiities of chemisorption versus coverage can be obtained. However in practice this tends to be quite difficuit since the beam refiectivity technique is not accurate for vaiues of So<0.05 thus making the area representing coverage iii-defined. !!-18 ))L RESULTS AND D!SCUSS!ON Figure 3 shows initia! the probabi!ity dissociative of chemisorption of oxygen on the !r(110)-(1×2) surface as a function of incident kinetic energy for Tg=150 and 300 K at norma! incidence. At !ow incident kinetic energies ( < 4 the initia! probabi!ity of chemisorption decreases with increasing E{. This kca!∕mo!) behavior is characteristic of a trapping-mediated mechanism for chemisorption and is comp!ete!y different from expected for direct chemisorption. that which is As previous!y mentioned, the trapping probabi!ity is expected to decrease with increasing Ej since the fraction of kinetic energy dissipation that is required in order to trap wi!! increase. fraction of incident Simp!y put, trapping wi!! occur when the kinetic energy impingement is greater than attractive we!! depth^4. demonstrated with trajectory that is dissipated upon Ej∕(Ej+ε), where ε is an effective This qua!itative feature has a!so been simp!e techniques26. hard-cube mode!s25 and c!assica! recent experiments Additiona!!y, investigating trapping of rare gases27-29 and mo!ecu!es3o-32 exhibit simi!ar trends. As mentioned ear!ier, a trapping-mediated dissociative chemisorption mechanism has previous!y been identified for C⅛ on Pt(111)12. Thefa!!offin So with increasing Ej for oxygen on Pt(111) in the trapping regime occurs over ∩ear!y the same energy range as that shown in Fig. 3. This fa!!off in So with Ej is quite rapid compared to other systems. The kinetic energy range over which H-1 9 trapping-mediated chemisorption occurs is sma)ier for oxygen on ιr(110)-(1x2) and Pt(111) than for f⅛ on W(1OO)14 or (⅛H6 on This is probabiy due to the depth of the weit in )r(110)-(1x2)i°. which trapping occurs being smaιier for the oxygen on ιr(110)-(1×2) and Pt(111) cases than the iatter cases, indeed for trapping- mediated chemisorption of C⅛ on Pt(111) the weit into which the partic)e traps appears to be the weit for physicai adsorptio∩20 which is approximateiy 3 kca)∕moi deep. Aiternativeiy, the weit depth for the physicai adsorption of C2H6 on ir(110)-(1x2) is approximateiy 9 kcai∕mo∏°. However, this expianation is by no means definitive. Prior to their identification of a physicaiiy adsorbed state of oxygen on Pt(111) Luntz, et ai. favored trapping into the moiecuiarty The reconstructed (110) surface of iridium is chemisorbed state^o. expected to be more reactive than either the (110) or (111) surfaces of Pt however the two metais show simitar behavior in their reactions with hydrogènes and hydrocarbons34. indeed Tayior, et a).21 mention a stabie moiecuiariy chemisorbed state of oxygen on ir(110)-(1x2) for Tg )ess than 100 K. The simiiarities between the two surfaces ieads us to be)ieve that the oxygen moiecuie is probabiy trapping into the physicaiiy adsorbed weii on the ir(110)- Untii additionai surface spectroscopic work is (1x2) surface a)so. performed with spectroscopy at C⅛ on ir(110)-(1x2), iow Tg, comparions with particuiariy vibrationai Pt are the oniy aiternative. Aiso shown in Fig. 3 is a rapid increase in So with increasing incident kinetic energy for Ej greater than approximateiy 4 kcai/moi. H-20 This behavior is typicai of that expected chemisorption via a direct mechanism. for dissociative A direct chemisorption mechanism has aiso been identified for oxygen chemisorption on Pt(111)i2 and on W(11O)13. The data shown in Fig. 3 are for two surface temperatures (150 and 300 K) and have virtuaιiy the same va!ue of So as a function of Ej indicating a tack of dependence on T$. Such behavior is expected for dissociative chemisorption via a direct mechanism but it is somewhat atypicaι of chemisorption in the trapping-mediated regime. Figure 4 shows the initia! probabi!ity of dissociative chemisorption as a function of surface temperature for incident conditions. two The upper data represented by a Δ are in the direct regime with E}=9.1kcat∕mo! and θ}=15° whereas the !ower data are in the trapping-mediated regime with Ej=980ca!∕mo! and θj=45°. This figure is further evidence of a tack of surface temperature dependence on the initia! probability of chemisorption. The decrease in So for Ts greater than 500 K for the data in the trapping-mediated regime (!ower data in Fig. 4) is most !ike!y due to a decrease in the trapping probabi!ity as wi!! be discussed short!y. As mentioned previous!y the probabi!ity of trapping-mediated dissociative chemisorption can increase with Tg, decrease with Tg or be independent of surface temperature, depending on the detai!s of the pote∩tia! energy surface. Once the mo!ecu!e is trapped it !oses a!! memory of its prior gas-phase conditions and accomodates fu!!y to the surface temperature. Thus the surface temperature ίί-21 dependence of So in the trapping-mediated regime arises from a kinetic competition between the dissociative chemisorption and desorption of the moiecu!e from the trapped state. mode! A simpie kinetic has been constructed that can account for this kinetic behavior for the reiativeiy faciie reactions of N2 on W(1OO)14.23, C⅛ on Pt(111)12 and C2H6 on )r(11O)-(1x2)io and the activated reactions of aikanes on Pt(110)-(1x2)ie and NF?. the activation However for systems where barrier to desorption from the trapped state is equivaient to the activation barrier to dissociation from the trapped state there is no surface temperature dependence. Using the kinetic mode) mentioned abovei°Ι4.i6.23 we estimate that the difference in the heights of the barriers to dissociate and desorb from the trapped state is approximate^ 100 ca)∕mo). As mentioned previously the decrease in the initia) probabi)ity of dissociative chemisorption for surface temperatures greater than 500 K as shown in Fig. 4 is probab)y due to a decrease in the trapping probabi)ity. Recent)y Rett∩er, et at/s have made detai)ed measurements of the trapping of N2 on W(100) for Tg from 300 to 1000 K. These studies indicate that the trapping probabi)ity does decrease s)ight)y (<_ 20%) over this temperature range, )ntuitive)y we wou)d expect that the amount of decrease in the trapping probabi)ity for a given system wou)d depend on the trapping wet) depth and the vibrationa) structure of the substrate. Additiona) work regarding the trapping of Ar on Pt(111) as a function of surface temperature29 support the weak but measureab)e dependence of the trapping probability on Tg. H-22 Figure 5 shows So as a function of incident angie for measurements in the trapping-mediated and direct regime. c!ear)y shown in As Fig. 5 the vaiue of the initia) probabiiity of dissociative chemisorption is independent of the incident ang)e for both the direct and the trapping-mediated regime. Aιso shown in Fig. 5 are predicted curves of the va)ue of So assuming norma) energy seating of the data, )t is dear that norma) energy seating does not ho)d and that the system sca)es with tota) energy. The nature of the angu)ar dependence on dissociative chemisorption as wet) as on trapping is poor)y understood at the moment. Tota) energy seating was found for N2 activated dissociative chemisorption on W(110)* and exptained by invoking a models which scrambtes the parattet and perpendicutar components of momentum via retativety smatt corrugations in the motecute-surface interaction potential· the activated dissociative chemisorption of However H2 on various Cu surfaces'* has been found to scate with normat energy. mediated chemisorption of ethane on tr(110)-(1x2) Trapping- is found to nearty scate with totat energy white the direct channet scates with ∩ormat energy*°. Oxygen chemisorption on W(110)i3 scates with normat energy but O2 dissociative chemisorption on Pt(111)i2 scates between norma) and totat energy seating. Ctearty much more expérimentât and theoreticat work is required to understand the nature of the angutar dependence on dissociative chemisorption. H-23 ÎV. SUMMARY The resu)ts of a molecu)ar beam investigation of the dynamics of the dissociative chemisorption of oxygen on ιr(110)-(1x2) in the ιimit of zero-coverage have been presented. dissociative chemisorption probabi)ity can We find that the initia) be described by a trapping-mediated mechanism at tow kinetic energies and by a direct mechanism at high kinetic energies. The vaιue of So is independent of the surface temperature except in the trapping- mediated regime for Tg greater than 500 K. Here the va)ue of the initia) probabi)ity of chemisorption decreases most )ike)y due to a decrease in the trapping probabi)ity. We a)so find that the va)ue of So is insensitive to incident ang)e in both the trapping-mediated and the direct chemisorption regimes seating with tota) energy. This work was supported by the Nationa) Science Foundation under grant number CHE-8617826. Acknow)edgme∩t is a)so made to the Donors of the Petro)eum Research Fund of the American Chemica) Society for partia) support of this research under grant number PRF 19819-AC5-C. H-24 REFERENCES 1. D.J. Auerbach, H.E. Pfnur, C.T. Rettner, J.E. Schiaeget, J. Lee and R.J. Madix, J. Chem. Phys. 81, 2515 (1984); H.E. Pfnur, C.T. Rettner, D.J. Auerbach, R.J. Madix and J. Lee, ∕b∕d. 85, 7452 (1986). 2. C.T. Rettner, H.E. Pfnur and D.J. Auerbach, Phys. Rev. Lett. 54, 2716 (1985). 3. S.T. Ceyer, Ann. Rev. Phys. Chem. 39, 479 (1988). 4. M. Batooch, M.J. Cardi!)o, D.R. Miι)er and R.E. Stickney, Surf. Sei. 46, 358 (1974). 5. M.B. Lee, Q.Y. Yang and S.T. Ceyer, J. Chem. Phys. 87, 2724 (1987). 6. M.P. D'Eveιyn, A.V. Hamza, G.E. Gdowski and R.J. Madix, Surf. Sei. 167, 451 (1986). 7. A.C. Luntz and D.S. Bethune, J. Chem. Phys. 90, 1274 (1989). 8. C.T. Rettner and H. Stein, Phys. Rev. Lett. 59, 2768 (1987); J. Chem. Phys. 87, 770 (1987). 9. G.R. Schoofs, C.R. Arumainayagam, M.C. McMaster and R.J. Madix, Surf. Sei. 215, 1 (1989). 10. C.B. Muiiins and W.H. Weinberg, J. Chem. Phys, (submitted). 11. A.V. Hamza, H.-P. Steinruck and R.J. Madix, J. Chem. Phys. 85, 7494 (1986). 12. A.C. Luntz, M.D. Wiιiiams and D.S. Bethune, J. Chem. Phys. 89, 4381 (1988). 13. C.T. Rettner, L.A. DeLouise and D.J. Auerbach, J. Vac. Sei. Techno). A 4, 1491 (1986); J. Chem. Phys. 85, 1131 (1986). 14. C.T. Rettner, H. Stein and E.K. Schweizer, J. Chem. Phys. 89, 3337 (1988). ][-25 15. W.H. Weinberg, in Kinetics of tnterface Reactions, edited by M Grunze and H.J. Kreuzer, Springer Series in Surface Science, Vo). 8 (Springer-Ver)ag, Ber)in, 1987), p. 94. 16. Y.-Κ. Sun and W.H. Weinberg, to be pubιished. 17. A.G. Sau)t and D.W. Goodman, J. Chem. Phys. 88, 7232 (1988). 18. C.T. Rett∩er, E.K. Schweizer, H. Stein and D.J. Auerbach, Phys. Rev. Lett. 61, 986 (1988). 19. C. Wang and R. Gomer, Surf. Sei. 84, 329 (1979). 20. A.C. Luntz, J. Grimbtot and D.E. Fowιer, Phys. Rev. B 39, 12903 (1989). 21. J.L. Tay)or, D.E. ιbbotson and W.H. Weinberg, Surf. Sei. 79, 3 49 (1979). 22. C.B. Mu))ins, Y. Wang and W.H. Weinberg, J. Vac. Sei. Tech. A 7, 2125 (1989). 23. D.A. King and M.G. We))s, Proc. R. Soc. London Ser. A 339, 245 (1974). 24. W.H. Weinberg and R.P. Merrit), J. Vac. Sei. Techno). 8, 718 (1971). 25. R.M. Logan and R.E. Stickney, J. Chem. Phys. 44, 195 (1966). 26. J.C. Tu))y, Surf. Sei. 111, 461 (1981). 27. C.T. Rett∩er, E.K. Schweizer and C.B. Mu))ins, J. Chem. Phys. 90, 3800 (1989). 28. C.T. Rettner, D.S. Bethune and D.J. Auerbach, J. Chem. Phys. 91, 1942 (1989). 29. C.B. Mu))i∩s, C.T. Rettner, D.J. Auerbach and W.H. Weinberg, to be pub)ished. 30. Π-26 E.W. Kuipers, M.G. Tenner, M.E.M. Spruit and A.W. Kieyn, Surf. Sei 205, 241 (1988). 31. S. Andersson, L. Wilzen and J. Harris, Phys. Rev. Lett. 55, 2591 (1985). 32. C.B. Muιiins and W.H. Weinberg, to be pubιished. 33. J.R. Engstrom, W. Tsai and W.H. Weinberg, J. Chem. Phys. 87, 3104 (1987). 34. P.D. Szuromi, Ph.D. Thesis Catifornia institute of Technoiogy, 1985. 35. J.W. Gadzuk and S. Ho!)oway, Chem. Phys. Lett. 114, 314 (1985); S. Hoιιoway and J.W. Gadzuk, J. Chem. Phys. 82, 5203 (1985). H-27 FιGURE CAPTtONS Figure 1. the Schematic of motecuiar beam apparatus for determining probabitity of dissociative chemisorpti<*',). Key: C1-C3, differentia) pumping chambers for nozz.e source (ns), GM, gas manifo)d, VG, va)ve for gas ftow, VP, view port, sa, skimmer aperture, F, f)ag, a, aperture, BC, beam chopper, UHV, u)tra-high vacuum scattering chamber, Ar+, argon ion gun, QMS, quadrupo)e mass spectrometer, AES, Auger e)ectron spectrometer, LEED, towenergy e)ectron diffraction optics. The sing)e crysta))ine samp)e is he)d on a maniputator which can coo) or heat the samp)e and rotates on an axis perpendicu)ar to the mo)ecu)ar beam. Figure 2. Schematic of the partia) pressure of oxygen in the scattering chamber as a function of time during a beam ref)ectivity measurement. At to the nozz)e is fitted with gas and at tι the ftag is opened so that the beam impinges on the crystat surface. The partiat pressure at tι is P2 greater than the basetine partiat pressure. As time proceeds the partiat pressure approaches a vatue P1 + P2 greater than the basetine and is steady after t2. This is because the surface is saturated with oxygen at this point and thus att oxygen motecutes in the beam are reftected from the surface. Finatty, at t3 the ftag is shut again. Figure 3. The initiât probabitity of dissociative chemisorption of O2 on tr(110)-(1x2) as a function of incident kinetic energy for Tg=150 K (o) and for Tg=300 K (X) at norma) incidence. ίί-28 Figure 4. The initiai probabiiity of dissociative chemisorption of O2 on ir(110)-(1x2) as a function of surface temperature for Ej=980 cai/moi and θj=45° (a) and for Ej=9.1 kcai/moi and θ⅛=15° (∆). Figure 5. The initiai probabiiity of dissociative chemisorption of O2 on )r(110)-(1x2) as a function of incident angie for Ej=9.1 kcai/moi and Tg=300 K (a) and Ej=2.6 kcai/moi and Tg=225 K (o). The X symboi and ∆ symboi represent the vaiues of So at Ej=9.1 kcai/moi and Ej=2.6 kcai/moi, respectiveiy, if norma) energy seating heid. These vaiues are obtained by interpoiating between points on Fig. 3. H-29 GM Figure 1 H-30 t, tl Figure 2 ιnitiat Probab iιity of Chemis orption U-31 Figure 3 H-32 ιnιtiat Probab itity of Chemiso rption 1.00 0.95 ∆ 0.90- A A A A A ∏ 0.85- π α α α α α a 0.80- 0.75a 0.70 — 1 00 —!----- *----- r* 200 300 —,----- '----- ι——1----- ι— 400 500 600 Surface Temperature, K Figure 4 700 initiai Probab iiity of Chemis orption ii-33 Figure 5 ίίί-1 CHAPTER )H A MOLECULAR-BEAM STUDY OF THE DiSSOC)AT!VE CHEMιSORPT!ON OF O2 ON )r(110)-(1x2) [The text of Chapter t!ι consists of an artic)e coauthored with Y. Wang and W.H. Weinberg, which has appeared in J. Mac. Sc/. 7ec∕7co∕. A 7, 2125 (1989).] Ht-2 A moιecu)ar-beam study of the dissociative chemisorption of (⅛ onir(110)-(1×2) C. B. Mullins/' Y. Wang, and ⅛V.H. Weinberg Dιιι⅛<oπ c/'C⅛emrst∕y oci/ CAemt∞∕ Fngineer/ng. Cc∕∕∕crn∕a ∕∕trmure q/Tec/tnc/ogy. Porodenc. Cα∕∕∕0rπ<c 97723 (Received t4October 1988; accepted31 October 1988) The zero-coverage probability of dissociative chemisorption of O2 on lr( IIO)-( I X2) has been measured using molecular-beam techniques for a wide range of incident kinetic energies, incident angles, and surface temperatures. The data indicate that a trapping-mediated mechanism is responsible for dissociative chemisorption at low energies, whereas at high energies a direct mechanism accounts for dissociative adsorption. Total energy scaling approximately describes the dissociative dynamics on the very corrugated Ir( H0)-( I X2) surface. Dissociative chemisorption on transition-metal surfaces is a key step in many industrially important heterogeneous cata­ lytic reactions. The dynamics of dissociative chemisorption are currently poorly understood and an active area of re­ search. In an effort to contribute further to an understanding of dissociative chemisorption, we have conducted a series of experiments to probe the interaction of molecular oxygen with the Ir( H0)-( I ×2) surface. Two studies of oxygen chemisorption dynamics employ­ ing molecular beams have appeared recently. Rettπer e/ a/. ' have studied the interaction of oxygen with a W( 110) sur­ face and concluded that the initial probability of adsorption depends strongly on the incident kinetic energy Æ, scaling with the normal energy Æ. = F, cos"/?,, where #, is the angle of incidence. They also found some evidence for the exis­ tence of a weakly bound precursor to dissociative chemi­ sorption at low incident kinetic energies. Williams et a/? have investigated the dynamics of dissociative adsorption of oxygen on Pt ( 111 ). They concluded that trapping-mediated chemisorption occurred for low incident energies and low surface temperatures, whereas at high incident kinetic ener­ gies a direct mechanism was likely the mechanism for chemi­ sorption of oxygen Taylor et a/? have investigated the inter­ action of oxygen with lr( 110)-( I x2) previously, and their work provided motivation for this study. They used thermal desorption mass spectrometry, low-energy electron diffrac­ tion, and work function measurements, and concluded that the initial probability of chemisorption was independent of surface temperature Γ,. They also concluded for 7*, between 300 and 700 K that the adsorption kinetics could be de­ scribed by a second-order precursor model. Zero-coverage dissociative chemisorption measurements have been made using the reflectivity method (of King and Wells*) employing an apparatus that will be described in detail elsewhere/ Briefly, the apparatus consists of a thrice differentially pumped supersonic molecular-beam source and ultrahigh vacuum scattering chamber. Low-energy elec­ tron diffraction optics. Auger electron spectroscopy, and an ion gun are mounted on the scattering chamber for obtaining and checking surface cleanliness and order. A quadrupole mass spectrometer is also mounted on the scattering chamber for thermal desorption mass spectrometry, reflec­ tivity measurements, and beam time-of-∏ight measure­ 2125 J. Vac. Sd. Techno). A7(3), May∕Jun1999 ments. The source chambers contain the nozzle, a high­ speed shutter, and a chopper for beam modulation as well as apertures for beam collimation. Both the scattering chamber and the third-beam chamber are pumped by turbomolecular pumps. The other two beam chambers are pumped by diffu­ sion pumps. The lr(H0)-(l×2) sample is mounted on a manipulator in the scattering chamber which provides pre­ cise alignment of the sample in the beam. The manipulator is liquid nitrogen cooled, providing rapid cooling of the sample to 80 K. The sample temperature is determined from a 0.003-in. W/5% Re-W/26% Re thermocouple spotwelded to the back of the crystal. As mentioned above, the initial probability of adsorption data reported here have been deter­ mined by a beam reflectivity method similar to that of King and Wells/ The partial pressure of oxygen in the scattering chamber is used as a measure of the 8ux of O, molecules that do not chemisorb. Initial chemisorption probabilities ⅜ (corresponding to chemisorption at zero coverage ) are thus determined by a comparison of the initial oxygen partial pressure, following the opening of the high-speed shutter, to the oxygen partial pressure from the beam scattering from the saturated surface. Beam energies are varied by a combi­ nation of seeding and variation of nozzle temperature, and are measured by tιme-of-8ight techniques. Figure I shows S„ as a function of incident kinetic energy for an angle of incidence of 15* with respect to the surface normal and surface temperatures of 150 and 300 K. As F, is increased from its lowest values, ⅜ decreases from a relative­ ly high value to a minimum at —4 kcal/mol. Note that this occurs for the measurements at ! 50 K as well as those at 300 K. The data displayed in Fig. 2 support the insensitivity of ⅜ to 7*,. This 6gure shows ⅜ vs 7*, for an incident angle of 45* and total incident energy of980 cal/mol, and an incident angle of 15* and F, of 9.1 kcal/mol. A decrease in ⅛ with increasing kinetic energy in the low-energy range is associat­ ed with a trapping- or precursor-mediated chemisorption mechanism, cf. Fig. I. A similar decrease in ⅜ with F, in the low F, regime for O2∕W(ΠO),' O,∕Pt(lll)∕ N,∕ W(l00)∕ C,H,∕lr(H0)-(!×2)∕ and C,H,,∕Ir(H0)( I X2)S has been observed before. However these systems displayed a strong sensitivity to surface temperature, with ⅜ decreasing with increasing surface temperature. This type of behavior can be interpreted as follows: the trapping proba- 0734-2101∕99∕032125-02!01.00 @ 1999 American Vacuum Society 2125 ί N-3 2i26 Mu)Hns, Wang, and Weinberg: Motecutar-beam atudy of the di⅛⅛octatιve ehemteorpt)on of Ο, of surface temperature for i^, = 980 c⅛l∕moi and kcal/mo! and #, = ! S*. FιC. ί. The mitiaι probabthty of dissociative chemisorption ⅛ as a function of incident kinetic energy for 7*, = i 50 and 300 K. and incident angle of 15*. biiity into the precursor state is a strong function of F, de­ creasing with increasing F,, but weakly influenced by 7*,.' Once a particle is in the precursor state it either desorbs or chemisorbs dissoctative!y, the temperature dependence of which depends on the kinetic parameters for desorption and dissociation?" For the system under discussion, O3∕ ir( Π0)-( ) X2), the relative insensitivity of F„ to 7*, at low F, suggests that the activation energy for dissociation is nearly equal to the activation energy for desorption from the precursor or trapped state. The slight decrease in ⅜ ob­ served from 400 to 600 K is probably due to a decrease in the trapping probability? The insensitivity of F„ to Γ, at high F, supports the assignment of a direct mechanism to dissocia­ tive chemisorption in this regime. Behavior of this type has been observed previously, ' * but one should note that in the study by Williams ef a/? F„ was found to have a surface temperature dependence even at high F,. The previous study of Oj ∕lr( 110)-( 1X 2) by Taylor ef a/? supports these con­ clusions in that F„ was determined to be insensitive to T, from 300 to 700 K. The fact that F„ for the O3∕Ir(!!O)(1x2) system is nearly insensitive to incident angle suggests near total energy scaling for the chemisorption dynamics of this system. This is not too surprising considering that the surface reconstructs into a very corrugated geometrical structure, providing a corrugated potential energy surface. Total energy scaling in dissociative chemisorption of a di­ atomic has been reported previously for the N3∕W( 110) sys­ tem' and the N,∕W (100) system? Chemisorption probabi- J. Vac. Sei. Tecħnol. A, Vd. 7, Mo. 3. May∕Ju∩ 1989 = 45*. and for Æ, = 9.1 lities F have also been determined as a function of coverage over a wide range of conditions. At low incident kinetic ener­ gies and low surface temperatures, F remains constant or decreases slowly for low coverages. This provides further supporting evidence for our assignment of a trapping-medi­ ated chemisorption mechanism for the low F, regime. A more complete description of the coverage dependence of the probability of dissociative chemisorption, as well as oth­ er details, will appear in a future publication? ,4c⅛πou½e<⅛,πeπh This work was supported by the Na­ tional Science Foundation under Grant No. CHE-8617826. " IBM Predoctora! Fellow 'C. T. Rettner, L. A. DeLouise, and D J. Auerbach. J. Chem. Phys. 85. 1131 (1986); J. Vac. Sei. Technol. A4. 1491 (1986). M D Williams, D. S. Bethune, and A C. Luπtz, J Chem. Phys 88, 2843 ( 1988); J. Vac. Sei. Technd. A 6, 788 ( 1988). *J. L. Taylor, D E. ibbotaon, and W. H. Weinberg, Surf. Sei. 79. 349 (1979). *D. A. King and M. G. Wells, Surf. Sei. 29,454 ( 1971 ). ^C. B. Mullins, Y. Wang, and W. H. Weinberg (to be published) *D. J. Auerbach, H. E. Pfnur, C. T. Rettner, J. E. Schi⅛egel, J. Lee. and R J. Madia, J. Chem. Phy*. gl. 2315 ( 1984). ?C. T. Rettner, H. Stein, and E. K. Schweizer. J. Chem. Phys 89, 3337 (1988). 'A. V. Hamza, H. P Steinruck, and R. J. Madia, J C⅛em. Phys, 83.7494 (1986). *C. T. Rettner. E. K. Schweizer, H. Stein, and D J Auerbach. Phys. Rev, Lett. 61.986 (1988). '°W. H. Weinberg, in Aments o∕^/ntez/hee AeacHtw. edited by M. Grunze and H. J. Kreuzer, Vol. 8 in Springer Series m Surface Science (Spππger, Berlin, 1987). p 94. iV-1 CHAPTER tV TRAPPtNG OF MOLECULAR ETHANE ON THE tr(110)-(1x2) SURFACE [The text of Chapter tV consists of an articte coauthored with W.H. Weinberg, which has been submitted to The Journa∕ of Chem/ca/ Phys∕cs.] tV-2 Trapping of a gas-phase particte at a surface is fundamental important and an etementary step in many physicaι and chemicat processes. Yet there is a paucity of data avaitabte concerning the kinetic energy and angutar dependence of trapping. Moreover, most of the extant data are from measurements of rare gas scatteringi^6 and extraction of thermatty averaged trapping probabitities from accommodation coefficients?. The retevant dynamics regarding atomic versus motecutar trapping are somewhat different, )n both cases the particie must transfer sufficient norma) kinetic energy to other degrees of freedom to remain bound. co))ision, transfer of kinetic energy to For a mo)ecu)e-surface interna) rotationa) and vibrationa) degrees of freedom is possib)e, as we)) as substrate excitations. Paral)e) momentum must u)timatety be dissipated a)so to insure complete trans)ationa) accommodation, and this is usua))y much slower than toss of norma) momentums. discuss qua)itative)y, an accurate mo)ecu)ar trapping dynamics A)though simp)e to quantitative description of is very comp)icated to construct. Dynamica) measurements of mo)ecu)ar trapping wit) increase our understanding of the ro)e p!ayed by the excitation of various degrees of freedom and provide a benchmark for predictive schemes. Our particu)ar interest in studying the trapping of mo)ecu)ar ethane on )r(110)-(1x2) was motivated by studies of trapping- chemisorption mediated dissociative system^. We present here the resu)ts of such a study emp)oying a molecular beam apparatus. in this same gas-surface )n particular, measurements of the trapping probability as a function of E,, the incident kinetic energy tV-3 (1.2 - 17.3 kca)∕mo)), and θ,, the incident angte (0° - 45°), at a surface temperature Tg of 77 K are presented and discussed. Measurements of the trapping probabitity in the timit of zero surface coverage, ζo, have been made using the refιectivity method of King and We!!s∏ empιoyi∩g an apparatus that wit) be described in detait etsewhere^. Briefty, the apparatus consists of a thrice differentiate pumped, supersonic motecu)ar beam source and uttra- high vacuum scattering chamber. Low-energy etectron diffraction optics, Auger etectron spectroscopy and an ion gun are mounted on the chamber scattering cteantiness and order. for obtaining checking and surface A quadrupote mass spectrometer is atso mounted on the scattering chamber for thermat desorption mass spectrometry, reftectivity measurements and measurements. beam time-of-ftight The source chambers contain the nozzte, a high­ speed shutter and a chopper for beam modutation, as wett as apertures for beam cottimation. Beam energies are varied by a combination of seeding and variation of the nozzte temperature (300-750 K), and are measured by time-of-ftight techniques. the beam-reftectivity With method∏, the partiat pressure of ethane in the scattering chamber is used as a measure of the ftux of C^H.that does not adsorb, tnitiat trapping probabitities are thus determined by a comparison of the initiât ethane partiat pressure fottowing the opening of the high-speed shutter, to the ethane partiat pressure from the beam scattered from a saturated surface. [V-4 At Tg=77 K alι mo)ecuιes that trap on the ctean surface wiιι remain physically adsorbed since the desorption temperature at iow coverages of ethane on ιr(110)-(1x2) is approximate)y 150 K12. This fact has been verified by also measuring ζo atTg=90 K and 110 K. There is no change in ζo for 77 K < Tg≤, 110 K, atthough there is a reduction in the saturation coverage with increasing Tg. Previous studies of the trapping of Ar on Pt(111)5 have shown ζ<, to be sensitive to background contamination on the surface, measurements coverage until here, reported near indeed, in the ζ was observed to increase with saturation^. The effect of background contaminants on the ζo measurements reported here are insignificant since the base pressure in our scattering chamber is -1x101° Torr, and the sampie can be cooιed from - 1620 K to 77 K in < 90 seconds. For norma! kinetic energies (E{Cθs⅜i) >_ 8 kca!∕mo!, there is a significant component of direct dissociative chemisorption. The measured vaιues of the direct chemisorption components.!° have been subtracted from the initia! adsorption probabi!ities for the re!evant measurements to obtain va!ues for ζo- Figure 1 shows ζo for ethane at three different incident ang!es as a function of Ejcos° ^θj. with increasing The initia! trapping probabi!ity decreases Ej since the mo!ecu!e must dissipate a !arger fraction of the normal component of its !inear momentum. We find empirica)!y that the data scale very wet) with Ejcos° ^θj, as is apparent in Fig. 1. This suggests a rather corrugated interaction potential consistent with which is the structure of this reconstructed surface^,15 corrugated geometrical tV-5 ft is possibte that the interna! degrees of freedom of the mo!ecu!e p!ay a significant rote in the dissipation of kinetic energy. The trapping probabi!ity of Xe interacting with Pt(111)6forθj=0° and Ej=1O kca!∕mo! (ζo,xe^O.1) is tower than the trapping probabitity for ethane on tr(110)-(1x2) for the same conditions (ζo,C2He^O.35), even though Xe is much more massive than CgH. and the wett depths for physicat adsorption of the two partictes are nearty equal· at tow Ej (-1 kcat/mot) ζo for both systems is However, nearty unity. Trajectory catenations for NO and N2 scattering from Ag(111) performed by Tutty and coworkersis-18 demonstrate that conversion of transtationat to rotationat energy during impact makes a very important contribution to trapping for these systems. Additional, Tutty and Cardites suggest that motecutes with tow-frequency vibrations can be expected to transfer vibrationat energy easity to and from rotationat, transtationat and surface vibrationat motion, tntuitivety, it is easy to visuatize conversion of transtationat energy to rotationat and vibrationat energy, but the re/af/ve rotes of the internat degrees of freedom of the impinging mo!ecute and the substrate surface excitations are difficutt to quantify. However, recent expérimentât and theoreticat work regarding NO impinging on Ag(1 11)19 suggests that increasing rotationat excitation is accompanied by decreasing energy transfer to phonons. Further expérimentât and theoreticat work is required in order to assess the significance of the observed seating of ζo with )V-6 E}cos° 5θj on )r(110)-(1×2) and the ro!e of interna) degrees of freedom. This work was supported by the Nationa) Science Foundation under grant number CHE-8617826. Acknow)edgment is a)so made to the Donors of the Petro)eum Research Fund of the American Chemica) Society for partia) support of this research under grant number PRF 19819-AC5-C. tV-7 REFERENCES 1. C.T. Rettner, E.K. Schweizer and C.B. Muttins, J. Chem. Phys. 90, 3800 (1989). 2. J.E. Hurst, L. Wharton, K.C. Ja∩da and D.J. Auerbach, J. Chem. Phys. 83, 1376 (1985). 3. H.P. Steinruck and R.J. Madix, Surf. Sei. 185, 36 (1987). 4. H. Schtichting, D. Menzet, T. Brunner, W. Brenig and J.C. Tutty, Phys. Rev. Lett. 60, 2515 (1988). 5. C.B. Muttins, C.T. Rettner, D.J. Auerbach and W.H. Weinberg, to be pubtished. 6. C.T. Rettner and D.S. Bethune, to be pubtished. 7. W.H. Weinberg and R.P. Merritt, J. Vac. Sei. Tech. 8, 718 (1971). 8. J.C. Tutty, Faraday Discuss. Chem. Soc. 80, 291 (1985). 9. The direct chemisorption component is taken as the vatue of the (temperature independent) probabitity of dissociative chemisorption at Tg=550 K where trapping-mediated chemisorption is negtigibte, as discussed in C.B. Muttins and W.H. Weinberg, to be pubtished. 10. Probabitities of direct chemisorption as a function of kinetic energy for ethane interacting with tr(110)-(1x2) are atso presented in A.V. Hamza, H.-P. Steinruck and R.J. Madix, J. Chem. Phys. 86, 6506 (1987); and H.-P. Steinruck, A.V. Hamza and R.J. Madix, Surf. Sei. 173, L571 (1986). 11. D.A. King and M.G. Wetts, Proc. Royat Soc. London A 339, 245 (1974). )V-8 12. T.S. Wittrig, P.D. Szuromi and W.H. Weinberg, J. Chem. Phys. 76, 3305 (1982). 13. H.C. Kang, C.B. Muιιins and W.H. Weinberg, J. Chem. Phys. (submitted). 14. C.-M. Chan, M.A. Van Hove, W.H. Weinberg and E.D. Witiiams, Surf. Sei. 91, 440 (1980). 15. M.A. Van Hove, W.H. Weinberg and C.-M. Chan, Low-Energy E/ecfron D∕7fracf∕on, Springer-Verιag, Bertin, 1986. 16. C.W. Muhihausen, L.R. Wi!ιiams and J.C. Tuity, J. Chem. Phys. 83, 2594 (1985). 17. J.C. Tuiiy, C.W. Muhihausen and L.R. Ruby, Ber. Bunsenges. Phys. Chem. 86, 433 (1982). 18. J.C. Tu!!y and M.J. Cardi))o, Science 223, 445 (1984). 19. J. Kimman, C.T. Rettner, D.J. Auerbach, J.A. Barker and J.C. Tuiιy, Phys. Rev. Lett. 57, 2053 (1986). tV-9 F)GURE CAPTtON Figure 1. Trapping probabitity as a function of Ejcos° ^θj (kcat/mot) for ethane on tr(110)-(1x2) at Tg=77 K for Θ; of 0°(α), 22.5°(+) and 45°(Δ). Expérimentât uncertainties^ in ζo range from ±.0.020 to +.0.040 with an average uncertainty for att 23 measurements of ±.0.029. The azimuthat orientation of the sampte is such that an [001] vector tying in the ptane of the sampte surface is -15° from the axis of rotation of the maniputator (i.e., the potar axis). The inset shows the same data ptotted versus the totat incident kinetic energy Ej. ÎV-10 Ejcos° 5θi(kcat∕mot) Figure 1 V-1 CHAPTER V TRAPPtNG-MEDiATED DiSSOCiATιVE CHEM!SORPTtON OF ETHANE ON )r(110)-(1x2) [The text of Chapter V consists of an articιe coauthored with W.H. Weinberg, which has been submitted to T∕7e Joαrπa∕ of Cbem/ca/ Pbys∕cs.] V-2 ABSTRACT Evidence is presented to support a trapping-mediated dissociative chemisorption mechanism for ethane interacting with an ιr(110)-(1x2) surface. The data were obtained from supersonic motecutar beam measurements with an incident kinetic energy E⅛ ranging between 1.2 and 24.1 kca)∕moι, a surface temperature Tg between 154 and 500K, and an incident angte θj between 0° and 45°. For Ej iess than approximate^ 13 kcat/mo), the probabi)ity of dissociative chemisorption So decreases rapidiy with increasing Tg. For a surface temperature of 154 K, So decreases with increasing E, for 1.2 <, Ej<. 13.4 kcat∕moι, consistent with a trapping-mediated chemisorption mechanism, indeed, the data support quantitativeιy a kinetic mode) consistent with a trapping-mediated chemisorption mechanism. The difference in the activation energies for desorption and chemisorption from the physica)iy adsorbed, trapped state Ed-Ec is 2.2+0.2 kcat/mo). tn the trapping-mediated regime, So is found to be rather insensitive to incident angte, seating with Ejcos° 5θj. V-3 L ÎNTRODUCTÎON Aιthough dissociative chemisorption of a gas-phase molecule on a metal surface is one of the "simpler" surface chemical processes and an important step in many industrial catalytic reactions, the dynamics and energetics of this class of chemical reactions are currently only poorly understood. This is due in part to the inherent theoretical difficulties in describing molecule-surface interactions accurately, and also because there have been relatively few fundamental chemisorption. studies of the dynamics of dissociative Current understanding of dissociative chemisorption on a bare metal surface involves two distinct mechanisms: a direct mechanism^ and a mechanism mediated by trapping of the incident particle in a motecu)arly adsorbed statei-s-9. In the trapping- mediated case, the molecule must lose sufficient energy to become trapped in a molecularty bound, adsorbed state after a collision with the surface. until it The molecule remains in this molecularly bound state either desorbs from the surface or reorientates into an appropriate configuration for dissociation to occur. Trapping- mediated chemisorption is thought to be an important mechanism in industrial catalytic processes, which are often activated, because of the relatively low kinetic energies of the gaseous reactants. In the case of the direct mechanism, dissociation occurs upon impact with the surface if the molecule has sufficient energy to surmount any activation barriers extant. V-4 Heretofore, most mo)ecuιar beam studies of chemisorption dynamics have focused on dissociative chemisorption by the direct The primary thrust of this earιier work has been to mechanism. study the effect of incident kinetic energy and a∩gte on the initia) probabiιity of dissociative chemisorption activation barrieri^4,ιo-i3. These for systems studies have with an contributed sig∩ifica∩t)y to a dearer understanding of the nature of the barrier to chemisorption for the systems studied. For these systems the probabi)ity of dissociative chemisorption increases with increasing kinetic energy. Variation of the incident angιe a)so provides information regarding the potentia) energy surface of interaction by providing, for examp)e, a systemization of the seating reιatio∩ship that exists between the probabi)ity of dissociation and the incident kinetic energy. These studies demonstrated a)so that the probabi)ity of direct dissociative chemisorption is typica))y a weak function of surface temperature. The microscopic phenomena that are expected to be re)evant for trapping-mediated dissociative chemisorption are rather different, and variations in the incident kinetic energy and ang)e wit) provide quantitative information rejecting these differences, tn particu)ar, a variation of these two parameters wi)) probe the dynamics and energy transfer processes by which the mo)ecu)e traps, permitting it to enter and remain in a weak)y bound potentia) we)). )ntuitive)y, one wou)d expect that the trapping probabi)ity ζo of a partic)e impinging on a bare surface wou)d decrease with increasing kinetic energy Ej, since a greater fraction of the incident V-5 energy must be dissipated, indeed, several expérimentai studies of the trapping dynamics of rare gases and molecuies have shown this to be the case^-17 A moiecuie wouid be expected to trap more readiiy than a rare gas atom with the same heat of phystcai adsorption, since tra∩siationai energy can be converted to rotatio∩ai and vibrationai substrate energy in the excitations^.18-20. moiecuiar case, addition to in Additionaiiy, there is frequency a strong dependence on surface temperature Ts in trapping-mediated chemisorption due to the kinetic competition between desorption and dissociative chemisorption from the trapped state. How the trapping probabiiity scaies with incident angie is iess intuitive, atthough severai expérimentai studies with both rare gases and moiecuies have shown it to be between normai (EjCθs^θj) and totai Aithough expérimentai studies of these seating^4-ι7 energy phenomena are of great practica) and fundamentai interest, o∩iy recentiy have reievant dy∩amicai data appeared in the iiterature2i- 28. Traditionaiiy, the evidence for precursor-med/afed chemisorption has been the insensitivity of the probabiiity of chemisorption to adsorbate coverages. This expérimentai observation is expiai∩ed by assuming that an impinging moiecuie can trap efficientty on an cccαp∕ed empty sites.29 site with subsequent migration to an ]∩ this work we distinguish between on the one hand phenomena occurring within the traditio∩ai framework of precursormediated chemtsorption and on the other hand dissociative chemisorption in the iimit of zero-coverage, mediated by trapping of the moiecuie in a moiecuiariy adsorbed state with subsequent moiecuiar desorption or reorientation into a configuration more V-6 favorabιe for chemisorption (i.e., trapping-mediated chemisorption), indeed, we concentrate here on the tatter point of view. in a study of the dissociative chemisorption of C⅛ on W(110) empioying an effusive moiecuiar beam source with Tg ranging from 20 to 1000K, Wang and Gomer? demonstrated that the probabiiity of chemisorption was insensitive to the adsorbate coverage at iow surface temperatures. A moiecuiar beam study by Rettner et aι.2s of both the initiai and the coverage-dependent dissociative adsorption for this system has not supported assignment of either a trapping- mediated or a precursor-mediated mechanism of chemisorption for Ts between 280 and 800K. in fact, the initial probabiiity of dissociative chemisorption So increases rapidiy with increasing Ej from 2.3 to 11.5 kcai/moi and was expiai∩ed to be due to a direct Oniy weak indirect evidence for frapp∕'πp- dissociation mechanism. medfaied dissociative chemisorption was found for kinetic energies ιess than 1.1 kcai/mot. The N2∕W(IOO) system is recognized^.30-32 as a system exhibiting precarsor-med/afed chemisorption, ι∩ a weιι-documented dynamicaι study, Rettner et a).26 found strong supporting evidence for both trapping-mediated and precursor-mediated chemisorption of N2 on W(100). Fi∩aι)y and particu)arty retevant to our work are the moiecuiar beam results of Hamza et al.21 regarding propane and butane dissociation on lr(110)-(1x2). Evidence for trapping- mediated chemisorption at low values of Ej and a direct mechanism V-7 at high Ej was presented, similar to the results reported here for ethane. This paper is a presentation and discussion of data concerning the interaction of ethane with the )r(110)-(1x2) surface employing a In an earlier thermal desorption study33, molecular beam apparatus. evidence was presented for molecular adsorption of ethane on lr(110)-(1x2) at 100K. Subsequent heating caused C-H bond cleavage at approximately 130K in some of the adsorbed ethane with the remainder desorbing. The activation energy for C-H bond cleavage was estimated to be approximately 6-7 kcal/mol with respect to the bottom of the well for physical adsorption, the depth of which is In a recent molecular beam study of approximately 8-9 kcal/mol. ethane on activation lr(110)-(1x2) using a reflectivity beam technique, Steinruck et al.34 concluded that there was no evidence for dissociative trapping-mediated temperatures between 300 and 1400K. we find strong indicating evidence chemisorption for surface In the work we report here, that a trapping-mediated mechanism dominates the dissociative chemisorption of ethane on lr(110)-(1x2) for surface temperatures between 154 and 250K and for Ej between 1.2 and 13.4 kcal/mol. mediated reaction, So decreases approximately 13 kcal/mol). As expected for a trapping with increasing Ej (up to We also find for trapping-mediated chemisorption that So is relatively insensitive to incident angle, nearly scaling with total energy; whereas the direct chemisorption channel scales with normal energy34,35 V-8 II. EXPERIMENTAL METHODS Measurements of the probabitity of dissociative chemisorption in the limit of zero surface coverage have been made using the reflectivity method of King and Wellses, employing an apparatus that will be described in detail elsewhere^s. Briefly, the apparatus consists of a thrice differentially pumped, supersonic molecular beam source and ultrahigh vacuum scattering chamber. Low-energy electron diffraction optics, Auger electron spectroscopy and an ion gun are mounted on the scattering chamber for obtaining and verifying surface cleanliness and order A quadrupole mass spectrometer is also mounted on the scattering chamber for thermal desorption mass spectrometry, reflectivity measurements and beam time-of-flight measurements. The source chambers contain the nozzle, a high-speed shutter, a chopper for beam modulation, and apertures for beam collimation. Both the scattering chamber and the third beam chamber are pumped by turbomolecular pumps; the other two beam chambers are pumped by diffusion pumps. The single crystalline sample was cut from a boule that was oriented with a Laue diffractometer and then polished using standard techniques. The lr(110)-(1x2) sample is mounted on a manipulator in the scattering chamber, which provides precise alignment of the crystal in the beam. [001] vector approximately The azimuthal orientation of the sample is such that an tying in the plane of the sample surface is 15° from the axis of rotation of the manipulator (i.e. the polar angle). The crystal is cleaned in vacuum by occasional argon ion bombardment and routine annealing in oxygen. Sharp LEED patterns and negligible spectroscopy are V-9 contamination obtained easily with as judged by Auger this procedure. The manipulator is liquid nitrogen cooled, providing rapid cooling of the sample to below 80K. The sample temperature is determined by a 0.003 in. W∕5%Re-W∕26%Re thermocouple that is spot welded to the back of the crystal. As mentioned above, the data reported here for the initial probability of adsorption were determined by a beam reflectivity method similar to that of King and Wellses. With this method the partial pressure of ethane in the scattering chamber is used as a measure of the flux of C2H6 molecules that do not chemisorb. There is no significant steady-state accumulation of physically adsorbed ethane for any of the experimental conditions reported here. Initial probabilities of chemisorption are determined by a comparison of the initial ethane partial pressure, following the opening of the high-speed shutter, to the ethane partial pressure from the beam scattering from the saturated surface. The kinetic energies of the beam are varied by a combination of seeding and variation of nozzle temperature, and are measured by time-of-flight techniques. III. RESULTS AND DISCUSSION Figure 1 shows the initial probability of dissociative chemisorption of ethane on the lr(110)-(1x2) surface as a function of incident kinetic energy, parametric in surface temperature. For a given surface temperature (Tg< 250K), So decreases as E⅛ increases, V-10 which is consistent with different compieteiy from trapping-mediated chemisorption that which is expected and direct for chemisorption. As mentioned eadier, the trapping probabiιity shou)d decrease with increasing E; since the fraction of kinetic energy dissipation that is required to trap will increase. trapping wi)ι occur when the fraction of iitustrates the idea: incident kinetic greater than deptt)37. A simp)e mode) energy that is dissipated upon impingement is E}∕(Ej+ε), where ε is an effective attractive we!i The shape of the So(EJ curve for Tg=154K and E}≤,1 0 kca)∕moi is very simiiar to that for the trapping of moiecuιar ethane on )r(110)-(1x2) at 77K^. Aιso apparent in Fig. 1 is the strong dependence of So on Ts, especiatiy at relatively iow incident kinetic energies. For Ej greater than approximately 13 kcaι∕moι, So increases with increasing Ej, and there is a significantly weaker dependence on surface temperature, !n this higher energy regime, the mechanism primary (but not exciusive) for dissociative chemisorption is the direct mechanisr∩34,35. As may be seen in Fig. 1, the va!ue of So is a function of Tg at each vatue of the impact energy. The functionaι dependence is stronger at !ower kinetic energies, since this is the regime in which the trapping-mediated cha∩∩eι of chemisorption is dominant. The probabiiity of trapping-mediated dissociative chemisorption can either be independent of temperature, increase with Tg or decrease with Tg, depending on the detaiιs of the potentiat energy surface. A simpie model serves to explain the temperature dependence of the trapping-mediated chemisorption of ethane on tr(110)-(1x2). A gas- V-11 phase molecule which impinges on the ctean surface traps with a probability of ζo- Once the ethane molecule is trapped in the physically adsorbed state, it loses all "memory" of its prior gasphase conditions and fully accommodates to the surface A particle in this weakly bound, trapped state will temperature. eventually either desorb or chemisorb. Here we are concerned with chemisorption in the zero coverage limit, and there is negligible accumulation of physically adsorbed ethane at the temperatures at which measurements were carried out. surface Thus, the rates of chemisorption and desorption from the trapped molecular state are extremely rapid compared to the ethane impingement rate. In this case the probability of dissociative chemisorption for the trapping-mediated component in the zero-coverage limit can be written as So = ζokc∕(kc + kd), (1) where kd is the rate coefficient of desorption, and kc is the rate coefficient of adsorbed state. dissociative chemisorption from the physically Equation (1) can be rewritten as So = ζo(Ej,θj,Ts) ∕[1 + (⅛∕^)exp{(Ec-Ed)∕kBTs}], (2) where k⅛ and k°è are the preexponential factors of the two rate coefficients, and E^ and Ec are the corresponding activation energies. In Eq. (2) the trapping probability is written explicitly as a function of Ej, θ, and Tg. However, we assume surface-temperature V-12 independence here since a recent study by Rettner et a!.3Sof N2 trapping on W(100) indicates that varying Tg from 300 to 1000K decreases ζo by ιess than 20 %. A verification of this kinetic mode) wouιd be provided if a pιot For such a case of ι∩[(ζo∕So) - 1] as a function of 1/Tg were tinear. the stope of the tine woutd be equat to (Ec-Ed)∕kg, the difference in the activation desorption from energies the Bottzma∩n constant. for dissociative chemisorption physicatty adsorbed wet) divided and by the Figure 2 is such a ptot of data for severat vatues of E⅛ and θj, for which the measured va)ues of ζo^are used in evatuati∩g the ordinate, tt is dear that a straight tine provides a good fit to att of the data over an extremety wide range of measurements (the ordinate spans a factor of more than two orders of magnitude and the abscissa a ∆Tg of 350K), providing strong support for the proposed trapping-mediated chemisorption mode). The stope of this tine corresponds to a vatue for Ej-Ec of 2.2+0.2 kcat∕mot39. Since the depth of the physicat adsorption wett for ethane is approximate^ 8.0 to 9.0 kcat/mot, the activation barrier to reaction from the physicatty adsorbed state is approximate^ 5.8 to 6.8 kcat/mot, in excette∩t agreement with the eartier study by Wittrig et at.33 The vatue of the ratio of preexponentiat factors is equat to ½d∕½°c-390+,1 00.39 is expected since, A ratio of ½d∕½*⅛ entropicatty, that is greater than unity desorption is favored over dissociation due to the timited phase space in which dissociation can occur compared to desorption. V-13 Figure 3 shows So as a function of incident angle θj, measured with respect to the surface normal, for two combinations of kinetic energy and surface temperature. For trapping-mediated dissociative chemisorption, the scaling is Eicos°-5θb whereas the direct channel of dissociative chemisorption scales with normal energy. For Ej=10.0 kcal/mol and Tg=154K, Fig. 3 shows the trapping-mediated incident component of So to increase slightly with increasing As clearly shown by the upper curve in Fig. 3 (labeled angle4°. "E}Cθs^θj Scaling"), normal-energy scaling does not hold; whereas Eicos° 5θj scaling describes the measured data very well. The increase in So with increasing θj is due to an enhancement in the trapping probability, which also scales as Ejcos°-5θj.i4 Indeed, the trapping-mediated component of dissociative chemisorption scales as Ejcos°-5θj because the trapping probability scales in this manner. This conclusion is verified by the excellent fit of the model embodied by Eqs. (1) and (2) to the experimental data, as shown in Fig. 2. In these equations the "dynamics" are contained in the ζo(Ej,θi,Ts) term and the "kinetics" in the kc∕(kc + k<j) term. Thus we would expect the scaling law to be contained in the trapping probability. This relative insensitivity to consistent with other recent studies of trapping. incident angle is In trapping studies of argon conducted on hydrogen precovered W(100)is and clean Pt(111)16 surfaces, it was also found that the trapping probability did not scale with normal energy. Indeed, the trapping of ∣∖⅛ on W(100) has been found to scale with total energy26. V-14 Also shown in Fig. 3 are data for purely direct dissociative chemisorption (Ej=2O.9 kcal/mol and Tg=500K). which support a normal-energy scaling assignment34,35 The predicted values of So shown in Fig. 3 for both normal energy scaling (the tower + symbol) and Ejcos° 5θj scaling (the lower x symbol) at E}=20.9 kcal/mol and Tg=500K are based solely on data displayed in Fig. 1. At this kinetic energy and surface temperature, the trapping-mediated component is negligible. It is clear from Fig. 3 that normal-energy scaling accurately describes the direct chemisorption channel· energy is scaling chemisorption in very for common activated direct, systems3.4; Normal dissociative although the direct dissociative chemisorption of N2 on both W(110)2 and W(100)26 scales with total energy. IV. SYNOPSIS Evidence, obtained from an experimental study employing a molecular beam apparatus, has been presented which supports a trapping-mediated mechanism of chemisorption interacting with an tr(110)-(1x2) surface. for ethane At low incident kinetic energies, the probability of dissociative chemisorption decreases with increasing E}, as would be expected for a trapping-mediated mechanism. kcal/mol, Moreover, the form of So(Ej) for Tg=154K and E,^1 0 is very similar to that for the trapping of molecular ethane on lr(110)-(1x2) at 77K14 The value of the probability of dissociative chemisorption depends on the surface temperature at all measured values of incident kinetic energy. However, the V-15 functional dependence is stronger at lower impact energies, since this is the regime in which trapping-mediated chemisorption is dominant The data support quantitatively a kinetic model that embodies a trapping-mediated mechanism of chemisorption. The measurements indicate that the activation barrier to dissociation from the trapped (physically adsorbed) state Ec is approximately 5.8 to 6.8 kcal/mol, whereas the depth of the physical adsorption well Ed is approximately 8.0 to 9.0 kcal/mol. The difference between these two energies, Ed-Ec, was measured accurately and found to be 2.2+0.2 kcal/mol. The trapping-mediated component of dissociative chemisorption increases with increasing incident angle and scales as Ejcosθ.5θj. The increase in So with increasing θj is due to an enhancement in the trapping probability, which also scales as Ejcoso.5θj.i4 Indeed, the trapping-mediated component of So scales as Ejcoso.5θj because the trapping probability scales in this manner. higher kinetic energies direct dissociative chemisorption At was observed, the probability of which scales with the normal component of the kinetic energy. ACKNOWLEDGMENTS This work was supported by the Office of Basic Energy Sciences of the Department of Energy under grant number DE-FG03- V-16 89ER14048. Acknowtedgment is a)so made to the Donors of the Petroιeum Research Fund of the American Chemicaι Society for partia) support of this research under grant number PRF 19819-AC5- C. We wish to thank Mr. Y.-Q. Wang for expérimentai assistance and Dr. J. R. Engstrom for many usefuι discussions. V-17 REFERENCES 1. J.A. Barker and D.J. Auerbach, Surf. Sei. Rep. 4, 1 (1984). 2. D.J. Auerbach, H.E. Pfnur, C.T. Refiner, J.E. Schιaeget, J. Lee and R.J. Madix, J. Chem. Phys. 81, 2515 (1984); H.E. Pfnur, C.T. Refiner, D.J. Auerbach, R.J. Madix and J. Lee, ½)∕'<d. 85, 7452 (1986). 3. C.T. Rettner, H.E. Pfnur and D.J. Auerbach, Phys. Rev. Lett. 54, 2716 (1985). 4. S.T. Ceyer, Ann. Rev. Phys. Chem. 39, 479 (1988). 5. L Langmuir, Chem. Rev. 6, 451 (1929); J.B. Taylor and I. Langmuir, Phys. Rev. 44, 23 (1933). 6. G. Ehrlich, J. Phys. Chem. Solids 1, 3 (1956). 7. C. Wang and R. Gomer, Surf. Sei. 84, 329 (1979). 8. D.A. King and M.G. Wells, Proc. Royal Soc. London, Ser. A 339, 245 (1974). 9. W.H. Weinberg, in Kinetics of Interface Reactions, Eds., M. Grunze and H.J. Kreuzer, Springer Series in Surface Science, Vot. 8 (Springer-Verlag, Berlin, 1987), p. 94. 10. M. Balooch, M.J. Cardi!lo, D.R. Miller and R.E. Stickney, Surf. Sei. 46, 358 (1974). 11. M.B. Lee, Q.Y. Yang and S.T. Ceyer, J. Chem. Phys. 87, 2724 (1987). 12. M.P. D'Evelyn, A.V. Hamza, G.E. Gdowski and R.J. Madix, Surf. Set. 167, 451 (1986). 13. A.C. Luntz and D.S. Bethune, J. Chem. Phys. 90, 1274 (1989). 14. C.B. Mullins and W.H. Weinberg, J. Chem. Phys, (submitted). 15. V-18 C.T. Rettner, E.K. Schweizer and C.B. Mutiins, J. Chem. Phys. 90, 3800 (1989). 16. C.B. Mutiins, C.T. Rettner, D.J. Auerbach and W.H. Weinberg, to be pubιished. 17. C.T. Rettner and D.S. Bethune, J. Chem. Phys, (in press). 18. J.C. Tuιιy and M.J. Carditto, Science 223, 445 (1984). 19. J.C. TuHy, C.W. [yluh!hausen and L.R. Ruby, Ber. Bunsenges. Phys. Chem. 86, 433 (1982). 20. C.W. Muhthausen, L.R. Ruby and J.C. Tutty, J. Chem. Phys. 83, 2594 (1985). 21. A.V. Hamza, H.-P. Steinruck and R.J. Madix, J. Chem. Phys. 85, 7494 (1986). 22. M.D. W⅛!!iams, D.S. Bethune and A.C. Luntz, J. Chem. Phys. 88, 2843 (1988). 23. D.A. King, CRC Crit. Rev. Sotid State Mater. Sei. 167, 451 (1978). 24. K.C. Ja∩da, J.E. Hurst, C.A. Becker, J.P. Cowin, L. Wharton and D.J. Auerbach, Surf. Sei. 93, 270 (1980). 25. C.T. Rettner, L.A. DeLouise and D.J. Auerbach, J. Vac. Set. TechnoL A 4, 1491 (1986); J. Chem. Phys. 85, 1131 (1986). 26. C.T. Rettner, H. Stein and E.K. Schweizer, J. Chem. Phys. 89, 3337 (1988). 27. C.B. Muιιins, Y. Wang and W.H. Weinberg, J. Vac. Sei. Tech. A 7, 2125 (1989). 28. C.B. Mut!i∩s and W.H. Weinberg, to be pub)ished. 29. E.S. Hood, B.H. Toby and W.H. Weinberg, Phys. Rev. Lett. 55, 2437 (1985). 30. R. Ciavenna and L.D. Schmidt, Surf. Sei. 22, 365 (1970). V-19 31. S.P. Singh-Boparai, M. Bowker and D.A. King, Surf. Sei. 53, 55 (1975). 32. P. Atnot and D.A. King, Surf. Sei. 126, 359 (1983). 33. T.S. Wittrig, P.D. Szuromi, W.H. Weinberg, J. Chem. Phys. 76, 3305 (1982). 34. H.-P. Steinruck, A.V. Hamza and R.J. Madix, Surf. Sei. 173, L571 (1986). 35. C.B. Mu)ii∩s and W.H. Weinberg, to be pub!ished. 36. D.A. King and M.G. Weιιs, Surf. Sei. 29, 454 (1971). 37. W.H. Weinberg and R.P. Merrit), J. Vac. Sei. Techno). 8, 718 (1971). 38. C.T. Rettner, E.K. Schweizer, H. Stein and D.J. Auerbach, Phys. Rev. Lett. 61, 986 (1988). 39. The expérimentai uncertainties correspond to one standard deviation. 40. For θj=O° and 22.5°, components of direct chemisorption (0.05 and 0.035, respectiveiy) were subtracted from the totat vaiue of So to obtain the trapping-mediated component of chemisorption. V-20 FIGURE CAPTιONS Figure 1. The initial probability of dissociative chemisorption of ethane on lr(110)-(1x2) as a function of incident kinetic energy for Ts=154, 182, 250 and 500K. AH measurements were made at normal incidence. Figure 2. A plot of !∩[(ζo∕So)-1] as a function of 1/Tg for numerous different incident kinetic energies and angles. Only the trappingmediated component of So is used for the ordinate (i.e., the measured direct component of dissociative chemisorption is subtracted from the measured value of So). Figure 3. The initial probability of dissociative chemisorption of ethane on lr(110)-(1x2) for Ej=20.9 kcal/mol and Tg=500K (O symbol) and the trapping-mediated component of dissociative chemisorption for Ej=10.0 kcal/mol and Tg=154K (∆ symbol) as a function of incident angle. Also shown, for both cases, are the predicted values of dissociative chemisorption for normal energy scaling (E}Cθs^θj, + symbol) and Ejcos°-5θ} scaling (x symbol). The experimentally measured values are represented by the Δ and O symbols, and the calculated values are represented by the + and x symbols. V-21 0.8 ο ω c ο CL ⅛Ο ω *Ε ω Ο ο J3 (ΰ Χ3 Ο ⅛ÛΈ ο 0.7 ο 0.6 + 0.5 Ο Ts=154K + Ts=182K A Ts=250K X Ts=500K ο 0.4 0.3 0.2 0.1 0.0 0 5 10 15 Ei, kca!∕mo) Figure 1 20 25 V-22 1∕T, Figure 2 initiai Probabi)ity of Chemisorption, V-23 Figure 3 V)-1 CHAPTER V) VARtAT)ON OF THE TRAPPιNG PROBABιHTY OF Ar ON Pt(111) WtTH KiNET)C ENERGY AND ANGLE OF ÎNCÎDENCE: THE CHANGÎNG ROLE OF PARALLEL MOMENTUM W)TH SURFACE TEMPERATURE [The text of Chapter V) consists of an artic)e coauthored with C.T. Rettner, D.J. Auerbach and W.H. Weinberg, which wiιi be submitted to Cbem/ca/ Phys/cs Letters.] Vt-2 ABSTRACT The trapping probabitity of Ar on Pt(111) has been measured as a function of incident kinetic energy Ej and ang)e θj for surface temperatures Tg of 80, 190 and 273 K. This probabi!ity decreases with increasing Ej in a manner that depends on both θj and Tg. For each surface temperature the measured trapping probabitity at different a∩gtes can be reduced to a common curve if it is ptotted as a function of the variabte Ejcos"θi- For Tg=80, 190 and 273 K we find optima! vatues of n of approximate^ 1.5, 1.0 and 0.5, respectivety. These resutts indicate that parattet momentum dissipation becomes i∩creasingty increases. more important to the trapping dynamics as Tg Vι-3 Understanding the dynamics of energy exchange and the equitibration of gases impinging on so)id surfaces is of both fu∩damentat and practica! importance. These phenomena are key steps in the physica! adsorption of gases at surfaces, chemisorption in heterogeneous catatytic chemistry, crysta! growth, and other industrial important processes. Atthough there is a targe body of data concerning the i∩etastic scattering of rare gases from meta! surfaces^, o∩ty recentty have the dynamics of the trapping process been studied specificatty2-8. Hurst, et at.9 have demonstrated that the f)ux of Xe atoms )eaving a Pt(111) surface can be reso!ved into both direct and trapping-desorption components. A !ater study by the same group focused on Ar scattering from Pt(111)i°, and it was shown that the va!ue of the trapping probabi!ity ζo was tess than for Xe on Pt(111) for simitar incident beam conditions. These studies enhanced the avai)ab!e data base considerabty, but they did not provide comprehensive answers to questions regarding the detai!ed dynamics of trapping. Late!y, these questions have been addressed more fu!)y for both rare gas atoms2-s arid mo!ecu!es6-s interacting with we!!-characterized meta) surfaces. An important goat of these studies has been to assess the retative rote of parattet and perpendicutar momentum transfer in the trapping process through measurements of the effect of angte of incidence on the rate at which trapping decreases with increasing energy. A considerabte amount of data concerning inetastic scattering of rare gases from metat surfaces indicates that momentum Vt-4 exchange is much more efficient in the norma! than in the tangentia) Hence, we might expect the trapping probabitity to be direction!. primari)y a function of norma) momentum, or the so-ca)!ed "normat- energy" Ejcos2θ∣. Two groups have recent!y measured the trapping probabi!ity of Xe on Pt(111) at surface temperatures of 85 K3 and 95 K4 as a function of incident kinetic energy and a∩g!e to examine this issue. Both groups observed a decrease in ζo with Ej but found significant deviations from the simpie norma!-energy seating. tn one cases resutts obtained at different angtes seated as Ejcosi-6θj white in the other4 Ejcosi-°θj seating was reported, the same as that found in an eartier investigation of Ar trapping on W(100)-2H2. These studies suggest that parattet momentum is more important in trapping than previousty thought. White these measurements have begun to etucidate the physics of the microscopic energy transfer, much remains to be understood, tn particutar, the effect of surface temperature on the detaited trapping dynamics has yet to be examined. This is parity due to the difficultés inherent in making such measurements. Most previous studies have been carried out at temperatures at which the residence time of the trapped species is sufficiency tong to permit the trapping to be inferred from the surface coverage or the adsorption of incidence ftux. temperatures that are sufficientty high to atter However, at the trapping probability significa∩tty, residence times witt usuatty be much too smatl for such methods, tn these cases the trapping probability can be deduced instead from the intensity of the trapping-desorption component of the "scattered" ftux. This approach was used in the previousty mentioned study of Ar trapping on W(100)-2H at a surface Vι-5 temperature of 85 K^. Here we report an extension of this work to the case of Ar trapping on Pt(111), where we have been able to determine the probability of trapping as a function of incident energy and angle for different surface temperatures. The results reveal that the energy-scaling relation is not constant with surface temperature thus indicating the varying role of parallel momentum transfer in the trapping process as a function of surface temperature. A detailed description of the apparatus has been presented previously2.ιι, and thus we present only a brief account here. Generally, beams of Ar generated by a supersonic nozzle are directed at a single crystal of platinum and the scattered flux is detected. The platinum sample is contained in an ultrahigh vacuum (UHV) scattering chamber and is mounted on a manipulator which provides accurate control of the incident angle and the surface temperature between approximately 80 and 1500 K. the preparation of the sample. Particular care was taken in Alignment and polishing were accomplished on a carefully machined sample mount that allowed both orientation adjustment from Laue photographs and polishing. The surface is within +0.2° of the (111) plane. The crystal has been mounted on the manipulator such that the [121] vector lies in the scattering plane. The sample was cleaned routinely using standard oxygen cleaning and Ar ion sputtering techniques. Sharp LEED patterns are obtained and contamination levels (mainly carbon) are below 1%, as determined by Auger electron spectroscopy. In addition, the scattering of a 30 mev beam of He from the cooled V)-6 (Tg=80 surface K) gives a peak specutar with a width indistinguishab)e from the instrumenta) reso)ution of 1.6° and which contains greater than 80% of the incident f)ux. Beam energies are varied by changing the nozz)e temperature T∏oz of a 75 μm nozz)e and by seeding in He. For the )owest energies, the nozzιe is coo)ed to approximate^ 100 K, and the pressure of Ar in the nozz)e is hetd at about 500 condensation. are used. Torr, in order to avoid For T∏oz above 250 K, Ar pressures of up to 1500 Torr The insensitivity of the resu)ts to the beam conditions suggests that c)ustering in these beams is neg)igib)e. The beam energies are determined from f)ight times between a high-speed chopper and a differential pumped rotatabιe mass spectrometer. This device is aiso used to measure the time-of-ftight and angu)ar distributions of distributions have the scattered been and obtained A∩gu)ar desorbing Ar two schemes. using The distributions of the tota) scattered and desorbing ftux can be recorded using phase sensitive detection, referenced to the chopper and processed with a )ock-in amp)ifier. Atternative)y, distributions can be constructed by a∩a)yzing time-of-ftight spectra for specific scattering angιes so that ftux distributions can be obtained, rather than the density distributions that resutt from the tock-in method. Detaits of the time-of-ftight anatysis are given in Ref. 2. Time-of-ftight distributions have been recorded over a wide range of conditions and show bimodatity in many cases. Although bimodat distributions have not been reported previously for the Vt-7 Ar∕Pt(111) system, qua)itativety simitar distributions have been obtained for the Ar∕W(100)-2H2 and Xe∕Pt(111)9 systems, which can be separated into direct-inetastic desorption components. component and trapping- Thus, we use the ftux-weighted signa) of the trapping-desorption component as a measure of the re/af/ve trapping probabitities for differing incident beam conditions. A separate mass spectrometer is used to measure the Ar partiat pressure rise in the scattering chamber in order to account for differences in incident beam intensity. We have assumed that, for a given Tg, att Ar atoms that trap and thus accomodate to the surface desorb with energy and angutar distributions that are a function of surface temperature onty, i.e., the partictes have no "memory" of their incident conditions. Hence, for att the measurements of the re/af/ve trapping probabitities, the differentiatty pumped quadrupote mass spectrometer was positioned 5° from the surface normat in order to measure the appropriate desorbing ftux consistency for att incident beam component. conditions and to minimize the direct-inetastic The technique described above for measuring retative trapping probabitities has the advantage of a!!owing trapping to be studied as a function of surface temperature. However, if the surface temperature is too high (for this study 2i300 K), two effects combine to etiminate the separabitity of the direct-inetastic and trapping-desorption components. The direct-inetastic tobe broadens and the trapping-desorption component is faster. Absotute trapping probabitities can be estimated from angular distributions, where this probabitity is taken to be the fraction of Vt-8 scattered species not contained in the direct-i∩etastic iobei2. Here it is necessary to estimate the form of the out-of-pia∩e distribution for the direct-ine!astic tobe. a∩gutar distributions, but We have not measured out-of-ptane experiments on Ar scattering from Pt(111) by Hurst et at J 3 suggest that the out-of-ptane a∩gutar width is between 0.5 and 1 times the in-ptane width. The accuracy of this method increases rapidty as the trapping probabitity approaches unity, where the ftux contained in the direct-inetastic tobe becomes negtigibte. As a resutt of the uncertainty in the out-of-ptane angutar width, the absotute va)ues of the trapping probabitity reported here shoutd be considered accurate to onty ±10%. retative vatues of ζo are accurate to ±5%. The The resutts reported here atso agree with an of estimation of the absotute trapping probabitity that is based on comparing the absotute intensities in the desorption component with the diffuse scattering from a nominatty fiat part of the sampte hotder that is atigned at the crystat position, t∩ this case we observe onty a singte scattering component that peaks at the surface normat, which we take to represent 100% trapping^. We find that the trapping probabitity has a strong dependence on both incident kinetic energy and angte, increasing with angte of incidence for a given kinetic energy. For Tg=80 K, the angutar dependence is in quatitative agreement with normat-energy seating, as can be seen in Fig 1 where ζo is ptotted as a function of EjCos^Oj. However, it is apparent that this seating taw is far from being satisfactory. To describe these resutts more accuratety, we have considered a seating variable of the form Ejcos∏θ,, which attows for a Vt-9 continuous variation between norma) energy seating (∩=2) and tota) energy seating (∩=0). Figure 2 shows the trapping probabitity ptotted as a function of Ejcosi $6; which is much ctoser to an idea) seating function. This vatue of n is actuatty noticeabty better than ∩=1.6 or 1.4, but considering systematic errors this and other vatues quoted betow shoutd be taken as accurate to onty ±0.2. The chosen vatue of n=1.5 is remarkabty simitar to that obtained recentty for trapping of Xe on Pt(111) at nearty the same surface temperature (Tg=85 K)3, where a vatue of n=1.6 was estimated based on data obtained using a different technique. We atso note that there is no obvious physicat basis for this form, but contend that such a seating taw may be generatty appticabte to trapping. The decrease of ζo with Ej is expected, since the fraction of this energy that must be tost increases rapidty as it is raised. The quatitative form of this curve has been reproduced using hardcube^, "washboard"is and futt ctassicat trajectory methods^, the défaits of which witt be reported etsewhere. Simitarty, this form is in good basic agreement with previous catenations for this system by Tuttyi6 and LeuthausseH?, and with an expérimentât curve obtained by Hurst, et atjo who used detaited batance arguments to estimate its form scattering from from time-of-ftight Pt(111) at Tg=100 K. measurements of Ar Furthermore, a simitar detaited batance anatysis^To, apptied to the present time-of-ftight spectra gives vatues of the trapping probabitity in good agreement with the direetty measured vatues for att three temperatures. Additio∩atty, a detaited batance a∩atysis of the direetty measured Vt-10 trapping probabiiities predicts the correct degree of transιatio∩a) coo)ing in the desorbing fiux. For exampie, at θj=5°, we estimate effective temperatures for the desorbing Ar of approximate^ 70, 145 and 196 K for Tg=80,190 and 273 K, respective))/, which are within 10% of the fitted temperatures in a)most a)) cases. Again, detai!s are deferred to a future pub)ication. Figures 3 and 4 show experimental determined trapping probabiiities for Ar interacting with Pt(111) temperatures of 190 K and 273 K, respective)y. with surface in these cases we find that rather different seating taws appty, corresponding to n=1.0 and 0.5. Fig. 3 shows the trapping probabitity as a function of Ejcosi 0θ} at Tg=190 K. Simitarty, in Fig. 4 the trapping data at Tg=273 K are shown with n=0.5. As mentioned eartier, these seating taws are merety for convenience. However, the changes in the seating retationships with Tg can be used to gain further insight into the retative rote parattet momentum transfer ptays in trapping as a function of surface temperature and witt be discussed tater. The vatue of ζo at the two higher surface temperatures atso decreases with increasing Ej. However, a c)oser inspection of the tow E; data for Figs. 2-4 atso reveats a decrease in ζo with increasing Ts, white at higher energies the reverse is true. This behavior reftects the fact that surface atoms that are moving towards the incoming Ar atoms witt reduce the trapping probabitity of stow incident species, white at high energies, where trapping is tow, cottisions with surface atoms that are moving away from the incident species have a net effect of increasing the number that trap. This effect is Vt-11 readiιy seen in simp)e hard-cube modets^.i8, and has been discussed, for exampιe, for the NO∕Ag(111) system^. )t is apparent that the optimum vaiue of the exponent n systematicatty decreases with increasing Tg as can be seen in Fig. 4, which summarizes this behavior. This p)ot further suggests that normai-energy seating (n=2) may be appticabte as Tg approaches zero, white trapping at 500 K or so may be expected to be targety independent of angte of incidence. The dynamicat origin of this effect is by no means obvious. Ctearty parattet momentum becomes increasingty more important in the trapping process as the surface temperature is increased, suggesting a corresponding increase in the effective corrugation of the interaction potential· However, the actuat surface roughness due to the therma) dispιacement of surface atoms is aimost certainty too sma)t for this behavior to be accounted for by simpte singteco)ιision modets. Specificatty, we estimate that the tocat norma) at the point of impact deviates from the true normat by tess than 3° at 273 K. Catenations based on a "washboard" models, which is simitar to a hard-cube mode) but incorporating surface corrugation, reveat this to have onty a weak effect on the dependence on angte of incidence. Thus, we betieve that a muttipte cottision picture is most tikety required. A key issue here is the fate of atoms that are initiatty trapped by virtue of toss of normat momentum, but which retain sufficient totat energy to escape. Calcutations based on a semi-classicat treatment of the forced oscittator modet have shown V)-12 that the effect of a∩gte of incidence on the trapping of Ne on Cu(100) depends sensitivety on the fate of these atoms, and even decreases with increasing angte of incidence (n<0) if it is assumed that the initia) coι)ision must )eave the atom with a net negative tota) energy for trapping to occur^o. C)assicaι trajectory catenations for Ar and Xe trapping on Pt(111)ie suggest that para))et momentum accomodation is re)ative)y stow and that a significant fraction of species that tose their normat energy in the first cottision escape before their parattet momentum is futty accomodated, tt may be that as Tg is increased, thermat roughness serves to coupte parattet and perpendicutar momenta at a rate that can compete with the rate at which these are dissipated. Detaited catenations are required to further address this question. Finatty we note that the vatue of ζo for Tg=80 K appears to extrapotate to near unity as E; approaches zero as seen in Fig. 1. This is in contrast to recent resutts for Ne and Ar adsorption on Ru(001)5.2i. The physicat adsorption probabitity tends to vatues far tess than one with decreasing Ej for both Ne and Ar indicating the importance of quantum effects for these rare gas atoms interacting with Ru(001). These quantum effects appear to be comptetety accounted for when the sotid is treated quantum mechanicatty and the gas atom ctassicatty. The apparent absence of quantum effects for Ar trapping on Pt(111) appears to be due to the differences in vibrationa) structure for the Pt(111) and Ru(001) surfaces. Vι-13 tn summary we have presented measured vaιues, obtained using a nove! motecutar beam technique, of the trapping probabiiity of Ar on Pt(111) as a function of incident energy and angie and surface temperature. The expérimenta) resu)ts demonstrate an increasing)y more important rote for para))e) momentum in the trapping process with increasing surface temperature, which is manifested by a decrease in the exponent n in the seating retation Ejcos"θj with increasing Tg. to be a shift Thus, as the surface temperature rises there appears in the retative rate of dissipation of parattet momentum versus scattering from the surface via a coupti∩g of parattet momentum to normat momentum. Additional, our resutts do not show any significant quantum effects with regard to the trapping process. The trapping increases both decreasing with probabitity of Ar on Pt(111) surface temperature and with decreasing Ar kinetic energy. We are indebted to J.E. Schtaeget for assistance in the maintenance and upkeep of the apparatus, and to D.S. Bethune for hetpfut discussions. One of us (W.H.W.) thanks the Nationat Science Foundation for partiat support of this research under grant number CHE-8617826, and the Donors of the Petroteum Research Fund of the American Chemicat Society for partiat support under grant number PRF 19819-AC5-C. V)-14 REFERENCES 1. J.A. Barker and D.J. Auerbach, Surf. Sei. Rep. 4, 1 (1984). 2. C.T. Refiner, E.K. Schweizer and C.B. Mutti∩s, J. Chem. Phys. 90, 3800 (1989). 3. C.T. Rett∩er, D.S. Bethune and D.J. Auerbach, J. Chem. Phys. 91, 1942 (1989). 4. C.R. Arumainayagam, R.J. Madix, M.C. McMaster, V.M. Suzawa and J.C. Tuιiy, to be pubιished. 5. H. Schιichting, D. Menzeι, T. Brunner, W. Brenig and J.C. Tutιy, Phys. Rev. Lett. 60, 2515 (1988). 6. C.B. Muιιins and W.H. Weinberg, to be pubiished. 7. E.W. Kuipers, M.G. Tenner, M.E.M. Spruit and A.W. K)eyn, Surf. Sei. 205, 241 (1988). 8. S. Andersson, L. Witzen and J. Harris, Phys. Rev. Lett. 55, 2591 (1985). 9. J.E. Hurst, C.A. Becker, J.P. Cowin, K.C. Ja∩da, L. Wharton and D.J. Auerbach, Phys. Rev. Lett. 43, 1175 (1979). 10. J.E. Hurst, L. Wharton, K.C. Janda and D.J. Auerbach, J. Chem. Phys. 83, 1376 (1985). 11. C.T. Rettner, L.A. DeLouise and D.J. Auerbach, J. Chem. Phys. 95, 1131 (1986). 12. C.T. Rettner, E.K. Schweizer, H. Stein and D.J. Auerbach, Phys. Rev. Lett. 61, 986 (1988). 13. J.E. Hurst, L. Wharton, K.C. Janda and D.J. Auerbach, J. Chem. Phys. 78, 1559 (1983). 14. R.M. Logan and R.E. Stickney, J. Chem. Phys. 44, 195 (1966). Vi-15 15. J.C. Tuιιy, to be pubtished. 16. J.C. Tu)ιy, Surf. Sei. 111, 461 (1981). 17. U. Leuthausser, Z. Phys. B 50, 65 (1983). 18. W.H. Weinberg, Adv. Coii. inter. Set. 4, 301 (1975). 19. G.D. Kubiak, J.E. Hurst, H.G. Rennage!, G.M. McCietiand and R.N Zare, J. Chem. Phys. 79, 5163 (1983). 20. M. Persson and J. Harris, Surf. Sei. 187, 67 (1987). 21. D. Menzei, private communication. Vι-16 FtGURE CAPTtONS Figure 1. The experimental determined trapping probabiιity of argon on Pt(111) as a function of Ejcos^θ} at Ts=80 K. Figure 2. The same data as in Fig. 1 showing the experimental determined trapping probabi)ity of argon on Pt(111) at Ts=80 K pιotted as a function of Ejcosi 5θj rather than the normat-energy. Figure 3. The experimental determined trapping probabi)ity of argon on Pt(111) as a function of Ejcosi-°θ, at Tg=190 K. Figure 4. The experimental determined trapping probabi)ity of argon on Pt(111) as a function of Ejcos° 5θj atTg=273 K. Figure 5. The exponent n in the seating retation Ejcos"θj as a function of surface temperature. V)-17 1.0 Argon/ Pt(111) Ts=80 K 0.8 -û σ 0.6 .a o ⅛CL· CD çz *CL 0.4 Ûc -=30° ^=45° .=60° 0.2 A *Aw 0.0 - 0.00 0.05 H A 0.10 Ej cos⅞} (eV) Figure 1 A " AA * 0.15 0.20 Vt-18 1.0 .Argon/, Pt(11 1) A Ts=80 K 0.8 A -JÛσ 0.6 -Οο C∏ er <30.4 σCL Η -=30° A =45° .=60° ⅛ 0.2 *ί 0.0 0.00 0.05 0.10 Ε] cos^'⅛-, (eV) Figure 2 0.15 0.20 Vt-19 Figure 3 Vι-20 Figure 4 V)-21 Surface Temperature (K) Figure 5 A-1 APPEND)XA MOLECULAR ADSORPHON OF ETHANE ON THE )r(110)-(1X2) SURFACE: MONTE-CARLO StMULATiONS AND MOLECULAR BEAM REFLECTιViTY MEASUREMENTS [The text of Appendix A consists of an articte coauthored with H. C. Kang and W. H. Weinberg, which has been submitted to The Joorπa∕ of Chem/ca/ Phys∕cs.] A-2 ABSTRACT Experimental results, obtained using a molecular beam reflectivity method, for the probability of molecular physical adsorption of ethane on the Ir(110)-(lx2) surface are presented. We analyze these results using Monte-Carlo simulations and show that molecular adsorption can occur either "directly" or through a precursor state in which an ethane molecule is trapped in a second layer of molecularly adsorbed ethane with subsequent migration to a vacant site. From the Monte-Carlo simulations, we are able to establish that the energy barrier for the desorption of an ethane molecule from the precursor state is approximately 4.5 kcal/mol. We also find that the energy barrier for diffusion of an ethane molecule on top of a monolayer of ethane molecularly adsorbed on the Ir(llO)- (1x2) surface is approximately 3.7 kcal/mol. A-3 1. Introduction The adsorption of a molecule from the gas phase onto a solid surface does not necessarily occur in one direct step. The impinging molecule may be trapped temporarily in a weakly bound precursor well before it reaches the adsorption well. For example, a physically adsorbed molecule could be a precursor to either molecular or dissociative chemisorption (l). Likewise, condensed molecules in a second layer could be precursors to adsorption on the surface, depending on the relative rates of migration in the second layer and desorption from the second layer (1). The concept of such precursor states, first proposed by Langmuir (2,3), is rather important in understanding the kinetics of a variety of surface phenomena. If, for instance, chemisorption is mediated by a precursor state in which the adsorbing molecule accommodates thermally to the surface, then the probability of chemisorption would be strongly affected by the surface tempera­ ture. This would not be true if chemisorption occurs directly, i.e., without passing through a precursor state. Descriptions of the kinetics of precursor-mediated ad­ sorption have been presented by Becker (4,5), Ehrlich (6) and Kisliuk (7); and they have been reviewed by Weinberg (1). There have been important refine­ ments of these macroscopic approaches by treating the microscopic configurations of the adsorbed layer in a statistical manner (8-14). Monte-Carlo modeling, tak­ ing into account the microscopic configuration of the adsorbed layer around each adsorbed molecule, and hence the lateral interactions by which each molecule is influenced, has also been formulated (13). In this paper we present evidence demonstrating that the molecular physical adsorption of ethane on the Ir(110)(1x2) surface can be understood by drawing upon the concept of a precursor to molecular adsorption. We show that molecular adsorption can occur directly when an ethane molecule impinges upon the bare iridium surface or indirectly when an ethane molecule impinges upon already molecularly adsorbed ethane. The "extrinsic" precursor in this case is the mobile ethane molecule trapped on A-4 top of a layer of molecularly adsorbed ethane (15). The experimental results, which are obtained from reflectivity measurements performed using a molecular beam apparatus, are shown to be well described by Monte-Carlo simulations of a model which takes into account the two pathways delineated above. 2. Experimental and Algorithmic Details Measurements of the probability of molecular adsorption of ethane on the Ir(110)-(lx2) surface have been made using the reflectivity method of King and Wells (12), employing a molecular beam apparatus that will be described in detail elsewhere (16). Briefly, the apparatus consists of a thrice differentially pumped, supersonic nozzle molecular beam source and an ultrahigh vaccuum scattering chamber. Low-energy electron diffraction optics, Auger electron spec­ troscopy and an ion gun are mounted on the scattering chamber for obtaining and checking surface cleanliness and order. A quadrupole mass spectrometer is also mounted on the scattering chamber for thermal desorption mass spectrometry, reflectivity measurements and beam time-of-flight measurements. The source chambers contain the nozzle, a high-speed shutter and a chopper for beam mod­ ulation, as well as apertures for beam collimation. Beam energies are varied by a combination of seeding and variation of nozzle temperature, and are measured by time-of-flight techniques. Both the scattering chamber and the third-beam chamber are pumped by turbomolecular pumps; the other two beam chambers are pumped by diffusion pumps. The iridium crystal is mounted on a manipula­ tor which provides precise alignment. The manipulator is liquid nitrogen cooled, providing rapid cooling of the sample to below 80 K. The sample temperature is determined by a 0.003 in. W∕5⅞Re-W∕26⅞Re thermocouple spot welded to the back of the crystal. The partial pressure J⅛ of ethane, at time f, in the scattering chamber is used as a measure of the flux of ethane molecules that do not adsorb. Hence, A-5 the probability of adsorption at time t is (7⅛ * 7⅛)∕-F⅛] where is the partial pressure of ethane in the chamber due to scattering from a saturated surface. Both 7⅛ and J⅛ are corrected for a small effusive component of the flux (16). This technique is particularly useful in measuring the adsorption probability as a function of time or coverage. The surface temperature was maintained constant at approximately 77 K for all the measurements reported here. The experiments were performed with beam energies E⅛ ranging from approximately 2.2 kcal/mol to 10 kcal/mol, and with incident angles ⅞ ranging from 0° to 45°. The incident energy and angle for each experiment are summarized in Table 1. For normal kinetic energies E,cos^^ greater than approximately 8 kcal/mol, there is a non-zero component of direct dissociative chemisorption in the adsorp­ tion at 77 K (16). As shown in Table 1, experiment y has a normal kinetic energy of approximately 8.5 kcal/mol and hence has a small component of direct dissociative chemisorption. The initial probability of dissociative chemisorption under these conditions is approximately 0.03, and this has been ignored. We can also ignore desorption of the molecularly adsorbed ethane, which only occurs at temperatures above 100 K (17). The Monte-Carlo simulations are carried out as follows. The surface is modelled by a square lattice of size 400x400. Each site is allowed to be in one of three possible states: the site can be vacant; the site can be occupied by a molecularly adsorbed particle; or the site can be occupied by a molecularly adsorbed particle and a particle trapped in the precursor state. A site on the lattice is selected at random. If it is occupied by a trapped particle, then the particle is allowed either to hop to a nearest-neighbor site or to desorb. The choice is determined as follows. The trapped particle has a probability of 1/5 of attempting to move in each of five directions: the four lattice directions in the plane of the surface and the direction into the gas phase. If one of the lattice directions is chosen, the particle successfully moves to the nearest-neighbor site A-6 in that direction with a probability of p^. If the direction into the gas phase is chosen, the particle successfully desorbs with a probability of p^. If the site chosen is vacant, or is occupied by only a molecularly adsorbed particle, then a particle from the gas phase impinges upon it with a probability ofpγ. If the site chosen is vacant, then the impinging particle molecularly adsorbs with a probability po. If the site chosen is occupied by a molecularly adsorbed particle, then the impinging particle is trapped with a probability pi. Upon completion of this procedure, the time is increased by one Monte-Carlo step, another site is selected at random, and the procedure is repeated. The quantity measured in the reflectivity experiments is the partial pressure of ethane as a function of time. From the Monte-Carlo simulations, we obtain the values of τ⅛ the number of particles impinging upon the lattice, and τ⅛r the number of particles "reflecting" from the lattice in each Monte-Carlo step/site. The quantity includes both the particles that impinge upon and fail to be adsorbed molecularly or trapped, and the previously trapped particles that desorb from the lattice. The quantity that is obtained from the reflectivity experiments is (τ⅛ — n,-)∕7⅛. It should be noted that this is equal to the quantity τ⅛∕τ⅛, where is the number of particles molecularly adsorbing per Monte-Carlo step/site, only when the number of trapped ethane molecules in the second layer at any instant of time is vanishingly small. There are five parameters in the simulations, namely po, Pi, Pm, pj and p∕. The value of po used in the simulations is obtained from the zero-coverage limit of the experimentally measured probability of molecular adsorption (16). This is the probability with which an ethane molecule accommodates and adsorbs molecularly on the bare Ir(ll0)-(lx2) surface. At a particular surface tempera­ ture, it is strongly dependent on the energy of the incident beam of ethane. For the results that are presented here, the beam energy varies from 2.2 kcal/mol to ^- 10 kcal/mol, and the corresponding values of po vary from 0.95 to — 0.45 A-7 at a surface temperature of 77 K. The value of pi, the probability of accommo­ dation of an ethane molecule on top of a physically adsorbed layer of ethane, is clearly larger than po due to a better match of masses, facilitating momentum transfer, and the existence of a softer potential than in the case of a collision with the bare iridium surface (18-20). In principle, both po and pi are depen­ dent on the beam energy. On the other hand, in our model p∏, and pj must not be dependent on the beam energy because the trapped particle is assumed to be completely accommodated thermally to the surface. In the experiments the surface temperature is kept constant at approximately 77 K. Hence, for the simulations to be consistent physically, it is necessary that the same values of p^, and pj successfully descπhe the experimental results at all beam energies. We choose p^, to be unity, and vary the ratio of pj∕pm- As a result, we have set the time-scale of the simulation to be the time-scale for a trapped ethane molecule to hop Hom one site to the next. The parameter p/ is the probability, in each Monte-Carlo step, that a a gas-phase molecule impinges upon each site. This parameter can, in principle, be determined from an accurate measurement of the flux of ethane in the molecular beam used in the experiments. If each Monte- Carlo step corresponds to a real time of then the incident Hux would be p∕∕τ*,- site^⅛-ι. A crude estimate of this dux in the experimental measurements is 0.1 σ^-*s^^ι, where <r is the area occupied by one adsorbed ethane molecule when the surface is covered by one monolayer of ethane at 77 K. This is because the experiments take approximately 10 s, and, as shown below, the final fractional coverages are only slightly greater than one monolayer. In the simulations, how­ ever, we used a range of values of py, and chose the value that best fits the experimental results, cf. Table 1. A-8 3. Results and Discussion Before we compare the simulations with the experimental results, we dis­ cuss the simulation in the limiting case where py is set equal to zero, i.e. the limit of infinitesimally small flux. In this case the simulation cannot be done as we have discussed above. This is because with p/ = 0, the procedure discussed above will not allow the number of particles on the surface to increase. When the flux is infinitesimally small then any particle that impinges on the surface will have already adsorbed molecularly or desorbed before the next particle im­ pinges on the surface. Hence, particles will have to be introduced one at a time onto the surface. Then the time in the simulations is not measured as before. In this case, the time is linearly proportional to the number of particles that are introduced onto the surface, regardless of the number of hops that each trapped particle takes before it desorbs or adsorbs. This is in contrast to the simulation procedure for the case in which p/ is not zero, where the time is increased by one Monte-Carlo step each time an attempt is made to hop, desorb or trap a particle. The simulations for p/ = 0 are done as follows. First, we determine the probability of molecular adsorption for a fixed coverage on the surface, and then we increase the coverage on the surface. The probability of molecular adsorp­ tion is determined by introducing test particles one at a time, and counting the fraction that adsorbs molecularly. If a test particle impinges on a vacant site, then molecular adsorption occurs with a probability of po. If the test particle impinges on a site which already has a molecularly adsorbed particle, then it traps with a probability of pi- A trapped particle is allowed to attempt a move in one of the four lattice directions or in the direction into the gas phase. Each of these directions is picked with a probability of 1/5. An attempted hop to a nearest-neighbor site is successful with a probability of p^,. If the test particle attempts to move into the gas phase, it desorbs with a probability of pj. A test particle is allowed to migrate on the lattice by nearest-neighbor hops until it A-9 either desorbs or reaches a vacant lattice site whereupon it immediately adsorbs molecularly. Then the next test particle is introduced. In order to obtain the probability of molecular adsorption at a constant coverage the configuration on the lattice is not upoated even if a test particle molecularly adsorbs. For each datum point we introduce 2000 test particles. The configuration on the lattice is updated as follows. A certain number of test particles are introduced, again one at a time. However, now the positions of the test particles which adsorb molecu­ larly are recorded. A test particle which adsorbs molecularly into a site at which an earlier test particle had adsorbed is not counted. After TV particles have been introduced, the configuration is updated by using the recorded positions of the molecularly adsorbed test particles. Time is increased by 7V∕FT^, where F is the flux from the gas phase in units of site"*s^ι and L? is the number of sites on the lattice. For our simulations, we used JV = 1600, so that we obtain the adsorp­ tion probabilities for every fractional surface coverage increment of 0.01. The procedure is then repeated with the new configuration. It should be noted that the time increment is not dependent upon the actual number of migration steps that are taken by the test particles, since the time for each migration step is in­ finitesimally small compared to that between subsequent particle impingements on the surface when py — 0. The physical situation to which simulations with py = 0 can be applied is when the rate at which particles from the gas phase impinge upon the surface is much smaller than l∕*r, where τ is the lifetime of a trapped particle. When the fractional coverage of molecularly adsorbed particles is not high, a trapped particle can find a vacant site in a time interval τ* such that l∕τ* is much smaller than the rate at which particles impinge on the surface. In this case the quantity (τ⅛ — τt,-)∕τ⅛ obtained from the simulations in which py is not zero would be equal to the fraction of molecularly adsorbed test particles in the simulations in which py = 0. In the simulations using py = 0, the lifetime of the trapped particles is A-10 infinitesimally small compared to the time between which particles impinge on the surface from the gas phase. Therefore, there is no accumulation of trapped particles. This means that the simulations in which p/ ÷= 0 model the situation when a second adsorbed layer is not allowed. As a result, by comparing such simulations with experimental results it is possible to infer the accumulation of a second adsorbed layer. Since both desorption and migration occur infinitely fast on the timescale of the particle flux onto the surface, it is not possible to establish, via comparison with experimental results, the absolute rates of desorption and migration. This, however, can be done with the simulations in which p/ is not zero. The results of the reflectivity measurements are plotted as crosses in Figs. 1 and 2. The ordinate has been labelled as the probability of adsorption although for the experimental data it really corresponds to the quantity (τ⅛ — τι,-)∕τ⅛. As mentioned earlier, this is equal to 7⅛∕τ⅛ only when the incident flux is extremely low. The abscissa is the time in units of 0.1 s. The experiments were not all performed with the same value of the incident angle 0,. Thus, even though the beam intensities are the same for all the experiments, the incident fluxes are not. In order to correct for this we have scaled the time by cos⅞. Although it would be interesting to perform the experiment at lower surface temperatures (this would give us an independent estimate of the difference between the des­ orption and migration barriers discussed below), we were not able to cool below the temperature of liquid nitrogen. However, it should be noted that for this par­ ticular system, the temperature of 77 K is optimal. At higher temperatures the fractional coverage of the second layer and hence the 'tail' in the experi­ mental data would not be so large, and at lower temperatures it might become necessary to consider the accumulation of more than two layers of ethane on the surface. A smaller 'tail' would lower the accuracy of the Monte-Carlo analysis, and the accumulation of multilayers would render it necessary to introduce ad­ A-11 ditional parameters to describe the third or fourth layers and hence complicate the simulations. The parameters that were fitted to the data are pi, pj∕p^, and py∕pt∏. As explained earlier, po is obtained directly from the experimentally measured adsorption probability at zero coverage. The parameters which best fit the ex­ perimental data are summarized in Table 1, where each row refers to the corre­ spondingly labelled curves in Figs. 1 and 2. The results of simulations in which py = 0 are shown by the lines in Fig. 1, while the results of simulations in which py is not zero are shown by the circles in Fig. 2. Since the second layer does not accumulate until the first layer is close to saturation, the optimal values of pi and pj/pfn can be determined from simulations in which p/ is zero. The values of pi used for each beam energy are shown in Table 1. It should be noted that these values are all rather close to unity, although there is a dependence on the beam energy as would be expected. Having determined the values for pi and P<*∕Pτ∏ that best ht the experimental data except for values of the coverage close to saturation of the first layer, the same values of pi and p^∕p∏, are used in simulations in which py is varied. In this systematic way, we have determined that the value of p<^∕pτn is 0.005 and the value of py is 6 × 10"^. Note that the relationship between the incident hux and py is given by JF — py∕τ,-. Thus, the simulations in which py=0 are simulations of experiments in which the incident Hux is infinitesimally small on the time scale for migration or des­ orption. These simulations are sufficiently sensitive that significant deviations between the experimental data and the simulation results are observed if pa/pm is set less than 0.001 or greater than 0.01. Similarly, deviations between the experimental 'tail' and the simulation results are apparent if py is set greater than 1 × 10*4 or less than 1 × 10"^. From Fig. 1 it can be inferred that at 77 K a second layer of ethane accumulates on the surface (at least at coverages where the first layer is nearly saturated). This is indicated by the 'tail' in the A-12 experimental data which is not seen in the simulation results in which py = 0. It is possible that such a 'tail' can result from an equilibrium between the first monolayer of ethane and the gas phase rather than the second layer with the gas phase. However, the desorption barrier for a molecularly adsorbed ethane molecule Hom the Ir(ll0)-(lx2) surface is approximately 7.5 kcal/mol (17). As will be seen below, this 'tail' indicates equilibration between the gas phase and a layer of adsorbed ethane with a desorption barrier of 4.5 kcal/mol. Hence, the 'tail' in the experimental data is much too long to be accounted for by desorption Hom the first monolayer. From Fig. 2, it can be seen that by simply allowing p/ to be non-zero, the simulations can produce the 'tail' observed in the reflectivity experiments. The accompanying accumulation of particles in the second layer in these simulations is shown in Fig. 3. The equilibrium second-layer coverage for the simulation parameters used is approximately 0.057. By comparing the experimental data with the simulations in which p/ is not zero, it is possible to determine the real time that corresponds to each Monte-Carlo step/site. In many simulations, it is convenient to use f*^*, the reciprocal of the frequency of the frustrated translational degree of freedom par­ allel to the surface, as the time scale so that each Monte-Carlo step corresponds to an attempted hop. However, we have set p^, to be unity so that the time scale in these simulations is not the time scale of the frustrated translational motion. Rather, it is the time scale at which successful hops occur. Therefore, τγ can be much larger than y"* of which the latter has a typical value of approximately 10^*3s. The rate of desorption of a trapped particle is p^/5 per Monte-Carlo step. The factor of 1/5 arises from allowing a trapped particle to attempt to move with equal probability in each of five directions, as discussed above; and desorption is allowed in only one of these directions. Hence, the rate of des­ orption from the second layer of ethane is given by = p½∕5r,-. Similarly, as mentioned earlier, the flux incident upon the surface is given by py/r^. For each A-13 value of the beam energy, the values of the incident flux mid the values of the energy barrier for desorption from the second layer F7j are summarized in Table 1 (21). The simulation results presented in Fig. 2 are from simulations in which py = 6 × 10*s. As mentioned earlier deviations between the experimental 'tali' and the Monte-Carlo 'tail' begin to be apparent for py greater than 1 × 10"^, the simulation results showing a longer 'tail' than is observed experimentally. This analysis indicates that the fluxes used in the reflectivity measurements are constant to approximately 20 percent, which is physically reasonable. This un­ certainty in the value of flux is from the uncertainty in the value of τ*,-. This same 20 percent uncertainty in τ^r implies an uncertainty of approximately 50 cal/mol in the value of F⅛ but, as noted earlier, there is a much greater uncertainty due to the lack of information concerning the preexponential factor for desorption (21). For the simulation results shown in both Figs. 1 and 2, the value of p^∕p∏, is 0.005. This is the ratio of the probability of success of desorption to the probability of success of migration. Hence the difference between the respective energy barriers ½F7 = F⅛ — is approximately 800 cal/mol, assuming that the preexponential factors for these two processes are equal. If the preexponential factor for desorption is larger than that for migration, as might be expected, then FÆ would be somewhat larger than 800 cal/mol (22). Simulations were also performed for pj∕p∏, = 0.01 and pj∕p∏, = 0.001 to estimate the sensitiv­ ity of the ht. These runs allow us to conclude that the uncertainty in &Ë is approximately 200 cal/mol. 4. Conclusions We have analyzed molecular beam reflectivity measurements by Monte- Carlo simulations of a model in which ethane adsorption can occur through two routes: a direct molecular adsorption pathway, and a precursor-mediated adsorption pathway in which the precursor is an ethane molecule trapped above A-14 a previously adsorbed ethane molecule. We have performed simulations for the limiting case in which the dux incident upon the surface is infinitesimally small, and also for cases in which this dux is not zero. It is possible to infer from the discrepancy between the 'tail' of the experimental data and the simulations with indnitesimally small dux that some accumulation of a second layer of ethane occurs. The simulations with non-zero dux allow us to determine the real time corresponding to each Monte-Carlo step in these simulations. Thus, it is possible to determine the rate of desorption of ethane from the second layer and hence the energy barrier for this process. This is approximately 4.5 ±0.3 kcal/mol, and the barrier to migration of trapped ethane molecules is approximately 3.7 ± 0.2 kcal/mol. The saturation coverage for the second layer is approximately 0.06 relative to the saturation coverage of the drst monolayer at 77 K. It is didicult to imagine a simpler and yet more intuitively reasonable model that would give rise to the experimentally observed maximum in the probability of adsorption. The essential ingredient in the model is the physically reasonable assumption that an incident ethane molecule accommodates better if it impinges on a previously adsorbed ethane molecule than if it impinges on the bare irid­ ium surface. It should be noted that the use of a square lattice to simulate the Ir(ll0)-(lx2) surface, which is a rectangular lattice, is acceptable. The probable anisotropy of the Ir(110)-(lx2) surface to diffusion need not be considered be­ cause the hopping of trapped ethane molecules occurs on top of a layer of molecularly adsorbed ethane, which probably presents an isotropic surface to a trapped ethane molecule. The validity of using a square lattice is also indicated by the good agreement obtained. It is gratifying that using values of po obtained from experiments, values of p^ that are consistent with the beam Aux, and assuming (reasonably) that pj 1, it is possible to describe the probability of adsorption for the entire range of surface coverage from zero coverage to saturation using only a single value of pj∕p^,. This single value of pj/p-m also describes the data A-15 obtained using beams of different incident energies, and is physically consistent with the idea that the trapped ethane molecules are completely accommodated thermally to the surface. Acknowledgment: This research was supported by the National Science Founda­ tion under Grant No. CHE-8617826. Acknowledgment is also made to the donors of the Petroleum Research Fund for partial support of this research under Grant No. PRF 19819-AC5-C. A-16 References 1. W.H. Weinberg, in Kinetics of Interface Reactions, Eds., H.J. Kreuzer and M. Grunze, Springer-Verlag, Heidelberg, 1987, p. 94. 2. I. Langmuir, Chem. Rev. 6, 451 (1929). 3. J.B. Taylor and I. Langmuir, Phys. Rev. 44, 423 (1933). 4. J.A. Becker, in Structure and Properties of Solid Surfaces, Eds., R. Gomer and C.S. Smith (Univ. of Chicago Press 1953) p. 459. 5. J.A. Becker and C.D. Hartman, J. Phys. Chem. 57, 157 (1953). 6. G. Ehrlich, J. Phys. Chem. 59, 473 (1955). 7. P. Kisliuk, J. Phys. Chem. Solids 3, 95 (1957); 5, 78 (1958). 8. P.W. Tamm and L.D. Schmidt, J. Chem. Phys. 52, 1150 (1970); 55, 4253 (1971). 9. L.R. Clavenna and L.D. Schmidt, Surf. Sei. 22, 365 (1970). 10. C. Kohrt and R. Gomer, J. Chem. Phys. 52, 3283 (1970). 11. D.A. King and M.G. Wells, Surf. Sei. 23, 120 (1971). 12. D.A. King and M.G. Wells, Proc. Roy. Soc. (Lond.) A339, 245 (1974). 13. E.S. Hood, B.H. Toby and W.H. Weinberg, Phys. Rev. Lett. 55, 2437 (1985). 14. O.M. Becker and A. Ben-Shaul, Phys. Rev. Lett. 61, 2859 (1988). 15. Since the product of the adsorption reaction is physically adsorbed ethane, there can be no "intrinsic" precursor to adsorption in the usual sense on the bare surface (1). 16. C.B. Mullins and W.H. Weinberg, J. Chem. Phys, (submitted). 17. T.S. Wittrig, P.D. Szuromi and W.H. Weinberg, J. Chem. Phys. 76, 3305 (1982). A-17 18. C.B. Mulhns, C.T. Rettner, D.J. Auerbach and W.H. Weinberg, to be published. 19. R.J. Hamers, P.L. Houston and R.P. Merrill, J. Chem. Phys. 88, 6548 (1988). 20. L. Wilzen, S. Andersson and J. Harris, Surf. Sei. 205, 387 (1988). 21. The values of -E⅛ are calculated from Rj = ⅛^°^exp(-R^∕⅛gT), where ⅛⅛°) is assumed to be 10*3 s*!. If is assumed to be lO^ s^∖ the corresponding values of F⅛ would be about 0.4 heal∕mol greater than the values indicated in Table 1. 22. For example, if the ratio of the preexponential factors were equal to 100, then ½F7 would be equal to 1500 cal/mol, and J‰t would be 3.0⅛0.2 kcal/mol. A-18 Table 1 Set Jk A Po a 2.2 0 0.95 b 3.7 45 c 5.2 d Ea Flux 1.00 4.55 0.087 0.90 1.00 4.57 0.077 45 0.83 0.97 4.53 0.102 6.6 0 0.68 0.96 4.52 0.107 c 10.0 45 0.50 0.95 4.53 0.104 f 10.0 22.5 0.45 0.95 4.55 0.085 Values of parameters used in simulations and experiments. The energy of the beam E, is in units of kcal/mol and the incident angle ⅞ is in degrees. The energy barrier for desorption is in units of kcal/mol and the fluxes in units of s"i σ^^* for each beam energy, where <y is the area occupied by a molecularly adsorbed ethane molecule in a saturated first layer on the Ir (llθ)-(lx2) surface at 77 K. The probability of adsorption on the clean surface (in the limit of zero coverage) is po, and the probability of trapping on top of an adsorbed ethane molecule is pi. The value of pa∕p∏,, where pa is the probability of success of an attempted desorption event and p∏, is the probability of success of an attempted hop by a trapped particle, is 0.005 for all the simulation results shown here. A-19 Figure Captions Figure 1. The crosses indicate experimental data, and the lines indicate results from simulations in the zero-Aux limit. The abscissa is time in units of 0.1 s. For each experiment, the time has been multiplied by cos#;, where #; is the angle of incidence in order to normalize the area of the surface upon which ethane molecules are incident. The parameters for each figure can be found in the corresponding rows of Table 1. Figure 2. The crosses indicate experimental data as in Fig. 1. The circles indicate results from simulations in which the value of py, and thus the Aux which is equal to p∕∕τ,-, is not zero. The abscissa for the Monte-Carlo data has been converted from Monte-Carlo steps/site to real time. The conversion factor τ^τ- gives the number of seconds in real time to which each Monte-Carlo step/site corresponds. This, as explained in the text, enables an absolute determination of the desorption rate from the second layer. The units for the abscissa are 0.1 s. Figure 3. Simulations in which p/ is not zero enable a second layer of particles to accumulate. The fractional coverage of this second layer is shown here as a function of time which, as in Fig. 2, has been converted from Monte-Carlo steps/site to real time. Figure 1a A-20 Figure 1b A-21 A-22 ο ω ⅛- en Μ O <U E ∣ τ m e∕ o.ι s 100 150 Figure 1d 200 250 A-23 jim e∕ o.ι s Figure 1e A-24 lim e/ o.ι s Figure 1f A-25 Tim e/ o. ι s Figure 2a A 26 A-27 ο -Q <N ω ⅛Z3 Ο) LL Μ O <u Ε Tim e/ o. ι s Figure 2c A-28 ×)O li m e/ o. ι s 100 150 Figure 2d 200 250 A-29 Figu re 2e A-30 Tim e/ o. ι s Figure 2f A-31 Tim e/ o. ι s 100 150 Figure 3a 200 250 A-32 A-33 ο u^) CM Ο ο CM J2ι CO ω ⅛- Ζ3 σ) Ο ίθ Μ Ο Ε ο ο ο L∏ ί'Ο 80Ό 90Ό -ίΟ'Ο ΖΟ'Ο auoμ^ paddojj^ ;ο s6ojaλθQ [ouo'joojj A-34 A-35 A-36 ο ο C∖! Φ CO Φ tn ,, L^0 800 90 0 ίΟΌ 20Ό auom3 peddojj^ j.o 96ojaΛθQ )ouo∣^3θjj . Φ † 0 Tim e/ o. ι s Figure 3f A-37 B-1 APPENDι× B DESORPTiON AND TRAPPiNG OF ARGON AT A 2H-W(100) SURFACE AND A TEST OF THE APPUCABiUTY OF DETAtLED BALANCE TO A NONEQUiUBRiUM SYSTEM [The text of Appendix B consists of an articte coauthored with C. T. Rettner and E. K. Schweizer, which has appeared in J. Cham. Rhys. 90, 3800 (1989).] B-2 Desorption and trapping of argon at a 2H-W(100) surface and a test of the appiicabiiity of detaiied baia∩ce to a ∩onequiiibrium system C. T. Refiner, E. K. Schweizer, and C. B. Mullins*' ∕½,t∕.4∕f∏cd'f∕! Afjforc∕ι Center. ^33∕⅛07. /ere. Ca∕{∕or∏ic 96720-6099 (Received 3 October 1988; accepted 8 December 1988) Molecular beam techniques have been used to probe the dynamics of the trapping and trapptng-desorptioπ of Ar at a hydrogen-saturated W( 100) surface. Trapping probabilities have been measured as a functton of mcideπce energy F,, and angle f?, for a surface temperature Γ, of 85 K. We find that this probability scales approximately with F, cos rather than F, or the so-called "normal energy" F, cos^ P,. Trapping probabilities approach unày for low energies, falling to 0.5 and 0.05 for F, cos —30 and 100 meV, respectively. The time-of-flight distributions of scattered Ar are clearly bimodal in many cases, having both direct-inelastic and trapping-desorption components. The latter component has been characterized over a wide range of conditions to provide information on the desorption of Ar from this surface. We 6nd that desorbing species emerge with a near-cosine angular distribution for 7*, ≈ 85 K. However, these distributions become increasingly noncosine as 7*, is raised, becoming substantially broader than cosine. In addition, at the lowest temperature employed ( — 85 K), the velocity distributions of the desorbing atoms are well described by Maxwell-Boltzmann distributions characteristic of the surface temperature. At higher temperatures, these distributions are still approximately Boltzmann, but the characteristic temperature falls below 7*,. The "lag" between this effective temperature and 7*, increases with 7*, and is most pronounced for atoms desorbing at angles close to the normal. We show that the desorption results are very close to the predictions of a model in which angular and velocity distributions for desorption are synthesized by applying detailed balance arguments to the trapping data. Similarly the trapping results are close to trapping curves extracted from the desorption data. t. INTRODUCTION Trapping and desorption are two key steps in gas-sur­ face chemistry and are arguably two of the most elemen­ tary. ' The dynamics of the trapping process in particular would appear transparent. Here a species must simply lose sufficient energy in colliding with the surface to be unable to escape the gas-surface potential. Desorption is the reverse of this, occurring when the gas-surface coordinate accumu­ lates energy in excess of the well depth. In reality, the dy­ namics of such systems is deceptively complex. Current de­ scriptions of the trapping process are far from adequate, and "in spite of being investigated both experimentally and theo­ retically for almost a century, the mechanism of desorption is still not understood in detail."^ In the case of trapping, the dynamics are complicated by the fact that while the initial step requires only loss of mo­ mentum normal to the surface, parallel momentum must also be lost before trapping is assured. This is often a relative­ ly slow process/ so that the dependence of the trapping probability on incidence angle, for example, is difficult to predict * This is true even in the absence of appreciable chemical interactions, as for the adsorption of the rare gases on smooth metal surfaces. Again, modeling of the energy transfer dynamics strictly requires accurate knowledge of the full hyperpotential surface of the system. Approximate 3800 J.ChemPhys.90{7).lAprι)1989 methods can be of value but rely heavily on experimental input to set parameters. Thus cube models require knowl­ edge of the elective cube mass, but since this mass may vary with collision energy, they provide little more than a consis­ tency check for measurements. In fact there have been re­ markably few detailed experimental studies to provide a ba­ sis for such modeling. Thus while some aspect of rare gas scattering has received considerable attention, yielding de­ tailed knowledge of inelastic scattering dynamics/*" the trapping process has rarely been addressed specifically. The few exceptions include the extraction of thermally averaged trapping probabilities from accommodation coefficients," and a small number of measurements for Ar on Pt( 111)," and Ir( 110).'* Relatively detailed measurements ⅛αue been reported recently for the trapping of Ne on Ru (001 ) " Al­ though these results are of great fundamental interest, they may not be typical as the low mass and weak binding of Ne serve to restrict trapping to very low energies, where the dynamics are strongly affected by quantum effects. The primary obstacle to developing a detailed theoreti­ cal treatment of thermal desorption is the rare nature of the relevant events. In principle, one would like to picture the entire process by which desorbing species acquire energies of many times ⅛ 7* before escaping into the vacuum. However, this may require, say, 10'° vibrations in the gas-surface well. If initial conditions are chosen such that desorption is rela­ tively prompt, it is difficult to be sure that the chosen trajec­ tories correspond to a representative sample. This makes 0021-9606/89/073800-14302 10 @ 1989 American institute of Pħys,cs B-3 "brute force" computation impractical, requiring instead the appiicatton of speciat techniques for deaitng with infre­ quent events? '6 Again, comparison with experiment is vital. Of the detailed studies reported to date, the majority have concerned recombtnative desorption, where the desorbing species is the product of a surface reaction? Here the dynam­ ics appear to be dominated by a poteπtiai barrier which serves, for exampie. to focus the desorbing flux ciose to the surface normai. fπ contrast, the desorption of Xe from Pt ( t i t ) is found to fo∏ow a cosine distribution, " white Ar desorbs from this surface into a shit broader distribution. " Faced with the paucity of data and the difficulties in modehng these two processes, tt is reasonabfe to exploit their complementary nature by correlating dynamical results for the two. At the most basic level, microscopic reversibility allows desorption to be studied computationally by follow­ ing trajectories of species which trap?"* thereby circum­ venting the sampling problems encountered in the forward direction. More generally, for systems at equilibrium, the principles of detailed balance and/or reciprocity can be used to rigorously characterize the desorption process given knowledge of only the trapping dynamics, and vice versa? '*^2* A, will be discussed later, these principles are based on time-reversal invariance and serve to maintain the steady properties of equilibrated systems. However, it is not immediately obvious how to treat nonequilibrium systems. While it is reasonable to imagine that detailed balancing will not lead to useful predictions for highly nonequilibrium sys­ tems or systems that are incapable of reaching equilibrium, detailed balance does appear to hold, at least approximately, for many nonequilibrium systems of interest.^ " In this paper we present a study of the Ar∕2H-W (100) system in which we have measured the dependence of the trapping probability on incidence kinetic energy F, and an­ gle #, and have determined the velocity distributions of de­ sorbing Ar under conditions that facilitate a direct compari­ son of the two processes. Specifically, trapping probabilities have been measured for a surface temperature 7*, of 85, for F, from —29 to 250 meV and for #, = 30*, 45*. and 60*. Time-of-Hight distributions have been obtained for angles from — 25* to 85* from the normal and for 7*, from 85 to 275 K. ( Here the minus sign indicates an angle on the opposite side of the normal to specular. ) The 2H-W( 100) surface is obtained by saturating a clean W( 100) surface with and was chosen for study because, unlike clean W(∣00), it is possible to clearly resolve the direct-inelastic and trappingdesorption scattering components in the same manner as reported previously for the Xe∕Pt(III) system.'? We find that trapping probabilities approach unity for low energies, falling to 0.5 and 0.05 for F, cos 6, — SO and 100 meV, re­ spectively. Desorbing species are found to emerge with a near-cosine angular distribution at 7^, = 8 5 K. However, an­ gular distributions are observed to become increasingly non­ cosine as F, is raised, becoming substantially broader than cosine. Similarly, at the lowest temperatures employed ( — 85 K), the velocity distributions of the desorbing atoms are well-described by Maxwell-Boltzmann distributions characteristic of the surface temperature. At higher tem­ peratures, these distributions are still approximately Max­ wellian, but the characteristic temperature falls below 7*,. It is shown that the desorption results are very close to the predictions of a model in which angular and velocity distri­ butions are synthesized for desorption by applying detailed balance arguments to the trapping data. Similarly the trap­ ping results are close to trapping curves extracted from the desorption data. This nonequilibrium system is thus found to behave in a manner similar to what would be predicted at equilibrium. )l. EXPERIMENTAL A. General The apparatus and techniques have been described in detail elsewhere?^*" but will be summarized here for com­ pleteness. Supersonic beams of Ar are directed at a sample mounted in a UHV chamber on a manipulator which pro­ vides accurate control of the incidence angle and surface temperature in the range —85 to 2500 K. The sample con­ sists of a W( 100) crystal that has been prepared and cleaned using standard techniques. Sharp LEED patterns are ob­ tained and contamination levels are below I % as determined by Auger electron spectroscopy. In addition, He scattering gives a specular peak with a width indistinguishable from the instrumental resolution of 1.6* and close to 100% of a 1.6 eV Ar beam can be accounted for within a 12° wide specular lobe. While the coherent He scattering indicates that the crystal has well-ordered regions, they do not readily provide an indication of the fraction of the surface area that might be "damaged" or rough. The Ar scattering data thus comple­ ments the He measurements by allowing us to account for the entire scattered Hux. The clean crystal is saturated with hydrogen by dosing with — 1000 L provided over a 100 s period from an auxiliary beam. During the dosing period the sample typically cools from 275 to 150 K, after which it is annealed to 275 K. Beam energies are varied by changing the temperature T,*,, of a 75 ∕ιm nozzle and by seeding in He. For the lowest energies, the nozzle is cooled to — 100 K, for which the pres­ sure of Ar behind the nozzle is held at about 500 Torr, to ensure against condensation. For 7*^, > 250 K, Ar pressures of up to 1500 Torr are employed. The insensitivity of results to beam conditions suggests that the level of clusters in these beams is negligible. The beam energies are determined from Hight times between a high-speed chopper and a differential­ ly pumped rotatable mass spectrometer."*" This device is also used to record the time-of-∏ight and angular distribu­ tions of the scattered and desorbing species. Angular distri­ butions are obtained in two basic modes. The distributions of the total scattered and desorbing Hux can be recorded using phase sensitive detection referenced to the chopper and pro­ cessed with a lock-in amplifier. Alternatively, distributions are constructed by analyzing time-of-∏ight spectra for spe­ cific scattering angles in such a way as to select out either the direct-inelastic or trapping-desorption scattering channel. This latter approach also permits velocities to be calculated so that distributions can be obtained, rather than the density distributions that result from the lock-in method. Details of the time-of-flight analysis are given below. J Chem. Phys.. Vol· 90, No. 7, 1 Apriι 1989 B-4 B. Time-of-H)ght analysis Time-of-Hight (TOF) distributions recorded at differ­ ent scattering angies, can be analyzed to yieιd detaifs of the velocity and angular distributions of both the scattered and desorbing Ar. as well as reιative trapping probabilities for different incident beams. The TOF spectra are treated in much the same manner as descπbed previousfy by us.^^ using an approach which follows closely that described byHurstetα∕."'"'' The basic strategy that we employ is as follows. First we prepare the data for Atting by subtracting the background level and the 23 /is ion Hight time, and then normalizing the largest chan­ nel to unity. The resulting array is then fed into a nonlinear least-squares fitting subroutine that Hts a model function to the data by adjusting a number of carefully chosen fitting parameters. Information on the velocity distributions of the scat­ tered molecules is restricted to the determination of just two parameters: y, which is a correlation parameter relating the incoming velocity v, to the outgoing velocity v^, and Oγ which is a width parameter, where our analysis and notation is the same as that of Ref. 10. The beam distribution is first fit to a Hux-weιghted velocity distribution formula of the form: F* (v, )Jv, = exp[ — (y — (I) to yield the stream velocity v, and the width parameter a,, via a nonlinear least squares fitting routine. Then each inci­ dent velocity v, is assumed to yield a distribution of scattered velocities of the form: F*(v,,v, )<Λγ = C∕"i"t^exp[ - ( - v,,)-∕αj]iΛγ . (2) where v∏ = yv, and we are assuming zero surface residence time. We note that even at the lowest surface temperature of 85 K, the surface residence time is believed to be negligible. For example, assuming a binding energy of 0.1 eV^* com­ bined with a preexponential of. say, IO'^ would give a resi­ dence time of only — 2 /rs at this temperature. The normalization parameter in Eq. (2), C"°"", can be written as ___________________________________________________ ) "/(vo + <zj)exp( - vg∕tz^) + (τ"i∕2) αγt⅛(3<z^ + 2v⅞ ) [ I + erf(v„ ∕αγ) ] (3) I---------------------------------------------------------------------------- in order to uncouple the parameters αy and y from the over­ all normalization parameter iW*"". Thus [A""*"" = J^ F* ( v,)<fv, ]. Equation ( 3 ) can be approximated or set to a constant in many cases, providing a considerable decrease in running time without unduly hindering the Siting conver­ gence. This is particularly true for cases where it is possible to provide the Siting routine with good initial guesses for the various parameters. The corresponding parameter in Eq. ( I ), Cis given by a similar expression with v, replacing v,, and α, in place of α, . It will be apparent from Eq. ( 2 ) that the scattered mole­ cules are assumed to have velocities corresponding to a shift­ ed Boltzmann distribution, of the same form as for a super­ sonic expansion, but shifted to a stream velocity of ytι,, and with a width related to an effective temperature by αγ = (2⅛Γ∕m) "2. Here perfect elastic scattering would yield y= 1.0 and αγ = 0.0, while complete accommodation at the surface temperature would yield y = 0.0 and = (2⅛Γ,∕m) "2. The Htting procedure includes a convolu­ tion of F* ( y,, v, ) with the velocity distribution of the inci­ dent beam, F* ( v, ) from Eq. ( I ), so that each incident veloc­ ity also has a different arrival time at the sample. In practice, the Siting routine compares the intensity or signal in each TOF channel with that predicted by the mod­ el. The total TOF, is the sum of the Hight time from the chopper to the surface, = ⅛,∕v, and from the surface to the detector = r*, Λγ — r,°, — . Here ⅛, and are the chopper-surface and surface-detector distances of 10.7 and 10.0 cm, respectively. For each data point, with a given t,„, the model subroutine integrates over time r, setting v, =⅛,∕t, andvγ = ⅛,∕(r,o, — r) . The contribution for each v, is then weighted according to Eq. ( I ) with v = v,. and for each according to Eq. ( 2 ). The result is then compared to the intensity or signal associated with . After stepping through all data points, the Sttmg routine adjusts the various free parameters and repeats the whole process until a satis­ factory St is obtained to the TOF distribution. Our analysis program also includes provision for a con­ volution over the shutter function of the chopper, but this was not found necessary here, as the —5-10 ∕zs FWHM chopper function is negligible compared to the > 300 ∕rs Ar Hight times. In cases where it is needed, convolution over the chopper function is included by means of an "outer loop" over time, where weighting is set for each time step to match the measured chopper function, which is contained in an array in tabulated form. As will be seen later, in most cases bimodal TOF distri­ butions were obtained, corresponding to a combination of direct-inelastic and trapping-desorption channels. This re­ quired a modification of our model function to represent a sum of two components, i.e., to include a contribution from the trapping-desorption channel. Thus we assume that each v, yields two contributions to the total Hux F^,°, : f,<M (f/A, )<fvγ = [ ( * — <y)F*d, ( V,.v, ) + <yf^", (v^) ]iΛγ , (4) where the direct-inelastic and trapping-desorption compo­ nents. F*^, and F*,^, are given by Eq. (2). with relative contri­ butions St by tr. In order for the Siting to converge with the addition of these extra parameters, care must be taken to accurately represent the denominator of Eq. (3). and. if pos­ sible. to carry out preliminary Sts to different sections of the J Chem Phys . Vo) 90. No 7. 1 Apri) 1989 B-5 TOF distribution with reduced numbers of parameters, which are then fed in to the fuff Ht as initial guesses. )tι.RESULTS Time-of-fhght distributions have been recorded for Ar scattering from 2H-W(!00) over a wide range of condi­ tions. These are cieariy bimoda) in many cases, as is iffustrated by Fig. f. This shows spectra recorded for "scattering" in the direction of the surface normal, = 0*. for a surface temperature of 85 K for two different incident beams. The upper spectrum was obtained for F, = 0.085 eV, #, = 60*, and the tower one for F, = 0.180 eV. = 60°. It is apparent that they are composed of a retativety narrow fast peak that is superimposed on the teading edge of a much broader peak. It is also apparent from Fig. t that the TOF of the "fast" peak varies with incident energy. In addition, we atso 6nd that it depends s)ightty on <9, and 7^, and reaches a peak intensity ctose to the specular angle. We therefore assign this feature to direct-inelastic scattering. In contrast, the shape of the broad peak is independent of the incident beam energy and angιes, being sensitive primarily to the surface tempera­ ture. This is illustrated in Fig. 2 where the two distributions given in Fig. I are scaled to show how the "slow" peaks can be superimposed. Consequently we assign this peak to a trapping-desorption component. Qualitatively similar dis­ tributions have been reported once previously in a study of Xe scattering from Pt( 11 ι where the same assignments were made and supported by subsequent modeling.^"*" From Figs. I and 2 it can be seen that the ratio of directineiastic to trapping-desorption peaks changes in favor of the direct-inelastic channel as the incident energy increases. In fact, the effect is even stronger than suggested by these Hgures, as the anguiar distribution of the direct-inelastic Time of Flight (ms) FIG 2. Time-of-Hight distributions for Ar scattering from 2H-W( 100) in the direction of the surface normal for the two different incident beams The data points for the ή*, = 0.083 eV case have been connected to distin­ guish them from the ⅜ =0.180 eV data which has been rescaled to b∏ng the two slow (trapping-desorption) components into coincidence. Note = 0.180 eV case this feature is off-scale. channel narrows with increasing F,. so that the direct-in­ elastic flux scattered in the direction of = 0° is reduced. This is a manifestation of a general trend; the trapping-de­ sorption intensity falls quite sharpfy with increasing F,, indi­ cating that the trapping probabiftty, ∕7(F,, <9,, 7^,), faffs from values ctose to unity at the fowest energies empfoyed to zero as F, exceeds —0.25 eV. We can in fact use the integrat­ ed signaf in the trapping-desorption peak as a measure of the I--------- '--------- 1--------- '---------]— —I-------- - — 1.0 Argon∕2H-W(100) ί­ T. = S5K o.a 0.6 *=30' * 45 .=60' 0.4 - ο. P . . - * 0.2 i * 1 !t 0.0 ' ) Time of Flight (ms) 0.0 FIG. 1 Time-of-Hight distributions for Ar scattering from 2H-W( !00) in the direction of the surface normal for two different incident beam energies of 0 083 and 0.18 eV, with #, = 60*. The surface temperature in both cases was83K. <1 o.l 0.2 Kinetic Energy (eV) ! 0.3 FIG 3. Trapping probabilities of Ar on 2H-W( 100) at 7^, = 85 K for dif- J. Chem. Phys., VoL 90. No, 7, 1 April 1909 B-6 K). This latter measurement in turn agrees with a separate method of estimation of absolute trapping probabilities based on comparing the absolute intensities in the desorp­ tion component with the diffuse scattering from a nominally Hat part of the sample holder aligned at the crystal position. In this case we observe only a single scattering component that peaks at the surface normal, which we take to represent 100% trapping?' e, (degrees from normal) FIG. 4. Angular distributions for the scattering of Ar from 2H-W( 100) at 7*, = 85 K for an incident angle of 60*. These distributions have been taken directly from the mass spectrometer signal, and as such represent the mea­ sured density, rather than dux The lobes close to specular belong to the direct-inelastic channel, while the "background" level at large angles from specular arise from trapping-desorption. Also apparent are sharp features on the direct-inelastic lobe due to Ar diffraction. relative trapping probabilities for different incident beam conditions. This has been done for a range of conditions, and the results are displayed in Fig. 3 for 7*, = 85 K. Trapping probabilities can also be estimated from aπguiar distributions, where this probability is taken as being the fraction of scattered species nor contained in the direct-in­ elastic lobe, as described previously?' Figure 4 displays some representative angular distributions recorded for #, = 60* and 7) = 85 K. Here it is necessary to estimate the form of the out-of-plane distribution for the direct-inelastic lobe We have not measured out-of-plane angular distribu­ tions, but comparison with the Ar∕Pt(HI)'" system sug­ gests that the out-of-plane angular width for the Ar∕2HW(!00) system should be between 1/2 to I times the in­ plane width. In fact, the accuracy of this method increases rapidly as the trapping probability approaches unity, where the Hux contained in the direct-inelastic lobe becomes negli­ gible. The data of Fig. 3 have been normalized to a single point at F, - 29 meV, #, = 60*. 7) = 85 K) = 0.90, ob­ tained assuming that the out-of-plane width is half the in­ plane value. Taking the two widths as equal, lowers this to 0.82, so that the scaling of Fig. 3 should be considered accu­ rate to only — ⅛ 10%. The resulting normalized data of Fig. 3 also agree within experimental uncertainties with absolute trapping estimation at /?( F, = 63 meV, #, = 60*, T, = 85 Analysis of the TOF distributions can also provide in­ formation on the desorption dynamics. In pπnctple. we can obtain the velocity distributions of species desorbing at each final angle and surface temperature. In practice it is not pos­ sible to separate the full trapping-desorption signal from the direct-inelastic component for angles close to specular, where the latter fraction dominates the TOF distributions. This is illustrated in Fig. 5 which shows TOF distributions at different angles for 7*, = 85 K. However, it F possible to separate enough of the distributions to determine that at Γ, = 85 K the shape of the trapping-desorption part of the observed TOF spectra is essentially independent of #/. Thus we Hnd that the slow tails of these spectra (f>400∕ts) are well described by the same functional form that 6ts the full distributions in the region of the surface normal. This distri­ bution in fact is very close to a Boltzmann at the surface temperature, i.e., to Eq. (2) with y = 0.0 and αγ — (2½7^,∕m ) '^. This is illustrated in Fig. 6(a) which shows a TOF distribution for = 60* and 7*, = 85 K. Also shown is a 6t to the data, following the approach described in Sec. FIG 5 Time-of^(fight distributions for Ar scattering from 2H-W( !00) at different scattering angles with respect to the surface normal In each case #, == 60* and 7*, = 85 K. Also displayed for the 60* case is the tail of the distribution multiplied in intensity by a factor of 10. The data for — 25* was ι⅛r. J. Chem. Phys.. Vo). 90. Nc7.1Apπ)1989 B-7 Rettner, Schweizer, and Mutιiπs FIG 7. Angular distributions for Ar atoms desorbing from 2H-W ( 100) ai Γ, = Si K The taro sets of symbols refer to two dιCerent experiments with F, = 63 tneV and with an incidence angle of 60*. however, the sampιe was tiited out of piane by i 3* in the case of the died circles. The solid line repre­ sents the results of a best 6t to a cos* yie)ding a = 0 6 The dashed curve shows a simpie coainc distribution for comparison T!m⅛ of F!⅛h⅜ (ms) FÏG. 6. (⅛) Ηme-of-9ight distribution for Aj scattering from 2H-W( Î00) at = 60* for Æ, = 35 tneV, # = IO* ⅛od 7^,=⅛3K- A⅛o dtsp⅛yed is ⅛ &t to the data where the intensity is assumed to contain a trapping-desorption component corresponding to a Boltzmann distribution appropriate to the surface temperature of 85 K, together with an arbitrary direct-ineias⅛c contribution, (b) Actual 6tied form shown in (a), together with the fitted direct-ineiastic and trapping-desorption contributions Notice that the ⅛t H, which, although not perfect, gives a fairly good represen­ tation of the behavior. This 6t also contains a smaii directineiastic component, as can be seen in Fig. 6(b) which displays the fitted curve and its decomposition into trappingdesorption and direct-ineiastic contributions. The St can be stightiy improved in this case by allowing the 7*, to be a free parameter, but the improvement is almost certainty not sta­ tistical stgndicant. Slightty diH⅛rent best-6t temperatures are obtained for different values. These range from — 75 K at the normal to atmost 90 K at = 80*, but again it is not obvious that this variation is sigπiScaπt compared to esti­ mated uncertainties of — ⅛ 10 K (due primarily to ambigu­ ities in subtracting the direct-inetastic component, which vanes drasticatty with #γ). The trapping-desorption components of theTOF distri­ butions at different 6/s can be integrated to provide a mea­ sure of the totat desorbing ftux as a function of angle. Where overlap with the direct-inelastic intensity is severe this inte­ gral is estimated by fitting to the slow tails, away from the direct-inelastic contribution Figure 7 displays the resulting angular distribution for Ar desorbing from 2H-W( (00) at 7*, = 85 K. The data have been fit to a form F (0γ) <χ cos" for which we obtain optimum agreement with rt — 0.6. This curve is given as the solid line in Fig. 7, while the dashed line displays a cos'"#, curve for comparison. A qualitatively similar broader-than-cosine angular distribution has been reported previously by Hurst ft a/. fot⅝Ar desorbing from Ft(lll) at H0K. FIG. 8. Angular distributions for Ar atoms desorbing from 2H-W ( !0O) at Γ, — igθ K. Here F, = 90 meV. the m-piane incidence angie was bθ*. and thesampiewas tiited out of piane by i3*. giving an# = 6i*. Thesohd i'ne represents the resuit of a best 6t to a cos" #^. yιeidmg π = 0 3 The dashed curve shows a simpie cosine distribution lor comparison J Chem Phys.. Vof. 90. No 7. 1 Aprii 1989 B-8 FIG 9. Angu⅛rd⅛stπbutions for Ar atoms desorbing from 2H-W(!00) at 7^, = 273 K. Here f, ⅛= 90 meV, the in-piane incidence angle was 60*. and the sampte was tilted out of p⅛t∖e by tS*. giΑng an 6, = bΓ. The dashed Figures 8 and 9 show corresponding piots for 7*, = 180 and 273 K. It is apparent that the anguiar distributions of desorbing Ar deviate even more noticeabiy from a cosine at these higher temperatures. A best-tit of a cos" to the data for 180 K yieids π = 0.3, but it is cieariy questionabie whether such a form is reasonabie. No attempt was made to ht such a form in the 273 K case for which the distribution appears to peak at an angie other than zero. These distribu­ tions were obtained without aliowing for possibie variations in the velocity distributions of the desorbing species, which cou!d serve to cause shght distortions in the forms. At higher surface temperatures, the desorption TOF"s become faster and are harder to distinguish from the directinelastic channel· Thus bimodai TOFs art no ionger ob­ served even at 180 K While it is possibie to cieariy separate and unambiguousiy ht the TOF^s obtained under such con­ ditions for ang)es far from specuiar, htting becomes increasingiy difhcuit as the specuiar is approached This so called "abasing" has been encountered and described previously for the case of N, scattering from poiycrystalliπe tungsten.^ Figure iθ disρiays TOF data for three surface temperatures for f7χ = — 25*. Here Æ, = 35 meV and #, = 60*. so that for these data is 85* from specuiar. minimizing the directineiastic component. While it is ciear that the corresponding veiocity distributions become substantially faster on raising 7",, they do not in fact "keep pace" with Γ,. Although the data are weii described by Boitzmann distributions, the ef­ fective temperatures 7*,, obtained as a best ht fall further and further beiow 7^, as this temperature is raised. Thus the re­ suits of Fig. i0for7*, =85, 160, and 273 K, are described by best ⅞r values of 80, 130, and 182 K, respectiveiy. These hts are of comparable quaiity to the sotid curves in Fig. iθ, aithough these actually refer to a modei based on detaiied bai- FIG !0. Time-of-Hight distributions for Ar scattering from 2H-W( 100 ) at different surface temperatures, for Æ, = 33 meV, #, = 60* and 6, = — 23*. i.e., with the detection axis on the opposite side of the norma) to the specu­ iar. The soιid curves refer to a ht to a mode[ based on detailed balancing, as presented in Sec. IV ance arguments presented in Sec IV C 3. Similar results are obtained at other final angles. Thus for 6, = 60* we find for T, = 85, 200, 230, and 273 K. ⅞, = 90. 150, 170, and 192 K, respectively. Again, there is a substantial and qualitative­ ly similar "lag" between 7⅛ and Γ, at higher temperatures. J Chem Phys . Vol. 90. No 7. 1 April 1989 B-9 Renner. Schweizer, and Mu!iiπs Trapping of argon at a surface I------------- ------------- Γ In this case (?, — 10°, giving 50° between (7χ and the specuιar. These results are summarized in Fig. 1 1, together with data for 5/ = 80*. This figure atso shows a iine corresponding to full accommodation, with ‰ = 7**,. It iscιearthat devi­ ates )ess from Γ, as (7^ increases. >- °'S )V. DISCUSSION -g °-6 1.0 Argoπ∕2H-W(100) T. = 85K * = 50* ' = 45* a = 60* A. Trapping va incidence energy and ang)e There have been remarkably few reported measure­ ments of /?(F,, Γ,). We recently reported such data for the N,∕W( i00) system, where a surprising insensitivity to was observed.^ As with that system, we find that the trapping probability of Ar on the H-saturated W( ι00) sur­ face falls with increasing F, (see Fig. 3). Such behavior is intuiti vefy reasonable, since as F, is raised a larger and larger fraction of the collision energy must be fost for trapping to occur. The actual form of this faff-off is fess obvious. Let us first consider a highfy simpiified picture of the trapping dy­ namics, appropriate to a cube mode)? For trapping to occur, there must of course be a potentiaf weff, say of effective depth c. Then we can say that trapping will occur when the fraction of the incidence energy that is fost to translation exceeds F,∕(F, + c). For a stationary cube (équivalent to a surface temperature = 0 K), we woufd then predict a trapping prob­ ability of unity for alf energies up to some critical ("nor­ mal" ) energy F„ above which it will fall to zero, giving a step function form to/?( F,, T, =0). For a hard cube model we obtain _ ⅛(m∕Λi)— [l-(m∕,M)]' where AT and ∕π are the cube and incident atom masses, re­ spectively. Any thermal motion of the cube will serve to smooth this step function. This motion may be included in the modeling by adding a component to the collision energy consistent with a one-dimensional Maxwellian distribution of velocities, but weighted by the actual relative velocity." Such a distribution has a most probable value close to zero velocity, with a roughly equal probability of 6nding faster and slower velocities. The result is that in this simple picture, we expect a trapping probability of about 0.5 at F, = F,, going smoothly to ! at lower energies and to zero at higher energies. This behavior is reminiscent of the density of states form predicted by the Fermi function, where F, is equivalent to the Fermi energy. In fact we find that /?( F,, 7*, ) values obtained from cube models can be well described by a expression similar to a Fermi function, but with in place of Γ.** That is we suggest that it may be useful to describe /?( F, , 7^, ) by the form F(F,, Γ,) = ____________ 1____________ exp[(F, -F,)∕x7ψ]] + 1 (6a) ________ 1________ exp(xF, — F, ) + 1 (6b) J 0.⅛ n. o. o *^ 0.2 0.0 0.0 0.1 0.2 E∣ eoa^St (eV) EfG 12. Trapping probabilities of Ar on 2H-W( 100) at 7*, = 83 K for CO$i It may be apparent that such a form can be used to de­ scribe the trapping vs energy curves of Fig. 3, for any fixed value of #,. However, the data clearly do not scale with F,, so that different curves are required for each (7,. In the limit where parallel momentum is conserved, we might reasona­ bly expect results to scale as the so-called "normal energy," F„ — F, cos^ #,. This is not the case here, as is apparent in Fig. 12, which shows the data of Fig. 3 replotted against F„ Comparison of Figs. 3 and 12 suggest that the true scaling parameter is somewhere between total and normal energy. or equivalently, F(F,, Γ,) = treating F^, x, and y as adjustable parameters. A rather simi­ lar form has been employed by Hurst ere/. " in modeling the FIG !3 Trapping probabilities of Ar on 2H-W(100) at Γ, = 85 K for trapping-desorption of Ar on Pt( 111 ). J. Chem Pħys.. Vo!. 90, No. 7, 1 Apπ! 1989 B-10 We 6nd empirically that the data scale rather well with cos as is apparent from Fig. [3. This suggests a rather corrugated interaction, which is supported by our observa­ tions of scattering rainbows in this system for example for Æ, —100 meV and #, = 30*, as will be described elsewhere." Also shown in Fig. 13 is the result of a best 8t to the data using Eq . 6 ( b ) with F, cos f7,, in place of Æ,. Here we obtain X = 78eV^' andy = 2.65, giving ⅜ = 34 tneV. This is con­ sistent with Eq. 5 with A/ equal to about 2 W atoms, taking r — 80 meV." We stress that the intention here is not to ex­ tract accurate dynamical parameters, so much as to gain a qualitative understanding of the trapping vs energy behav­ ior. B. Desorption We have shown that the TOF distributions of Ar scat­ tering from the hydrogen-saturated W( 100) surface contain a clearly resolvable trapping-desorption component. As such, the form and spatial distribution of this component reflect the velocity and angular distributions of Ar desorbing from this surface. Analysts of data obtained over a wide range of conditions leads us to several important conclu­ sions. At the lowest temperatures employed ( —85 K), the velocity distributions of the desorbing atoms are well de­ scribed by Maxwell-Boltzmann distributions characteristic of the surface temperature. Moreover, these species emerge with an angular distribution in quite close agreement with the Knudsen Cosine Law (see Fig. 7). At higher tempera­ tures, velocity distributions are still approximately Maxwel­ lian in form, but the characteristic temperature falls below Γ,. The lag between 7⅛ and 7*, increases with 7*, and is most pronounced for desorbing angles close to the normal (Fig. H). Similarly, angular distributions become increasingly noncosine as the surface temperature is raised. Considering 6rst the velocity distributions, we note that there have been surprisingly few previous studies of this kind Although velocity distributions of desorbing species have been reported for a number of systems, going back to 1971," these pertain mostly to the desorption of reaction products? In these cases strong deviations from Max­ well-Boltzmann distributions are observed, but in most cases in the opposite sense to that reported here. For exam­ ple, the energy of D, desorbing from Cu(100), Cu(H!), and Ni(!ll) surfaces following permeation through the bulk is found to be much higher than 2⅛7*,." The exception here is the desorption of the D,O product of the oxidation of D, on Pt( 111 ), where mean energies equivalent to only 7*,∕2 have been reported." Energies in excess of 2⅛7*, are thought to arise due to the presence of a potential barrier which rises above the large-separation asymptote, or zero level, and pro­ vides the departing species with energy comparable to the barrier height? Conversely, systems with (E^) less than 2kΓ, are unlikely to be associated with such a barrier. The closest previous studies to the present work in­ volved the scattering ofXe" and Ar" from Pt( 111 ). Here Hurst er a/, were able to identify trapping-desorption com­ ponents in the TOF distributions of Xe and Ar scattering from Pt ( 111 ). In the former case, the velocity distribution of Xe desorbing from a 185 K surface was found to ft a Max­ well-Boltzmann with 7⅛ = 185 ⅛ 3 K." In the Ar study, data were obtained for several desorption angles with 7*, — 100 K. Mean energies of between 70% and 90% of 2⅛7*, were observed, increasing with increasing angle from the normal. A casual comparison with the present study sug­ gests that the Ar∕2H-W( 100) and Ar∕Pt( III) systems be­ have similarly. Shortly after the Xe study, Tully reported the results of stochastic trajectory calculations for the desorption of Xe and Ar from Pt( 1II)? which are in fairly good agreement with the limited measurements available. Of particular inter­ est here is that Tully was able to examine the effect of 7*, on the desorption energies. He found that at temperatures somewhat above those employed by Hurst er u∕., these ener­ gies lag behind 2k7*, in a qualitatively similar manner to the results reported here (Fig. 11). This behavior was again ob­ served in calculations by Muhlhausen, Williams, and Tully" for the case of NO desorbing from Ag( 111 ). The dynamical origin for this effect is not immediately obvious. There are at least two distinct ways in which a subtheπnal velocity distribution could result. If we consider those atoms destined to desorb after one further collision with the surface, which we will term "protodesorbers," we expect that the velocity distribution of these species after desorption will reflect both the dynamics of the last collision and their energy distribution prior to this collision. Thus translational cooling might arise if: (i) the energy distribu­ tion of the protodesorbers is out of equilibrium with Γ,, de­ caying faster than exp( — F∕k7*,), or (it) the final collision does not just add a 6xed amount ofnormal momentum to the desorbing species, but a variable amount with a probability that falls off with increasing energy transfer. Effects due to (i) must certainly exist, as the constant loss of energetic species has to preferentially deplete the high energy tail of the distribution of protodesorbers. However, as we will see below, our results are in good agreement with detailed bal­ ance, which strictly applies to a system at equilibrium. Un­ der such conditions, the protodesorbers should indeed be equilibrated to 7*,, since "evaporative cooling" is perfectly matched by "condensation." This suggests that the velocity distributions of desorbing atoms rather reflects (ii), the dy­ namics of the last collision. Here the idea is simply that the Boltzmann term. P(E) =exp( — E∕%7*, ), appropriate to the surface temperature will be replaced by a function based on the product of P(F) with a transition probability func­ tion" that decays with increasing energy gain. The resulting function will then fall off with increasing Æ faster than P(F), and therefore will correspond to a subthermal velocity distri­ bution. At the lowest temperatures, the fall off associated with P(E) must dominate the dynamical term, so that the effective translational temperature approaches 7*,. Finally we note that the extra normal momentum needed for desorp­ tion may be acquired either by the redirection of parallel momentum by surface corrugation or by energy transfer from the substrate. Since the elective surface mass is consid­ erably higher than the departing atoms, direction changing collisions may be more important. There have been more studies of the angular distribu- J. Chem. Phys.. Vo!. 90. No. 7.1 Apn! 1989 B-11 tions of desorbing species? Here the situation is rather simi­ tar to the case of velocity distributions, m that most work appties to reaction products? " In this case, angutar distributions are found to deviate from a cosme by being sharper, rather than broader than this function. Steinrück et u∕. found that H, desorbs from Nt(tll) withacos^ ^distri­ bution?' white the CO; product of the oxidation of CO on Pt( t H ) emerges with a remarkabty peaked cos'*' t), distri­ bution?^ Again the presence of a potentia) barrier is inferred, which serves to "focus" the emerging species along the sur­ face normal, by increasing their normal energy. The desorption of Ar from Pt( t H ), by contrast, was found to yietd a broader-than-cosine distribution? consis­ tent with trajectory studies for this system. This behavior is consistent with the "translational cooling" effects discussed above. The rate of desorption in a given direction must con­ tain a factor corresponding to the rate of leaving the surface, i.e., a velocity factor. Lower velocities along the surface nor­ mal then lead to smaller fluxes. This effect must, of course, "compete" with the phase-space weighting factor of cos which dominates to give a cosine distribution at low Γ, where the perpendicular momentum approaches that of the parallel, at which point the mean energy is close to 2⅛7^,. C. Test of detailed balance 7. Geπe∕τ,∕ A valuable aspect of the current study is that results have been obtained simultaneously for the complementary processes of trapping and desorption These data can there­ fore provide the basis for a test of the extent to which detailed balancing arguments can be applied to relate such processes for a system that is not in strict equilibrium. Before proceed­ ing, it is useful to review the thinking in this area. The principle of detailed balance rigorously applies to systems at equilibrium, for which it relates the (transition) probabilities or rates for all distinguishable processes in such a way that the steady properties of the system are main­ tained. ft may be expressed as" 7*(A-B, C-D) = Γ(B-A. D-C), (7) where 7*( A — B, C — D ) is the probability per unit time that a state A of the system will make a transition to state B, while at the same time C goes to D (in order to conserve energy and momentum). Averaging over an equilibrium distribu­ tion of initial states C, and summing over final states D, leads to the simpler relationship" JV(A-B)=A(B-A), (8) where Λ'( A — B ) is the number of particles in state A making transitions to state B per second. These expressions are based on time-reversal invariance and invariance under reflection of spatial coordinates.** As a result, they do not strictly apply to the gas-surface interface, as the collision represented by the right-hand side of Eq. ( 7 ) does not exist at a surface where the collision represented by the left side exists?" Thus if A is a momentum state of an atom moving towards the surface, the collision might cause a surface atom to move in towards the bulk in a manner given by state D (Alternatively we may view this as creating a phonon with momentum directed away from the surface.) The incoming atom will change to a state B. having momen­ tum directed away from the surface, so that B and D are both moving away from the interface (and each other). In order to collide. B would need to strike a surface that has been reflected about the surface plane, where D (or the phonon) would be moving towards the surface. In more general terms, detailed balance cannot be applied here as the gassurface interface is not invariant under reflection of spatial coordinates. Instead it is reciprocity, which is expressed as Γ(A-B,C-D) = 7*(B^-A^,D,-C,) (9a) A(A-B) = y(B,,-A',) (9b) which may be used to derive the relationships relevant to the gas-surface interface. Here A, is in every way identical to state A, except for a reversal of its momentum, i.e., it is tra­ velling in the opposite direction. Actually two extra key as­ sumptions are needed to permit these relationships to be ap­ plied to the gas-surface interface. First we must assume that the surface is capable of maintaining a canonical distribution appropriate to 7*,, i.e., that the impinging Hux of gas does not disturb this distribution. In addition, we must specifically preclude any sorting of trajectories by a Maxwell's demon­ type entity?" 2 7⅛⅛Mo∕M∕⅛ie ⅛wfwwπ frapp/np and daso∕pM,π Equation 9(b) then expresses a rather basic idea that we might have about a gas-surface system at equilibrium: at equi/ibrium t⅛e/Tux o/ea<rb d⅛rι∏gu⅛⅛αb∕e incident rtate iτ murcbed by on e%tια∕ but oppositey?ux q∕^titαf state ieouing t<⅛e sur∕⅛ce. Thus it can be shown that the cosine distribution of incident species that must apply at equilibrium gives rise to a cosine distribution of particles leaving the surface?" Similar­ ly, a Maxwell-Boltzmann distribution of velocities is de­ manded for species leaving the surface under these condi­ tions. However, note that these constraints apply only to the sum of the scattering channels, so that ι∕ιe md<utdua∕ c⅛αnne⅛q∕^ιne∕asric sconcing and ( trapping- ) desorption may de­ mote ar⅛ιtroπ7y/bom cosine and ι⅜faxιoe∕∕-βo∕tzmonn distri­ butions, prσoided die ot∕ιer cAonne/s bebaue in a comp/ementaty manner. Only if one channel dominates the scattering does it become constrained. For example, the an­ gular distribution for desorption at 7*, must approach a co­ sine as the trapping probability for all species in the Boltz­ mann distribution at this temperature approaches unity. We might consider at this point that it is possible to equate the individual channels of adsorption and desorption, arguing that whatever species are lost to the scattering chan­ nel in trapping, have to be made up for in desorption, in order to preserve the detailed sums. Then we would write'* y^(v„<?,.r„...) =A*(v, = v,,e, = ^,τ,,...). (io ^hce∕aα and∕^,, are the differential flux for desorption and adsorption, respectively, which may depend in addition on. say, surface coverage, etc. However, acceptance of such an expression requires a further assumption that the differential fluxes for the scattering channel must be the same before and after encountering the surface, i.e., the scattering must be- J Chem Phys . Vo!. 90. No7.1Apr,l!989 B-12 have on average in a manner that is indistinguishable from elastic scattering at the specular. Thus it ccκ∕J be that inelas­ tic scattering somehow, on average, slows down or speeds up the scattered species, or focuses them towards or away from the normal. If this occurred, the desorption distributions would need to "compensate" by assuming the complemen­ tary behavior. In fact Eq. ( 10) is likely to be rigorously true, due to the fact that reciprocity may be applied to the scatter­ ing channel alone," permitting Leuthàusser to arrive at an expression equivalent to Eq. ( 10)." Equation (10) can thus be written as ∕⅛,(f.0γ.7*,) o: COS 3 exp( — f∏v *∕2⅛7*, )/?( V = v,,0y = 0,,7*, ), (ID where the cos 0γ and v ^* exp( — mv ^∕2kΓ, ) terms corre­ spond to the incident cosine and ( Hux-weighted ) Boltzmann distribution, respectively. Angular distributions can be ob­ tained simply by integrating over velocities, giving <x cos 0, J v/ exp( — m v?/2%7*,)E(v,,0,,Γ,)<fv, , (12) where the angle brackets signify an average over all desorb­ ing velocities. Thus, for a system at equilibrium we are able to synthesize desorption properties from a knowledge of the trapping probability as a function of angle and energy, and tιice ner∞. We will now test the extent to which these expres­ sions hold for the Ar∕2H-W( 100) system under molecular beam conditions. Finally, we should make a semantic point. We have stat­ ed that detailed balance as such does not strictly apply to the gas-surface interface and that Eqs. ( 10)-( 12) are based on the related principle of reciprocity?" However, the practical application of these equations to the gas-surface interface involves on exercise in detailed balancing, where the differ­ ential fluxes associated with adsorption and desorption art compared or balanced. This may account for the adoption of this term in many previous studies. " " '9'"'" We will there­ fore sacrifice rigor in favor of consistency and refer to the process of comparing our results with the predictions of Eqs. ( 10)-( 12) as one of testing the applicability of detailed bal­ ance, in the sense accepted by the surface science communi­ ty J. ^pp∕∕eaf∕oπ /o A∕*∕2∕≠-Mγ7a%∕ FIG. t4. Construction illustrating the prediction of the energy distribution of desorbing species based on detailed balance. ( a ) Boltzmann distribution corresponding to a (surface I temperature of 300 K. together with the trap­ ping vs energy function estimated to apply to the case of Ar trapping on 2HW ( 100) for F, - 30*. ( b ) The result of multiplying the two curves in I a ) to obtain the velocity distribution predicted for desorption at (i.. assuming de fits to our discrete observations, as discussed in Sec. IV A above. The process by which Eq. (II) is used to predict velocity distributions is illustrated in Fig 14. The top panel displays a Boltzmann at 300 K, corresponding to the V 3 exp( — mv *∕2⅛ Γ, ) term of Eq. (II). and a curve of the trapping function^ v,, 0, = 30*). Thedistribution function for the velocities of those particles that trap for 0, = 30° nd 7*, = 300 K is given by the product of these two curves, which is then the distribution function for desorption, as predicted by Eq. (II) For the purposes of this example, we have taken the function ' While Eqs. ( IO)-( 12) are valid forasystem at equilibri­ um, there is no 0 ∕"roπ reason why they should hold for systems even slightly out of equilibrium. However, "detailed balance" constructions based on these relationships do seem to work, at least approximately, for many systems of interest under conditions comparable to the present study.We are thus encouraged to apply them here. Equations (II) and ( 12) can be used to predict both the velocity and angular distributions of Ar desorbing from this surface, provided that the trapping function /?( v,. 0,, T, ) is known. We are indeed able to estimate this function based on exp(78E, cos 0, — 2.65) -)- I as discussed in Sec. IV A, where E, is in eV. Actually this applies to trapping data obtained at 7*, = 85 K. rather than 300 K, but for now we will neglect this difference. The lower panel of Fig. 14 shows the resulting velocity distribution, which would be predicted for Ar desorbing at 0, = 30°. Also displayed is a curve corresponding to a Maxwell-Boltzmann at 180 K, the "best ht" temperature for this case. This construction has been incorporated into the TOF analysis procedure described in Sec. II, so that the trappingdesorption components can be ht not to a simple MaxwellBoltzmann, but to the product of this function and the trap­ ping function for the particular value of 0/, generated using J. Chem Phys.. Vd. 90. No. 7.1 Apπt 1989 B-13 Eq. (13) with = The Maxwell-Boltzmann tempera­ ture is teft as a free parameter that is optimized by the nonlinear [east-squares fitting program. We find that for = 25°, (i.e., the data of Fig. 10) the fitted temperature ts virtuaity identical to 7*,. in fact the soιid curves in Fig. [0 show the resuiting 6ts obtained for data recorded at 7^, = 273, [60, and 85 K. We obtain best-fits vaiues of ⅞, = 267, !53, and 84 K, which are in remarkab)y good agreement with the actual surface temperatures, ft is also possible to take the opposite approach here: to deduce the trapping function from desorption spectra. In this case our aπaiysis program again Sts TOF spectra to the product of a Maxwell-Boltzmann distribution and a trapping function, but fixes the Boltzmann temperature at 7*,, and Sts the data by optimiz­ ing the parameters of the trapping function. This function is assumed to be of the form given in Eq. 6(b), withx andyas Siting parameters. With this approach, however, there is a substantial uncertainty in the abso)ute trapping values, as the TOF Siting program also Sts a normalization constant, which is coupled to the abso[ute scaling of the trapping func­ tion. Figure 15(a) displays the result of one such St, to the 273 K data of Fig. 10, compared to the results of a St of Eq. 6(b) directly to the trapping data of Fig. 2 for #, = 30*. Con­ sistent with the above, the St only yields the shape of the desorption function, while the overall normalization has been adjusted here for the purposes of presentation. Only the shape of the curve is actually St. Again we note that this comparison is not quite perfect, as the trapping data refer to 7', = 85 K, while the desorption temperature ts 273 K. Nev­ ertheless, the agreement is within experimental uncertain­ ties. Carrying out similar analysts at other angles leads to qualitatively similar results, but with less satisfactory agree­ ment at larger 6γ(6, ) values. Thus for = 60°, the Siting procedure described above with 7*, as a free parameter gives for Γ, = 273, 200, and 85 K, best Sts of 7^,, = 220, 185, and 91 K. Similarly, the trapping function obtained from the 273 trapping-desorption TOF spectrum is in only modest agree­ ment with the measured function, as shown in Fig. 15 ( b ). ft may be that variation of /7 with temperature is greater at grazing angles, ft is clear that a precise test of detailed bal­ ance requires trapping data over a range of temperatures to match desorption data. As a Snal comment on the form of the observed TOF distributions, we have noted that these can always be well described by a Maxwell-Boltzmann distribution, albeit at a temperature other than the surface temperature. This im­ plies that the trapping function must be close to exponential. Thus if /7(2?) = rf exp( — F/7*„) is substituted in Eq. (11), we will obtain ∕^,(F)<χ2sexp(-F∕⅞r) where 1∕7*,, = 1∕7^ + 1/T„, as was recognized by Hurst er a/.ft should also be apparent from Fig. 13, for example that the actual trapping function is indeed likely to be close to expo­ nential. In our opinion, the TOF data are not precise enough to distinguish between Eq 6(b) and an exponential. Continuing this line of reasoning, we may also use the trapping results to predict angular distributions assuming detailed balance to apply Here we can build up distributions by use of Eq. 12 to give the desorbing 8ux at each value of 0γ Figure 16 displays predicted distributions for T, = 100,200, and 300 K, based on Eq. 12 with ^(v,#, ) as given by Eq. Desorption Angla (degrees from normal) Ei(.V) FIG 16. Angular distributions predicted for Ar desorbing from 2HW( !00) st 7* = i00. 200. and 300 K. assuming detatied balance holds for thts system (solid curves). The curves are based on the trappmg function described in the test. Also shown for reference is a simple cosine distπbu tιon ( dashed curve ). J. Chem. Pħys.. Vol. 90. No. 7. 1 April 1989 B-14 Rθttner. Schweizer, and Mullins: Trapping of argon at a surface ( 13). It is seen that these distributions become increasing!y noncosine as 7^, is increased, in good qualitative agreement with direct observations for 7", = 85, 180, and 273 K, shown in Figs 7-9, respectively. Here we have not attempted to best-fit the detailed balance synthesis to the data, feeling that the added complexity of the modeling is not merited at this point. We do note, however, that both the synthesized and measured angular distributions for 7*, — 300 K peak at an­ gles away from the normal. Several further observations are apparent here. We note that it is the#, dependence of the^(v, #,) teπninEq. (14) which is responsible for deviations from a cosine angular distribution. Thus systems for which trapping is a function of the total incidence energy, such as N,∕W( 100)^ would be predicted to always desorb with cosine angular distribu­ tions. At the other extreme, systems in which trapping scales with E, cos^ #,, would display even more pronounced devia­ tions than observed here for the Ar∕2H-W( 100) system, and would be expected to show clear minima at the surface normal. Further, distributions will be broader than cosine where the trapping function falls with increasing energy, but will be narrower for systems involving activated sticking, where adsorption increases with E,. It would appear that the trapping and desorption results obtained here are indeed consistent with detailed balance. That is to say the individual adsorption and desorption processes behave as if the system were at equilibrium. This may not be too surprising when it is appreciated that the surface conditions of the present study may well be indistinguishable from those attainable at equilibrium. In both cases the incident dux is likely to be too small to perturb the canonical distribution of states at the surface or to produce significant Ar coverage. Thus detailed balance is seen here to be more than a simple consequence of the need to maintain the steady properties of a system at equilibrium. That is to say the differential duxes of desorbing species (i.e., into each velocity, angular element, etc.) at equilibrium are not driven by any buildup of particles, ener­ gy, momentum, etc. Only the total yield of desorbing atoms is controlled by the incident dux. The ne½Mwe <⅛∕⅛feπd⅛∕ /faxes in <⅛stwptK7∏ mκMfnrAer⅛e gouemedhy ⅛trπ⅛nc∕),tyefT⅛τ q∕^fAe ud∞r⅛αfe-rur∕0ce rystetn. These properties also independently govern the trapping dynamics. The connec­ tion between adsorption and desorption may therefore be understandable, as the two processes share a common poten­ tial and are intimately related by the fact that the equations of motion are invariant to time reversal. Thus any given tra­ jectory can equally apply to either adsorption or desorption. It is only the relative occurrence of each element of differen­ tial dux that is, in principle, free to vary for a nonequilibrium system. The actual differential duxes in each direction may be thought of as arising from a combination of dynamical and statistical factors. For example in Eq. ( 1 ! ) we might identify the/?( v. #, 7*, ) term as "dynamical," and the cosine and Maxwell-Boltzmann terms as "statistical." The fact that we find that this equation seems to be approximately true even in the absence of the incident Maxwell-Boltzmann suggests that this factor arises naturally at the surface. In­ deed, this is again reasonable, since, as discussed in Sec. IV B, protodesorbers most likely have a Boltzmann distribu­ tion of energies (shifted by the well depth, which sunply introduces an exp( — r∕A7*, ) normalizing factor) prior to their final collision with the surface. V. SUMMARY AND CONCLUSIONS We have shown that Ar traps on a hydrogen-saturated W( 100) surface with a probability that scales roughly with E, cos #, and which falls from values close to 1 at the lowest energies to zero as E, cos #, approaches about 0.2 eV. This scaling law suggests that the interaction potential is quite corrugated. The falloff with increasing energy was shown to be consistent with a simple cube model for a cube mass of close to 2 surface atoms. Time-of-dight distributions of scat­ tered Ar are found to be clearly bimodal in many cases, hav­ ing both direct-inelastic and trapping-desorption compo­ nents; the latter of which has been characterized over a wide range of conditions to provide information on the desorption of Ar from this surface. We have found that desorbing spe­ cies emerge with a near-cosine angular distribution at Γ, = 85 K, but that these distributions become increasingly noncosine as 7*, is raised, becoming substantially broader than cosine. Similarly, velocity distributions of the desorb­ ing atoms are well described by Maxwell-Boltzmann distri­ butions at 7*, = 85 K. At higher temperatures, these distri­ butions are still approximately Boltzmann, but the characteristic temperature falls below 7*,. The lag between this effective temperature and T, increases with Γ, and is most pronounced for desorbing angles close to the normal. Analysis of these results has revealed that the desorption data are very close to the predictions of a model in which angular and velocity distributions are synthesized for de­ sorption by applying detailed balance arguments to the trap­ ping data. Similarly the trapping results are close to trapping curves extracted from the desorption data. Thus we have learned that detailed balance can be used quite reliably to extend trapping and desorption data by permitting one to be derived from the other. Finally, we have speculated that de­ tailed balance is applicable to a nonequilibrium system in this case because desorption and trapping produce comple­ mentary differential Bux distributions at equilibrium [Eq. ( 10) ] rather than being driven to do this by any feedback from one process to the other. ACKNOWLEDGMENTS We are indebted to J. E. Schlaegel for assistance in the maintenance and upkeep of the apparatus and to D. S. Beth­ une, D. J Auerbach, G. M. McClelland, and J. A. Barker for useful discussions. CB.M, would also like to thank IBM for a Predoctoral Fellowship and the NSF for partial support under Grant No CHE8617826. 'J. A Barter and D J Auerbach. Surf. Sei. Rep 4. ! ( 1984) ^G. Comaa and R Dav⅛d, Surf Sei Rep. 5, 145 ( 1985) 'J. C. Tully, Surf. Set ttt. 4⅛1 ( 1981 ). 'M Persson and J. Hams. Surf. Sei tS7. ⅛7 ( 1987) iτy (Academic, New York, !976). "D R. MiHer and R B Subb⅛rao, J Chem. Phys 52. 42$ ( !970). S M Lui. W E. Rodgers, and E. L Knuth, Pwffdι∕!gyo∕*f⅛eM∏i⅛ ∕πfer- J. Chem Phys.. Voi. 90. No 7.1 Apnι 1989 B-15 (1988) Gerτπany.!974bVol.H.pE.8-l. "C W. Muhlhausen. J. A. Sem. J- C. Tully. G E. Becker. and M. J Cardil)o, lsr. J. Chem. 22. 315 ( !982). T Engel and R. Gomer. J Chem. Phys. 52. 5572 ( 1970) Set.130.395 (1983). '^'J. E Hurst. L. Wharton. K. C Janda. and D. J Auerbach. J Chem Phys. 78. 1559 ( 1983). ' ' A. AmM⅛∖. M. J. Cardillo. P. L. Trevor, C Lim. and J C Tully. J. Chem. Phys. 87. 1796 ( 1987). '^W. H. Weinberg and R. P Merrill. J Vac. Sei. Technoi. 8. 7!8 ( 197! ). ' 'J E. Hurst. L. Wharton. K C. Janda. and D J. Auerbach. J. Chem. Phys. 83. 1376 (1985). "H. P Steinruckand R. J. Madix, Surf. Set 185. 36 ( 1987). '*H. Schl,cht!∏g. D Menzel. T. Brunner. W Bren,g, and J C Tully, Phys. Rev Lett. 60. 2315 ( 1988). "*E. K. Grimmelmann, J. C. Tully, and E. Helfand. J. Chem. Phys. 74. 5300 (1981). ' J E Hurst. C A Becker. 3 P Cowin. K C 3anda, L Wharton, and D. J Auerbach. Phys. Rev. Lett. 43. 1 !75 ( 1979). '*R. L. Palmer. J. N Smith. 3r. H Saltburg. and D R O'Keefe. J Chem Phys. S3. 1666 (1970). '**R. L. Palmerand D R. O'Keefe. Appl. Phys Lett !6. 529 ( 1970). -°E. P. Wenaas, J. Chem. Phys. 54. 376 ( !97! ). "!. Kuscer. Surf. Sei. 2S. 225 ( 1971 ). ^*M. J. Cardillo. M. Balooch, and R. E. Stickney. Surf. Set. 50.263 ( 1975). ^U. Lcuthäusser. Z. Phys. B50, 65 (1983). **H P Steinrück, K. D Rendu!ic, and A Winkler. Surf. Sei. 154, 99 ( 1985). *^H. E. Pfhür, C. T Rettner. J. Lee. R. J Madia, and D. J. Auerbach. 3. Chem. Phys. 85, 7432 ( 1986). ι*C T Rettner. L. A. DeLouisc, and D 3. Auerbach. 3 Chem. Phys. 85, 1 131 ( 1986). C T Rettner. H Stein, and E. K Schweizer. J Chem Phys. 89, 3337 ( 1985). '"3 C. Tully, Faraday Discuss. Chem. Soc. 80. 291 ( 1985). "C. T. Rettner. E K. Schweizer. H. Stem, and D. J. Auerbach. Phys. Rev. Lett 6!. 986 ( 1988). '*K. C. 3anda. 3. E. Hurst. C. A. Becker. 3. Cow,∏. L. Wharton, and D 3 Auerbach. Surf. Sc: 93. 270 ( 1980). 3'E. K. Grimmelmann, 3. C. Tully, and M 3. Cardillo. 3. Chem. Phys. 72. 1039 (1980). '*C. T. Rettner and 3. A. Barker ( to be published ). "E. K. Schweizer and C. T. Rettner ( to be published ) **A. E. Dabiri. T. 3. Lee. and R E. Stickney. Surf. Set. 26. 522 ( 1971) G. Comsa, R. David, and B 3. Schumacher. Surf. Sei. 85. 45 ( 1979); G Comsa and R David, ½nd. 117. 77 ( 1982). ^C. A. Becker. 3 P. Cowtn. L. Wharton, and D 3 Auerbach. 3 Chem Phys. 67. 3394 (1977). i*S. T. Ceyer. W. L. Guthne. T -H. Lin. and G A. Somoηai. 3 Chem. Phys 78. 6982 ( 1983); T -H Lin and G. A. Somorjat, ½nd 81. 704 ( 1984). W. Muhlhausen. L. R Williams, and 3. C Tully, 3 Chem Phys 83. 2594 (1985). *'H. P. Steinhick. A. Winkler, and K D Rendulic. 3 Phys C 17. L3! 1 (1984). ∏T. Matsushima, Surf Sei. 123. L663 ( 1983); 3. Catalysis 83.446 ( 1983); Surf. Sei. 127. 403 (1983). *iR. C. Cesser, S. R. Bare, S. M. Francis, and D A King. Vacuum 31. 503 (1981). University. Oxford. England. 1954). p. 412. J Chem. Phys.. Vol. 90. No. 7, 1 April 1989 c-1 APPENDtX C ELECTRON ENERGY LOSS SPECTROSCOPY OF AMMON!A ON Ru(001) [The text of Appendix C consists of an articιe coauthored with J. E. Parmeter, Y. Wang and W. H. Weinberg, which has appeared in J. Chem. Phys. 88,5225 (1988).] C-2 Etectron energy toss spectroscopy of ammonia on Ru(001) J. E. Parmeter,*' Y. Wang. C. B. Mullins,"' and W. H. Weinberg Zλu⅛κM! p∕^C⅛emu∕ry and CAemrca/ i'∕ty∏eerιΛ⅛ Ca∕r∕⅛∕7iM /rtru/p/e p∕^ 7⅛f⅛∏c∕o⅛y. Po∞deeA Ca∕r∕⅛r-Λ<d 9/723 (Received 23 October 1987; accepted 13 January 1988) The adsorption of ammonia on Ru(OO) ) has been studied using high-resoiution eiectron energy loss spectroscopy. Multilayer, second-layer, and monolayer ammonia have been characterized vibrationally. These three states desorb near IH, 130 and between approximately 150 and 350 K, respectively. The symmetric deformation mode of chemisorbed ammonia shifts down in frequency continuously with increasing coverage from approximately 1160 cm* ' in the low-coverage limit to approximately 1070 cm* ' at (monolayer) saturation. The frequency of this mode in coordination compounds of ammonia is sensitive to the charge on the meta) atom (increasing with increasing positive charge), and the frequency shift of this mode on the Ru(00, ) surface can be correlated with the work function decrease that this surface undergoes as the ammonia coverage increases. Off-specular EEL spectra allow the weak NH3 rocking mode and the frustrated translation of the ammonia perpendicular to the surface (i.e., the metal-nitrogen stretch) of chemisorbed ammonia to be resolved near 625 and 340 cm* ', respectively These modes have not been identified in previous EELS studies of chemisorbed ammonia on hexagonaHy close-packed metal surfaces. Second-layer and multilayer ammonia yield EEL spectra similar to those observed on other meta! surfaces. In agreement with previous results, the adsorption of ammonia on Ru(00l ) at 80 K, followed by annealing, leads only to reversible desorption. I. INTRODUCTION The adsorption of ammonia on weU-dedned single-crys­ talline meta) surfaces under ultrahigh vacuum (UHV) con­ ditions has received considerable attention in recent years. '^" This subject is of interest both because it relates to industrial catalytic processes such as ammonia synthesis," and because ammonia is a prototypical example of a weak donor ligand that bonds to metal atoms both in organome­ tallic compounds"*" and on surfaces via the electron lone pair on the nitrogen atom. As part of a continuing study of the chemistry of nitrogen-containing molecules on the hexago∏a)ly close-packed Ru(00l ) surface,"*" we report here the results of a high-resolution electron energy loss spectro­ scopic ( EELS ) investigation of ammonia adsorption on this surface. The interaction of ammonia with Ru(00l ) has been the focus of several previous studies. '*^" Danielson er a/. " em­ ployed thermal desorption mass spectrometry (TDMS), Auger electron spectroscopy, and low-energy electron dif­ fraction (LEED) to study this system, and concluded that for adsorption temperatures below 300 K the chemisorbed ammonia desorbed completely reversibly, with negligible dissociation occurring upon annealing. Dissociation could be induced only be electron beam bombardment of the ad­ sorbed ammonia, or by exposing the surface to large fluences of ammonia with the surface temperature maintained above 300 K. Benndorf and Madey' ' used LEED, TDMS, electron stimulated desorption ion angular distribution (ESD1AD) " AT&T Be!! L⅛bor⅛to∏es Predoctor*! Fdtow °' Link Predoctor*! FeHow !985-86. ,BM Predoctor⅛l Fei⅛w ,986-87. J Ch⅛m Phys a< (8). 15 April 1988 measurements, and work function change (Δ^) measure­ ments to study ammonia adsorption on Ru(00l ), and also concluded that for low adsorption temperatures (approxi­ mately 80 K) ammonia adsorption is completely reversible. They showed that large ammonia exposures at 80 K lead to ammonia desorption over a broad temperature range, with monolayer desorption between approximately 150 and 350 K (and displaying weak maxima near 180 and 270 K), de­ sorption of second-layer ammonia at 140 K, and multilayer desorption in a sharp peak at II5 K. Both ESDIAD and ∆d measurements supported the expected mode! of bonding for monolayer ammonia, namely via the lone pair of electrons on the nitrogen atom. The saturation monolayer coverage of chemisorbed ammonia was estimated to be approximately 0.25 monolayer (or 3.9X 10" molecule cm*') based pri­ marily on LEED data and in analogy to previous studies of ammonia adsorption on Pt( III P and Ni( I !0). " It was also suggested that ammonia adsorbs in threefold hollow sites on this surface. Electron energy loss spectroscopy has been used pre­ viously to study ammonia adsorption on Pt(lll),' Ni(HI),' Ni( 110).' Ag(110),' Ag(3l I )." and Fe( I 10)7' In all cases, ammonia adsorbs molecularly at low tempera­ tures (approximately 80-120 K) and only on Fe( 110)" and Ni( 110) " does significant dissociation occur as the surface is annealed. The EEL spectra of ammonia adsorbed in the monolayer are dominated in most cases by a strong loss fea­ ture occurring between approximately 1050and I !70cm^ ', which is due to the NH, symmetric deformation mode. Con­ siderably weaker loss features occur in the frequency ranges of 1580-1650, 3200-3320, and 3340-3400cm* ', which are due, respectively, to the NH^ asymmetric (degenerate) de­ formation, symmetric stretch, and asymmetric (degener- 0021-9606/88/085225-12S02.10 (t) 1933 American institute of Pħys*cs 5225 C-3 tified on only three of the metal surfaces studied, and there is a large variation in the reported values: 340 cm ^ ' on Fe(H0),'* 470 cm^' on Ag(3ll),' and 570 cm ' on Ni(H0⅛." It was partly in the hope of identifying these latter two modes clearly that we undertook this study of ammonia on Ru(00l ) As in several of the previous studies, we have measured EEL spectra of both NH. and ND, in order to confirm our mode assignments Several of the earlier studies have also characterized second-layer and multilayer am­ monia vibrationally. As might be expected, the formation of a second-layer, molecularly adsorbed state (desorbing be­ tween 120 and 160 K ) and of a multilayer state (desorbing between 110 and 120 K) appears to be a general feature of ammonia adsorption on these metal surfaces, and both the second layers and the multilayers show some common EEL features on all surfaces on which they have been studied. that are characteristic of the clean surface. The UHV chamber also contains a quadrupole mass spectrometer for performing TDMS expeπments, and for the analysis both of background gases and of the purity of the ammonia that was admitted into the chamber. The thermal desorption measurements performed tn connection with this work were obtained with a heattng rate of approximately 8 K s" '. The NH, used in these studies was obtained from Matheson (99.99% reported purity) and was purified by several freeze-pump-thaw cycles. The NH, exposures were performed by backfilling the UHV chamber through a leak valve, and the purity of the NH, was verified in situ using mass spectrometry The ND, was obtained from MSD iso­ topes ( 99 at. % D ). Due to H/D exchange of the deuterated ammonia either on the walls of the UHV chamber or in the stainless steel leak lines. ND, could not be admitted into the chamber cleanly through a leak valve and was therefore ad­ mitted into the chamber via a directional beam doser consist­ ing of a microcapillary array. The purity of the ND^ ad­ sorbed on the surface was best monitored by comparing the integrated intensities of the ND and NH stretching vibra­ tions measured in subsequent EEL spectra. All exposures were effected at a surface temperature of 80 K, except as noted. It. EXPEHtMEMTAL DETAILS HI. RESULTS ate) stretch. The characteristic frequency ranges for an NH, rocking mode and the frustrated trans)ation of the ammonia perpendicular to the surface [i.e., the metal-nitrogen stretch, v(M-NH,)≡7*, ] are much less certain, however The NHι rocking mode has not been identified prevtously for chemisorbed ammon,a on any metal surface, although a frequency of 720 cm _ ' has been reported for second-layer ammonia on Pt( ] ] I ) .'The v( M-NH,) mode has been iden­ The EEL spectrometer used in this study, and the UHV chamber housing it, have been described in detail else­ where?" The spectrometer employs 180* hemispheres as the energy dispersing elements in both the monochromator and the analyzer. The analyzer is rotatable within the scattering plane to allow off-specular EEL spectra to be measured. Typical parameters used in obtaining the EEL spectra pre­ sented here were the following: energy of the electron beam incident on the surface. 4 eV; resolution (full width at half­ maximum of the elastically scattered beam), 50-85 cm count rate tn the elastic peak. 1-3 X IO' counts per second; and (fixed ) angle of incidence of the electron beam, approxi­ mately 60* with respect to the surface normal. All EEL spec­ tra were collected tn the specular direction, except as noted. AH EEL spectra were recorded with the surface at a tem­ perature of 80-100 K. after annealing to indicated tempera­ tures. The stainless steel UHV chamber was pumped by a 220 /s ' Vartan noble vacton pump and a titanium sublimation pump. Typical working pressures during experiments were in the range of 1-3 X 10 ^ "'Torr The Ru(00l ) crystal could be cooled to 80 K using liquid nitrogen, and could be heated resistively to temperatures exceeding 1600 K. After each ex­ periment, the surface was heated to 1000 K prior to the next exposure in order to desorb any trace amounts of atomic nitrogen'" that might be formed as the result of ammonia decomposition at defects or the cracking of ammonia on hot filaments, as well as any residual hydrogen" and carbon monoxide" adsorbed from the chamber background Peri­ odic, extensive cleaning of the surface was accomplished both by argon ion sputtering and annealing the crystal in a background of oxygen. Surface cleanliness was verified by EELS, and by H, ' and CO" thermal desorption spectra A. Th⅛rm⅛t desorption m*** spectrometry Since a detailed study of the thermal desorption of am­ monia from Ru(00l) has already been carried out." this subject was investigated only briefly, primarily to determine the appropriate ammonia exposures and annealing tempera­ tures to obtain EEL spectra characteristic of multilayer, sec­ ond-layer, and ( various coverages of) monolayer ammonia, fn reasonable agreement with the previous results, we detect the desorption of second-layer ammonia at 130 K for expo­ sures in excess of approximately I L(l L=] Langmuir ≡ 10" Torr s), and the desorption of multilayer ammonia at 115 K for exposures exceeding approximately 2 L. While the two monolayer desorption maxima were not well re­ solved using our apparatus, our EELS results (cf. Sec HI B) indicated that monolayer ammonia desorbed between ap­ proximately 150 and 350 K. This is also in good agreement with the previous results. The possibility that some of the ammonia dissociates rather than desorbs as the surface is annealed was investigat­ ed briefly using H, thermal desorption The H, thermal de­ sorption spectra that were measured following saturation ammonia exposures were compared to those obtained fol­ lowing saturation hydrogen exposures on Ru(00l ) This al­ lowed the maximum possible amount of ammonia that dis­ sociates to be estimated by using the known saturation coverage of hydrogen adatoms (approximately 0.85 monolayer) on Ru(00l )." The amount ofhydrogen adatoms that desorbed was never greater than 0.03 monolayer, which cor­ responds at most to 0.0) monolayer of NH,. This could cor­ respond to the decomposition of a small amount of ammonta at defect sites. However, such a small amount of hydrogen desorption could also result from the adsorption ofhydrogen from the residual gas or from a small hydrogen impurity in J Chem Phys . Vo) 98. No 8. IS Aρπ) 1988 C-4 the ammonia sample Thus, we are in compiete agreement with the previous conclusion" that ammonia decomposi­ tion ts πegiigib)e on Ru ( 00 ι ) fot)owtng ammonia adsorption at 80-100 K. B. Electron energy )oaa spectroscopy ΠG ! TheEELs-cu*ths(rau)tfoUmha,,mvs)etM.mes<)fNH,to theRu(∞!)sMrfscesttOR (!n the este oftheOTLeuptmire. the surf⅛ce ιempeτsture wss ch-er to 90 R. sad the e⅛½c pe^ oouat rste ww ∞iγ 2×)f.) Eiectron energy loss spectra for several coverages of am­ monia on Ru(00!) at 80-90 K are shown in Fig. I. The exposures involved correspond to submonolayer coverages; no second-layer or multilayer ammonia is present in any of these spectra with the exception of spectrum (d), where a very small amount of second-layer ammonia is probably present. Following a 0 I L ammonia exposure, only a single loss feature is resolved at 1155 cm* '. For larger ammonia exposures, this loss feature shifts downward in frequency, reaching approximately 1085 cm*' following a I L expo­ sure. ft is by far the most intense loss feature for all submono­ layer coverages, and is identified easily as being due to <S, (NH,), the symmetric deformation mode of molecularly adsorbed ammonia. At all coverages, it is strongly upshιfted in frequency from its gas phase value of950 cm* '." Follow­ ing a 0.5 L exposure, a second loss feature is observed at 3265 cm* ' due to the (unresolved) symmetric and asymmetric NH, stretching modes Increasing the exposure to 0.7 L al­ lows these two modes to be resolved at 3270 and 3370 cm ^ ', respectively, and another loss feature becomes apparent at 1560 cm ^ ' which is assigned as the asymmetric deformation mode of ammonia, <5. ( NHy ). Following a I L exposure all of these loss features are again resolved, and a very wealt loss feature appears at2!45cm^'whichis assigned as the over­ tone of the NH, symmetric deformation. Another loss fea­ ture appears at 330 cm* '. While this feature may derive some intensity from the very wealt frustrated translational TABLE L Vibrations! frequencies (in cm') sad mode assignments for gssphstetmmtmit. tad foe chemisorbed ammonia αa several mets! turfsees Ru(00)) Mode ".(NH,) v.(NH,) ⅛.(NH,) <S.(NH,) p(NH,) v(M-NH,) NH,(t) (Ref 53) MD,(s) (Ref 33) 34)4 3337 23⅜ 2420 ∏91 747 )628 930 MH, ND, (Th⅛ 1*or⅛) 3370 3260 )38O )070U60 623* 340- Pt(!!!) (Ref. 3) Ni(!!!) (Ref. 9) Ni(U0) (Ref. 9) Ft(Π0) (Ref. 28) As(!!0) (Ref !) 23)3 2333 ∏63 833- 3S40 3240 )600 ))40 3360 3270 )38O Π40 3330 3200 )38O Π20 3370 3290 !640 3390 9)0 4M* 330* no? no no. no. no 370 j. Cham Phys.. Vo) 88. No 8 . 15 April ι⅛88 ))03H70 no. 340 Ag(3]!) (Ref. 2) 34)0 3320 !640 )030 no. no. )648 )!00 no. 470 C-5 FIG 2. The EEL spectrum that results ⅛Rer ⅛ Ru(OO! ) suri⅛cc with con* denied muM⅛yeτs of NDy is ⅛nne⅛ied to ! 30 K This spectrum ts ch⅛r⅜cteristic of monoi⅛yer ND, mode (normal to the surface) of the chemisorbed ammonia ( cf. Fig. 3 ), it is most likely due p∏mariiy to the presence of a very smali amount of second-layer ammonia, which exhibits an intense ioss feature due to a iibrational mode at 360 cm ' ' (cf. Fig. 6 and Tabie H). The vibrational frequencies of che­ misorbed ammonia on Ru(001 ) are summarized in Table 1, and compared to those of gas phase ammonia," and of am­ monia chemisorbed on several other single crystalline metal surfaces. (Information is also included in this table that is derived from o<Γ-specular data, cf. Fig. 3. ) It should be noted that while the weak loss features due to <S. ( NH, ), v, ( NH, ), and v. (NH,) occasionally were found to have frequencies differing by as much as ⅛ 40 cm ^ ' from the average values listed in Table 1, they showed no consistent frequency shifts as a function of coverage. Although such shifts would be difficult to detect due to the low intensities of these loss fea­ tures, all of these features could usually be detected in offspecular spectra following exposures as low as 0.5 L. A typical EEL spectrum of a saturated monolayer of ND, on Ru(00! ) is shown in Fig. 2. The surface was pre­ pared by exposing the Ru(001 ) crystal at 100 K to approxi­ mately 30 L ND,. followed by annealing to 150 K to remove multilayer and second-layer ND,. The intense <$, (ND,) mode occurs at 835 cm^'. As in the case of 6,(NH,), the frequency of this mode was found to shift up with decreastng coverage, and it occurred near 910 cm ^ ' in the low coverage limit. The <5. (ND,), v,(ND,). and v. (ND,) modes occur at frequencies of 1165, 2335, and 2515 cm" ', respectively. The frequencies of these modes are also listed in Table 1. and the observed frequency shifts upon deuteration clearly sup­ port the assignments. A very weak loss feature is observed at 1455 cm" ' in Fig. 2. A similar loss feature was observed in the EEL spectrum of second-layer ND, on Pt( 111 ) and was attributed to 6.(ND,H) of a small ND,H impurity? The slight asymmetry on the high energy side of the <5,(ND,) loss feature could also be due to the presence of a small amount of ND,H. The weak loss feature at 2930 cm ' ιs most likely due to v( CH ) of a small hydrocarbon impuri­ ty" Off-specular EEL spectra of both NH, and ND, al­ lowed two additional loss features to be resolved clearly in the low-frequency region. Figure 3 shows EEL spectra of nearly saturated monolayers of (a) NH, and (b) ND, ob­ tained w]th the analyzer rotated approximately 5* off-specular (towards the surface normal). In addition to the loss features observed in specular EEL spectra. Fig. 3(a) shows features at 625 and 340 cm"', and Fig. 3(b) shows features at 480 and 350 cm * '. Based on the observed frequency shifts and a comparison to vibrational data for metal-amine com­ pounds ( see Sec. IVA), the loss feature at 625 (480 ) cm ^ ' is assigned to the NH,(ND,) rocking mode∕,(NH,) and the loss feature at 340(350) cm" ' is assigned to the frustrated translation of the ammonia perpendicular to the surface v( Ru-NHj ). Both of these modes are absent in gas phase ammonia, where the "rock" corresponds to a free rotation and the "frustrated translation" to a free translation of the molecule. The low intensity of the v( Ru-NH, ) mode is con­ sistent with previous results for ammonia adsorption on hexagonally close-packed metal surfaces, in which no metalnitrogen stretching mode was resolved? * Due to the low intensities of the/, ( NH, ) and v( Ru-NH, ) loss features, it is impossible to say whether or not these modes undergo fre­ quency shifts as a function of coverage. Note that in the offspecular EEL spectra presented here the 6, (NH,)[6,(ND,)] loss feature has decreased greatly in intensity relative to the other loss features of chemisorbed ammonia. This indicates that this mode is strongly dipole enhanced in specular EEL spectra, while the remaining modes of chemisorbed ammonia are primarily impact excit­ ed." It should be noted that the∕,(NH,) mode was occa­ sionally observed as a very weak shoulder in specular EEL spectra, near 650 cm ^ ' for NH, and near 500 cm ^ ' for ND, Indeed, such shoulders are barely visible in Figs 1(d) and 2, but the frequency cannot be assigned accurately based on these spectra ENERGY LOSS.cm" FIG. 3 CMΓ-specul⅛τ EEL *pectr⅛ of monol⅛ycr (*) NH, ⅛nd {b) ND, on J Chem Phys . Vo! 86. No 0. 15 Ao∏! 1968 C-6 FIG 1 T⅛c frtqumcy of the ⅛, ( NH,) mode oi chem⅛o⅛ed ⅛mmoma oa Ru(OOI ) M e function of smmonis oown⅛t- T⅛e tpproximele ooveτaχes given are relative to the surface coooeπtratioa of ruthenium atoats. which ⅛ L!7 X )0" an ^ '. and they are based on the aasuatptioa that the saturation Figure 4 show! the variation in the frequency of the symmetric deformation mode of ammonia adsorbed on Ru(OO] ) at 80 K as a function of the ammonia exposure. The exposure of ! L corresponds to the coverage where the second-layer theπnai desorption feature just begins to ap­ pear. However, this coverage does not correspond to mooolayer saturation because the second iayer begins to form be­ fore the monolayer is saturated*'; it is estimated that this exposure corresponds to approximateiy 80% of monoiayer saturation or an absoιute coverage of approximateiy 0.2.** Note that the assumption of a constant probability of ad­ sorption. in agreement with the resuits of Benndorf and Ma dey*' and as expected for NH, at 80 K given that ammonia multilayers condense at this temperature, means that the coverages are directly proportional to the ammonia expo­ sure. The frequency of the <5, ( NH, ) mode decreases monotonicaliy with increasing coverage, from approximateiy 1155 cm ^ ' following a O i L NH, exposure. The frequency decrease is very nearly linear with coverage, although there may be a slight decrease in the absolute value of the slope at high coverages A iinear ieast-squares Ht to the data yie!ds <S,(NH,)=-69.6e+Π57. (l) where <5, ( NH, ) is the frequency of the NH, symmetric de­ formation mode in cm* ', and f is the exposure in Lang­ muirs. Figure 5 shows the variation with anneaiing tempera­ ture of the frequency of the symmetric deformation mode of ammonia adsorbed on Ru(001) for a number of different ammonia exposures at 80 K Foiiowing all exposures, the surface was annealed first to i 50 K to be certain that oniy monolayer ammonia was present. It may be seen by com­ parison to Fig 4 that annealing from 80 to i 50 K generally did not change the frequency of <5, (NH,) by more than iθ cm* ' for exposures of 1 L or less. Foiiowing a 0.1 L am­ monia exposure, the frequency of ⅞(NH,) is neariy con­ stant at i 150-1160 cm ^ ' regardless of annealing tempera­ ture. This is consistent wtth the fact that the coverage is not changing for annealing temperatures between 80 and 300 K for such a iow exposure (i.e,, coverage) the ammonia de­ sorbs only in a single peak near 315 K.*' On the other hand, the exposures of 2.0 and 8.0 L correspond to exposures well in excess of monolayer saturation, and annealing the surface to 150 K in each case produces a saturated monoiayer of ammonia. Thus, the measured frequencies in these two cases should be the same for annealing temperatures of 150 K and greater, and this is found to be true within experimental er­ ror ( + 10 cm ^ ' ). The measured frequencies increase mar­ kedly as the surface is annealed from 150 to 230 and 300 K. because for a saturated monolayer of ammonia, significant molecular desorption occurs in these temperature ranges?' This results in a decrease in the coverage and thus an in­ crease in the symmetric deformation frequency, as would be expected from an inspection of Fig. 4. For initial exposures of 0.5 and 1.0 L, the measured frequencies for any given annealing temperature fall between those obtained following 0.1 L or saturation exposures, and also increase ( though less dramatically than for saturation exposures) with increasing annealing temperature. These results are all consistent with the ⅜ (NH,) frequency being a function of ammonia cover­ age only; the annealing temperature appears to have no ef­ fect other than to change the coverage. It should be noted that the two data points obtained following annealing to 350 K correspond to trivial amounts ( 50.01 monolayer) of am­ monia readsorbed from the chamber background, since the thermal desorption of ammonia from Ru(001 ) is complete between 300 and 350 K The <5,(NH,) frequency of 1160 cm*' thus corresponds to the zero-coverage limit for am­ monia on Ru(001 ), in agreement with the data of Fig. 4 and Eq. (1). Electron energy loss spectra of second-layer NH, and ND, on Ru(001 ) are shown in Fig. 6, and the observed fre­ quencies are compared to those of second-layer ammonia on other metal surfaces in Table II. The EEL spectra were ob­ tained after the Ru(00!) surface at 85 K was exposed to large Hue∏ces of NH, and ND,, followed by annealing to 120 K to remove multilayer ammonia. The loss features due to ⅛,(NH,), έ. (NH,). V,(NH,), and v. (NH,) are assigned easily as for monolayer ammonia. The <5, ( NH, ) loss feature J Cham Phys.. Vot. 88. No 8.15 April 1988 C-7 FIG 6. The EEL spectra tint resuii when a Ru(00! ) surface with coo* denvd multilayers of (a) NH⅛ and (b) ND, ts annealed to 120 K These of second-layer NH, at 1145 cm ' shows a pronounced shoulder at 1070 cm^^ ', which is due to the parttally screened <5, ( NH,) loss feature of monolayer NH,. The corresponding shoulder in the case of ND, is not so well resolved because the peak frequencies are closer together ( i.e., approximately 30 rather than 75 cm*' splitting). The spectrum of NH, (NO,) contains an intense mode at 360 ( 280) cm*' as­ signed to the NH, twisting mode ι*( NH, ) ( equivalent to the frustrated librational mode ) Note that this mode is not observed for monolayer ammonia because in that case it cor­ responds to a free rotation about the ruthenium-nitrogen bond. Evidence for this free rotation in the monoιayer has been obtained from ESD1AD data?' This rotation is hin­ dered in the second layer due to hydrogen bonding of the second-layer ammonia molecules to the monolayer am­ monia. The EEL spectrum of Fig. 6(a) also shows loss fea­ tures at 545 and 225 cm ^ ' which we tentatively assign, re­ spectively, to an NH, rocking mode and the frustrated translation (perpendicular to the surface) Γ, of the secondlayer ammonia. The corresponding loss features are not well resolved in the case of ND, due to overlap with the r( ND3 ) loss feature at 280 cm * '. Both spectra of Fig. 6 show a weak loss feature near 1400 cm* '. the identity of which is uncer­ tain. As stated previously, a similar loss feature has been attributed to <5. ( ND,H ) of an ND^H impurity in the case of second-layer ND, 0nPt(lll)7ln thecaseof NH,, the 1415 cm*' loss feature is probably a combination band (i.e.. 360+ 1070 and/or 1145).'"* " We can think of no likely impurity that would give rise to a loss feature in this frequen­ cy range; <5(H,O) of water occurs at the considerably higher frequency of 1560 cm ^ Figure 7 shows EEL spectra of (a) NH, and (b) ND, multilayers on Ru (001 ), and the vibrational frequencies and their assignments are listed in Table H! along with data for solid ammonia and for NH, multilayers on other meta! sur­ faces. Both of the EEL spectra in Fig. 7 were measured fol­ lowing ammonia exposures of approximately 30 L. The loss features due to the v, (NH,), v„(NH,), <S,(NH,), and <5.(NH,) modes are again assigned in a straightforward manner Both spectra also show a weak loss feature due to the <5, (NH,) overtone/double loss. A strong loss feature at 200 cm * ' in both cases corresponds to an ammonia lattice mode (i.e., a frustrated translation of the ammonia mole­ cules within the multilayers). Two strong librational modes are observed, one at 540 (420) cm* ' and the other a shou!- TABLE !L Vibraboant t⅛qutno<* (in on ^ ' ) and mode a⅛ptmtnα for second-toyer NH, ⅛rsd ND, on Ru(00! ⅛. and for ⅛econdιayer MH, on ⅛ev eraj other met⅛J ⅛urface⅛ R⅛(00t) Pt()H) Aa(ttθ) Ni(H0) At<3ιl) (Ref 3) tRef 1) (Ref. 9) (M2) Mode MH, MD, (Th⅛ WOΓ⅛) τ<NH,)' p(NH,) <5,(NH,) <S.(MH,) v.(NH,) M0 545 ∏45 MM 3225 M0 no? 885 Π90 2340 350 720 !!90 1630 3!5O 340 n.o? !I00 1630 3320 360 Π.Ο !H0 1630 3200 *.(NH,) 3350 225 2503 Λ.Ο. 3320 3390 no. 3340 no. 260 no. !!00 ι65O 3410 Γ, "° no. ' tπ some EEL spectre a ahovider was present at approximate^ 450 cm ' that miχht be aaatn⅛ed top<MD,) *t — syτ∏∏seιrsc. a < asymtτsersc. n o - nor ohaerved. FIG 7 The EEL spectra ofcondensed multilayers of (a) MH, and ib) ND, tureof 85 K.. J. Chem. Pħys., Vo!. 86. No 8.15 Apoi 1988 C-8 Sol,dammoπn Ru(00!) NH, (Ref.38)* ND, (Ref.38)* ND, NH, (This wortc ) V,' !38-!84 323-386 ∏2-174 242-312 460 200 330 V,' d,(NH,) <5.(NH,) v.(NH,) v.(NH,) 4!2-⅜62 !O37-tO8O ι630-!679 3!6O-333O 3370-3378 3!3-4O6 8*3-836 HS3-!2!7 23ι8-2336 2503-23H 340 ! )00 !640 3230 3380 420 863 1220 2380 2320 Mode v,* 200 rt(∏n NH, (Ref.3} Ag(H0) As(3∏) NH, NH, (Ref.l) (Ref.2) no. !40 no 4t0 400 430 Π20 <640 no/ 3320 !070 Î63O 3320 3380 ι!30 !37O no. ... * A r⅛ng< of vtl.es it given for enoh frequency of solid NH, and ND,, b<αed on several fR end Raman smdiea nt temperatures ranging from 18 to 131 K !n some studies. more then one frequency was observed due to crystal splitting "v, ≡ frustrated translation or latttce mode. der at 460 ( 350) cm*' These frequencies are somewhat higher than those reported previously for both ammonia multilayers and solid ammonia/* although we note that the frequencies were lower if only a few layers of ammonia were present (eg., approximately 370 and 450 cm*' following an 8 L NH, exposure ) Both spectra show low intensity tails on the high energy side of the v, ( NH, ) loss features, the origins of which are not entirely clear, although they may be due to combination bands of v, (NH,) [ v. (ND,) ] and the libπttional and/or frustrated translational modes. The peak at 3370 cm* ' in the NT), multilayer spectrum is due primarily to the combination band <5,(NO,) + v.(ND,), although v( NH ) modes due to contamination of the ND, sample may also contribute to its intensity A loss feature ofsimilar intenstty was observed at 4480 cm* ' in NH, multilayer spectra that were scanned to higher loss energy, and this peak is assigned as <5,(NH,) + v. (NH,). IV. DISCUSSION A. Comparison to ammonia on otħar matai surfaces and In coordination compounds The EEL spectra that have been obtained for ammonia adsorbed on Ru(001 ) show strong similarities to those ob­ tained when ammonia is adsorbed on other hexagonally close-packed metal surfaces [i.e., Pt( 111 p and Ni( 111 )*], and on the pseudo-close-packed Fe( 110) surface/* On the more highly corrugated surfaces, the EELS data currently available do not provide a consistent picture. On Ag( 110),' the EEL spectra of adsorbed ammonia are very similar to the spectra obtained for ammonia on the more densely packed surfaces, with the 0, ( NH, ) loss feature at 1050 cm ^ ' being by far the most intense feature in the entire spectrum. The EEL spectra of ammonia adsorbed on Ag ( 311 ) are also sim­ ilar, although in this case a mode of moderate intensity was observed at 470 cm*' that was assigned as v(Ag-NH,)∕ The EEL spectra of ammonia on Ni( 110) are surprisingly different, with ⅛(NH,).<5.(MH,), and v.(NH,) loss fea­ tures of comparable intensity and an additional strong mode at 570 cm*' that was assigned as v(Ni-NH,)Z However, this mode assignment does not appear to have been vended by obtaining EEL spectra of chemisorbed ND, on this sur­ face. Despite the lack of a consistent pattern of mode intensi­ ties on the highly corrugated surfaces, the mode frequencies of 6,(NH3),<S.(NH,), v.(NH,). and v.(NH,) are quite similar on all of the metal surfaces where ammonia chemi­ sorption has been studied by EELS. This fact, along with the consistent pattern of mode intensities on Ru(001 ), Pt( 111 ), Ni( 111), Fe( 110),and Ag( 110), suggests some useful com­ parisons to infrared data for metal compounds containing ammonia ligands. Table IV provides a compilation of vibrational data for ammonia adsorbed on metal surfaces and in four different classes of metal-ammonia compounds The data for the met­ al compounds show some interesting trends with regard to the charge (both formal and actual) on the metal atom. Note that the "electron richness" of the metal atoms in these compounds is expected to follow the order M(NH,½÷ >M(NH,½÷ >M(NH,½÷. The frequency of the <5, ( NH, ) mode is very charge sensitive and shifts upward with increasing charge on the metal atom, having average values of 1158 cm ^ ' in the M ( NH, ) ⅛ com­ pounds, 1333 cm*' in the M(NH,)^ compounds, and in­ termediate average values in the case of the M(NH,)^ compounds. The changes observed in the frequency of this mode when the metal atom is changed and the charge held constant are minor compared to those resulting from charge changes/* The p(NH,) and v(M-NH,) modes are also charge sensitive and show marked upshifts in frequency in J Chem Phys . Vd. S8. No 8.15 Apπt 1988 C-9 Parmele? ⅛f a/. Ammonia on λυ(OOl ) pven in each c⅛⅛e ⅛ the average frequency for the spect6c mode in a certain type of compound, and is followed (k)v and htgh extremes) is hated for the same mode. Mode Ru(001) (Tht* work) *.(MH,)' 3370 v,(NH,) 3260 4.<NH,) !3S0 4.(NH,) Π!S 1070Π60 425 ∕,(NH,) HM-MH,) (-4,) M0 Other met*! 33430 33403400 3264(5) 32003320 1615(6) 1580!648 )H9(4)' 1050Π70 Not observed prevκ**⅛ly Reported vaiuαof 340,470, M(NH,H' M(NH,½* M(NH,)d (Ref. 44)" M(NH,)j* (Ref 44)' (Ref. 44)" (Ref. 64 )* 3249(3) 32403310 3183(3) 3)303350 !6!6(6) 15871630 1333(6) 12901368 SO4(4) 748857 4M(4) 442527 3336(7) 33003353 3200(7) 31603250 1601(8) 158S1610 ∏58(8) 10911220 454(S) 592749 332(4) 29S370 33)4(3) 32333354 3224(3) 31503267 )4OS(3) 15961617 !227(3) 11761253 4!3(3) 670493 409(3) 370432 3272(3) 32363327 3)45(3) 31563)70 )4)2(3) 15631669 )29S(3) )279)325 M7(3) 735849 4!t(3) 420524 570 *D⅛t⅛⅛ye⅛rAt(liO).At(3Π). Pt(Hl), N1(H1). Ni( 1101. and Fe( 110). * Data are for M — Cr, R⅛ Os, Co. Rh. and !r * Data are foe M — Mg. Mn, Fe. R⅛, Co. N‰ Zn, and Cd. Data are for M — Co, Zn, and Cd * Data art for M — Pd, Pt, and Cu the caee of Fe( HO), where 6, ( NH, ) was reported to shift in frequency as a function of coverage ( Ref 23), an average vaJue was used as the characteristic frequency of chem⅛or⅛ed ammonia *r — symmetric; a — asymmetric; n o. ≈ not obaerved going from the M(hfH,)^ to the M(NH,)^ met*! com­ pound*, although in these esses there is somewhat Larger overiap between the frequency ranges for these compounds and the M(NH,)^ compounds. The v.(NH3) mode is siight)y charge sensitive, and shifts upward in frequency with decreasing charge on the metai atom. The v, (NH,) and 6. (NH,) modes are not sensitive to the oxidation state of the metai atom and show no consistent frequency shifts. Comparing the EELS data for ammonia on metai sur­ faces to the data for coordination compounds reveals immediateiy that the 6,(NH,) frequencies observed on surfaces are very similar to those observed in the M(NH,)^ * com­ pounds, although the average frequency of ! 119 cm ' and the high and low extremes are all 40-50 cm" ' lower for the adsorbed ammonia, !t is expected that the <5, ( NH, ) frequen­ cies would be somewhat lower on metal surfaces, since the formal oxidation state of the surface metal atoms is zero. Thus, the observed frequencies of this mode for ammonta adsorbed on a number of metal surfaces appear to correlate well with the 1R data for coordination compounds. The same is true in the case of the v, ( NH, ) mode, the average frequency of which is higher on surfaces than in the coordi­ nation compounds. However, this is of only minor signifi­ cance because this mode is only slightly charge sensitive, and due to its low intensity and overlap with the v,(NH,) loss feature, its frequency on any given surface cannot be deter­ mined as accurately as that of J, ( NH, ). The fact that the <5, (NH,) frequencies of chemisorbed ammonia are similar to those of M(NH,)^ + compounds suggest that the frequencies of the other two vibrational modes that are strongly charge sensitive, p(NH,) and v(MNH,), might also be similar in these two cases. While very few data for these two vibrational modes of chemisorbed ammonia are available, our results provide strong support for this proposition. The measured /?(NH,) frequency of 625 cm " ' is well within the range of 592-769 cm ^ ' reported (or the M(NH,)2+ compounds, and is slightly lower than the average value of656 cm " ', just as our measured average <⅞, ( NH, ) frequency of 1115 cm* ' is slightly lower than the average value of 1158 cm"' in these compounds The ^7( NH, ) frequencies of the M ( NH) ÷ compounds are con­ siderably higher, ranging from 748 to 857 cm" ' In the case of v(M-NHj), the available data for am­ monia on metal surfaces present a more complex picture, but our measured frequency of 340 cm" ' is in good agreement with data for the M(NH,)^ compounds and identical to the value reported (or ammonia on the pseudo-close-packed Fe(l 10) surface?" The reported values of v(M-NH,) for Ag(3! ! )2 and Ni( 110)'' of 470 and 570 cm* ', respectively, are considerably higher and similar to those observed in the J Cham Phys Vo< 88. No 8. 15 Apn! 1988 C-10 M(NHj)⅛* compounds (462 to 527 cm*'), even though the observed <S, ( NH, ) frequencies on both surfaces are very similar to those observed on ctose-packed and pseudo-c)osepacked surfaces and in the M(NH,),^ compounds. White this difference is no doubt retated in some way to the highly corrugated structures of the Ag(3t 1 ) and Ni( t ιθ) surfaces, its exact explanation is unctear. ιn générai, however, the comparison of the v⅛bratioπai spectra of chemisorbed and coordinated ammonia appears promising, and also serves to point out the importance of obtaining o<Γ-specuiar spectra when EELS is used to study ammonia chemisorption. B. Origins of the fr*qu*ncy shift of <S.(MH,) Corrsistion with work function data The pronounced charge sensitivity of the <5, (NH,) fre­ quency in coordination compounds aiso provides an exp!anation for the observed shift in the frequency of this mode as a function of coverage for ammonia chemisorbed on Ru(00i ). The work function change of Ru(00] ) as a func­ tion of ammonia coverage at 80 K has been studied by Benn­ dorf and Madey?' and the work function was found to de­ crease monotoπicaiiy by approximately 1.8 eV as the cover­ age is increased from 0 to 0.25 moπoiayer. The work func­ tion change is saturated only after a total ammonia coverage of approximateiy 0.6 monoiayer, and the work function is then approximately 2.3 eV iess than on the cιean surface. The decrease in the work function is initially tinear with coverage, but it increases s)ightty tess rapidty with coverage above a coverage of approximately 0 15 to 0.20 monotayer, and much tess rapidty above 0.25 monotayer. The work function decrease resutts from a net transfer of electrons from the ammonia to the ruthenium surface atoms, so that the etectron richness of these surface atoms is increased as the coverage increases. The decreased "charge" on the metai atoms then teads to a tower <5, ( NH, ) frequency, just as in the coordination compounds of Tab)e IV The magnitude of this charge, or the number of e!ectrons donated to each surface metai atom by the ammonia adtayer for a given ammonia coverage, can be estimated ap­ proximately using the surface-ammonia dipo)e moment of 1.9 D calculated by Benndorf and Madey?' and assuming a surface-nitrogen bond distance of 2. ! 5 A.*° This suggests that approximately 0 18 electron is transferred to the surface per adsorbed ammonia molecule While the dipole moment of 1.9 D was based on the slope of the work function vs ammonia coverage curve in the limit of zero coverage, the curve is only slightly nonlinear below a coverage of0.25, and a large part of this nonlinearity is probabty due to the fact that some adsorption into the second layer occurs before the monolayer is completety saturated?' Our results showing that the frequency of 0, ( NH, ) decreases nearty linearly with coverage up to at teast a coverage of approximately 80% of monolayer saturation supports this point of view, and it is therefore reasonable to assume this same dipote moment for each adsorbed ammonia molecule up to monolayer satura­ tion. Thus, for a saturation monolayer ammonia coverage of 0 25. an average of approximatety 0.045 etectron is trans­ ferred to each surface ruthenium atom from the ammonia adtayer This interpretation of the frequency shift of ∣5,(NH,) has several important imptications regarding ammonia che­ misorption on Ru(00t) and other metai surfaces. First, it suggests that this downshift should occur as the ammonia coverage is increased on alt metal surfaces, and that the mag­ nitude of the shift shoutd scale approximately with the amount of charge transfer characteristic of ammonia ad­ sorption. This hypothesis cannot yet be tested fully due to a lack of experimental data, but it is interesting to note that on Fe(110) a comparable work function decrease (approxi­ mately 2 eV)"' and frequency shif⅛ (from 1170 to 1105 cm ^ ' ) 3* have been found. The Pt (111) surface exhibits a work function decrease of approximately 2.7 eV when a monolayer saturation coverage of ammonia is adsorbed? While the published EEL spectra for monotayer ammonia on this surface show <5, (NH,) frequencies varying from at teast ! 170 to 1090 cm"'? it is not clear whether there is a monotonic shift as a function of coverage, it must also be pointed out that a series of EEL spectra for various ammonia coverages on Ag(311) show no apparent shift in the fre­ quency of this mode? Work function data are not available for this surface. This interpretation suggests also that the <5,(NH,) frequency of ammonia chemisorbed on Ru(001) should be increased by the presence of electronegative ada­ toms such as oxygen, and decreased by the presence of elec­ tropositive adatoms such as alkali metals, !n the case of oxy­ gen, however, the effects of hydrogen bonding will compticate the analysis?^ Finally, we note that frequency shifts as a function of ammonia coverage might also be expected for the p(NH,) and v(Ru-NH,) modes of ammonia chemisorbed on Ru (001 ), since these modes are also charge sensitive for am­ monia in coordination compounds. We cannot address this question unambiguously because the toss features due to these modes are below our limit of detectability for ammonia coverages less than approximately half of monolayer satura­ tion. In addition, these low frequency modes are likely to be more affected than the <S, (NH,) mode by such effects as tilting of the ammonia at high coverages, a phenomenon which has been proposed to occur based on ESDIAD re­ sults?' C. Adaorption alte of MH, on Ru(001) The adsorption site(s) of ammonia on metal surfaces has been the subject of some disagreement in the literature Based on indirect evidence obtained using a variety of ex­ perimental techniques, it has been suggested that ammonia adsorbs with the nitrogen atom in threefold hollow sites on Ru(00!)?' Pt( 111 )? !r(ll 1 ),"<" and Nι( 111 )?and with the nitrogen atom in the long bridging sites on the pseudo­ close-packed Fe( 110) surface?* On the other hand, the ontop site was favored for ammonia adsorption on Al( 111 )?' and calculations for ammonia adsorption on Al( 1 It )*' and Cu( 111 )*ι support this model, showing the interaction be­ tween the nitrogen lone pair and the threefold hollow sites to be repulsive. The great similarities among the EEL spectra of ammonia chemisorbed on Ru(001 ), Pt( 111 )? Ni(It 1 )? Fe( 110)?* and Ag( 110) ' suggest (but do not prove) that a common adsorption site is occupied on all of these surfaces. J Chem Phys.. Vol 88. No 8. 15 April 1988 C-11 Parmeter a/ ; Ammonia on Ru(OOi ) For several reasons, we prefer the on-top site as the bonding site of chemisorbed ammonia on Ru(001 ). The pri­ mary reason is that the vibrational frequencies of adsorbed ammonia agree so well with those of ammonia ligands in hexamine compounds, which are "on-top" by definition. We would expect that the frequencies 0f∕7( NH, ) and especially v(Ru-NH,) would be substantially different if bonding oc­ curred in threefold hollow or twofold bridge sites. This hy­ pothesis cannot be tested because f⅛ere is no exomp/e ⅛πctoπ fo M in ∕nero∕ c∕⅛Mer c⅛em⅛iry o∕^ on ammonio ∕iganJ r⅛of occupies o r⅛nfe∕oi<f Ao/ioto or noq/o/J ⅛πi⅛ingsire, oronpsire of⅛er fAan on on-rop sire." The same is true of the closely related phosphine (PR,) ligands. In fact, to our knowledge, NH,, NH,(NR,), and NH(NR) ligands occupy only ontop, twofold bridging and threefold hollow sites, respective­ ly, in metal cluster compounds." We note also that the Fe( 110) surface contains no true threefold hollow sites, although it is possible that the pseu­ do-threefold sites on this surface might be sufficiently simi­ lar to true threefold sites to give rise to virtually identical EEL spectra for an adsorbed species. All of these arguments support the intuitive expectation that a strong <τ-electron donor such as ammonia should adsorb on the most electron deficient site on these surfaces, namely, the on-top site." While none of these points taken alone is compelling, we believe that together they provide considerable support for on-top bonding. Hopefully, additional data for ammonia ad­ sorption on surfaces and (especially) in metal clusters will clarify this issue. D. S*cond-⅛ay*r and mutt)layer ammonia As pointed out in Sec. Ill B. the EELS spectra of sec­ ond-layer ammonia on Ru(001 ) show strong similarities to those obtained on other metal surfaces. Most notably, the strong r(NH,) loss feature at 350cm*' is very similar both in frequency and intensity to the corresponding loss features on Pt(lll)∕ Ag(l!0),' and Ni(llO)? Surprisingly, the corresponding loss feature on Ag(311 )^ was observed at the substantially lower frequency of 260 cm * '. Nevertheless, it appears that a strong τ(NH,) loss feature near 350 cm* ' is characteristic of second-layer ammonia on most surfaces and may be used as a vibrational fingerprint for the second layer. Note that the librations! modes reported to date for multilayer ammonia (cf. Table III) occur at frequencies greater than 400 cm*'. The intensity of the r(NH,) loss feature, and the fact that it is strongly attenuated in offspecular EEL spectra indicate that it is dipole enhanced and thus belongs to the totally symmetric representation of the surface/second-layer ammonia complex This suggests that at least some of the second-layer ammonia is adsorbed with the molecular axts tilted significantly with respect to the sur­ face normal. This is in agreement with the ESD1AD results of Benndorf and Madey,*' who observed normal emission of H * for second-layer but not monolayer ammonia, and inter­ preted this in terms of tilting of the molecules within the second layer. We note also that 0ur∕7( NH, ) frequency of second-lay­ er ammonia of 545 cm ^ ' is substantially lower in frequency than that of 720 cm* ' reported for second-layer ammonia on Pt(lil).' The corresponding loss feature in the case of second-layer ND, on Pt( 111) occurred at 570cm*'? In our study, the corresponding mode in the case of ND, was not resolved unambiguously, although in some EEL spectra of second-layer ND, a shoulder was observed at approximately 450 cm* '. For both of these reasons, and because this mode has not been identified in EEL spectra of second-layer am­ monia on other metal surfaces, our assignment should be regarded as tentative. The vibrational frequencies of multilayer NH, and ND, on Ru(001 ) are in reasonably good agreement with those of solid ammonia, although the frustrated translational and !ibrational modes occur at slightly higher frequencies. We em­ phasize that even though all ammonia layers above the sec­ ond are indistinguishable in thermal desorption spectra, quite large ammonia exposures were necessary in order to obtain an EEL spectrum of true ammonia multilayers, i.e., a spectrum that showed no changes in the frequencies at the various vibrational modes with increasing exposure of the surface to ammonia at 85 K The main changes for lower exposures were in the frequencies of the librational modes, which shifted upward with increasing exposure until a total exposure of 15-20 L was reached. This would correspond, very approximately, to ! 5 "layers" of ammonia. It is possible that some of the previously reported librational frequencies of ammonia multilayers on metal surfaces would have been upshifted further if the exposures had been increased, and the fact that the observed librational frequencies of multi­ layer ammonia are in all cases higher than those of the strong librational mode of second-layer ammonia is consistent with our results. V. COMCLUS!OMS High-resolution electron energy loss spectroscopy has been used to study the interaction of ammonia with the Ru(001 ) surface. The results are generally in agreement with those of previous studies of the ammonia/Ru (001) sys­ tem utilizing other experimental techniques. In particular, we Und no significant decomposition of ammonia following exposures to the Ru(001 ) surface at 80-100 K, and our EEL spectra support the desorption temperatures for multrlayer, second-layer, and monolayer ammonia that have been re­ ported previously The principal new conclusions that may be drawn from the EELS results are the following: (1) Monolayer ammonia on Ru(001) exhibits EEL spectra with d,(NH,). d. (NH,). v,(NH,), and v„(NH,) frequencies very similar to those of ammonia chemtsorbed on other close-packed and pseudo-close-packed metal sur­ faces. In addition, the NH, rocking mode and the frustrated translational mode of the ammonia perpendicular to the sur face are observed in off-specular EEL spectra and occur at frequencies of approximately 625 and 340 cm * ', respective­ ly. Thep(NH,) mode has not been identified previously for chemisorbed ammonia, and the v(metal-NH,) mode has been identified only rarely. These results indicate the impor­ tance of obtaining off-specular spectra in EELS studies of chemisorbed ammonia. (2) The vibrational spectra of ammonia chemisorbed on Ru(001) and other metal surfaces of stmilar geometry JCħemPΠys.Vol88.No8 1⅛Ao∏l!988 C-12 Parm⅛!er a/ Arπrncr"3 on Ru(001 ) are very simitar to those of hexamine 2 + compounds. M(NH,)^*^, and agree much tess wett with those of M(NH,)i!+ andM(NH^)^ compounds, regardtess of the type of meta) atom and counter ions in the comp)exes. Thus, both on surfaces and in coordination compounds, the reat ( not formal ) "oxidauon state" of the meta) is of paramount importance in determmtng the vibrationa) frequencies of co­ ordinated ammonia. ( 3 ) The intense <5, ( NH, ) mode of chemisorbed NH, on Ru(OOt) shifts down continuousty and essentiaι)y )inear)y with increasing coverage, from 1160 cm ^ ' in the )ow cover­ age hmit to )070cm^ ' at monoiayer saturation. This parai!e)s the downshift of the frequency of this mode in coordina­ tion compounds as the charge on the meta) atom decreases This change can be corre)ated with the targe decrease ( — 2 eV) in the work function of the Ru(00) ) surface that occurs as ammonia is adsorbed,making the surface atoms, in ef­ fect. more "negativeiy charged." (4) The EELSresu)ts, and theorganometa))ic )iterature concerning NH, tigands. are more consistent with chemis­ orbed NH. bonding in on-top sites, rather than in threefo)d sites. ( 5 ) The vibrationa) characteristics of second-tayer and mu)tiιayer ammonia on Ru(00)) aregeπera∏y quite simitar to those of second-tayer and muttitayer ammonia on other meta) surfaces. The intense r(NH,) toss feature near 350 cm ^ ' appears to be a genera) feature of second-tayer am­ monia on most meta) surfaces. The fact that this mode is dipotar in nature suggests a titting of the motecutar axis of the second-tayer ammonia away from the surface normal, consistent with previous ESDtAD resu)ts7' White a)) )ayers above the second are indistinguishabte in therma) desorption spectra, substantiate )arger coverages (i.e., approximate)y 15 )ayers) of condensed ammonia are necessary to obtain a muttitayer EEL spectrum whtch shows no further changes in the vibrationa] frequencies of the various modes when additions! exposures are made. ACKNOWLEDGMENTS We are indebted to Dr Udo Schwa)ke for assistance in the earty stages of this work. Hetpfut discussions with Pro­ fessor J E Bercaw. Dr T E. Madey, and Professor W. A. Goddard are gratefully acknowtedged. This work was sup­ ported by the Nationat Sctence Foundation under Grant No. CHE-8617826. 'J L Gt*πd. B A Sexton. Md G E. MitchtH.Surf Sc. t!S. 623 ()982). 'S T Ccγer xnd 3 Ύ Y.tcs. Jr. Surf Sc. )SS. 584 ( I985). 'B A Sexton .nd G E Mitcheιι. Surf Sc. 4!. 523 ( t9S0) *B A Sexton .πd G E M.tcheH, Surf Set 99. 539 ( 1980) 'G B F.xher. Cheτn P⅛yx Lett 44.683 (<98ι) ^JLGI*nd.SurfSci7t.327 (]978) 'J L Gtend xnd V N Korchxι. J Cxtx) S3. 9 ( I978). 'J LG⅛ndxn<JEB Kot)ιn. Surf Set J04.478 ()98l). B Fisher and G E Mitchel!, J Electron Spectro⅛c Rel⅛t Phenom. 29, 233 (!983). "'C W Seabury, T N Rhodm, R J Purte!!. and R P MerτiH. Surf Set 93, !!7(!98O). "M. Grunze. M Go)ze. R K Dτiscol!. and P A Dowbeτ, J Vac Set Techno! M. 6!! (!98!) Jacobi. E. S. Jensen. T. N. Rhodin.and R P Memll. Surf. Sei. 10S. 397 (!98!). '*T. E Madey. J. E Houston. C W. Seabury. and T N. Rhodin, J. Vac Sei. Techno!. M.476 ( !981). "F.PNe:zerandT.E.Madey.Phys.Rev.Lett.47.928 (!98!). "F.P.Netzer aπdTE.Madey.Surf. Sei.∏9,422 (!982). "'M. Grunze. P A Dowber.andC. R. Brund!e. Surf. Set 128. 311 ( !983). "(a) R.J.Pu∏e!!.R.P.Mem!!.C.W Seabury,andTNRhodm.Phys. Rev. Lett. 44. !279 (1980); (b) W M. Kang.C. H. L,. S. Y Tang, C. W Seabury, K Jacobi, T. N. Rhodin, R. J. Purte!l.and R. P MerτiH, ibid. 47, 93!(!98!). "*L R Danielson. M J Dresser. E E Donaldson, and J T. Dickinson. Surf. Scι. 72.399 (1978). Surf.Scι.7L6]5(!978). ^*T E Madey and J T Yates. Jr., in Proceedings q∕* t⅛e 7ιh /ntemahona/ Pacuum Cdngr and 3rd /nrema/vbna/ Con∕⅛rence on So∕ιd Surtees, edit­ ed by R. Dobrozemsky, F Rüdenauer, F P. Viehböck. and A. Breth ( Vi. enna, 1977), p H83. ^'C Benndorf and T E Madey, Surf Set !35. !64 ( !983) ^*C Benndorf and T. E. Madey. Cheπt Phys Lett MH. 59 ( !983) Ύ E. Madey. C. Benndorf. D L Doering, and S Semancιk. Proc 8th lnt Congr.Catal.4.3t(!984). M Grunze. F Bozso. G Ert). and M Weiss. App! Surf Sei !. 24! (!97S). "Μ Drechs!er. H Hoinkes. H. Kaarmann, H Wiisch. G Ert!, and M Weiss. App!. Surf. Sei. 3. 2!7 ( !979). ^M. Weiss. G. Ertl. and F. Nitschke. Appl. Surf. Set 3. 614 ( 1979) Grunze and G. Ert!. in Ref. 20. p. ! !37 "W. Er!eyand H. Jbach. Surf. Set H9. L337 (!982). Kishi and M. W Roberts. Surf Sei. $2. 252 ( 1977). "T. D Gay, M Textor. R. Mason, and Y twasawa, Proc R Soc London Ser A 356, 25(!977). "F. P. Neuer and T E. Madey. Chem Phys Lett W. 3!5 (1982). 1962) P "J R. Hall and G. A Swile, J Organomet Chem 42. 479 ( 1972) **R P. Shibaeva. L P Rozenberg. R M Lobkovskaya, A E. Shdov. and G B Shu!pin. J Organomet Chem 220, 271 ( !98! ) '*L M Vallanno and S W Sheargold. Inorg Chim Acta 36. 243 ( 1979). t*W Rigby. J. A McC!everty. and P M Maitüs. J <?hem Soc Da!toπ Trans. 1979 382 P. Su!hvan. J A. Baumann. T J Meyer, D J Salmon. H Lehmann, and A Ludi, J Am Chem Soc 99, 7368 ( 1977). "H Lehmann. K J Schenk. G Chapuκ, and A Ludi. J Am Chem Soc 202.6197 (1979) Rehder, J. Organomet Chem. 37. 303 ( 1972) *"J E EHisand K L Fjare. Organomet L 898 (!982) ^'D Sellrnann. A Brandi, and R. Ende!!. J Organomet Chem 121. 303 (1976) *'M Heïberhoid. F Wehrmann. D Neugebauer. and G Hüttner. J Or­ ganomet Chem 252. 329 ( 1978) *'A B. Anton. N R. Avery, B. H Toby, and W. H Weinberg, J Electron Spectrosc.29.!S!(!983). ^*A. B. Anton. N R. Avery, T. E Madey. and W H Weinberg. J Chem Phys. M. 507 (!986). Schwalke. J. E Parmeter. and W. H Weinberg. J Chem Phys 94, 4036(1985). **U Schwalke. J E. Parmeter. and W H. Weinberg, Surf Sei 278. 623 ( !986). ^^J. E. Parmeter. U. Schwalke. and W H. Weinberg. J Am Chem Soc. !09.!876 (!987). **J E Parmeter. U Schwalke. and W H Weinberg. J Am Chem Soc 109. 5083 (1987). **G E Thomas and W H. Weinberg. Rev Sei !nstrum SO. 497 ( 1979) near 8OO-⅛5O K ( Ref. 21 ). curs below 450 K; H Shimizu. K Chhstmaπn. and G Ert!. J. Catai. 61, 412 (1980). !ow 500 K (a) T. E Madey and D Menze!. J App! Phys Supp!. 2. Part 2. 229 ( 1974); (b) H. Pfnür. P Feu!ner, H A Enge!hardt. and D. Menze!. (Them Phys. Lett. 59. 48! ( !978); (c) H Pfnür. P Feulner. and O Menze!. J. Chem. Phys. 79.4613 (!983). J Chem. Pħys.. Vo). 80. No 8.15 Apπ) 1988 C-13 Parmeter a/ : Ammon⅛a on Ru(OO1 ) ι*H. lbach and D. L. Mills. fFecrmn fnergy Lo≈ ⅛ecrrcncρpy and S⅛∕y⅛ce Pi⅛rononr (Academic. New York. 1982). p. 348. ing senes of M(NH, ), * compounds: Mn. 1146; Fe, 1156; Co. 1163; Ni. between Co(NH,½ + (1163) and Co(NH,⅛' (∏29). 3. *The estimate of 80% of monolayer saturation is arrived at from both the thermal desorption spectra of Ref 21. and by noting in Fig 5 that the minimum 6, (NH,) frequency for monolayer ammonta on Ru(001 ) is ap­ proximately 1070 cm ' '. and assuming the frequency decrease to be linear up to saturation '7p A. Thiel. F M Hoffmann, and W H Weinberg. J C⅛em Phys. 75, 5556(1981). **O. S Binbrek and A Anderson. Chem. Phys. Lett !5. 421 ( 1972). and J. Chem. Pħys.. '''K. Hermann. P S Bagus. and C. W Bauschlicher. Phys Rev S3t, 637! (!9S5). *'C. W Bauschlicher. J. Chem Phys M. 2619 ( 1985) non Cont∕v⅛ndr. 3rd ed. (Wiley. New York. 1978). pp. 197-202 8B. No. 6.15 April 19M D-1 APPENDtX D )NTERACTιON OF ETHYLENE WtTH THE Ru(001) SURFACE [The text of Appendix D consists of an artic)e coauthored with M. M. Hiιιs, J. E. Parmeter and W. H. Weinberg, which has appeared in J.ÆC.S. 108,3554 (1986).] D-2 Reprinted from the Joum⅛i of the Americ⅛n Chemic⅛i 9ociety, 1986, /0& 8556. Copyright Φ 1986 by the American Chemien! Society ⅛nd reprinted by peπnieeion of the copyright owner. Interaction of Ethylene with the Ru(001) Surface^ M. M. HH⅛, J. E. Parmeter,' C. B. MuUt∞.' and W. H. Conrri⅛u∕ιoπ ∕rorrt /Tie D,p<Jιoπ o/ C⅛e∕t<tsrry umf C⅛enuca∕ Engineering. Cu/i/ornia /nsri/u/e o/ Tec/tno/ogy, Pαsai∕enn, Cu∕i∕⅛miα 97 72! ∕!ecetoedt9cro⅛er⅛. ∕M5 A⅛strace The interaction of ethylene with the Ru(00! ) surface has been investigated via high-rtaoiutioπ eiectron energy loss spectroscopy and thermal desorption maaa spectrometry. Foiiowing desorption of an ethylene multilayer at HO K. di-<r-bonded molecular ethylene is present on the surface. Competing desorption of molecular ethylene and dehydrogenation to form adsorbed ethylidyne (CCH,) and acetylide (CCH) as well as hydrogen adatoms occur between approximately 150 and 260 K. The ethylidyne ⅛ stable to approximately 330 K. whereupon it begins to decompose to carbon and hydrogen adatoms. The desorption of hydrogen occurs in a sharp peak centered at 335 K, resulting from simultaneous ethylidyne decomposition and desorption of surface hydrogen. Further annealing of the overlayer to 380 K causes cleavage of the carbon-carbon bond of the acetylide. creating carbon adatoms and adsorbed methylidyne (CH). The methylidyne decomposes above 500 K with accompanying hydrogen desorption, leaving only carbon adatoms on the surface at 700 K. introduction The adsorption and reaction of ethylene on single crystalline rfaces of the groups 8-10 transition metals' '? have been the bject of intense study both as a prototype for ole(tn hydronation and dehydrogenation reactions'*'?' and to provide a basis comparing the bonding of olefins to surfaces with the bonding st has been observed to occur in multinuclear organometallic tster compounds. Spectroscopic studies of ethylene adsorbed these surfaces have shown that both the nature of the bonding molecular ethylene to the substrate as well as the thermal composition pathway of the adsorbed ethylene vary widely. For tmpie. ethylene rehybridizes to a di-o-bonded molecular species en adsorbed on Fe<U0).Fe(Ht).Ni(l 10), Ni(lll). Ni(S(lll) [H0)]. Pt(HI). and Pt(lOO),''? whereas molecularly adsorbed ylene on Co(00l ) at 115 K is χ-bonded? as is ethylene adsorbed Pd(lll)at 150 K and on Pd(IIO) at 110 K."-'?'? A mixed .HI —3andl3d" *" ** * 0O02-7863∕86∕!508-3554S0!.50∕0 overlayer of τ- and di-<r-bonded ethylene forms on Pd(IOO) at 150 K."-" Ethylene adsorption on the Ru(00l) surface has been (!) Erley. W ; Bsro. A- M.; McBretn. P ; Ibsch. H Sur/. Sei. t982 /70. 273. (2) Stip. U.; Tsai. Μ -C.; Kappen. J ; Ertl. G. Sw/ Sri t984. 747. 63 (3) Strosdo. J. A.; Bart. S. R.; Ho. W. Sw/ Sd. 1984. 748. 499 (4) Lebwald, S.; Ibsch. H. Sw/. Sri. t979. 89. 425. (5) Carr. R. G ; Sham. T K.; Eberhardt, W E Chem Phys Lett ]985. 773. 63. (6) Demuth. J. E. Sw/. Sc7. 1979. 84. 313. (7) Steininger. H.; ïbach, H.; Lehwald. S Sw/. Set t982. 777. 685 (8⅛ Albert. M. R.; Sneddon. L. G.; Plummer. E. W. Sw/ So 1984. /47. 127. (9) Gates. J. A.; Keamodel. L. L. Sw/ Set. !982. 720. L46L (10) Gates. J A.; Keamodel. L. L. Sw/ Sri )9⅛3. /24. 68 (11) Tyaœ. W T ; Nyberg. G. L.; Lambert, R M 7 Phyr. Chem 1984. 88. I960. (12) Ratajczyhows. L: Szymcraka. L Chem. Phyr. Lett. 1983. 9d. 243 (13) Keamodel. L. L.; Gates. J. A. Sw/. Sd. 1981. ∕∕∕. L747. (14) Stuve. E. M ; Madix. R. J. 7 Phyr. Chem !985. 89. 105 (15) Stuve. E. M.; Madix. R. J.; Brundie. C. R. Sw/. Sri !985. /52//33. 532. ( ! 6) Dubois, L. H : Cartner. D. G.; Somorjai. G A 7 Chem. Phyr 1980. 72.3234. (17) Gupta. N. M.; Kamble. V. S.; Iyer. R M. 7 Coin/. ]984. 88. 437 (18) Chesters. M. A.; McDougall. G S.; Pemblc. M E.; Sheppard. N rtpp/. Sw/ Sd. )985. 22/23. 369. (19) Boudart. M.; McDonald. M. A. 7 Phyr Chem. 1984. 88. 2185 (20) Henrict-Olivi. G ; Ohv⅞. s Chem . /ni 77 Ene/ t976. /3. 136 @ 1986 American Chemical Society /nrerocr/on q/ Æf/ty/ene wf/A ∕∕u7V9∕∕' Sur/dce D-3 studied receπtiy by Barteau and co-workers?* A detaiied com­ parison between our more compiete study and the preιimiπary resu)ts of Barteau et ai?* is presented in section !V. The thermai decomposition of ethy)ene (uitimateiy to hydrogen and surface carbon) on the surfaces mentioned above has been investigated by vibratioπai eιectron energy toss spectroscopy (EELS), thermai desorption mass spectrometry (TDMS). and UV photoeiectron spectroscopy The thermai decomposition inter­ mediates in ethyiene dehydrogenation on Fe(H0), Ni(ill), Ni[5()ii) × (iiθ)], and Ni(iiO) are acetyiene and acetyiide (CCH).'-" Madia and Stuve have postulated the formation of a v⅛nyi group from ethyiene adsorbed on Pd(i00).'**'* On Co(00i) and Fe(lιi) no stabie surface intermediates were observed; chemisorbed ethyiene evidently dehydrogenates compieteiy just beiow room temperature to carbon and hydrogen?** On Pt(ii i), Rh(i ] 1), Pd(i i i). and Pt(iOO), chemisorbed ethyiene dehydro­ genates to form ethytidyne beiow 300 Thus, ethylene adsorbed on the ciose-packed surfaces of each of the 4d and 5d groups 8-i0 metais studied previousiy forms ethyiidyne.** On the other hand, ethyiene adsorbed on all of the surfaces of the 3d groups 8-10 metals studied to date, inciuding the dose-packed surfaces, dehydrogenates more compieteiy to acetylene, acetyiide, or directly to carbon. Hence it is of fundamental interest to determine whether ethylene adsorbed on the Ru(001) surface behaves as it does on the other hexagonal 4d and 5d transitionmetal surfaces studied previously [Rh(l 11). Pd(ll 1). Pt(!OO), and Pt( i ∏ )] or whether it dehydrogenates more completely, as it does on the hexagonal Ni(!!!) and Co(001) surfaces. H. Expérimentai Procédâtes The experiments! measurements were conducted in two different uitrahigh-vacuttm (UHV) chambers, each with base pressures below 1 × iθ^'° ton* The first UHV chamber is equipped with a quadrupoie mass spectrometer, s single pass cγitndncal mirror electron energy ana­ lyzer with an intégrai electron gun for Auger electron spectroscopy, and LEED optics. Ail thermai desorption measurements were carried out in this chamber; data were collected by using an LSl-i 1 DEC laboratory computer, and linear heating rates of 3-i3 K/s were employed. A glass enclosure is placed around the ionixer of the mass spectrometer, and a small aperture in the front of the glass envelope permits sampling of gas that is desorbed only from the weii-oriented front of the Single crystal. Thus the effects of desorption from the edge of the crystal, the support lcsds. and the manipulator are excluded from the thermai desorption spectra.* The second UHV chamber contains both a quadrupoie mass spec­ trometer and a home-built EELS spectrometer of the Kuyatt-Simpson type.**** The energy dispersing eiements in the EELS monochromator and rotatable analyzer ate 180* hemispherical deflectors. The 'off. specutar* EEL spectra were measured with the analyzer rotated 7-10* from the specular direction, toward the surface normal Ai! EEL spectra were measured using s beam energy of 4 eV and with the incident beam approximately 60* from the surfaoe normai. The instrumenta! resolution ⅛full-width at half-maximum of the elastically scattered peak) varied between 60 and 80 cm ', while maintaining ⅛ count rate of approximately 3 × !(F cps in the elastic peak. A more extensive description of these UHV chambers as well as the procedures followed for cutting, polishing, mounting, snd cleaning the Ru(001 ) crystals has been disrussrrl in detail previously.**-* The cleanliness of the surfaces wss monitored with Auger spectroscopy in the first chamber and with EELS in the second Research purity hydrogen (99 9993%) and C P grade deuterium and ethylene (99.3%) were purchased from Matheson. Research purity tetradeuterated ethylene (99.99%) was obtained from Merck. The H,. D]. and C⅛D⅛ were used without further purification, whereas the C]H, wss subjected to freeze—thiw-pump cycles prior io use. The purity of (21) Disit. R. S ; Tsvisrides. L L /ad. Eng. C⅛**π Process Z⅛s. Dec 1933 22. 1. (22) Barteau. M A.; Broughton. J Q ; Mensel. D Ίρρ/. Surf. Set. !9B4. /9. 92 (23) lb*ch. H Presented st the Proceedings of the InternsLions! Confer­ ence on Vibrations in Adsorbed Lsyeτs. Juιtch. 1978; pspef 64. (24) The Pt(100) surfsce reconstructs tos slightly buckled, ciose-pscked (3 × 20) superstructure " (23) Heilmsπn. P ; Heins. K. Malier. K Sue/ Jef. 1979. 83. 487 (26) Feulner. P ; Mensel. D. 7. Poe Jci 7cc⅛ac∕. 1999 / 7. 662. (27) T⅛oπtιι. G. E.: Weinberg. W. H. f⅛ys. Esc. Lett 1978. 4/8 1181 (28) T⅛omss. G E.; Weinberg. W. H. Esc. Jef. ∕<tsrrum. 1979. 30. 497 (29) Wil)isms. E D : Weinberg. W H Joe/. Jef. 1979. 82. 93 (30) Wiil⅛ms, E D.; Went berg. W. H. 7. Poe. Jef 7⅛*ao∕ 1932. J0. 534 7. ,4"). C⅛em. See.. Po∕. ∕0⅞. Ao. /?. /Mb ∏gw* 1. Thermal deaorption spectra after C,H, adsorption on Ru(00) ) st 90 K: (a) CιH⅛ and (b) Hi desorption following a 1-Iangmuir expo­ sure and (c) Hi desorption following a 0.2-langmuir exposure ail gases was verified in situ with msss spectrometry. ΠΙ. Res*⅛s Typicai thermai deaorption spectra foliowing exposure of the Ru(00!) surface to i iangmuir (i tangmuir * i × i(P^* torr s) or more of i⅛H4 at 90 K are shown in Figure ia and ib?' Oniy hydrogen and ethyiene art observed to desorb from the surface in parttcuiar, no ethane, methane, or acetyiene is detected as judged by the a⅛eace of 30 and !6 amu peaks and by comparison of the 28,27, and 26 amu peaks to the cracking pattern of ethyiene. Figure ib estabiishes that most of the hydrogen, after an ethyiene exposure exceeding 0.6 iangmuir, desorbs in a sharp peak centered at 335 K (independent of coverage), with a smaii high-temperature shouMer above 400 K and a broad taii extending from 500 to 700 K of which the iatter represents iθ-i 5% of the desorbing hydrogen, it will be shown beiow that the major peak corresponds to the desorption of hydrogen enhanced by the de­ composition of ethytidyne (one of the decomposition products of ethyiene), the shouider corresponds to desorption-[united hydrogen from the surface, and the high-temperature taii corresponds to the dehydrogenation of surface methyiidyne (another decompo­ sition product). The thermai desorption spectrum of moiecuiar ethyiene (cf. Figure ia) shows that ethyiene desorbs in a sharp peak centered near i iθ K. foi)owed by a broad peak of which the taii extends to approximately 250 K As discussed beiow. EEL spectra of the surface on which ethyiene is adsorbed and which is anneated to various temperatures show that the higher-tem­ perature peak corresponds to the desorption of di-<τ-bonded ethyiene. whi)e muitiiayer ethyiene desorbs in the )ower-temperaturt peak. The 'muitiiayer* peak shown in Figure i a actually corresponds to oniy deaorption from a second iayer. This mui­ tiiayer peak does not saturate with increasing ethyiene exposure, however, and is suf∏cient]y intense foiiowi∏g a i 3-iangmuir ex­ posure of ethyiene that it obscures the desorption peak due to chemisorbed ethyiene. For iower exposures of ethyiene, beiow 0.6 iangmuir. the hy­ drogen theπnai desorption spectra are quite different. The thermai desorption spectrum of Hi after an ethyiene exposure of 0.2 iangmuir. shown in Figure ic, contains a high-temperature taii terminating beiow 600 K that is due to methyiidyne decomposition and a rather broad peak centered at 420 K that shifts to iower temperature as the initiai surface coverage of chemisorbed ethyiene increases. The iatter is essentiaiiy identicai with that which is observed after adsorption of hydrogen on the ciean Ru(00i) surface?* The maximum rate of Hi desorption shifts from 420 (32) Shimizu H.; Chfistmann. K ; Ertl. G. 7. Cam/. 1930. 6/. 4i2. D-4 ff,7∕r ci o/. y. Am. C⅛em. Soc. ko/. 30⅛. JVc. /J. 1M⅛ Ta⅛⅛ L Comparison of Vibration Frequencies of Multilayer CjH⅛ Adsorbed on Ru(001) at 80 K with C]H4(%)- C]H⅛(1), and C)⅛(s)^ C,H,(t)" no∕r^r. mod. Raman tR Raman IR C,H.(,)^ iR "I A, "082. ".(CH,) 3026 f f 2989 30t9 30t6 2983 2973 ",B,, r,B2, ".(CH,) 3t03 f f 3105 3075 3085 3075 C,H.(g)" multilayer C)H⅛ on Ru(001 ) 3000 3093 *2 A, "(CC) t623 f t⅛2t t620 16t⅛ t⅛30 **) A, ",t B,. CHï sets t342 f f 1444 t340 1339 t437 1336 t438 1350 1460 4 CHι twist f f f f 950 f 810 949 f "⅛ B,, CHj rock ι B,. ", B⅛ CH,wag -ιoto t239 828 960 t236 n.r. 827 970 860 970 943 440 ".(RuC) muttitayer C,D. ∣Mt Ru(00t) no./repr. mode C,D.(g)" "; A, "II B⅛ ".(CD,) 2251 2200 23 tθ ft B,, ",B⅛ ".(CD,) 2304 2345 n.τ. 2 A, "ICC) 15t5 1550 "t A, CD3 sets. 981 1078 tθt5 1175 '4 A, * CD: twist CD3 f<x⅛ 72S t009 5S6 n.r. a.r. <7 B,. ", B⅛ CDi wag 720 7S0 735 ^!2 R⅛ ".(RuC) * n i * nor reactved 425 f * forbidden K for an ethylene exposure of 0 2 tangmuir to 395 K for a 0.4tangmutr exposure and then drops to 355 K for atl ethytene ex­ posures exceeding 0 6 tangmuir. The decrease in peak temperature for exposures of ethytene betow 0.6 tangmuir is indicative of second-order desorption kineues of surface hydrogen. For these tower exposures of ethytene. betow 0.6 tangmuir, no desorption of motecuiar ethytene is observed. Figure 2a and 2b shows t⅛e EEL spectra of 4 tangmuir of C;FL and 3 tangmuir of C,D.. respectrvety, adsorbed at 80 K on the Ru(00t) surface. Consistent with the thermal desorption spectra, a comparison of the observed energy toss features to IR and Raman spectra of gaseous, tiquid. and sotid ethytene (cf. Tabte irrjs) demonstrates that the overtayers of Figure 2 correspond to motecuiar mu)tilayers. ħiote the intense CH, wagging mode at 970 cm ', which is the best fingerprint of motecuiar ethytene multilayers, and also the carbon-hydrogen stretching mode at 3000 cm '. The frequencies of both these modes are characteristic of an sp^-hybr⅛dized carbon atom. Table 1 also shows that the isotopic shifts for multilayer C,D, on Ru(00t ) art in good agreement with those of C,D,(g). Annealing these overtayers to H0 K desorbs the muttilayer, as shown in the thermal desorption spectra (cf. Figure ia). teaving di-<r-bonded ethytene which is stabte to i 50 K. The EEL spectra of this chemisorbed species are exhibited in Figure 3a for C2H< and Figure 3b for C^D^. The rehybridization of the carbon atoms to neariy sp* is reflected in the shifts of the CH, wagging mode and the carbon—hydrogen stretching mode to t !45 and 2985 cm ', respectivety. The shoutder at t040 cm ' is probabty due to the carbon-carbon stretching mode, but it is poorty resolved from the CH] twisting mode in C2H, and the CD2 wagging mode in C2D^ both of which are at 900 cm '. A carbon-carbon stretching frequency of t040 cm ' is consistent with the rehybridization of the carbon atoms of ethylene to sρ∖ Other modes of di-u-bonded (33) Shiπunouctu. T .VSAC⅛S-ΛΑ5 t972. 39. 74 (34) Brecher. C.- Ht)ford. R. S 3. C⅛em R⅛yj )96t. 33. [ [09 ENERGY LOSS, cm*' f⅛we 2. EEL spectra of motecuiar muttilayeτs of ethytene on Ru(00)): (a) 4 tangmuir of C2⅛ at SO K and (b) 3 tangmuir of C,D, at SO K. ethytene are the symmetric ruthenium-carbon stretching mode at 460 cm ' (420 cm^' for C2D^). the CH2 rocking mode at 775 cm ', the CH2 twisting mode at 900 cm^' (700 cm ' for C2D.). and the CH2 scissoring mode at t450 cm ' (t2tθ cm ' for C.D^) The CD2 rocking mode of di-<τ-boπdcd C,D, was not resolved in Figure 3b due to the poorer cutoff in the etastic peak. The symmetric ruthenium-carbon stretching mode of di-<r-bonded D-5 ∕π∕frac∏<9∏ o/ Λu(OO7J S^ζ∕ac^ y. CA^w. 5of.. Pb/. 70^. /Vo 7J, ?9#6 T⅛⅛⅛ ∏. Compariaon of Vibration*! Frequeacie⅛ of Di-σ-bonded C^Hg on Ru(00t) at !30 K with Other Chemisorbed Ethyienc Species. Gaseous Ethyiene Compounds. OrganometaUic Ethy!ene Compounds, and a Surface Mcthy!ene Species^ dι-<r-C,H. on Ru(00)) Ni("0)' NiOOfFeOlO)' mode CnHa (or CH?) '.(CM) 400 '.(CM) ∏.r. 775 CHιroc⅛c CHiïwist 900 CH^wag ((45 CH2 M1MOΓS !450 '.(CH,) 2940 '.(CH,) MM r(CC) t040 mode QD4 '.(CM) '.(CM) CD, rock CDιtwist CD,wag 420 ∏.r. 7(5 8M ((45 (435 MM n.r. n.r. 450 6(0 720 880 πto MM MM n.f t200 480 4!0 720 9!5 !!05 !4!0 2960 n.r. !25O 390 n.r. 6(5 725 925 ]235 2(70 2290 π.r. 420 590 6M 740 S70 (200 2170 MM n.f. 440 n.r. 635/540 700 8M t040 2(75 n.r. H60 '.(CD,) '.(CD,) '(CC) 420 n.r. n.r. 700 900 !2!0 22!0 2295 n.r. * f * forbidden π.r. * not rcaoived. Fe(tt()' 580 450 n.r. 870 n.r. !385 MM n.f. ∏!5 ENERGY LOSS, cm*' Flgatra 3. EEL spectra of dt-e-⅛onded ethyiene on Rυ(00i) (a) 4 ιangmutr of C,H, annea(ed to 139 K and (b) 3 ιangmuir of C,D, aπneaιed to (23 K. C,D. atso contains a sma)t contribution from v(Ru-CO) The symmetric and asymmetric carbon-hydrogen stretching modes of C,H. at 2940 and 3050 cm ' were resotved in spectra simitar to Figure 3a. Peak assignments for di-<r-bonded ethy)ene (both C,H. and C,D,⅛ on Ru(00t) are compared in Tabte 11 with those of other chemisorbed ethytene species as wet] as IR data for C,H.Br,(g). C,H.(g), Zeise s satt, and a tow-vatent nicket comptes [Ni(C,H.).].'*''-"-"-*' A comparison of the frequencies of the modes of di-e-bonded C,H< on Ru(00! ) to these data shows that ethytene undergoes rehybridization on the Ru(00t) surfaces as on Ni(i )0), Ni(l 11 ), and Fe(! 10 ⅛'ι- ' The existence of a rr-bonded ethytene admotecute can be exetuded by a comparison to the tR data for (33) Htrstsht. J Sperrroc8tm A."tu. fa/-.' A 14⅛4 23A. 749 (36) Oztn. G A. 7 Am C⅛em. Sor. (97g. ∕(M. 4730 Pt(!H)' C,H<(g)" gauche C,H,Br,(g)" 470 560 660 790 980 MM MM 3000 1060 826 f 949 !444 MM 3096 (623 848 Π04 1278 t420 2953 3005 M19 450 n.r. n.r. 600 740 ΠM 2!M 2250 900 586 f 720 MM MM 2304 t5(5 7t2 79! 947 ∏41 2t74 227] tθ(4 K[PtCI,(C,H,)]H,o" Ni,(C,H,)" CH^ on Ru(00!)** 376 π.r. n.r. 785 84] 9(0 975 (5t5 MM MM !243 (]80 (208 2880 2908 !488 !155 2965 n.r. Zeise's satt and Ozin's nicket comptexes. which have, among other differences, higher frequency CH, rocking modes and carboncarbon stretching modes. A comparison of EEL spectra of di-<τ-bondcd ethytene with spectra of diazomethane ted to an originat assignment of the spectrum of Figure 3a as a bridging methytene species.'^ However, a subsequent review of these and other spectra has shown that the reaction of CH,N, to form C,H, and N, may occur in the gas dosing tines prior to introduction into the UHV chamber. An assignment of EEL spectra of uncontaminated diazomethane, which produces urCH, groups on Ru(00!), is tisted in the tast cotumn in Tabte H. Additions) vibrationat data concerning bridging methytene may be found etsewhere "*" The adsorption of C,H⅛ on Ru(00) ) with annealing to t tθ K produces di-<r-bonded motecutar ethytene, which can be distinguished from μ,-CH, by the intense CH, twisting mode at 900 cm '. In agreement with this conctusion. the thermat desorption spectra of ethytene on Ru(00t) show desorption of motecutar ethytene up to 250 K. In an effort to describe further the character of the di-<τ-bonded ethytene on Ru(00t), isotopic exchange, thermat desorption ex­ periments of ooadsorbed C,H. and C,D< were carried out In ait cases, onty C,H, and C,D, desorbed from the muttitayer. How­ ever. alt five isotopicaiiy tabeted species (C,H.. C,H3D. C,H,D,. C,HD,. and C,D.) appeared in the di-e-bonded ethytene that desorbs motecutarty. Figure 4 shows the retative ratios of these five species that desorb motecutarty above ]50 K These ratios were obtained by correcting the areas under the thermat desorption peaks of the 26—32-amu spectra, both for the cracking patterns of the five species and for the reiative sensitivity of the mass spectrometer to each species. (These ratios exclude muttitayer C,H< and C,D..) Figure 4 shows that isotopic exchange is timited, and no mixed species (C,H,D. C,H,D,. or C,HD,) is favored over the other two. On the other hand, the corresponding H,∕HD∕D, thermat desorption spectra exhibited comptete isotopic exchange The above resutts suggest that the isotopic mixing observed for (37) George, P M.; Avery. N R.; Wctnbefg. W H Tebbe. F N. 7. Am C8rm Soc (983. ∕∣75. 1393. (38) EEL spectra of 2.5 (artgmutr of H,CN, doaed at 80 K. aoneated to (92-316 K. and coo(ed to 80 K prior to measurement (39) Fox. D J.; Schaefer. H. F 7. C⅛rm P⅛yr ]983. 78. 328. (40) TbeopokL K H.t Bergman. R G 7 Am C7tem Soc t98!. /03. 2489 (4() Oxtoo. ! A.; Powett. D B.; Sheppard. N.. Burgess. K ; Johnson. B F Gt Lewts. J C⅛em Soc . <7∕tem. Commun (982. 7(9 (42) McBreen. P. Ht Ertey. W t (bach. H Sur/ Set (984. /48. 292 (43) Chang. S -C.t Kafaft. Z H ; Hauge. R (31 Bittups. W E . Margsve. J L 7 Am C⅛rm ^oc (985. /07. (447 D-6 M∕∕j y. Ί/π. C⅛em. Soc.. Fc∕. ∕0⅞. /Vo. /3. ∕P⅛6 a/. TaMt Μ. Comparison of Vibrations! Frequencies of Ethy!idyne CCH) on Ru(00!) at 280 K CCH, on Pd(]∣l) at 300 K ' CCH3 on Pt(lll)at 300 K' CCH, on Rh(lll) at 300 K" (CO),Co,(e,-CCH,)" p,(CM) A, '.(CM) E P(CH,) E *<CC) A, 4.(CH,) A, 4.(CH,) E '.(CH,) A, '.(CH,) E 480* n.r.* 1000 1140 )37O 1450 2943 3045 409 n.r. n.r. 1080 s 1334 vs 1400 2900 m n.r. 435 vs 600 w 980 sh H30 vs !355 vs 1420 sh 2920 3050 450 n.r. n.r. !130 1330 π.r. 2900 3000 40t 555 !004 1163 ! 356 1420 2888 2930 '.(CM) A, '.(CM) E ι(CD,)E ΙCC) A, 4.(CD,) A, 4.(CD,) E 480 n.r. 800 1150 1000 nr. 2190 2280 nr. n.r. n.r. Π20 n.r. n.r. 218! n.r. mode '.(CD,) E Deuterated Species 410 -600 790 1160 990 1030 2080 2220 393 536 828 !!82 1002 1031 n.r. 2(92 * π.r * not resolved. *!denttfted from spectra simi!ar to that of Figure 3a but without CO contamination. C,0. C,O,H C,¾H,⅞DH,C,H, ENERGY LOSS, cm^' ISOTOPICALLY LABELLED SPEC!ES ∏gwa 4. Coadsorption of C,H. and C,D,. Relative coverages of C,H^ C-H}D, C,H]t⅛ C,HD⅛ and t⅛D. from thermal deaorptioo spectra: (a) O.⅛-iangmuir exposure of C,H. fo!!owed by 3 ιaag∞uir of C,D. at ! [0 K and (b) I-langmuir expoaure of C,H. followed by 3 !angmuir of C,D. atH0K. chemisorbed ethylene that desorbs molecular)y resuits from ex­ change between an ethylene admolecule and a hydrogen (or deuterium) adatom, since the onset of desorption of the mixed molecular ethylene species (C⅛H,D^ I < x ≤ 3) coincides with the initial decomposition of ethylene to ethylidyne. acetylide, and surface hydrogen via carbon-hydrogen bond cleavage. Annealing din-bonded ethylene to 250 K produces two new carbon-containing surface species as well as hydrogen adatoms. The modes due to surface hydrogen were not resolved in the corresponding EEL spectrum shown in Figure 5a. The weak losses of hydrogen adatoms, which occur at 845 and HU cm'.** were obscured by various carbon-hydrogen and carbon-carbon modes. However, the presence of hydrogen adatoms was con­ firmed by stoichiometric considerations and hydrogen postad­ sorption experiments which will be discussed later. The two hydrocarbon fragments present on the surface have been identified (44) Barteau, M A ; Broughton. J Q ; Mettre!. D -Sur/ -Sri t983. /33. 443. Ftgwv 5. EEL spectra of 4 langmuir of C,H, adsorbed on Ru(00l) st SO K and annealed to (a) 280, (b) 360, and (c) 300 K. Spectrum a exhibits the modes of both ethylidyne and acetylide. Spectrum b is characteristic of acetylide. Spectrum c corresponds to methylidyne. as ethylidyne and acetylide from the EEL spectrum of Figure 3a. the corresponding EEL spectrum of the deuterated species, offspecular EELS measurements, and EEL spectra measured fol­ lowing annealing to various temperatures Pealt assignments for CCH, and CCD, are compared in Table HI to fR data for a tricobalt ethylidyne complex as well as EELS results for ethylidyne adsorbed on various close-packed groups 8-10 metal surfaces?-^"-* A thruthenium ethylidyne complex has also been synthesized, but no relevant fR data have been published.* For all of the ethylidyne adspecies listed in Table !H. the carbon-carbon stretching mode produces a strong, dipolar enhanced peak- By analogy to the structure of the thruthenium and tricobalt organometallic compounds, and considering the relative intensities of the (dipolar-enhanced) carbon-carbon stretching modes, the carbon-carbon bond axis of each of the (43) Skinner. P.: Howard. M. W.; Oxton. 1 A.; Kettle. S F A : Powell. D B.; Sheppard. N. 7 C⅛ent. S⅛c.. Fere4ey Trent 2 I VS ! 77. ]2O3 (46) Sbeidrick. G M.: Yenr⅛owι⅛]. 3. p y C⅛ent. 5or. De/lon Trent 1973. ⅛73. D-7 /nreeacrtort // ^(⅛y∕eπe wir⅛ 7!⅛(0t)∕) Aur/oce Ta⅛)e ΓV. Comparison of Vibrations! Frequencies of Acetylide* mode CCH on Ru(001)at 360 K CCH on Pd(100) at 400 K" " ".(CM) ".(CM) 4(CH) "(CH) "(co 435 ∏.r. 750 2960 1290 n.r. n.r. 750 3000 1340 mode CCD CCD ".(CM) ".(CM) 4(CD) "(CD) "(CC) n.r. n.r. 550 2210 1260 n.r. n.r. 540 2220 1340 *n.r. * not resotved ethyiidyne adspecies is nearly perpendicular to the surface pιane. A comparison of the EELS losses for CCH] and CCD] on Ru(001) with 1R data for (CO),Cθ](u3-CCH,) and (CO),Co,(∕t]-CCD]) (cf. Tabie Hι) shows that the structure and bonding of the ethyiidynes in the two cases are quite simiiar. The acetyitde species is characterized by a carbon-hydrogen bending mode at 750 cm^'. a carbon-hydrogen stretching mode at 2960 cm*', and a carbon-carbon stretching mode at f290 cm*'. The vibrations) modes of acetylide art partiaUy obscured by the ethyiidyne modes because the ratio of ethy)idyne to acety)ide present in Figure 4a is approximately 3:2, on the basis of hydrogen therma) desorption measurements. Annealing this overlayer to 360 K decomposes the ethyiidyne. leaving acetylide, carbon, and a small concentration of hydrogen adatoms on the surface. Thus the modes of the acetylide are completely resolved in spectra measured after anneating the overlayer to 360 K (cf. Figure 3b). This acetylide also forms from the thermal decomposition of acetylene and is discussed in greater detail in the following paper. We merely note here that these assignments agree quite well with those of Kesmodel et al. for acetylide on Pd( 100) for which 6(CH) is 750 cm*'. r(CC) is 1340 cm*', and <fCH) is 3000 cm*'."" The EEL spectra of the deuterated acetylide show that the carbon-deuterium bending mode of CCD downshifts to 550 cm*' from 750 cm*' for CCH (cf. Table ΓV), which compares well with the value of 3(CD) of 540 cm*' for CCD on Pd(!00)." * We also observe a slight shift in r(CC) from 1290 cm*' in CCH to 1260 cm*' in CCD and the expected shift in "(CD) from 2960 cm*' in CCH to approximately 2210 cm*' in CCD. These losses persist up to 380 K where cleavage of the carbon-carbon bond of the acetylide occurs, forming surface carbon and methylidyne The adsorbed methylidyne is identified from the EEL spectrum of Figure 5c with y(RuC) at 465 cm*'. 3(RuCH) at 810 cm*', and v(CH) at 3010 cm*' of which the latter two are significantly higher than the corresponding modes of the acetylide. The dis­ appearance of the 129O-cm^' carbon-carbon stretching mode of acetylide upon annealing to 500 K also assists us in distinguishing acetylide and methylidyne The mode at 570 cm*' in Figure 5c is the carbon-metal stretching mode of carbon adatoms and/or carbon dimers The broad feature at 1100-1600 cm*' may be attributed to the carbon-carbon stretching mode of these dimers. The vibrational modes of the methylidyne are compared in Table V to those of high-temperature (7^ > 400 K) methylidynes on various transition-metal surfaces, as well as thcobalt and tri­ ruthenium U]-CH complexes Our assignments agree quite well with those of methylidyne adsorbed on the Fe(l 11). Ni(] 11). and Pd(l 11) surfaces and are consistent with those of the organometallic complexes The isotopic shifts of the vibrational (47) Farmeter. J E : Hills. M M.; Weinberg. W. H 7 Am. CAe*∏. Soc.. (4%) Kesmodel. L L Fee -Sr; 7⅛e⅛ao∕. A t⅝S4, 2. !O83 (49) Kesmodet. L L.: Wtdill. G D : Gate*. J A. Sw∕ So !9⅛4. /35. 4⅛4 (50) Demath. J E ; !t*c⅛. H Sv/ So. 197X. 78. L238 (51) Eriey. W.; McBreen. P H.; lb⅛ch. H 7 Cola/ 1983 54. 229 (52) Oxi or.. 1. A SperfrofAor oto.'o. Parr A 19X2 35. 18] (53) Howard. M W.; Keltic. S F : Oxton. 1 A.: Powell. D B : Sheppard. N ; Skinner. P. 7 Chew. Soc . Faraday Tram 2 19X1 77. 397 7. r<m. C⅛rrπ.Soe.. Fa/. ∕0⅞, .Vo ∕∕. 79⅛⅛ modes of methylidyne adsorbed on Ru(001) are also in agreement with those of (CO)<Cθ](urCD) All of the methylidyne modes are eliminated by annealing the Ru(001 ) surface above 700 K. After this annealation. weak modes near 600 and 1100-1600 cm*' are observed which are attributed to "(RuC) of carbon adatoms and dimers and v(CC) of carbon dimers. These modes arc present in the EEL spectra of both C⅛H< and QDi. No other features are present in the high-temperature EEL spectra, supporting the thermal desorption results which show complete desorption of hydrogen (and ethylene) below 700 K. Bearing in mind what we have learned from EELS concerning the decomposition of ethylene on Ru(001), we now return to a more detailed analysis of the thermal desorption spectra. In the case of ethylene adsorbed on the Ru(001 ) surface, three reactions generate surface hydrogen below 400 K. namely, (1) (⅛H< deh­ ydrogenation to CCH(a) and 3H(⅛). beginning at 150 K; (2) C⅛H< dehydrogenation to CCH](a) and H(a). also beginning at 150 K; and (3) CCH] decomposition to 2C(a) or C,(a) and 3H(a), be­ ginning at 330 K. As shown in Figure lb, a large ethylene exposure produces a sharp peak at 355 K with a shoulder near 400 K in the hydrogen thermal desorption spectrum, followed by a long, high-temperature tail. Since the higb-tetnperature tail corresponds exclusively to methylidyne decomposition and rep­ resents a small fraction of the total hydrogen that is desorbed (approximately 10%), a hydrogen mass balance requires that the desorption which occurs below 500 K corresponds to surface hydrogen from acetyitde and ethyiidyne formation as well as from ethyiidyne decomposition. For rather low exposures of ethylene, below 0.6 langmuir, the hydrogen thermal desorption spectra are quite different, although the EEL spectra of all coverages of ethylene adsorbed on Ru(001 ) are qualitatively the same. The thermal desorption spectrum of hydrogen following a low ethylene exposure contains a prominent peak (which shifts as a function of coverage) and also a hightemperature tail (cf. Figure )c). As shown by the EELS results, the tail corresponds to hydrogen desorption that is limited by methylidyne decomposition. The hydrogen desorbing in the major peak is due to surface hydrogen from ethylene decomposition to ethtylidyne and acetylide and ethyiidyne decomposition to surface carbon, as is the hydrogen desorbing in the 355 K peak following higher ethylene exposures. The shift in this peak as a function of ethylene coverage indicates that the desorption of this hydrogen is desorption-limited. This is confirmed by experiments conducted on the carbonaceous residue that remains after annealing to 700 K the ruthenium surface which had been exposed to 0 4 langmuir of C,K) at 100 K. Hydrogen was adsorbed on this carbonaceous residue at 90 K, and then a thermal desorption measurement was carried out. The thermal desorption spectra showed that the major hydrogen thermal desorption peak was repopulated, confirming that this peak is due to desorption of surface hydrogen The sharp peak at 355 K in the theπnaι desorption spectra of hydrogen following an exposure of ethylene exceeding 0.6 langmuir consists of surface hydrogen formed from ethylene decomposition at lower temperatures ( 150-280 K) and driven by the presence of that hydrogen from ethyiidyne decomposition. The maximum rate of the latter occurs at 355 K. The high-temperature shoulder on this peak (near 400 K) corresponds to desorption of residual surface hydrogen. That the sharp peak at 355 K and its hightemperature shoulder are derived from surface hydrogen has been confirmed by hydrogen postadsorption experiments First, the Ru(001) surface was exposed to 5 langmuir of C]H. at 90 K. annealed to 800 K, cool&i to 90 K, and exposed to 30 langmuir of hydrogen. A subsequent thermal desorption spectrum (Figure 6b) shows a peak at 355 K with a high-temperature shoulder. A comparison with the hydrogen thermal desorption spectrum after an exposure to 5 langmuir of C,H< (Figure 6a) shows that less hydrogen is present in the 355 K peak of Figure 6b and that the leading edge of the peak in this spectrum is not so sharp These differences are due solely to the presence of ethyiidyne decom- D-8 ∕∕'∕∕r er a/. y.y4m.C⅛ent.Soc.. Fol. 108, A'o. 13. /986 CHM Ru(00l) CH on Fe())))i CH on Fe())0)'^ CH on Ni())))'-" CH on Pd()H)''° (CO),H,Ru,(u,-CH)" (CO).Co,(μ,-CH)" v(CH) ⅛(MCH) (e) '.(MC) (a.) '.(MC) (e) 30)0 810 3015 795 3050 880 2980 790 3002 2⅛2 29⅛8 894 304) 830 '(CD) 3(MCD) (c) ".(MC) (a,) '.(MC) (e) 2250 613 HM 680 415 696 n.r. 410 mode 465 670 7)3 Π.Γ. 424 4!7 Flgwe 6. Hydrogen thtπna) desorption following (a) exposure of 5 Iangmuir of C,H, and (b) exposure of 5 Iangmuir of C,H. foUowed by annea!ing to MO K. cooling to 90 K. xnd exposure to 30 Iangmuir of H. position in Figure 6a and its absence in Figure 6b. in a second experiment, the ruthenium surface was exposed to 0.4 iangmuir of C,H,. foUowed by i )angmuir of at 90 K. A subsequentiy measured hydrogen therma] desorption spectrum (Figure 7a) was unlike that foUowtng an exposure of 0.4 iangmuir of C,H, (Figure 7b). Rather, it appears qualitatively stmiiar to that observed after an exposure of ! tangmiur of C,H, (cf. Figure i b) in that both spectra contain sharp peaks at 355 K. The major difference between the two spectra of Figure 7 is that more hy­ drogen adatoms are present in spectrum a. and this is reflected in the much more prominent high-temperature shouider. Fur­ thermore. more et⅛ybdync is formed reiative to acetyiide foUαwing the postadsorption of hydrogen, suggesting that this branching ratio is a function of hydrogen coverage. This wiiι be discussed in greater detai) in section [V To summarize, at a)) coverages ethy)ene adsorbes in a di-<rbonded configuration that decomposes to ethy)idyne. acety)ide, and surface hydrogen above 150 K. The ethy)idyne dehydroge­ nates above 330 K. generating additional surface hydrogen. The surface hydrogen desorbs at a temperature which decreases with increasing coverage foι)owing ethy)ene exposures be)ow 0.6 )angmuir and at 355 K fo∏owing higher exposures of ethy)ene. The acety)ide decomposes to carbon adatoms and methy)idyne near 380 K. FinaUy, methyιidyne decomposes, evoιving hydrogen, after annealing above 500 K. For exposures of ethy)ene be)ow 0.6 tangmuir, complete decomposition of methylidyne occurs be)ow 600 K- For higher ethy)ene exposures, the carbon adatoms (which are present in a higher concentration on the surface) stabilize the methylidyne, such that methylidyne decomposition extends up to 700 K. A plot of ethylene coverage as a function of ethylene exposure as well as a plot of the fractional coverage of ethylene which desorbs molecularly as a function of ethylene exposure is presented in Figure 8 F⅛we 7. Hydrogen thermeι desorption foιiowiπg exposures of (a) 0 4 Iangmuir of C,H, foUowed by 2 Iangmuir of H; and (b) 0.4 tangmuir of C,H,at]00K. . Sa *0.03h-itt F⅛wv 8. (a) Fractional coverage of chemisorbed C,H, as a function of exposure and (b) fractional coverage of chemisorbed C,H, that desorbs molecularly as a function of exposure. The temperature of adsorption isMK. ΓV. Discussion As described in section H!. ethylene chemisorbs on the Ru(00l) surface in a di-o-bσnded configuration (at a surface temperature below 150 K) and at 80 K condenses into a molecular multilayer that resembles the free ethylene molecule (cf. Table I). As may be seen in Figure 2. all five fR-active modes of ethylene appear in the EEL spectra of the molecular multilayer In addition, the carbon-carbon stretching mode, the CH, rocking mode, and the asymmetric CH, scissoring mode were resolved in some spectra. These modes are excited via an impact scattering mechanism Chemisorbed ethylene, which is stable below 150 K. is di-σbonded to the Ru(00l ) surface, and assignments of the observed vibrational modes are listed in Table H. This molecularly chemisorbed ethylene on the Ru(00l) surface appears to undergo a somewhat greater degree of rehybridization than it docs on the D-9 /nreracrion o/ Er7ιy∕eπe wirTi Ru(007) Surface Ni(i iθ), Ni(i ί ί), and Fe(i iθ) surfaces.'-*^' The differences in frequency between the CH; twisting and scissoring modes of C⅛⅛, on Ru(00i) and on Fe(i ί )) are not unexpected since ethyiene on Fe( Hi ) is severe)y ti]ted with respect to the surface p)ane such that two hydrogens are subject to mu)ticenter interactions, which are mainfested by a softened <*(CH) mode at 2725 cm '? This distorted geometry downshifts both the CH? wagging and scis­ soring modes. Rather iitde can be said coccituiveiy concerning the symmetry of the di-<τ-bonded ethyiene on Ru(00i). Appticadon of the dipoiar seiection ruie wouid impiy that the symmetry of the adsorbatesubstrate comptes is C,. since both the Cf⅛ rocking and twisting modes appear in figure 3a. However, EEL spectra measured off-specuiar show that these modes are targety impact excited, and therefore, the dipoiar seiection ruie does not appiy. Hence, the symmetry of di-<r-bonded ethyiene on Ru{00!) remains indeterminant. A near-edge X-ray absorption fine structure (NEXAFS) study of ethyiene adsorbed on Pt(i i i) at 90 K by Horsiey and co-workers" has shown that ethyiene ⅛ symmetτicaiiy di-<r-bonded to two p)atinum atoms with the carbon-carbon bond axis paraiiei to the surface and a carbon-carbon bond )ength of i.49 ± 0.03 A. We expect that di-<r-bonded ethyiene wouid be adsorbed simiiariy on the Ru(00i) surface. Barteau et ai.** have reported a di-a-bonded ethyiene species on Ru(00i ) with a carbon-carbon stretching mode at ] 330 cm '. Our resuits indicate that their assignments are incorrect, however, for the foiiowing reasons. First, they adsorbed ethyiene at i70 K, a temperature at which we have shown that di-<τ-boπded C^H, has begun to decompose forming a mixed overiayer of C⅛H<(a), CCH)(a). CCH(a), and H(a). Thus the modes they identify as resuiting from di-<r-bonded C⅛H4 are in fact a combination of di-<τ-bonded C^⅛. CCH3, and CCH modes. The )33O-cm*' toss which they assign to v(CC) is actuaiiy the <i,(C!i,) mode of ethyiidyne. This feature becomes more intense with further anneaiing which decomposes the C;H.,(a) and produces more ethyiidyne Second, our EEL spectra of C,D⅛ on Ru(00!) anneaied at i 70 K show cieariy that the previous assignment** is incorrect, because the i33O-cm^' )oss downshifts to i000 cm ' in the deuterated spectra, and there are no modes of comparabie intensity to the ]300-cm^' mode between i ]50 and 2i90 cm*' in xthe deuterated spectra. Barteau and co-workers** did not measure any EEL spectra of deuterated ethyiene and thus couid not dis­ tinguish carbon-carbon vibrational modes from hydrogen modes. A comparison of the EEL spectra of C;H< and C2D, is essentia) to the correct identification of these vibrationa) modes. As shown in Tabies H-fV. EEL spectra of C]D< on Ru(00i) anneaied between i i 0 and 400 K confirm our mode assignments for the three adspecies. di-<τ-bonded ethyiene. ethyiidyne. and acetyiide. Fmaiiy. we note that a carbon-carbon stretching frequency of )040 cm*' is more reasonabie than one of i330 cm*' for an sp'-hybridized hydrocarbon species, and it is consistent with the car­ bon-carbon stretching mode observed at i !35 cm*' for acetyiene chemisorbed on Ru(00i) " !t wouid be expected that r(CC) of chemisorbed ethyiene on Ru(00!) wouid be iower than this vaiue, ruiing out the assignment of the ]]45-cm*' mode to v(CC) of C;H, Furthermore, the i i45-cm^' mode shifts considerabiy (to 900 cm*') in the spectra of C⅛Dι and certainty cannot be due to <-(CC). No LEED patterns other than the ( i × i ) of the substrate were observed for the ethyiene overiayer between 90 and 300 K. Hence LEED measurements cannot aid in a determination of the ethyiene surface structure or absoiute coverages. However, thermai de­ sorption resuits for C-H. and H, have been used to estimate the ethylene coverage using the known saturation fractionai coverage of hydrogen (0 85) " The ethyiene coverage (excluding the muitiiayer) is presented as a function of ethyiene exposure in Figure 8a. From this figure, we see that the saturation (fractionai) coverage of chemisorbed ethyiene is approximateiy 0.30. Figure 8a was used aiso to obtain the inittai probabiiity of adsorption 7. Am. C⅛em. Soc.. To/. 706. 7Vo. 73. 798⅛ of ethyiene at i 00 X. which was found to be unity within the iimits of expérimentai uncertainty. The activation energy of desorption (equai to the heat of adsorption) of di-<τ-bonded C^H^ was esti­ mated from the thermai desorption measurements. When the method of Redhead** is used and a preexponentiai factor of ]0"-i0" s^' is assumed, a vaiue of approximateiy !i.6 ± i kcai/moi is obtained. Considering the changes in bond strengths due to rehybridization of the carbon atoms, we have aiso estimated that the binding energy of dj-a-bonded ethyiene is between i05 and !35 kcai/moi. Thus the observation q/ a 7ow hear q/ ad­ sorption /or chemisorbed et⅛y∕ene on Ru(007) does not impiy that the rut⅛enturrt-carhon bond is weak. When the saturated overiayer of chemisorbed ethyiene is an­ neaied to 250 K, approximateiy 20% of the ethyiene desorbs, whiie the remainder dehydrogenates to ethyiidyne, acetyiide, and surface hydrogen. The stoichiometry of the ethyiidyne formed by ethyiene decomposition on Pt(i ! i) was confirmed by hydrogen thermai desorption spectra in which approximateiy 25% of the totai hy­ drogen desorbed from the surface at 300 K, the same temperature at which ethyiidyne was shown to form via EELS.** Unfortunateiy, the hydrogen from ethyiidyne formation remains adsorbed on the Ru(00i) surface, ultimate)y desorbing with hydrogen from ethyiidyne decomposition. Hence we are unabie to confirm <7irecr7y the stoichiometry of the ethyiene decomposition products from hydrogen thermai desorption spectra. The ethyiidyne on Ru(00i) begins to decompose at approximateiy 330 K, whereas the ethyiidyne formed on Ft(i i i) is stabie to 400 K? The reduced stabiiity of ethyiidyne on Ru(00i) is undoubtediy due to the stronger metai-hydrogen and metal-carbon bonds formed on the ruthenium surface, which makes the decomposition of ethyiidyne via metai-hydrogen and metai—carbon bond formation more fa­ vors b)e both thermodynamicaiiy and kineticaiiy. The identification of ethyiidyne and acetyiide as the decom­ position products of di-σ-bonded ethyiene on Ru(00i) can be contrasted to the resu)ts of Barteau et ai?* who oniy identified ethyiidyne. We also note that Barteau assigned the p(CH,) mode of ethyiidyne to a ioss at 870 cm*'. We observed no such mode in our EEL spectra and cannot account for this discrepancy but mere]y mention that our assignment of the i000-cm*' ioss to p(CH;) and the observed isotopic shift to 800 cm*' for p(CD3) are in agreement with the frequencies observed for other ethyiidyne species. Our EEL spectra of the thermai evoiution of ethyiidyne and acetyiide on Ru(00i) and compιementary thermai desorption spectra show that virtuai)y aii of the ethyiidyne dehydrogenates to surface carbon beiow 360 X, ieaving oniy CCH(a) and H(a) on the surface. Thus an EEL spectrum measured after annealing to 360 K (Figure 5b) contains oniy the ioss features of acetyiide, permitting unambiguous identification of this intermediate. The observation of the carbon-hydrogen bending mode of acetyiide at 280 K, prior to the onset of ethyiidyne dehydrogenation, impiies that acetyiide is not a decomposition product of ethyiidyne. Further proof of this assertion comes from CO and C2H4 coad­ sorption experiments?' By analogy to aii reievant organometaiiic ethyiidyne compiexes synthesized to date, it is aimost certain that the carbon-carbon bond axis of ethyiidyne on Ru(00i) is very neariy perpendicuiar to the surface. This configuration is aiso supported by the strong intensity of the v(CC) mode of ethyiidyne at i i40 cm*' in the EEL spectrum of Figure 5a and its predominantiy dipoiar character. Furthermore it is very probabie that ethyiidyne is bonded to the Ru(00i) surface in a threefo)d hoiiow site both by analogy to the tπmetai U3-CCH, compiexes'*** and because ethyiidyne has been observed oniy on hexagonai surfaces?"' A further indication of this bonding of ethyiidyne on Ru(00i) is provided by NEXAFS resuits of ethyiidyne on Pt( ] i i )." which showed that the ethyiidyne is symmetricaιly bonded to three (56) Rcdbcad. P A Poruum t962. 203 (5S) Horsley. J A . St6hr. J 3)46 Koeslπcr. R J ∕ Chem P6γj ]9S5. #3. (58) Hiib. M M.. Parτneter. J E.I Weinberg. W H . unpubhshed resuits D-10 ∕∕t∕∕r er α/. y. .4rn. C⅛em. -Sr⅛γ. P⅛∕. ∕0⅛. Λο .'? ∕9<M pιatinum atoms with the carbon-carbon bond axis essentiaιiy perpendicuiar to the surface. Nest, we consider the effects of simυ!taneous ethyιidyne de­ composition and hydrogen desorption on the observed shape of the hydrogen thermai desorption peak. When the Ru(OOt ) surface on which ethylene is adsorbed (exposures above 0.6 langmuir) is annealed to 300 K. hydrogen desorption is observed. In these measurements, hydrogen desorbs at this low temperature due to its higher surface coverage. As the overlayer is annealed to 330 K. additional hydrogen desorbs, and ethylidyne begins to dehy­ drogenate, replenishing the supply of surface hydrogen. Thus adjacent to the decomposing ethylidyne. an area of high local density of hydrogen adatoms is formed, which accelerates the rate of desorption of hydrogen and gives the hydrogen thermal de­ sorption peak a sharp leading edge. Comparing Figure 6a and 6b, it is clear that the high-temperature shoulder on the 355 K hydrogen thermal desorption peak results from residual surface hydrogen. For these higher exposures of ethylene, the temperature of the hydrogen thermal desorption peak remains at 355 K independent of coverage. Ethylidyne decomposition dictates this hydrogen desorption since the desorption of hydrogen occurs at a lower temperature in a broader peak at higher hydrogen adatom cov­ erages (from the adsorption of hydrogen). The adsorption and decomposition of ethylene at high coverages is otherwise identical with that for lower coverages with three exceptions. First, a multilayer of ethylene forms which desorbs at HO K. Second, some of the di-<r-boπded ethylene desorbs (cf. Figure 8b). Finally, the ratio of ethylidyne to acetylide that is formed is increased. Recall that annealing to 400 K the Ru(00l) surface on which ethylene is adsorbed not only decomposes the ethylidyne but also causes deavage of the carbθιr-c⅛rbon bond of the acetylide. ieaving methylidyne and carbon adatoms. The high-temperature tail of the hydrogen thermal desorption peak after ethylene adsorption corresponds to dehydrogenation of methylidyne. Consequently, the ratio of ethylidyne to acetylide that is formed from ethylene can be obtained from the relative areas of the 355 K peak and the tail. The acetylide coverage is equal to the coverage of the hydrogen desorbing in the high-temperature tail. The ethylidyne coverage is obtained by subtracting 3 times the acetylide coverage from the coverage of hydrogen desorbing in the 355 K hydrogen thermal desorption peak and its high-temperature shoulder and dividing this number by four. We Und that a saturation exposure of ethylene decomposes to ethylidyne and acetylide in the ap­ proximate ratio of 66h40, whereas a lower exposure. 0.4 langmuir of C;Hi. yields a ratio of 50:50. Consequently ethylidyne for­ mation is favored at higher surface coverages. This result may be interpreted in terms of the number of hydrogen adatoms generated by ethylene decomposition to ethylidyne (one per QU,) vs. acetylide (three per (⅛H)). At higher surface coverages, more threefold sites, required for hydrogen adatoms, will be either occupied or blocked. Also, the coverage of hydrogen will be greater as ethylene dehydrogenation proceeds. Thus, the decomposition product that requires fewer vacant surface sites for formation and is composed of more hydrogen atoms, ethylidyne. is favored The dependence of the ratio of ethylidyne to acetylide formed upon the hydrogen coverage has also been confirmed by hydrogen thermal desorption experiments measured following a saturation ethylene exposure st 350 K. At this temperature, the hydrogen adatom concentration on the surface is reduced. This lower hydrogen coverage caused the acetylide coverage to increase by approximately 50% compared to the coverage of acetylide formed following a saturation ethylene exposure at 80 K, and annealing to 350 K. as judged by the relative intensities of the high-temperature tails in the hydrogen thermal desorption spectra. Finally, we discuss briefly the mechanism of dehydrogenation of ethylidyne- The EEL spectra measured immediately following the decomposition of ethylidyne show no enhancement of the carbon-hydrogen bending mode of acetylide and provide no ev­ idence for methylidyne formation. Thus, we can rule out both acetylide and methylidyne formation from ethylidyne decompo­ sition and conciude that ethylidyne must dehydrogenate completely to either carbon-carbon dimers [with v(CC) at I I00-l⅛00 cm^' in the EEL spectra of Figure 5b and 5c] and hydrogen or carbon adatoms and hydrogen. The observed compιete dehydrogenation of ethylidyne at a temperature at which acetylide is stable on the surface is quite important. It implies that the reaction coordinate that results in the loss of the first hydrogen atom from ethylidyne is not the one which would lead to the stable surface acetylide. A plausible but most certainly speculative scenario for the deh­ ydrogenation of ethylidyne would involve interaction with an adjacent threefold hollow site, whereas the stable acetylide (almost certainly not oriented parallel to the surface plane) has a structure that is rotated 60° with respect to this reaction coordinate. V. Conclusions Ethylene chemisorbs on Ru(00l ) in a dt-σ-bonded configuration at temperatures below approximately 150 K. When heated, competing molecular desorption, dehydrogenation to acetylide. and dehydrogenation to ethylidyne occur. The resulting thermal decomposition scheme for a saturation coverage of chemisorbed ethylene (⅛ = 0.30) may be depicted as :ow ------ C^⅛<9) tsθ-!40 * M⅛ (4-0 3) CCt⅛<a! + nt.) sso-sss s I *SM V t C.(α)(.-t.2) + 3H<al ' 2H(a) — M,⅛! sow ------- CCH(α) + 3H(a) soo * j C<a) + CH(a! soo-rooxi Cla) + '∕,⅛ta) The ethylidyne formed on Ru(00!) is less stable than on other groups 8-10 metal surfaces; e g., it begins to decompose to carbon and hydrogen adatoms near 330 K compared to the decomposition temperature of approximately 400 K observed on Pt( 111 ) and Rh(t II). Carbon-carbon bond cleavage of the acetylide occurs at 380 K, producing surface carbon and methylidyne of which the latter dehydrogenates between 500 and 700 K. Following ethylene exposures exceeding 0.6 langmuir. desorption of hydrogen occurs in a sharp peak with a maximum rate of desorption at 355 K, which is limited by ethylidyne decomposition, in a shoulder at approximately 400 K on this peak due to desorption-limited hydrogen and also in a high-temperature tail due to methylidyne decomposition. Hydrogen desorption following lower ethylene exposures becomes desorption-limited, except for the hydrogen that is evolved from the decomposition of methylidyne To summarize, ethylene adsorbed on Ru(00l ) produces both ethylidyne and the more extensively dehydrogenated acetylide Thus, the behavior of ethylene adsorbed on Ru(00! ) appears to be intermediate between the more complete dehydrogenation observed on Ni(IH) and Co(00l) and the exclusive formatton of the leas dehydrogenated stable ethylidyne species found on the hexagonal surfaces of rhodium, palladium, and platinum. Acknowledgment- This work was supported by the National Science Foundation under Grant CHE-8516615. Acknowledg­ ment is also made to the donors of the Petroleum Research Fund, administered by the American Chemical Society, for the partial support of this research. Rtt⅛try Ne. C,H<, 74-85-1; Ru, 744O-]8-8- CCH,. 67624-57-]; CCH. 29075-95-4; CH. 3315-37-5.