Studies in Surface Science and Catalysis 9 Proceedings of the Symposium on PHYSICS OF SOLID SURFACES Edited by M.L8znitka, Institute ofPhysics, Czech. Acad. Sci., Prague ELSEVIERSCIENTIAC PUBLISHING COMPANY Amsterdam - Oxford - Nnw York 1982 Published 10 co-edition with SNTL Publisher..01Technical Literature. Prague, Czechoslovakia and Society 01Czechoslovak Mathematicians and Physicists. Prague. Czechoslovakia Distnbulion of this book is being handled by the lollowing publishers lor tho U.S.A. and Canada Elsevier North-Holland. Inc. 52 Vanderbilt Av.mue New York, N.Y. 10017 lor the East European Countries. China. Northern Korea. Cuba. Vietnam and Mongolia SNTL Publishors 01Technical Literature, Prague. Czechoslovakia and Society 01Czechoslovak Mathematiciansand Physicists. Prague. Czechoslovakia for all remaining areas ElsevierScienufic Publishing Company Molenwerf 1, 1014 AG Amsterdam PO. Bo. 211. 1000AE Amstmriam. ThnNetherland!! ISBN0-444-99716-4 (Vol. 9) ISBN0-444-4HiOt-6fSt>rics) e Soc,lf1ty01C/echoslovak Mathematicians and PhYSICIstS.Prague. Czechoslovakia All "(Jhts Il'se",·d. No part of thIs publwalion may be I..produced. stored ,n a retrieval system, or transm,'''-ri '" any 'IlIrn Itr by any m.-ans. et.'ctrOOlC. mCf:hanll:al, phlttocupylOg. recording or oth," WISt!. without ltlt! plior permrsston 01tho publisher Prinh!d 10 CZedl')slov.I~la QUANTUM THEORY OF PHONON MEDIATED ADSORPTION AND DESORPTION CeroId Doyen Institut fur Physikalische Che.ie, Sophienstr. II, 08000 Munchen Z, Federal Republic of Germany For a theoretical understanding of surface reactions it is necessary to investigate the quantum mechanical movement of the reactants on given potential energy surfaces. The solid surface provides inelastic channels, which consist of electronic or vibrational excitations. Metastable helium atoms show an abnormally large de-excitation probability at surfaces as compared to gas phase scattering. The explanation of this is a dimensionality effect, which reduces the three dimensional elastic scattering fTom small particles to a quasi one dImensional one in the case of a flat surface. This dimensionality effect is ~lso operating for phonon inelasticity and leads to unit sticking probability for zero kinetic energy of the incoming particle and zero substrate temperature. If the coupling to the phonon field is sufficiently strong, many phonon transitions are dominating. In this case the gas atom is temporarily trapped in a transient resonance state, in which it exchanges energy with the phonons before decaying into the possible final channels. Time reversal yields a close relationship between sticking and desorption. Assuming perfect accommodation of the parallel motion of the adsorbate on the flat metal surface, a relation between the mean energy, the normalized speed ratio and the angular distribution of desorbing particles is derivcd, which agrecs well with the available experimental data. The developed formalism is also applied to inelastic reflection from surfaces. 1. Introduction The application of the adiabatic approximation to the electronic motion leads to the concept of potential energy surfaces. It i~ the task of chemisorption theory to calculate them for the gas - surface interaction. Surface chcmistry investigates the motion of the particles on these potential energy surfaces and the rates of transitions between different states on the same or on different potential energy surfaces. Elas\ic scattering is a (physical) part of this problem and the theory of clastic scattering from surfaces appears to be in rather good share (I) • AithouJh the problem of cal- culating or esti_ating the potential energy surface is largely un- solved, it oight be very helpful for surface chemistry to investi- gate, how the particles move under the influence of assumed forces. 1 A detailed understanding of this scattering process could even lead to a successful reversal of the theoreticai attack, i.e., deducing important properties of the potential energy surface from the scattering data provided by atomic and molecular beam experiments. A wealth of experience and results is available from the theory of reactive gas phase collisions. An important question is, if qualitatively different effects can be expected for inelastic surface scattering. One such effect has recently been observed experimentally {l} • Although it does not Involve phonons, it is the starting point of this paper, because it demonstrates a principle which is also of importance for phonon mediated inelasticity, but is harder there to observe directly in experiments. II. Reflection coefficients for metastable Ite-atoms The scattering of electronicaily excited helium atoms from gaseous atoms and molecules has been used for a long time to obtain detailed information about the interaction potentials of the colliding particles (l}.The cross sections can often be calculated from the experimental data. In the case of lte(ZlS) and He(lJS) scattering from gaseous CO the clastic cross section at a relative kinetic energy of 25meV has been found to be more than hundred times larger than the inelastic cross sectIon for the dc-excitation process{4}. The Munich group H. Conrad, C. [rtl, J. KUppers and W. Scssclmann studied the scattering of metastable Ite-atoms from a dense layer of CO-molecules adsorbed on a Pd(lll)-surface in extremely carefully performed experJments{l}. 4• They found a reflection coefficient of the order of 10- A radiative decay of the metastable lie-atoms due to a long range interaction with the surfdce was excluded by time-of-fllght measurements. One therefore has to conclude that the large inelasticitr is due to Auger dc- excitation processes, whIch are very similar to those occurin~ in the gas phase, because carbon monoxide is known to adsorb oon- dlssoclatively on Pd with the gross features of Its electronic structure being unchanged. Oeing confronted with this discrepancy between y~s phase and surface scattering behaviour, it Is useful to recdll the different nature of clastic scattering in hoth cases. lor an absolutelr flat surface clastic scattering means Sp~cuidr reflection and this is a uniquely determined final state. On thc other hand qUdntum mechanlcai elastic scattering from a potentlaJ JocdliLcd in all three dimensions is possible in anr direction. Using ho~ normallLdtlon for counting, 14 one could say that there arc 10 eutuoLnn st.ates avatLabLe for ,1.1S 2 phase elastIc scatterIng but only one state for surface elastIc scatterIng. Hearly all the eidstic channels opcn in gas phdse scatterIng are closed in the surface case and a drastic change of the clastic reflectIon coefficIent should not be surprIsIng. For a theoretical investigation of this effect the HamlJtonIan is split into five parts. This splitting turns out to be very convenient for dny kind of Ineiastic scattering from surfaces (~) and .ill be used below to study phonon scattering. (1) H =Hin + Vin_loc + Hloe + Vloc-out + Hout H describes the contInuous set of incoming scatterIng states in decaying exponentially at the surface. In the notation of second quantizatIon it can be written in thd for~l (2) Hin = Ik t k0nk0 t is the kinetic' energy, k denoles the ~omentum of the incomIng ko particle, the zero refers to inncr '1uantum states,i.e., the electronic confiljuration of the he11U1.1 olnd lhe solId before the collIsion In the case of Auger de-e~cltation dnd th~ vibratIonal configuration in the case of phonon scat terJuu, lero kInetIc energy and the i,dt LaI cOllfl'JlJr.tllon ,h'fin('s the cllergy zero. n Is the eccepetLon nur-,Iler ko ouerator for the :;LIl,· /1.;0>. " compriscs the states, .herr the loc 1','rti"lc I" loealLzcu 'W.lr the solid surface. IIcrc.1 simplr model ls .I:;SIJ""·,,, where for ('''"'ry internal confilJuration there 15 only one locdlllrd stolle IA> for the partlclcl The suuserlpt '\0 reLers lo till! localIzed lJartlclc with the total sysl"m beIu. In U,c luIt loI conf Iyuratlon "0". Tht· subscrIpt Am drnntes that tIu- 511 II 1\J('.11 lzed particle has Induccd a trdnsJtion to tlo.· cIJnfilfur.ltilln ",I". i.e., has untlerlJon" a ,\uIJer de-cltcJtatioll In llo(' 1'.151' 111101,,1' di scuss Len, Thl' finotl confIlJur.ltiun Is cloaractrrlzed "II"" I~ t d hy lilt' total « ,,,' .)1)<1 by the transltJon amplJtude Wm• aAPo,alld Am oIrc the creat lou .11". oI;slrlJellnn ouerators for the eorrc~pondJn!l stetes, Lrausi t lons h"llu','" dill .1IIt! IIl0c' I.e., penetrat Lon or the Inc..,,,in,! p.lrlil'I,· iuIo 1I", Loc.sI lied state IA> .llItI eLrst fe rrrJ~ctJon ·,rl' 'h'5criL"d II)' ll.<: lerL' 'ill-Inc' In :if'contl 'IualltlLdtion this r...dsl Magnitude and .o.entum dependence of the matrix elements V in Ak competition with the de-excitation matrix elements W. determine the reflection coefficient R. H describes the continuum of scattering out states of the particle after the inelastic transition. The hopping between these states and the localized states is given by the term Vloc-out. The latter two ter.s, Hout and Vloc-out' need not be specified for the moment. for the present purpose they can be absorbed into the Creen function C describing the effective inelasticity in A the localized state. It should be noted that eqs.(l) to (4) refer to many body states in contrast to the .ore co••on use of one particle states in connection with the notation of second quantization. Hence the described model Ha.iltonian characterizes interactions between configurations. The before_.en~loned special nature of surface elastic scattering is now contained in the density of states of the involved .any body states. The density of e!~enstates of H is one di.ensional, whereas H in loc and H have three dimensional density of states, because energy out and .o.entu. can be transferred into all three dimensions. The scattering .atrix ele.ent for elastic reflection is given byl (5) <ko-Iko+> = 1 + ilCA 1m qA C is the exact Creen function for the state lAo> and the embedding A function qA describes its interaction with the continuum of H in• It is convenient to assume that C and qA are regular functions of A the energy. This means, however, that the limit of infinite nor.ali- zation volu.e has been taken, in which case the singularities of C A and qA transfor. fro. isolated poles to branch cuts. Now the imaginary parts do no longer tell, whether they arise from one or three di.ensional densities of states and this involves a danger in cal- culating the elastic scattering probability. It can be shown that this .atters only for the states degenerate with the incoming scattering .ave function and they can be taken into account properly by introducing an additional resonance at the incoming energy (6). The final result is thenl The superscript -r- indicates that C and qA are regular functions A in the above mentioned sense. 6N is the percentage of the wave function k Iko> contained in the neighbourhood of the energy level (ko. If 6W k is unity then the scattering matrix element eq.(6) will be zero and the reflection coefficient given by as well. The physics depends now critically on the penetration potential Vin-loc. If the penetration strength is smail, i.e., if V is small, the state Iko> has a narrow width in the local conti- Ak nuum and nearly all the spectral density Is concentrated in an infinitely small energy interval around (ko. There will be nearly no coupling to other scattering states of H but s~rong mixing with the local in, continuum, because the local density of states is much higher than the density o( scattering states. The interaction of a metastable He-atom with a CO-molecule is Z._ very weak. The binding energy has been measured to be meV {_}. 6H will be practically unity, because the weakness of the potentiai k Vin-loc allows only coupling to deqenerate states. In gas phase scattering there are equally many states available for elastic and inelastic transitions at every energy, i.e., there arc no restrictions by a dimensionality effect. Eq. (5) can therefore be applied with C and qA taken as rc~ular functions. The iMagInary part of the A embedding functIon is given by! ThIs becomes very small if V gets small and the scatterinll matri. Ak element eq.(5) tends then to unity. This e.plains the large clastic reflection coefficient found in the gas phase experiments. How assume that the CO-molecules form a surface with nothing else being changed. TI,en clastIc processes w1ll be reduced to specular reflection and the dimensionality effect comes Into play. There will be infinitely m~ny states of Il degenerate _ith /ko> but only one loc state of Hi"' Therefore all the intensity has to go into thc inelastJc channels. lhis is e~pressed by the additional factor in eq.(6). lhus the proposed model gives a simple ~nd natural explanatIon of the initially surprising c~perimental findings. It should be noted that the dimensIonality effect operates only for weak cuupling. 'or strong coupling the width uf the state Iko> wIll be large and 6H small. In thIs lImIt eq.(6) becomes identIcal k wIth eq, (S). lIence for sufficiently strong interactlon non vanishIng elastIc reflectIon coeffIcIents from surfaces can be eKpected and have been found {7}. III. StIcking coefficIents In the foilowiny the vibrationai degrees of freedom of the solid surface are InvestIgated In their Influence on a scatterIng partIcle. Eqs.(l) to (8) remaIn valId wIth the necessary change in interpretatIon. H has essentIally the same meaning, eKcept that in the subscrIpt "0" now refers to the Initial vIbrational configuratIon of the solid. ~hen appiyIng eqs.(l) to (&) this InItIal phonon confIguration Is always assumed to be eKactiy specIfIed. Uefore comparIng the calculated results,e.g. the elastIc reflectIon coeffI- cIent eq.(7), wIth eKperiMent one has, of course, to perform the thermal average over all InItIal states. ThIs (non trIvIal) problea Is brIefly treated In the neKt two sections. The eIgenfunctIons of "loc dre gIven by the set{IA>lnd>} , where Ind> Is a product state of phonons modifIed by the presence of the partIcle In the locailled state IA>. This Ileans that H descrIbes the Lnteraet Lnq particle- loc phonon sytell when the forller Is localized near the surface. "out Ircludes the backscattered particle when It has eKcIted or de- eKclted one or several phonons. The eigenstates of H are the out final states for InelastIc reflection, .hIch .111 be cKamined In scction V. Surface chcmlstry cKperiments do ver~ often not distInquish between elastIc and Inelastlc reflectIon. in thIs case the relevant quantity to eKamine is the st LekLnu coeffIcIent, "hlch is defIne•• b)'1 where Hel Is gIven by eq,(7) dnd iilnel Is the pereenta"e of the fluK scattered Into "out. Assume thdt thc substrate temperaturc Is zero, I.e., before the SCdtterin~ event there 15 onl) the Lero-polnt motIon of the Lat t tce vibratIons. There Is a fal.lOUS !.roblt',. of hoI' the stlckln, coefficient wIll behave In this cdse, If the ~Inetic enery) of the lncorolnq l'i1rticle tends to zero It'}. riO(, eoufuslon erosc, because the lonl' estabLlshed ulstcrtcu-wavc dorn d:.;.ro"",.ltlon (V,,,,\)!')} predicts d vdnishinrJ st Lckln.] eoef f LcIr-ut ill LIds Llntt , which is ,,,It' clearly a'Jdlnst c"i,er!u;elltal trl,ncls llltuitilln. 6 For the here discussed model the sticking rate in the DWUA Is given by: (10) with Prho the phonon density of states. This formula has an intuitive interpretation: The particle hops into the localized state IA> with the transition amplitude YAk and then excites a phonon with the transition amplitude W • For a wave function IA> decaying m 2 Iv exponentially away from the surface , has the following behaviour Ak ncar zero kinetic energy: (II) V is a number characterLs tLc of tile shape and depth of the adsorption well, H is the mass of the sticking particle and l is the normalization length for the one dimensional continuum of "in. Inserting into eq.(IO) and dividin~ by the incident flux k/(Ml) shows that the OlGA - sticking coefficient tends to zero proportional to k. Instead of applyln'j this convcntional procedure, one could use eqs.(~) and (6), since at zero kInetic energy and zero substrate temperature H is zero, hecause lncLest ie ref Lect don would require inel eneryy transfer from the metal to the ~as atom and there arc no phonons available to supply it. Therefore the st LckLnq coefficient Is Just S = I - Hel• The .ipproxLmat Lons enter now as approxin,atiolls to the Cre"ll funet,Ion C1\. D;Ili!\ corresponds to nejIeet in<) the tel'us 1\ru,)";'\.0aAIn in III.ofor caleuLect Jnq the scat,terLnu weve funetIon.,kc.'o. nJln'J to "'I.(n) the 11"d~Jlnd'Y ,)dl't of the embedding function 'I~, 15 ZCI'O at zero kinetic eneryy dnd hence the scatterinu matrix clement eq.(5) would be unity. It 15 obvious that e'l.(5) has to he used in the D\\I;,\. because neqLectLnq the \'1"1 - terms nieens that the Green function does not contain the eouplinq to the three dlmcnsLonaI phonon continuum. Usinq thc full JoeaI :r.tuilt.Ollldll cq.(J) to calculate the Green functIon requires, however , the U~H' of eq, (il) for evaluat Lnq the forw,trd scatt.er iuu "latri, cLencnt , AccordInn to eq, (11), Vin-loc tends to zero r'Jr k t.endln.j to zcro .md this ImplIes that cll tends k to un Ity, Till, UIl..t stlckfn., prnbabfLl ty at zero kinetIc energy is now cxplalne:f as d conscquence of the dlmensLonalIty effect, Ifltleh comes into opcrat.Lon, because thr- behavLuur of Vin-Ioc ncar threshold It''dch is ofl."n referred t.o .15 "trililsmiss.ion l'robll,nl" {.IO} ) makes 7 the effective p.rtlcle surface Inter.ction weak. For this deduction of unit sticking coefficient it has been tacitly .ssuMed that the lower edge of the local continuum lies below the threshold given by zero kinetic energy. This Might .ppe.r obvious, If the w.ve function IA> is thought of .s a ordinary bound vibr.tion.l .t.te. The restriction to Ju.t one such state In the HaMiltonian Is not obviously a good appro_1Metlon. It can, however, be Justified, If the st.te fA> Is given a special nature. In order to clarify this point, the followtng for. of the locai Ha.l1toAlan Is lAvestlgated lA SOMe dchUI rw r Att,(q)C~Ct,(bqt (ll) Hloc • tf£tnt + q qbqtbq + t5' + bq) C~ and c are the creation and destruction operators for the gas atoM t In the one p.rtlcle b••ls functions, which are localized In the t .dsorptlon well. b and b are the phoftoft operators. The atOM - .olld • q q InteractloA Is lincar in the phonoft operators, which Mcans that only a••ll dlsplace..nta of the Metal atOMS are considered. The coupling const.nts define a local phofton ~de vi•• tr,' (1) m. ( l I l tt,(q»-l/l tt,(q)bq tl' The spectral resolution po(w) of the local phonon .ode will u.ually pe.k at .... phOfton frequency w • For siMplicity aSSUMe that It has o a 6 • function ahape. If the other ph."on .odes arc now orthog_.lhed tilt thc _de 0, the)' _Ill decouple and lhe local H••lltonlan can be handled by deflnln~ a. a dlaplaced local phonon ~ode r (I') Ao• trAtt.(q)<ctCt ,> All .pprod....te self-consistent version of .lloc Is Lhen "hen b)' {Ll] I 8