Electron emission from molybdenum under ion bombardment

Transcription

1 J. Phys. D: Appl. Phys., 14 (1981) Printed in Great Britain Electron emission from molybdenum under ion bombardment J Ferrht, EV Alonso, RA Baragiola and A Oliva-Floriot Centro htbmico Bariloche, Comisibn Nacional de Energia Atbmica, and Instituto Balseiro, Comision Nacional de Energia Atbmica and Universidad Nacional de Cuyo, 8400 San Carlos de Bariloche, Argentina Received 18 February 1981 Abstract. We report measurements of electron emission yields of clean molybdenum surfaces under bombardment with H+, Hz+, D+, Dz+, He+, N+, NZ+, 0+, Oz+, Ne+, Ar+, Kr+ and Xe+ in the wide energy range kev. The clean surfaces were produced by inert gas sputtering under ultrahigh vacuum. We compare our results with those predicted by a core-level excitation model of Parilis and Kishinevskii. The disagreement found when using correct values for the energy levels of MO is traced to wrong assumptions in the model. A substantially improved agreement with experiment is obtained using a model in which electron emission results from the excitation of valence electrons from the target by the projectiles and fast recoiling target atoms. 1. Introduction Energetic ions bombarding solid surfaces can lead to electron emission primarily through two formally distinguishable mechanisms. At low ion kinetic energies only potential electron emission (PEE) can occur through the transfer of potential energy to an electron upon neutralisation of the incident ion (Hagstrum 1954), if the neutralisation energy exceeds twice the work-function of the surface. For swift ions one can envisage kinetic electron emission (KEE) processes similar to those occurring in energetic gas-phase ionisation collisions, suitably modified to take into account differences in electronic states. Thus, one could consider electrons excited directly by the projectile or indirectly through Auger de-excitation of inner shell holes. Frequently in the past, measurements of electron emission yields y (average number of electrons ejected per incident ion) at projectile energies in the kev range, have been compared with the KEE theory of Parilis and Kishinevskii (1960) (hereafter referred to as PK). These authors assume KEE to proceed in two stages: (i) the ejection of an innershell electron of a target atom into the conduction band and (ii) the Auger recombination of the resulting hole with the emission of an electron into vacuum. This theory was applied to the specific cases of MO and W bombarded with heavy ions, but then extended by Kishinevskii and Parilis (1962) to the case of light ion impact. We have shown recently (Baragiola et a1 1979a) that KEE under low-velocity light ion (H+, He+) impact can be understood as resulting from direct binary excitations of valence electrons of the target. We have also shown (Alonso et a1 1980) that for A1 targets under t Present address: INTEC, Guemes 3450, 3000-Santa Fe, Argentina. $ Deceased /81/ $ The Institute of Physics. 1707

2 1708 J Ferrdn, E V Alonso, R A Baragiola and A Oliva-Florio low-energy light and heavy ion impact, the emission of electrons is not consistent with an inner-shell excitation model. It is however possible that the PK mechanism is important for heavier targets like MO, to which the theory was first applied. Ryazanov and March (1975) also proposed in a model for KEE an initial stage of excitation of the inner Nz,~ shell of MO under Ar bombardment, but with a very unlikely second stage: the photoionisation of the projectile with a soft x-ray from the decay of the core hole. Their prediction of a lower y for Ar2+ than for Ar+ impact has been contradicted in a previous experiment by Perdrix et a1 (1969b). In this paper we report our measurements of electron yields of clean MO under bombardment with kev HZ+, D+, D2+, He+, N+, NZ+, Of, OZ+, N+, Ar+, Kr+, and Xe+ ions and compare them with data of other workers taken under different experimental conditions. The results for Ar- and Xe+ are also compared with the predictions of the PK theory using modern data on target properties. We discuss the PK theory in detail and show that a substantially improved agreement with observations results if we consider instead excitation of the valence band of the solid by the projectile and by fast recoiling target atoms. 2. Experimental procedure The apparatus has been described in detail before (Baragiola et a1 1979a). In brief, it consists of an ion accelerator equipped with a radio-frequency ion source and a mass sorting magnet with resolution MjAM- 150, connected to the UHV target chamber via a differential pumping stage at 10-8 Torr. The pressure in the UHV target chamber during bombardment is in the Torr range. The target used was molybdenum of purity %, and was shown by x-ray diffraction analysis to be textured 75 % in the (100) direction and 25 % in the (111) direction. The sample was degreased normally, backed at 200 C during out-gassing of the UHV target chamber, and cleaned by sputtering with 30 kev argon ions until we observed that the ion-electron yield reached a value independent of further bombarding dose, as shown in figure 1. Figure 1. Dose dependence of the electron yield of MO during sputter cleaning by 30 kev Ar+ ions.

3 , Electron emission from MO under ion bombardment 1709 All measurements reported here were made at normal incidence of the ion beam on to the target surface. The results were insensitive to changes of a few degrees in the incidence angle, indicating the absence of channelling effects as observed by Cook and Burtt (1975) for single-crystal MO. The ion energy is known to within f (0.1 "/o + 30 ev) whereas the total uncertainties in y are less than 5 %, and include errors due to the presence of neutral atoms in the beam, and of back-scattered and sputtered ions in the target collector current. We did rough tests for the influence on the electron yields of long-lived excited ions in the beam (which would affect mainly PEE) by varying the operating parameters (power, gas pressure) in the ion source within working limits, but observed no changes in the yields within errors. 3. Results and discussion 3.1. Results We present our experimental results in figures 2-10, together with data of other workers. Of the large number of experiments with MO samples reported in the literature, we have chosen to show only those performed with mass-analysed ion beams and with clean targets under UHV. In many of these experiments the targets were subjected to high 3 c oc1i ~,,,,, I I IC0 Energy ikevi Figure 2. Electron yieldsof MO under H+ and D+ impact versus projectile energy. H+: 0 this work,... Arifov et a2 (1962), Mahadevan et a2 (1965), 0 Perdrix et a1 (1969a), V Losch (1970); D+: 0 this work,... Arifov et a1 (1962). temperature flashing for cleaning, with the possible outcome of surface contamination by carbon impurities segregated from the bulk, as pointed out by Vance (1967, 1968a). However, this overiayer would probably have been easily removed, in most cases, by sputtering during the measurements. One can observe in the figures that there is in general excellent agreement between our results and those of other workers in the regions where there is overlap. The exceptions are the heavier rare gas ions in the limit of low energies, where the data of Hagstrum (1956) lie above ours, the more so the lower the ground-state neutralisation energy Ei

4 1710 J Ferrdn, E V Alonso, R A Baragiola and A Oliva-Florio L t J i 1 i t 4 Figure 3. Electron yields of MO under Hz+ and Dzt impact versus projectile energy. Hat: 0 this work,.... Arifov et al(1962), Mahadevan et a1 (1965), A Vance (1968a, b), B Losch (1970); Dd: 0 this work,... Arifov et a1 (1962) l Energy (kew Figure 4. Electron yieldsof MO under Het impact versus projectile energy. 0 this work, -- Hagstrum (1956), - Brunk (1957), Mahadevan et al (1963), A Vance (1967, 1968a), 0 Perdrix et a1 (1968, 1970). of the ion. This is most likely caused by differences in surface structure. Hagstrum (1956) and Vance (1967) measured work-function values around 4.33 ev, or about 0.2 ev (Berge et az1974) lower than the value expected for surfaces preferentially oriented in the (100) direction as used by us and the other workers. Therefore, if one recalls the sensitivity of PEE to work changes (Baragiola et a1 1979b), particularly for ions like Kr+ and Xe+

5 Electron emission from MO under ion bombardment 1711 Nf,I' I i Figure 5. Electron yields of MO under N+ and NZ+ impact versusprojectile energy. N+: 0 this work, --- Mahadevan et a1 (1965); NZ+: 0 this work, --- Mahadevan et az(1965), A Vance (1968a, b). 0.l 1' 10 Energy (kev) Figure 6. Electron yields of MO under O+ and OZ+ impact versusprojectile energy. Of: 0 this work, Mahadevan et a1(1965), 0 Perdrix et al(1969a); OZ+: 0 this work, Mahadevan et al(1965), A Vance (1968a, b). 100 of low Ei, one can account for the observed discrepancies and also for the lack of agreement between Vance (1967, 1968a, b) and Mahadevan et a1 (1963, 1965) in terms of differences in the surface structure of the targets. From the results of the pairs H+(D+)-Hzf(D2+), N+-N2+, and one can infer that the yields for molecular ions are slightly less than twice those for atomic ions, when

6 1712 J Ferrdn, E V Alonso, R A Baragiola and A Oliva-Florio Figure 7. Electron yields of MO under Ne+ impact versus projectile energy. 0 this work, -- Hagstrum (19561, Vance (1968a), 0 Perdrix et U/ (1968, 1969b) I, I,ll,I, t $ I l Ill Energy i kev) Figure 8. Electron yields of MO under Ar+ impact versus projectile energy. 0 this work, -- Hagstrum (1956), Mahadevan et ai(l963, 1965), A Vance (1967, 1968a), 0 Perdrix et a2 (1968, 1969b, 1970), 0 Oeschner (1978). compared at equal impact velocities. This indicates that the two atoms in the molecular ion do not behave independently. This effect, which is only in part due to differences in PEE was discussed by Baragiola et a1 (1978) for hydrogen ions. For nitrogen (oxygen) ions the existence of this molecular effect is yet unclear since the results could have been affected by an unavoidable, though probably small, fractionfof N22+ (022f) ions in the beam which would increase the measured atomic yields by a factor N 1 + 2j

7 Electron emission from MO under ion bombardment 1713 Figure 9. Electron yields of MO under K.r+ impact versus projectile energy. 0 this work, " Hagstrum (1957), 0 Perdrix et al(1968). 1 L 1 - O'lI -"" I8 1 LLLi Energy l kevi Figure 10. Electron yields of MO for Xe+ impact versus projectile energy. 0 this work, " Hagstrum (1957). No isotope effectwas found, within the uncertainties in our measurements for H+(Hz+) and D+(Dz+) ions of equal velocity, in agreement with the observations of Arifov et a1 (1962) in this velocity range The theory of Parilis and Kishinevskii (PK) The PK kinetic emission model is based on an internal Auger process started by a collision between the projectile and a target atom in which a core electron of the latter

8 1714 J Ferrdn, E V Alonso, R A Baragiola and A Oliva-Florio is excited to the conduction band. The recombination energy of this inner-shell hole may then be used in the ejection of another electron from the valence band into vacuum (figure 11). PK calculate the cross-section for production of the electron-core hole pair using the friction model of Firsov (1959) for inelastic energy transfer, modified to take into account nonrectilinear trajectories of the colliding atoms. In a later paper, Kishinevskii and Parilis (1962) use instead of Firsov s model (which is in principle valid for 1 /4 < 21/22 c 4, where Z1 and 22 are the atomic numbers of the projectile and the target, respectively), an extension by Kishinevskii (1962) to any 21 and ZZ. Vocuum level Fermi level Core level Figure 11. Schematic electron energy diagram showing the Auger process in the model of Parilis and Kishinevskii (1960). v= work-function; U=binding energy of the core level. The minimum energy transfer for inner-shell excitation is U, the binding energy of the shell with respect to the Fermi level which, for Auger EE to be possible, must be larger than cp, the work-function of the surface. Therefore, the use of Firsov s model or of it s derivations, none of which include fluctuations in the energy transfer for given initial conditions during the collision, leads to a velocity threshold for emission, Uth, at which the maximum inelastic energy transfer in a collision equals U. Once the inner-shell hole is produced, PK approximate the problem of the transport of the excited electron to the surface by using an exponential attenuation law with an average mean free path L for decay of the electron energy below the vacuum level and an empirical relation W=0.016 (U-cp) (1) for the probability that the Auger process causes electron ejection in the absence of attenuation, i.e. W includes the probability of surmounting the potential barrier at the surface. With these assumptions, the electron yield y is expressed in this model as y = N W jr.(v) exp [-(x/l)] dx (2) where N is the target number density, xn the depth from the surface at which the velocity of the projectile drops to uth and.(v) the excitation cross-section a=-jp*p&(p) 1 dp J O where &(p) is the energy transfer at impact parameter p, J the average energy spent in

9 Electron emission from MO under ion bombardment 1715 producing an electron with energy above the vacuum level through the Auger process, and p1 is the impact parameter at which &(p) = U, the minimum energy required to excite the inner shell. Through the dependence of U on v, the theory predicts yields to be proportional to v at low velocities and then linear with velocity for v > 108 cm S-1. We will now evaluate critically the assumptions made in the PK theory. The first problem appears with their assumption that the total electronic energy transfer in a collision goes into exciting an inner-shell level of the target atom. (In this context it is of interest to note that Harrison et al (1965), in their very simple model of electron emission, noticed this pitfall of the PK model and used Russek's (1963) theory for total ionisation instead, but considering all emitted electrons to originate from the projectile,) The values of U used by PK where those derived by Harrower (1956) from measurements of energy losses of reflected electrons. Lynch and Swan (1968) and Schubert and Wolf (1979) found later that these values of U correspond not to single excitation of bound electrons but to plasmon losses. Thus the value of U for the shallowest core level of molybdenum (4p) is 35 ev (Schubert and Wolf 1979), and not 16 ev-g, as used bypk. To show the effect of the value of U on the theory, we compare, in figures 12 and 13, the Figure 12. Electron yields of MO for Ar+ impact versus projectile velocity. 0 this work (experiment);-.-.- PK theory with U= 11.5 ev (Parilis and Kishinevskii 1960); _" PK theory with the correct value of U=35 ev; - this work (theory). 131 Figure 13. Electron yields of MO for Xer impact versus projectile velocity. 0 this work (experiment); PK theory with U= 11.5 ev; --- PK theory with the correct value of U=35 ev; - this work (theory).

10 1716 J Ferrdn, E V Alonso, R A Baragiola and A Oliva-Florio results of the PK theory with experiment for Ar+ and Xe+ ions on MO, using the two mentioned values of U and using PK s values 9 =4.3 ev, L = , and J=23 ev. One can notice that the PK theory with the correct value of U is in poor agreement with experiment, notably in the predicted threshold velocities which are too large. A more fundamental question is the PK s neglect of direct excitation of valence electrons. This neglect was based on a simplified picture in which these electrons are free (Petrov 1960), and that therefore they will receive only a small energy transfer in direct binary encounters with the atomic projectile due to the large mass mismatch. For the case of H+ and He+ impact, Baragiola et a1 (1979a) have recently shown that at energies below 50 kev, electron emission by the decay of inner-shell holes is negligible, and that a simple binary model involving free valence electrons in the target was not inconsistent with experiment. We have also shown (Alonso et a1 1980) that for a light element target like aluminium, and for a variety of projectiles ranging from H to Xe, there is no proportionality between inner-shell excitation cross-sections and electron emission yields. For the heavier projectiles, significant kinetic electron emission occurs at impact velocities below the threshold calculated from the model of binary collisions with free electrons. The reason for this is that in the case of a heavy ion moving slowly through a solid, the picture of free valence electrons loses its meaning locally due to the strong perturbation set near the ion; this causes collisions to resemble those occurring between free atoms in the gas phase. It thus seems reasonable to consider, based only on energy conservation, that the minimum inelastic energy transfer required for electron emission is just the work-functions of the surface; equations (2) and (3) now become where pa is the impact parameter such that.(p$ = F, and where it is now justified to use Firsov s model to calculate &(p) since we include all the electrons of the atom. The threshold velocity 0th is now that at which emax =v. B is the mean electron escape probability through the surface barrier. Using this approach we have calculated values of y for Ar on MO using B/J as a fitting parameter. The result is compared in figures 12 and 13, after adding the PEE yield, with our experimental data. It is worth noting here that the derived threshold energy Eth does not depend on the fitting constant and that it is in good agreement with the threshold that can be derived by extrapolating experimental data. Similar agreement in Eth is found for other heavy projectiles whereas the PK theory predicts excessively large threshold values (table 1). One may ask whether the use of Firsov s theory is adequate for estimating threshold energies. In the context of our model we are interested in the direct excitation of valence electrons, and therefore the inclusion of these electrons in the calculation is consistent. On the other hand the threshold energies are in general sufficiently low such that distances Table 1. Calculated threshold velocities. Collision t th (106 cm S-l) from Uth (106 cm S-1) from partner TFF potential Moliere potential Ar+-tAl 7.3 Ar++Mo 6.2 Ar -+Au

11 Electron emission from MO under ion bombardment 1717 of closest approach are larger than 1 A at which the Thomas-Fermi-Firsov (TFF) interatomic potential used may not be adequate. These distances are, however, about the same as impact parameters in channelling studies (Morgan 1973) where it has been found that electronic energy losses are, on the average, consistent with the results of Firsov s theory. To test the influence of the type of interatomic potential used in this model, at energies near the threshold, we compare the values for 0th obtained with the TFF potential with those obtained using the approximation of Moliere to the Thomas-Fermi potential (Wilson et a1 1977) and electronic densities which represent better the potentials at large distances (2 1 A). We found differences in 0th below 10 % (table 1). This suggests that the use of the TFF potential is satisfactory as regards to mean energy loss values even at the low energies with which we are concerned. Very close to Eth, however, the use of Firsov s theory is not proper, since it neglects energy loss fluctuations (straggling) in individual collisions. It is expected that proper inclusion of energy straggling by use of a more developed theory will smear the threshold The influence of elastic collisions Elastic collisions with target atoms will cause the projectile to lose kinetic energy and therefore ionising ability. This fact was taken into account by PK by including in equation (2) a law for the slowing down of the projectile as a function of depth. This law uses a straight-path approximation which, although acceptable in many cases, is not of general validity and should be modified considering that elastic collisions affect the path of the projectile and that sufficiently energetic recoiling target atoms can produce ionisation and contribute to electron emission. We have included all the above mentioned effects in a calculation performed with the aid of two Monte Carlo simulation programs with which we follow the trajectories of the projectile and of the recoiling target atoms. In these programs the trajectories are calculated using Moliere interatomic potentials and the inelastic energy transfers are evaluated with the Firsov theory as modified by PK, with the factor B/J obtained by fitting theory to the data. We note here that for heavy projectiles, J has been found not to depend much on the projectile-target pair in gas phase collisions (Macdonald and Figure 14. Electron yields of MO for Art impact versus projectile velocity. 0 experiment. The curves are results from Monte Carlo simulations: PK model, our model, total yields, our model, contribution from recoils.

12 1718 J Ferrdn, E V Alonso, R A Baragiola and A Oliva-Florio Sidenius 1969), and that B should be essentially a target property influenced only slightly by the type of projectile and its energy through variations in the angular and energy distribution of electrons in the excitation cascade within the solid. One should thus expect B/J to be roughly constant for the cases under study. This was tested by evaluating y for Ar+ on MO with B/J derived from a fit to the Xe+ on MO data. The agreement is fairly good as shown in figure 14; one should expect Iarger deviations for lighter projectiles. Velocity (10' cm S") Figure 15. Electron yields of MO for Xe+ impact versus projectile velocity. 0 experiment. The curves are results from Monte Carlo simulations: PK model, our model, total yields; our model, contribution from recoils using v, -.._..- our model, contribution from recoils using U as minimum energy loss for electron emission. It can be observed in figures 14 and 15 that the contribution of recoils to the electron yield is significantly smaller for Ar than for Xe projectiles. This can be understood on the basis of larger cross-sections for elastic energy transfer in the latter case. It is of interest to remark that when the projectile is heavier than target atoms the threshold and near threshold behaviour should be dominated by recoil effects, as shown by Alonso et al (1980) for Kr and Xe ions on aluminium. 4. Conclusions We have discussed the theory of kinetic ion-electron emission of Parilis and Kishinevskii (1960) in the frame of our electron yield measurements for heavy noble gas ions in molybdenum, the projectile-target combination for which this theory was originally applied, We find that this theory, based on excitation of inner-shell electrons, is physically incorrect and that its past success in reproducing observations was due to the fortuitous use of wrong values of the energy levels of the target electrons. We show that a basically different excitation model is more consistent with the present physical understanding, and with experiment. In this model, electron emission results from the excitation of valence electrons in collisions between the projectile and target atoms, and between recoil atoms and other target atoms.

13 Acknowledgments Electron emission from MO under ion bombardment 1719 We wish to thank Lic MM Jakas for helpful discussions and his aid in adapting Monte Carlo programs. We are also indebted to Lic J L Spino for performing the x-ray diffraction analysis of our MO target, and to Ing H Raiti for his help in the setting up of the experimental apparatus. This work was partially supported by the International Atomic Energy Agency through contract 1928/RB. References Alonso E V, Baragiola R A, Ferron J, Jakas M M and Oliva-Florio A 1980 Phys. Rev. B Arifov U A, Rakhimov RR, Abdullaeva M and Gaipov S 1962 Bull. Acad. Sci. USSR Phys. Ser Baragiola RA, Alonso EV, Auciello 0, Ferron J, Lantschner G and Oliva-Florio A 1978 Phys. Lett. 67A Baragiola RA, Alonso EV and Oliva-Florio A 1979a Phys. Rev. B Baragiola RA, Alonso EV, Ferrbn J and Oliva-Florio A 1979b Surface Sci Berge S, Gartland P 0 and Slagsvold B J 1974 Surface Sci Bruned C 1957 Z. Phys Cook N and Burtt RB 1975 J. Phys. D: Appl. Phys Firsov OB 1959 Sov. Phys.-JETP Hagstrum HD 1954 Phys. Rev Phys. Rev Harrison D, Carlston C and Magnuson G 1965 Phys. Rev. 139 A Harrower GA 1956 Phys. Rev Kishinevskii L M 1962 Bull. Acad. Sci. USSR Phys. Ser Kishinevskii LM and Parilis ES 1962 Bull. Acad. Sci. USSR Phys. Ser Losch W HP 1970 Phys. Stat. Solidi a Lynch M J and Swan JE 1968 Aust. J. Phys Macdonald J R and Sidenius G 1969 Phys. Lett. 28A 5434 Mahadevan P, Layton J K and Medved D B 1963 Phys. Rev Mahadevan P, Magnuson GD, Layton JK and Carlson CE 1965 Phys. Rev. A Morgan D V 1973 (ed) Channeling-Theory, Observations and Applications (London: Wiley Interscience) Oeschner H 1978 Phys. Rev. B Parilis ES and Kishinevskii LM 1960 Sov. Phys.-Solid St Perdrix M, Baboux JC, Goutte R and Guillaud C 1970 J. Phys. D: Appl. Phys Perdrix M, Paletto S, Goutte R and Guillaud C 1968 J. Phys. D: Appl. Phys a J. Phys. D: Appl. Phys b Phys. Lett. 28A Petrov NN 1960 Sov. Phys.-Solid St Russek A 1963 Phys. Rev Ryazanov M I and March NH 1975 Phys. Lett. 53A Schubert W K and Wolf EL 1979 Phys. Rev. B Vance D W 1967 Phys. Rev a Phys. Rev b Phys. Rev Wilson WD, Haggmark LG and Biersack JP 1977 Phys. Rev. B