Collision induced processes at the gas–surface interface

Colloids and Surfaces

A: Physicochemical and Engineering Aspects 208 (2002) 187–198

Collision induced processes at the gas–surface interface

Micha Asscher a,*, Leonid Romm b, Yehuda Zeiri c

a Department of Physical Chemistry and The Farkas Center for Light Induced Processes, The Hebrew Uni�ersity,Jerusalem 91904, Israel

b Department of Chemistry, Uni�ersity of California, Berkeley, CA 94720, USAc Department of Chemistry, Nuclear Research Center Nege�, P.O. Box 9001, Beer-She�a 84190, Israel

Received 27 June 2001; accepted 5 December 2001

Abstract

The collision of energetic gas phase particles with adsorbed species can induce a variety of processes. Events of thiskind can play an important role in the mechanisms governing heterogeneous catalysis at high pressures and elevatedtemperatures. Two collision induced processes (CIP) are described in this article. The first process discussed iscollision induced desorption (CID). The CID of N2 from Ru(001) is considered at both low and high coverage ranges.The interpretation of the experimental data using molecular dynamics (MD) simulations leads to the introduction ofa new desorption mechanism involving surface corrugation and adsorbate frustrated rotational motion. The secondprocess is collision induced migration (CIM), an event that has never been considered before neither experimentallynor theoretically. It is demonstrated, using MD simulations, that following energetic CIM, very long distances ofmore than 100A� can be covered by the adsorbates at low coverages. At high coverages, on the otherhand, thesedisplacements become considerably shorter due to surface collisions with neighbors. © 2002 Elsevier Science B.V. Allrights reserved.

Keywords: Gas–surface dynamics; Collision induced desorption; Collision induced migration

www.elsevier.com/locate/colsurfa

1. Introduction

The collision of energetic gas phase particleswith adsorbates on solid surfaces can induce sev-eral different processes. The attempt to under-stand the detailed mechanism of collision inducedphenomena has been a focus of interest for morethan a decade. Early molecular dynamics (MD)simulations demonstrated that detailed study ofcollision induced desorption (CID) may be used

as a direct probe of adsorbate–substrate interac-tion potential [1]. Following this numerical simu-lation, numerous experiments have shown thatenergy transfer from a projectile to the adsor-bate–surface system can lead to desorption [2–13]and dissociation [12–17] of the adsorbed species.In addition, experiments have demonstrated thepossibility for unimolecular [18] and bimolecular[19] reactions on solid surfaces induced by hyper-thermal projectiles. More recently, it was alsodemonstrated that collision of rare gas atoms withthe H–Ni(111) system results in the transition of

* Corresponding author.

0927-7757/02/$ - see front matter © 2002 Elsevier Science B.V. All rights reserved.

PII: S0927 -7757 (02 )00145 -0

M. Asscher et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 208 (2002) 187–198188

the hydrogen from its adsorbed state to a sub-surface site [20–22]. Moreover, both experiments[23–28] and simulation [29,30] demonstratedthat ‘hot’ atoms generated in the photo-dissocia-tion of an adsorbate can lead to CID or dissoci-ation of a neighboring adsorbate. The initialconditions for the projectile–adsorbate scatteringevent in these systems are dictated by the ar-rangement of the adsorbates on the surface.Hence, this type of events was termed localizedatomic scattering (LAS) [23,24,30]. A variety ofnumerical simulations, including simple hardcube models [5,6,8] as well as moderate [8,30–34] and large-scale [35,36] classical MD simula-tions, accompanied these experiments. Theinterest in collision induced surface processes islargely due to the possibility to relate ultra highvacuum (UHV) results to the outcome of mea-surements at high pressures. In particular, acti-vated collision induced processes (CIP) , thathave a reasonable rate only at high pressures,can be simulated in UHV using energetic projec-tiles to overcome the pressure gap [37].

The number of particles colliding with thesurface increases linearly as a function of pres-sure. However, only a small fraction has highenough energy to induce most of the phenom-ena described above. The flux of gas particlescolliding with an adsorption site on the surfaceis given by F=P/(NS(2�mkBT)1/2), where P isthe pressure, m mass of gas particles, Tg the gastemperature, kB is Boltzman constant and NS

number of adsorption sites in a unit surfacearea. For most metal surfaces NS is of the orderof 1015 sites per cm2, hence, at room tempera-ture and P=760 Torr the flux of argon atomscolliding with a given site on the surface will beF=2.44×108 s−1. Assuming that the kineticenergy of the gas atoms is governed by Boltz-man distribution at Tg=300 K, only one parti-cle in the flux cited above has energy of 0.5 eVand only one out of 10 particles has energy of 1eV. The number of energetic particles hitting agiven site on the surface can be compared withthe ‘turnover number’, TN (defined as the num-ber of product molecules generated on a unitarea of catalyst in a unit time).

For most heterogeneously catalyzed reactions

under typical conditions of 400–800 K and upto100 atmospheres of reactant pressure TN variesin the range 10−2–102 s−1 [38]. This range oftypical TN values is similar to the number ofparticles with energy upto about 0.75 eV thatcollide with an adsorption site during 1 s (usingtemperature and reactant pressure ranges de-scribed above). Hence, CIP may be an impor-tant route to obtain reaction products inheterogeneous catalysis, provided that the mag-nitude of the activation energy associated withthe rate determining step is low enough and canbe supplied by the projectile. The large value ofF suggests that the probability that low activa-tion energy processes will actually take placemay be strongly pressure dependent. One suchprocess is surface diffusion of adsorbates. Theenergy barrier for diffusion (Ediff) spans a widerange from zero upto about 1 eV, but for manyindustrially important systems Ediff is of the or-der of 0–0.3 eV. MD simulations show that forsuch low activation energies it is expected thatthe high collision rate of gas particles on thesurface at atmospheric pressure will have a pro-nounced influence on the diffusion process [39].

The purpose of this article is to describe ourapproach to CIP. The emphasis in the discussionbelow will be on the detailed mechanism underly-ing these CIP. The details of the involved mecha-nisms will be based on both experimental findingsand computer simulations of the different events.We shall start by describing the CID of N2 fromRu(001) surface [9,32]. Unlike the case of spheri-cally symmetric adsorbates, where the CID re-sults can be explained using hard cube models,here, the desorption mechanism, involves a morecomplicated sequence of energy transfer eventsamong different modes. The same system will beused to discuss a new type of CIP, namely, colli-sion induced migration (CIM) [33]. In all theseexamples, we shall not discuss the details of theexperimental or theoretical procedures. The em-phasis will be on a description of the basic molec-ular-atomic level mechanism involved. For acomprehensive description of the experimental ortheoretical details the reader is referred to thereferences cited above.

M. Asscher et al. / Colloids and Surfaces A: Physicochem. Eng. Aspects 208 (2002) 187–198 189

2. Collision induced desorption: theAr–N2–Ru(001) system

In the following we discuss the details of CIDin the system Ar–N2–Ru(001). The discussionwill start by examination of the CID event at thelow N2 surface coverage limit. In this limit exper-imental data is available as well as computersimulations where a single adsorbate is consid-ered. The second part of this section will discussthe CID process when surface coverage is in-creased. The high coverage regime is describedtheoretically by employing MD simulations.

2.1. Low co�erage

The basic quantity measured experimentallyand calculated in the simulations is the crosssection, �des, for the CID process. It is obtainedfor a set of incidence energies (Ein), projectileangle of incidence (�in) and surface coverage value(�). The cross section for CID is defined as waspreviously suggested by Beckerle et al. [4] as anarea on the surface within which an impact of

rare gas atom yields a CID event per one ad-sorbed nitrogen molecule. The experimental crosssection is derived from the exponentially decayingCID signal obtained when low coverage 15N2 isexposed to a supersonic beam of Ar atoms (orKr) seeded in He [9]. The experimental and theo-retical results obtained for normal approach ofthe collider are presented in Fig. 1.

Due to experimental limitations, Ar was used asprojectile for Ein upto 2.25 eV (open up triangles)while for larger Ein values Kr was used (opencircles). This data should be compared with thecorresponding calculated results with Ar and Kras collider (filled up triangles and circles, respec-tively). Both experiment and simulation indicatethat the CID process has a threshold energy,Ein=Ethr, below which no desorption is observed.For the N2–Ru(001) system both experiment andsimulation yield Ethr=0.5 eV. This value of Ethr isabout twice the magnitude of the adsorbate–sur-face binding energy. Moreover, that the magni-tude of Ethr is independent of the incidence angle[9], see discussion below. Comparison between the

Fig. 1. �des As a function of incidence collider energy: experimental (open symbols), calculated for normal adsorption geometry(filled symbols), calculated for parallel adsorption geometry (filled symbols). The dotted line through the experimental data pointsis based on an expression described in [9–11].


Fig. 2. Experimentally measured (open squares) and calculated (up triangles for normal adsorption geometry and down triangles forparallel adsorption) �des as a function of incidence angle for the four indicated kinetic energies.

experimental and calculated results for �des(Ei,�in=0°) shows an excellent agreement for inci-dence energies upto �2.5 eV. Above this Ein valuethe increase of the experimental �des is faster thanthe calculated one. To examine the dependence of�des on the adsorption geometry, simulations wereperformed at three Ein values using parallel ad-sorption. The calculated �des for this adsorptiongeometry are also shown in Fig. 2 as filled downtriangles. It is clear that in the Ein range examinedhere the cross section for CID is independent ofthe adsorption geometry at �in=0°.

Another comparison between experimental andcalculated results is shown in Fig. 2. Here, therelationship between �des and the collider inci-dence polar angle (measured from the surfacenormal), �in, �i is presented for four Ein values.Again, very good agreement between the experi-mental (open squares) and calculated (filled trian-gles) data is observed for the case of normaladsorption geometry. For all energy values �des

exhibits a small increase as a function of �in upto�in=40°. For larger incidence angles a rapid in-crease in the magnitude of the CID cross section


is observed. Calculated cross section values forEin=2.25 eV using parallel adsorption geometryare shown in Fig. 2 for three �in values. It is clearthat for off normal incidence angles the �des val-ues corresponding to normal adsorption are muchlarger than those for parallel adsorption.

The good agreement between experimental andcalculated results indicates that the semi-empiricalPES used in the simulation allows a reliable de-scription of the Ar–N2–Ru(001) system [9].

The threshold energy (Ethr) for desorption isdefined as the minimum energy of the colliderrequired to induce desorption. As it follows fromthis definition, Ethr is closely related to the bind-ing energy of the adsorbate. Levis and co-workers[5,6,40–42] proposed a new method to establishthe binding energy of an adsorbate based on theexperimentally measured Ethr for desorption. Em-ploying the hard sphere-hard cube (HSHC) modelfor CID, the binding energy was calculated by thefollowing equation, as suggested by Kulginov andcoworkers [8]:

Ebinding=Ethreshold

4mcolmads

(mcol+mads)2[1−4madsmM

(mads+mM)2]

×cos 4��in

2�

(1)

where mcol and mads are the collider and adsorbatemasses, respectively, mM is an effective substratemass, which is equal to a few times of the mass ofa surface atom.

The simplified HSHC model provides goodagreement with the experimentally measuredquantities, Ethr (�in=0°) and adsorbate–substratebinding energy (Ebin) for N2–Ru(001) system, as-suming mM=1.5×mRu. However, it cannot ex-plain the experimental observation that Ethr isindependent of the angle of incidence that wasfound for the N2–Ru(001) system [9]. Moreover,as follows from Eq. (1), Ethr is expected to in-crease as �in increases.

An examination of the activation energy fordesorption derived from MD simulation, Edes=0.25 eV in our case, as obtained from the defini-tion of Ethr for CID, reveals two differentexperimental values to relate to. Based on TPDdata, Menzel and coworkers [10] and Jacobi [11]

claimed that the activation energy for desorptionis 0.44 eV, while later work by Romm et al. [9]reported Edes=0.25�0.05 eV. Discrepancy in thedetermination of quantitative values for the acti-vation energy for desorption extracted from TPDdata alone are well documented in the literature.Most problematic is the definition of the preexpo-nential factor that often cannot be measured inde-pendently. We believe that when CIDmeasurements coupled to MD simulations suggesta value for the activation energy that is abouthalf of the Ethr for CID it can be considereddependable. The higher value obtained by Menzeland Jacobi [10,11] is clearly too close to the Ethr

for CID. Therefore, it seems as if their value forEdes is shifted too high by their choice of a ratherlarge preexponential factor. In general, one mayuse the Ethr for CID as an upper limit for theactivation energy for desorption, with the laterexpected to be near half of the Ethr. Edes= (0.5�0.2) Ethr.

The total cross section for CID, �des, wasshown by Beckerle et al. [4,5] to increase with �in.This behavior was related to the faster increase ofthe geometrical cross section (correlates withcos�in) versus the decrease of the normal energycomponent (correlates with cos2 �in), consideredto be relevant quantity for CID within the HSHCmodel. The magnitude of the increase, however, isfar too small to explain the results observed in theN2–Ru (001) system. Moreover, the HSHCmodel predicts the same results for any adsorbedmolecule regardless of the specific details of themolecule–metal interaction potential. This isshown to be incorrect in our case where we com-pare the two model adsorption configurations ofN2-the normal and the parallel ones. The strongdependence on �in is observed only in the case ofthe normal adsorption while the parallel geometryreveals practically no dependence on the angle ofincidence, as seen in Fig. 2. The limited ability ofthe HSHC model to treat polar angle dependenceof the CID cross section is further demonstratedin the O2–Ag(100) system [12,13]. Here, �des in-creases by a factor of 40 as �in increases fromnormal incidence to 60°. This cannot be explainedby any version of the HSHC model.


The variation of the average rotational energyof the nitrogen molecules, �Erot�, as a function ofthe incidence angle for five Ein values correspond-ing to normal adsorption and one to paralleladsorption are shown in Fig. 3. In the case ofnormal adsorption geometry �Erot� exhibits a lin-ear decrease for increasing values of �in. The rateof �Erot� decrease varies as a function of Ein,namely, larger incidence energy corresponds to afaster decrease of �Erot� as a function of incidenceangle. A quite different behavior is observed forparallel adsorption geometry. In this case, �Erot�exhibits a slow increase when the incidence angleincreases. Thus, �Erot� at �in=60° is larger byabout 25% than the corresponding value at �in=0°. These characteristics of the dependence of�Erot� on �in are closely related to the CID mech-anism and will be discussed below.

2.2. High co�erage

The high coverage simulations were performedusing the potential function employed in the lowcoverage calculations except for the addition ofadsorbate–adsorbate interaction term. The mag-nitude of this additional interaction was estimated

from the changes of temperature programmeddesorption (TPD) spectra as a function of theinitial N2 coverage on Ru(001). These measure-ments suggest that, at high coverage, the adsorbedspecies repel each other [6]. The magnitude of thisrepulsion was estimated from the shift in the TPDto lower temperatures as the coverage increases. Itwas found that at monolayer coverage the repul-sion among adsorbates results in the reduction ofdesorption barrier by approximately 1 kcalmol−1. This repulsion was modeled as a sum ofpair-wise interactions between Nitrogen atoms be-longing to different adsorbates. The N�N pairpotential was, thus, described by an exponentialfunction of the form

V(RN�N)=Arep e−�repRN�N (2)

The magnitude of the parameters Arep and �rep

were determined by requiring that the repulsiveenergy corresponding to a full mono-layer of Ni-trogen molecules will reproduce the experimen-tally observed reduction in adsorbate–substratebinding. The values of these parameters as used inthe simulations were: Arep=1 eV and �rep=0.715Bohr−1. The adsorbate–adsorbate interaction,using Eq. (2) indeed results in a decrease of 1 kcal

Fig. 3. Average rotational energy �Erot� as a function of �in for different Ein. The results obtained for parallel adsorption case areshown for comparison (crosses).


Fig. 4. Variation of desorption yield as a function of incidenceprojectile energy for five different coverages (designated by thenumber of adsorbates, Nad). Nad=12 correspond to mono-layer coverage.

increases less rapidly and tends to converge to asaturation value. The value of Ydes at saturationincreases as a function of initial coverage. Theincreased desorption yields as a function of cover-age, for a given Ein value, is due to three reasons.1. The repulsive interaction among the adsor-

bates results in a decreased adsorbate–sub-strate binding. Thus, a given amount of energytransferred from the collider to an adsorbate isexpected to result in a larger desorption prob-ability when the coverage is increased. Itshould be noted that based on the experimen-tal findings the adsorbate to substrate bindingdecreases by about 1 kcal mol−1 when cover-age is increased to full mono-layer. As a resultwe expect that the Ethr for CID at mono-layercoverage would be somewhat lower than thatat the low coverage limit. Indeed, the resultspresented in Fig. 1 show that Ethresh for mono-layer of adsorbates is smaller by about 10% ascompared with the value obtained for lowcoverage (0.5 eV, see [9–11,34]).

2. The effective corrugation along the surface feltby an adsorbate increases as a function ofcoverage. Increase in the potential corrugationresults in a more efficient transfer of adsorbateenergy from its translational modes along thesurface to the one normal to the surface,hence, desorption probability increases.

3. At high Ein values, the probability to obtainmultiple desorption events as a result of asingle collision increases with the projectilekinetic energy, incidence (polar) angle and theinitial surface coverage.

These features are closely related to the desorp-tion mechanism and will be further discussedbelow.

Next we examine the response of Ydes to achange in the incidence polar angle, �in, at whichthe collider approaches the surface. The resultscorresponding to Ein=4 eV are shown in Fig. 5for the same five coverages as in Fig. 4. It is clearthat at the low coverage limit only a very weakdependence of the desorption yield on �in is ob-served. When off normal incidence angles areused Ydes increases by a few percent only. Itshould be noted that the cross section for CIDwas found to increase, in this �in range, by a

mol−1 in the adsorbate–surface binding energy atthe completion of a mono-layer (12 adsorbedmolecules on the slab in the present simulations).

The cross sections for CID at low coverage,�des, were calculated using the opacity functionobtained from the MD simulations. The calcula-tion of �des is meaningful only if there exists animpact parameter value, bmax, above which theopacity function becomes zero. This requirementwas fulfilled in the study of the low coverage limitwhere a single adsorbate was considered. Thevalue of bmax was found to depend on the inci-dence angle of the projectile but in all cases waslarger than 3.5 A� . This value is of the order of thenearest neighbor adsorbate–adsorbate distance athigh coverage. As a result, the opacity function isnot expected to decrease to zero and cannot beused to calculate the cross section for the CIDprocess at high coverage values. Hence, in thepresent study rates of CID will be related to thedesorption yield, Ydes, defined by the ratio be-tween the number of desorbates, Ndes, obtained inthe calculation of all trajectories Ntraj (i.e. numberof projectiles considered).

The variation of Ydes as a function of projectiletranslational energy at normal incidence is shownin Fig. 4 for five different coverage values. For allcoverage values desorption yield exhibit a rapidinitial increase as a function of Ein, upto approxi-mately Ein=1.75 eV. At higher energies, Ydes


Fig. 5. CID yield as a function of incidence angle at fivedifferent surface coverages.

results shown in Fig. 5 clearly demonstrate thenon-uniform dependence of Ydes on �in for differ-ent initial coverage values. For example, Ydes formono-layer coverage increases by 60% as com-pared with the corresponding value at the lowcoverage for normal incidence, while for �in=60°the ratio between the desorption yields at mono-layer and low coverage increases by more than afactor of two.

The sequence of collision events of the projec-tile at off normal angle of incidence is quitedifferent. In this case the collider normal energycomponent is much smaller, hence, its turningpoint correspond to larger Ar-substrate distances.Similar to the case discussed above, the interac-tion between the rare-gas atom and the nearesttarget adsorbate results in the deflection of theprojectile from its original trajectory. If thisdeflection results in a sign change in the projectilenormal velocity component it will proceed to thegas-phase without additional collisions with otheradsorbates. Such trajectories can be viewed asglancing collision of the collider from the adsor-bate layer. The amount of energy transferred be-tween the projectile and the target adsorbate isdetermined by the magnitude of the impactparameter, b. Kinematically, for near zero b theprojectile may lose about 90% of its energy to thetarget N2, however, for large impact parametersthe amount of energy transfer becomes muchsmaller. While the glancing collisions constitute asmall fraction of the trajectories studied, in mostcases the projectile interacts with a number ofadsorbates before it is scattered back to the gas-phase. For such trajectories one expects the col-lider to lose a large fraction of its initial energy tothe adsorbates layer. Again, the interaction be-tween the adsorbates should result in some degreeof redistribution of the energy among the targetadsorbates and their nearest neighbors.

2.3. CID mechanism

In order to gain insight into the details of theCID mechanism a large number of individualtrajectories with different impact parameter valueswere carefully examined for the energy rangeEin=0.8–5.5 eV and both normal and off normal

factor of approximately four [9–11,32,33]. At thelow coverage range, variation of �in result mainlyin the increase of successful CID events at largeimpact parameters. But at the same time a corre-sponding decrease is found in the yield of CIDevents at the small impact parameter regime.These two trends almost exactly cancel eachother, thus, leading to the observed insensitivityto �in. At the highest coverage, on the otherhand,Ydes increases by 50% when �in changes from 0 to60°. Moreover, at mono-layer coverage the des-orption yield is larger than unity for �in=60°. Inthis case, a significant fraction of the trajectoriesend up ejecting more than a single adsorbate. The

Fig. 6. A typical CID trajectory. The two ends of the nitrogenmolecule are shown as open and filled circles.


angle of incidence. A typical example for Ein=4eV is shown in Fig. 6. As the projectile ap-proaches the adsorbed nitrogen molecule, therepulsion between the collider and upper N atomrises and causes the molecule to tilt and bendtowards the surface plane approaching a parallelgeometry. The adsorbate acquires the largesttorque when the collision geometry is not line-of-centers, but with the impact parameter in therange 1–1.5 A� . Here, the collision between projec-tile and adsorbate results in a large amount ofenergy transferred into the frustrated rotationalmode of the adsorbed molecule as well as intotranslation parallel to the surface (note the largepolar angle at which the desorbate leaves thesurface, �out). Part of the energy in these twomodes is transferred into kinetic energy in thedirection normal to the substrate that in turnleads to desorption. The energy transfer into mo-tion along the surface normal is possible due tothe coupling of this mode with the frustratedrotation and parallel motion modes by the corru-gation of N2–Ru PES. Detailed dynamic pictureof the kind discussed above is of course impossi-ble within the HSHC model, where parallel mo-mentum is assumed constant.

For off normal incidence the motion ofmolecule parallel to the surface prior to its des-orption becomes more probable. The parallel mo-mentum transfer from the incoming Ar atom intotranslational and rotational modes of the ad-sorbed N2 molecule leads to the tumbling of theadsorbate along the surface [32]. This motion isagain coupled to the motion normal to the surfaceby virtue of the PES corrugation.

Thus, the dominant mechanism of the CID isdirect impulsive bimolecular collision, in whichcollider energy is transferred into the frustratedrotation of the adsorbate, its kinetic energy alongthe surface plane, and into the surface. Althoughthe amount of energy transferred into each ofthese channels is dictated by the collision geome-try, the energy acquired by the adsorbate uponcollision is effectively channeled by the corrugatedmolecule-surface PES into the motion normal tothe surface. At normal incidence as a result ofsignificant excitation of the frustrated rotationthis degree of freedom is kept by the molecule all

the ways to the gas phase following desorption.At off normal incidence the frustrated rotation isless important in the CID sequence, and desor-bate leaves the surface rotationally colder. Mean-while, at off normal incidence the kinetic energyof the collider more effectively channels into thekinetic energy of the desorbate, and the latterleaves the surface translationally more excited rel-ative to the normal incidence case.

At high coverages, the effect of neighboringmolecular interactions on the desorbate kineticenergy, rotational energy and angular distribu-tions were determined. Vibrational excitations fol-lowing the CID event were not observed in theMD simulations, however, efficient rotational en-ergy excitation is predicted which depends onboth incident energy and angle of incidence. Polarand azimuthal angular distributions were found tobe strongly dependent on the incidence angle andenergy of the colliders. These distributions arevery similar to those obtained at the low coveragelimit, indicating the dominance of the local poten-tial over interactions with neighboring adsorbates.In general, neighboring molecules act to focus thedesorbing ones closer to the normal to the surfaceand broaden the energy distributions.

3. Collision induced migration (CIM)

The effect of energetic colliders on surface pro-cesses such as desorption (CID) [1,3,6,8,9,31,43]and reaction (CIR) [4,19] have been demon-strated. These processes were considered as possi-ble new routes for surface reactivity in industrialcatalysis, where energetic gas phase molecules inthe tail of the Boltzmann distribution can affectthe heterogeneous catalytic processes [1,3,31].

With this background of collision induced sur-face processes that have already been investigated,it is rather surprising to realize that the far moreprobable CIM has never been considered neithertheoretically nor experimentally. The results dis-cussed below are based on MD simulations. Thedata correspond to CIM of adsorbed nitrogenmolecules on Ru(001) at 90 K following collisionswith gas phase argon atoms. The complimentarystudy of CID of N2 from Ru(001) has been dis-cussed above.


In the following, we define as target adsorbate(TM) the molecule directly hit by the collider. Inall the simulations the TM was chosen as theadsorbate positioned at the shortest distance fromthe slab center. In Fig 9 a–c the average migra-tion distance (AMD) of the target molecule (TM)is shown for increasing impact parameters atEin=1.45 eV, �in=0, 40, 60°, for the indicatedinitial coverage values. We define the AMD as thedistance between the position of the TM at t=0and its position after 10 ps, averaged over impactparameters within a given range. The AMD isdefined here for all the trajectories that werefound to be non-desorbing during 10 ps. Theresults clearly show that at normal incidence max-imum displacement of the TM is reached at non-zero impact parameter, bimp�1°. At this collisiongeometry the energy transferred from the projec-tile to the adsorbate is channeled most effectivelyinto lateral migration of the adsorbate. At otherangles of incidence (40° and 60°), trajectories hav-ing bimp�0 A� are the most effective to inducelong migration distances. A strong dependence ofAMD on �in is observed. The migration distancesshown in Fig. 7 reflect the remarkable efficiencyof the CIM process in this system.

As coverage increases the AMD significantlyshortens while the CIM process attenuates. This isa direct consequence of the multiple inter–adsor-bates collisions between the TM and its neighboradsorbates, which block the original direction ofmotion of the TM on the surface. For all theseangles of incidence, the AMD decreases by nearlyan order of magnitude when the coverage in-creases from 1 (�=0.018) to 12 (�=0.21)molecules on the slab.

To estimate how the CIM process can be fol-lowed experimentally, one has to integrate theAMD values (IAMD) shown in Fig. 7 over theentire impact parameter range. The values ofIAMD obtained this way were then calculated fordifferent coverage values as a function of Arkinetic energy. The results are shown in Fig. 8 fortwo angles of incidence, �in=0 and 60°.

These results clearly show that the integratedAMD is about a factor of 5 shorter than thecorresponding AMD. Moreover, the energy de-pendence is quite modest as compared with that

Fig. 7. AMD vs. impact parameter.

obtained for AMD. The reason for this behavioris that there are many more Ar trajectories atlarge impact parameters which result in smallAMD than those with small impact parametersand large AMD. As the collision energy increases,AMD values at small impact parameters increase,but at the same time the number of trajectories atlarge impact parameters also increase. The netresult is a compensation effect, process whichdiminish the overall efficiency and energy depen-dence of the CIM.

It was found that single adsorbed molecule canmigrate over 150 A� following collisions at highenergies and large angles of incidence. As cover-


age increases, inter-adsorbate collisions efficientlyquench the migration distance. At high energies,the competing CID becomes dominant, leavingbehind only low energy adsorbates which migrateto relatively short distances. This results in anoptimum collision energy for the most efficientCIM process near 2.0 eV. It was observed that thetarget molecule migrates for long distances due tothe fact that its center of mass is found to residemore than 0.5 A� above its equilibrium adsorptionposition for 2–5 ps, during which it has a verysmall barrier for diffusion [33]. An interestingopen question that arises from this study andneeds to be addressed in the future is the concep-tual similarity and difference between CIM andthermal diffusion.

4. Conclusions

A number of different CIP were described andanalyzed. The first process considered was theCID of N2 from the Ru(001) surface. The experi-mental data suggested that, in the limit of lowcoverage, the CID cross section increases as afunction of projectile incidence energy and seemsto converge at high Ein to a saturation value. Inaddition, the CID process was found to exhibit aEthr below which no desorption is detected. Themagnitude of the Ethr was found to be about twicethe binding energy of nitrogen to Ru(001). Theexperimental results also indicated that the Ethr isindependent of the incidence angle of collidersuggesting that normal energy scaling is not appli-cable in this case. At a given Ein, the CID crosssection was found to increase as a function of �in

yielding values larger by a factor upto four atincidence angle of 60° as compared with normalincidence.

The detailed understanding of the CID resultswas obtained using MD simulations. Very goodqualitative (and semi quantitative) agreement wasobtained between the experimental and theoreticaldata, suggesting that the PES used to describe thesystem is reliable. The MD calculations indicatedthat the desorption mechanism involves energytransfer among desorbate modes due to potentialcorrugation and coupling between translationaland frustrated rotational motions. Similar desorp-tion mechanism was found to operate in the caseof high coverage. In the case of high coverage thedesorption yields increased and for grazing colli-sions, �in=60°, we obtained desorption yieldslarger than unity. These results suggest that someof the projectiles lead to desorption of more thana single adsorbate.

The MD simulations also exhibited the occur-rence of CIM process. It was found that at lowcoverage the CIM process could result in ex-tremely long migration distances of the targetadsorbates. This migration distance decreasesmarkedly when surface coverage is increased to amono-layer. It is argued that the CIM process canplay an important role in various surface reac-tions where the mixing between two (or more)reactants is required.

Fig. 8. Variation of the integrated AMD as a function of thecollider’s incidence kinetic energy for five different coverages(noted on the upper panel) and two angles of incidence:�in=0° (upper panel) and �in=60° (lower panel).


Acknowledgements

This work has been partially supported by agrant from the Israel Science Foundation and bythe German Israel Foundation. The Farkas Cen-ter for Light Induced Processes is supported bythe Bundesministerium fur Forschung und Tech-nologie and the Minerva Gesellschaft fur dieForschung mbH.

References

[1] Y. Zeiri, J.J. Low, W.A. Goddard, J. Chem. Phys. 84(1986) 2408.

[2] W.J. Hays, W.E. Rogers, E.L. Knuth, J. Chem. Phys. 56(1972) 1652.

[3] J.D. Beckerle, A.D. Johnson, S.T. Ceyer, Phys. Rev. Lett.62 (1989) 685.

[4] J.D. Beckerle, A.D. Johnson, S.T. Ceyer, J. Chem. Phys.93 (1990) 4047.

[5] G. Szulczewski, R.T. Levis, J. Chem. Phys. 98 (1993)5974.

[6] G. Szulczewski, R.T. Levis, J. Chem. Phys. 103 (1995)10238.

[7] L. Romm, T. Livneh, M. Asscher, J. Chem. Soc. FaradayTrans. 91 (1995) 3655.

[8] D. Kulginov, M. Persson, C.T. Rettner, J. Chem. Phys.106 (1997) 3370.

[9] L. Romm, Y. Zeiri, M. Asscher, J. Chem. Phys. 108(1998) 8605.

[10] P. Feulner, D. Menzel, Phys. Rev. B 25 (1982) 4295.[11] K. Jacobi, Phys. Status Solidi, A 177 (2000) 37.[12] L. Vattuone, P. Gambardella, F. Cemic, U. Valbusa, M.

Rocca, Chem. Phys. Lett. 278 (1997) 245.[13] L. Vattuone, P. Gambardella, U. Burghaus, F. Cernic, A.

Cupolillo, U. Valbusa, M. Rocca, J. Chem. Phys. 109(1998) 2490.

[14] J.D. Beckerle, Q.Y. Yang, A.D. Johnson, S.T. Ceyer, J.Chem. Phys. 86 (1987) 7236.

[15] J.D. Beckerle, A.D. Johnson, Q.Y. Yang, S.T. Ceyer, J.Chem. Phys. 91 (1989) 5756.

[16] E. Por, M. Asscher, J. Chem. Phys. 90 (1989) 3405.

[17] C. Akerlund, I. Zoric, B. Kasemo, J. Chem. Phys. 109(1998) 737.

[18] P. Lohokare, E.L. Crane, L.H. Dubois, R.G. Nuzzo, J.Chem Phys. 108 (1998) 8640.

[19] C. Akerlund, I. Zoric, B. Kasemo, J. Chem. Phys. 104(1996) 7359.

[20] K.L. Haug, T. Burgi, T.R. Trautman, S.T. Ceyer, J. Am.Chem. Soc. 120 (1998) 8885.

[21] Q.Y. Yang, K.J. Maynard, A.D. Johnson, S.T. Ceyer, J.Chem. Phys. 102 (1995) 7734.

[22] A.D. Johnson, S.P. Daley, A.L. Utz ans, S.T. Ceyer,Science 257 (1992) 223.

[23] I. Harrison, J.C. Polanyi, P.A. Young, J. Chem. Phys. 89(1988) 1498.

[24] E.B.D. Bourdon, C.C. Cho, P. Das, J.C. Polanyi, C.D.Stanners, G.Q. Xu, J. Chem. Phys. 95 (1991) 1361.

[25] Q.S. Xin, X.Y. Zhu, Surf. Sci. 347 (1996) 346.[26] X.Y. Zhu, Surf. Sci. 390 (1997) 224.[27] I. Harrison, V.A. Ukraintsev, A.N. Artsyukhovich, SPIE

285 (1994) 2125.[28] A.N. Artsyukhovich, I. Harrison, Surf. Sci. 350 (1996)

L199.[29] Y. Zeiri, J. Chem. Phys. 112 (2000) 3408.[30] J.C. Polanyi, Y. Zeiri, in: H.ai-L.ung Dai, W. Ho (Eds.),

Laser Spectroscopy and Photo-Chemistry on Metal Sur-faces, World Scientific, Singapore, 1995 and referencestherein.

[31] Y. Zeiri, Surf. Sci. 231 (1990) 404.[32] L. Romm, M. Asscher, Y. Zeiri, J. Chem. Phys. 110

(1999) 3153.[33] L. Romm, M. Asscher, Y. Zeiri, J. Chem. Phys. 110

(1999) 11023.[34] J.T. Kindt, J.C. Tully, J. Chem. Phys. 111 (1999) 11060.[35] Y. Zeiri, R.R. Lucchese, J. Chem. Phys. 94 (1991) 4055.[36] Y. Zeiri, R.R. Lucchese, Surf. Sci. 264 (1992) 197.[37] S.T. Ceyer, Science 249 (1990) 133.[38] G.A. Somorjai, Introduction to Surface Chemistry and

Catalysis, Wiley, New York, 1994.[39] Y. Zeiri, J. Chem. Phys., 113 (2000) 3868.[40] D. Velic, R.J. Levis, J. Chem. Phys. 104 (23) (1996) 9629.[41] D. Velic, R.J. Levis, Chem. Phys. Lett. 269 (1996) 59.[42] D. Velic, R.J. Levis, Surf. Sci. 396 (1998) 327.[43] C. Akerlund, I. Zoric, B. Kasemo, A. Cupolillo, F. Bu-

atier de Mongeot, M. Rocca, Chem. Phys. Lett. 270(1997) 157.

Documents

Collision induced processes at the gas–surface interface