360 Nuclear Instruments and Methods in Physics Research B58 (1991) 360-364

North-Holland

Skipping motion of Si off Cu( 111)

D.J. O’Connor

K. J. Snowdon Fachberezch Physik, Universitiit Osnabriick, Box 4469, W-4500 Osnabriick, FRG

The origin of a series of discrete peaks in the scattered ion spectra of Si scattered off a Cu(lll) surface at glancing angles has been

described by the authors as transient adsorption or skipping motion. Various experimental tests have been applied to the effect and

subsurface channelling or other crystallographic effects have been eliminated as possible explanations [K.J. Snowdon, D.J. O’Connor

and R.J. MacDonald, Phys. Rev. Lett. 61 (1988) 1760; Appl. Phys. A47 (1988) 83; Radiat. Eff. Defects Sol. 109 (1989) 251. This

leaves the trapping in the surface binding potential as the preferred model. Of the likely trapping mechanisms which could be

responsible, one not seriously addressed so far is a kinetic process involving scattering from either steps on the surface or atoms

experiencing extreme thermal displacements. These deflected projectiles could be trapped in a surface binding trajectory which results

in an oscillatory path perpendicular to the surface.

1. Introduction

The energy spectrum of scattered Si- off a Cu(l11)

surface at grazing incidence [l] exhibits a sequence of discrete loss peaks (fig. 1). The energy loss for each peak is proportional to energy (fig. 2) and the energy loss intercept at zero beam energy can be related to energy loss processes associated with charge exchange mechanisms. This effect has been identified as the tran- sient adsorption of Si into a surface binding state. It has

Si- off Cu( 111) I-- 2 keL_.___._._.

Fig. 1. The evolution of the energy spectra of specularly

scattered Si- off the Cu(ll1) surface as a function of the angle

of incidence for 2 keV Si+ incident ions. The angle of inci-

dence (in degrees) is represented by the degree of offset from zero on the _Y axis and the curves have been normalised so that

the integrated area under each is a constant.

been shown that alternative explanations for this effect which included special trajectories in the surface semi- channel or subsurface channelling are not supported by the observations [l]. The elimination of these exptana- tions was achieved by using inert gas ions as a control for the channelling effects and the observation that the relative yields for the different loss peaks is azimuthal angle independent.

Two different approaches could be used to explain the observed skipping effect. One is to explain it within a unified model in which the process responsible for trapping into and detrapping from the binding state is intimately linked to the process responsible for the binding potential. The other is to consider the trapping and detrapping processes as a different mechanism to the binding state. The existence of the surface binding state itself is not in question as there exist stable Si-Cu compounds indicating that an attractive potential exists. The trapping and detrapping mechanism is still a prob- lem in the interpretation of observations made so far as under some experimental conditions a large perpendicu- lar component of momentum needs to be eliminated or stored. This paper will explore the possibility that the trapping process is initiated by the elastic scattering of the projectiles through an angle large enough to reduce the perpendicular component of momentum to the point where trapping is feasible. It is important to state at the outset that so far no one model of the skipping process has been able to successfully describe all the observa- tions and this applies equally so for the kinetically trapped mechanism described.

0168-583X/91/$03.50 0 1991 - Elsevier Science Publishers B.V. (North-Holland)

D.J. O’Connor, K.J. Snowdon / Skipping motion of Si off Cu(I II) 361

An effective energy associated with the perpendicu- lar motion of the ion beam, E,, is defined as E&n% where Es is the incident energy and a is the angle of incidence of the beam to the surface. To explain the observations in [1,2] the trapping mechanism or mecha- nisms must satisfy the following conditions:

(i) It/they must be able to store or eliminate, large amounts of perpendicular momentum.

(ii) Both trapping and detrapping processes must be possible.

(iii) The energy loss or storage process must involve small quanta as the skipping effect can be seen over a continous range of incidence angles and incident en- ergies. For a particle to be trapped in a surface potential well either the initial perpendicular component of energy of the incident ion beam must be lost or stored. It has been observed that the scattered projectiles do lose at least the initial perpendicular component of energy from the total initial beam energy (fig. 3), though this alone cannot be used as evidence that a simple energy loss process will explain the observations as there are both

!a

0 0 500 a00 1500 2Of

b-ladalt Energy 6s)

Fig. 2. The energy difference between the loss peaks and the incident ion energy as a function of the incident ion energy for

Si+ off Cn(ll1) for an angle of incidence of 4O and a scatter- ing angle of 8O.

$

- 1 / a

2 4 6 g 10 12 1~0

ANGLE OF INCIDENCE Fig. 3. The ratio of energy loss to initial perpendicular compo- nent of energy (R = AE/E, ) as a function of the angle of incidence for Si’ off Cuflll). The near horizontal line at

R = 0.85 is the ratio of the energy loss for an elastic double scattering process relative to the initial E,

elastic and inelastic energy loss mechanisms operating simultaneously. Of equal importance is the observation that the scattered particles are concentrated close to the specular direction (on the surface side of specular) implying some ‘memory’ of the incident conditions. Potential mechanisms are:

(1) Energy may be lost to the electronic degrees of freedom of the solid in the form of the creation of plasmons and electron-hole pairs [Z] . This form of energy loss satisfies the characteristic that the quanta must be small but it is necessary to propose a plausible mechanism whereby a part of this excitation is returned to the projectile.

(2) The perpendicular component of energy can be lost through the deflection of the projectile by the atoms of the surface. This process is unlikely from an ideal flat surface at low energies and grazing incidence as it is not possible for the projectile to get close enough to the target atoms to experience a large angle deflection. This process has been extensively investigated in computer simulations by Otsuki et al [3].

(3) The energy can be stored in the potential energy associated with excited states of the projectile or by charge exchange to a binding state between the par- ticipants [2]. This is a particularly attractive model as the effect relies on the charge exchange process which is a selective process between the projectile and target, providing one reason why the skipping effect is not observed for all combinations of target and projectile.

In earlier papers [1,2] on this effect attention was focussed on mechanisms (1) and (3) incorporating en-

II. PARTICLE SCATTERING

362 D. J. O’Connor, K.J. Snowdon / Skipping motion of Si off Cu(l I I)

ergy storage. However the energy needed to be stored at higher beam incidence angles (up to 16”) for keV ions is of the order of 100 eV which appears uncomfortably high to be accounted for by these mechanisms.

A detailed investigation of mechanism (1) by Kato et al. [4] in the weak coupling limit demonstrated a cou- pling between the parallel and perpendicular compo- nents of momentum of the beam for surface scattering. This coupling arises because the surface-normal momentum component of the surface electrons is not a well-defined parameter due to the existence of the dis- continuity in this direction. This coupling provides a potentially promising trapping detrapping mechanism. It is expected that a similar coupling of momentum components of the beam will occur in the strong cou- pling limit which is more appropriate to this case. It is, however, not clear how this mechanism could retain a memory of the initial beam direction, and therefore explain the observed near specular scattering.

An alternative mechanism to explain the capture is that of a hard collision with atoms of the surface resulting in deflection into a direction near parallel to the surface in which the perpendicular component of energy of the beam is negative. This cannot happen on a perfectly smooth surface for Si off Cu(ll1) even if thermal vibrations are included [5]. Such deflection however is possible at a surface defect, either point or extended, which scatters the projectile parallel to the surface. Detrapping could also occur at a defect, how- ever there is no immediate explanation how this mecha- nism could retain a memory of the incident beam direction and thus explain the near specular scattering observation, Computer simulations of this process are being further pursued [5,6]. In This paper we explore whether the observed scattering cross-section supports trapping and detrapping via this mechanism.

2. Experiment

The scattering experiments were performed in an UHV chamber (base pressure 2 X lo-” Torr) using 200-2000 eV Si+ beams incident upon a Cu(ll1) face of a crystal with a purity of 99.999%. The surface was cleaned and flattened in situ by glancing incidence ion bombardment [l] and the cleanliness was monitored by negative ion recoil spectroscopy which is particularly sensitive to electronegative contaminants. The flatness was monitored by measuring the angular distribution of scattered Si ions when the ion beam was incident at 4O to the surface.

The scattered ions were analysed by a MO0 electro- static analyser with an energy resolution (AE/E) of 0.5% and an angular resolution of 0.5”. The analyser was mounted on a computer controlled two axis go- niometer allowing rotation in the plane defined by the

incident ion beam direction and the surface normal, and in a plane at right angles to the first plane. The target was mounted on a two axis goniometer which was also under computer control. The target could be heated to 1ooo”c.

3. Results

The evolution of the energy spectra of specularly scattered Si- off the Cu(ll1) surface (for 2 keV Sit incident) as a function of the angle of incidence is shown in fig. 1. The angle of incidence (in degrees) is signified by the degree of offset from zero on the y axis and the curves have been normalised so that the in- tegrated area under each is a constant. The oscillations are clearly seen at small angles of incidence but the peaks broaden for increasing angles. The oscillations become less distinct above about 10” and there is some evidence for a redistribution of relative yields in the different components. Similar trends have been ob- served for 1 keV Si- and for the Si+, however the yields in the loss peaks for the positively charged projectiles are insufficient to follow in this fashion. The significant difference in the yields for the positively and negatively charged ions may be indicative of the existence of an intermediate state from which the Sii ion is more likely to form than the Si+. The most problematic aspect of fig. 1 is that ions with discrete losses are seen to angles of incidence of 16” or more.

Further information can be derived from the ab- solute integrated yields of the scattered ions. The scattered ion yield is given by

Y* a Ni(N,/sin a)(da(e)/dQ)P*dQ, (I)

where N, is the number of incident ions, N,/sin cy is the number of scattering centres per unit area perpendicular to the ion beam direction, da(B)/dO is the scattering cross-section of the incident ion (in any charge state) and P* is the probability of being in a charged state (positive or negative). The angle of incidence is 1y and in the specular scattering geometry the scattering angle 13 = SLY. In the experiments in fig. 1 the scattering geom- etry involved specular scattering which means that the fraction of the incident ion beam seen by the analyser is independent of (Y, provided that the beam width on target and the area of the target seen by the analyser does not exceed the target width. The scattering cross section for a scattering event with a single atom can be calculated if the interatomic potential is known. To be trapped in a binding potential the first scattering event must be through an angle close to (Y so that the pro- jectile’s perpendicular component of energy becomes negative. In this case the number of particles trapped will be proportional to the scattering cross-section for an angle equal to the angle of incidence. The require-

D.J. O’Connor, K.J. Snowdon / Skipping motion of Si off Cu(ll1) 363

ment that the scattered particle has a perpendicular component of energy within some narrow band of en- ergies places a limit on the solid angle for this scattering event such that

dD = 2n sin (Y de d+, (2)

where d+ is the acceptance angle in the plane of the surface. For the detection of the projectile it may have to undergo a detrapping collision with a defect after completing one or more oscillations and in this case the scattering cross section through an angle of a will be a second parameter. Here the scattering cross-section for the incident particle is used as an approximation, how- ever the detrapping occurs at a smaller energy and hence the cross section for the second event will be larger. The ion yield would then be described by

Y* aN,No(do(a)/dL?)2P* de d+. (3)

The squared dependence on the cross section is char- acteristic of a double scattering process and eq. 3 is similar to the expression for the ion yield observed for double scattering observed in low energy ion scattering (LEIS). One significant difference is that in LEN the distance between collisions is known from the crys- tallography and a more explicit expression can be established.

The only parameters in eq. 3 which are not neces- sarily constant as the angle of incidence (and total scattering angle in a specular scattering geometry) is changed is P* and the cross section. It is not possible to directly estimate the dependence of P* on geometry, however from the azimuthal dependence of the negative ion yield reported in fl] the angular variation in yield was attributed largely to scattering processes and little

1 ooooo:-

1 .OE-20 1 .ooE-19

cross Section

if any of the variation was thought to relate to an azimuthal dependence of the charge exchange process.

In the following the ion yield is the integrated yield between 1.5 keV and 2 keV. The logarithmic compari- son between the negative ion yield (uncorrected for charge exchange, P*) and the scattering cross-section (using the ‘Universal’ interatomic potential [7]) in fig. 4(a) reveals that the power dependence for the cross section is 1.76 f 0.07 which is indicative of the double scattering process described above. The two departures from linearity at large cross sections correspond to angles of incidence of 4O and 6 O. At these angles for the approximately 1 mm ion beam diameter and a 10 mm target width the projection of the incident ion beam on the target is greater than the target width resulting in a reduced flux on target. The near constant ion yield has been observed to continue to lower incidence an- gles. As the log-log plot can disguise significant depar- tures in the linear relationship between the parameters a linear plot [fig. 4(b)] comparing the [scattering cross- section]‘.76 to the ion yield (not including the 4O and 6 o results) is shown. It reveals that the linear relationship is well supported and the intercept is at the origin to within statistical uncertainty. To test this conclusion there are two studies in progress. The first is the use of a molecular dynamics simulation [6] to model the interac- tion of the projectile with the surface to establish whether defects can explain the trapping process ob- served. The second is the development of a system which will use an STM to monitor the topographical condition of the surface while under grazing incidence ion bombardment. The second study is aimed at de- termining how the ion beam preparation modifies the surface defect density and whether there is a correlation between the ion yield in skipping trajectories and the flatness of the surface.

/’ /

/’

9 /’

f 40001 i” .A,“’

3000. ,, ”

0 cross section ’ ‘6

(Times 1 OE-34)

Fig. 4. (a) The variation of the Si + as a function of the elastic cross-section for increasing incidence angles for 2 keV Si+ incident.

The slope is 1.76 50.07. (b) The linear comparison between the Si-.- yield and the cross section to the power of 1.76.

II. PARTICLE SCATTERING

364 D.J. O’Connor, K.J. Snowdon / Skipping motion of Si off Cu(I I I)

4. Conclusion

The capture into a binding potential initiated by elastic scattering from a surface defect has been pro- posed as a trapping mechanism for the observed skip- ping motion of Si off a Cu(ll1) surface. While the comparison between the model and data is supportive of this model provided P* is not a strong function of the angle of incidence and it can satisfactorily explain why slopping is observed for high angles of incidence to the surface it is still necessary to explain why the exiting particles are observed to preferentially come out close to the specular direction. This is an anomaly for all but E, storage models as it implies a limited (or degraded) memory of the initial value of E, . Neither the kinetic deflection nor inelastic coupling processes afford an immediately obvious explanation of this feature of the results so far. Current experiments and simulations un- derway should help clarify the role of elastic scattering and inelastic processes.

References

[l] K.J. Snowdon, D.J. O’Connor and R.J. MacDonald, Phys. Rev. Lett. 61 (1988) 1760; Appl. Phys. A47 (1988) 83; Radiat. Eff. Defects Sol. 109 (1989) 25.

[2] K.J. Snowdon, D.J. O’Connor and R.J. MacDonald, Surf. Sci. 221 (1989) 465.

[3] Y.H. Otsuki, K. Koyama and Y. Yamamura, Phys. Rev. B20 (1979) 5044.

[4] M. Kato, K.J. Snowdon, D.J. O’Connor and R.J. Mac- Donald, Nucl. Instr. and Meth. B48 (1990) 319.

[5] B. Cooper, personal communication. Simulations carried out using the SAFARI code.

[6] D. Danailov and K.J. Snowdon, to be published. [7] J.P. Biersack and J.F. Ziegler, in: Ion Implantation Tech-

niques, eds. H. Rysseland and H. Glawischnig, Springer Series in Electrophysics Vol. 10 (Springer, Berlin, 1982) pp. 122-156.