Explanation of the surface peak in charge integrated LEIS spectra

Nuclear Instruments and Methods in Physics Research B 203 (2003) 218–224 www.elsevier.com/locate/nimb

Explanation of the surface peak in charge integrated LEIS spectra M. Draxler a, R. Beikler b, E. Taglauer b, K. Schmid R. Gruber a, S.N. Ermolov a,c, P. Bauer a,*

a,b

,

a

Institut f€ur Experimentalphysik, Johannes Kepler Universit€at Linz, A-4040 Linz, Austria Max-Planck-Institut f€ur Plasmaphysik, EURATOM Association, D-85748 Garching bei M€unchen, Germany Institute of Solid State Physics, Russian Academy of Sciences, Chernogolovka, Moscow district 142432, Russia

b c

Abstract Low energy ion scattering is very surface sensitive if scattered ions are analyzed. By time-of-flight (TOF) techniques, also neutral and charge integrated spectra (ions plus neutrals) can be obtained, which yield information about deeper layers. In the literature, the observation of a more or less pronounced surface peak was reported for charge integrated spectra, the intensity of the surface peak being higher at low energies and for heavy projectiles. Aiming at a more profound physical understanding of this surface peak, we performed TOF-experiments and computer simulations for He projectiles and a copper target. Experiments were done in the range 1–9 keV for a scattering angle of 129°. The simulation was performed using the MARLOWE code for the given experimental parameters and a polycrystalline target. At low energies, a pronounced surface peak was observed, which fades away at higher energies. This peak is quantitatively reproduced by the simulation, and corresponds to scattering from
2 atomic layers. Analyzing the contributions of the individual outermost atomic layers, one finds that the surface peak is due to binary collisions of projectiles with atoms in the first and second layer, while the contribution from deeper layers is dominated by multiple scattering. Ó 2003 Elsevier Science B.V. All rights reserved. PACS: 34.50.Dy; 34.50.Bw Keywords: Low energy ion scattering; Energy loss; Binary collisions; Multiple scattering

1. Introduction Low energy ion scattering (LEIS) is a standard tool for quantitative surface analysis [1,2], closely related to Rutherford backscattering (RBS) [3]. In

*

Corresponding author. Tel.: +43-732-2468-8516; fax: +43732-2468-8509. E-mail address: [email protected] (P. Bauer).

both techniques, ions are used as projectiles, and energy spectra of scattered projectiles are measured at a large scattering angle h. Thus, the energy spectra contain two different types of information: (i) the energy transfer from the projectile to the recoil atom in the close collision yields information on the mass of the scattering center; (ii) the energy lost by the projectile along its path due to electronic interaction yields information on the depth, in which the scattering took place [4].

0168-583X/03/$ - see front matter Ó 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0168-583X(02)02220-6

M. Draxler et al. / Nucl. Instr. and Meth. in Phys. Res. B 203 (2003) 218–224

The interpretation of the ion spectra along these lines is based on the single scattering (ss) model, which assumes that the projectileÕs path consists of two straight lines (incoming and outgoing trajectories) that intersect at the position where the large angle scattering takes place. Within the ss model, the interpretation of the energy spectra is very simple: the spectrum has a well defined onset, kE0 , with the kinematic factor k and the incident energy E0 . In the target, the energy loss dE of the projectile per path length dx is described by the electronic stopping power S ¼ dE=dx [5,6]. Projectiles impinge at an angle a with respect to the surface normal, are backscattered at a depth Dx, leave the target under an angle b with respect to the surface normal, and are detected at an energy Ex , which can easily be calculated within the surface energy approximation [4]: Ex ¼ kðE0  SðE0 ÞDx= cos aÞ  SðkE0 ÞDx= cos b: ð1Þ Thus, the energy difference DE  kE0  Ex is proportional to the scattering depth Dx,   kSðE0 Þ SðkE0 Þ DE ¼ þ Dx  ½S Dx: ð2Þ cos a cos b The ss model is most appropriate in the regime of high energies and small depths. The lower the energy is, the more important becomes multiple scattering (ms). The half width angle of the ms distribution, a1=2 , is inversely proportional to the particle energy and depends on the reduced target thickness s via a power law, being roughly proportional to s in a certain range of thicknesses [7]. Due to ms, the trajectories are not well defined anymore, with the following consequences [8]: (i) the mean path length increases, (ii) the large angle scattering is smeared out and (iii) along the trajectory the ion loses energy due to interaction with electrons (electronic stopping) and to ms (nuclear stopping). Recently, a comprehensive treatment of the influence of ms on the shape of an RBS spectrum was presented by Smit [9], showing that the influence of ms vanishes close to the surface. In LEIS, Ex and Dx are not as well defined as in the RBS regime. Only for noble gas ions, detected at kE0 , the ss model is still valid, since these projectiles are scattered in the outermost atomic layer.

219

Noble gas ions that are scattered in deeper layers, are detected as neutrals, unless they are reionized in a collision with a surface atom when leaving the surface [10], and detected at lower energies. If charge integrated spectra are measured, i.e. if all projectiles are detected that are scattered into a given solid angle (without charge selection), no surface sensitivity was expected. Nevertheless, also charge integrated LEIS spectra exhibit a more or less pronounced surface peak, the intensity of which increases with increasing atomic number of the projectile, [11], and with decreasing energy (see below). The origin of this surface peak in charge integrated LEIS spectra has not yet been explained; in [11] an explanation in terms of channeling was suggested, despite the fact that even computer simulations for amorphous targets yielded a surface peak. In the following we are going to explain the physical origin of the observed surface peak, on the basis of time-of-flight (TOF)-LEIS measurements and computer simulations (MARLOWE) for He ions and a Cu target.

2. Experiment and simulations The experiment was performed using our TOFLEIS set-up, which has been described elsewhere [12] and is just summarized here: the set-up consists of an ion source, a UHV target chamber with base pressure is in the low 1010 mbar range, and a target preparation chamber attached to the main chamber. The ion source produces a beam of monoatomic Heþ ions in the energy range from 500 eV to 10 keV, the beam is mass analyzed by means of a velocity filter. An electrostatic chopper is used to produce the required beam packets with a time width P 40 ns. Two Einzel-lenses focus the beam onto the target to a diameter <2 mm. The detection electronics consists of standard components. Between target and stop detector, positive ions can be post-accelerated by a negative potential applied to the drift tube, thereby separating ions and neutrals in the spectrum. Polycrystalline Cu was used as a target. The surface purity was checked by means of Auger electron spectroscopy (AES). After cleaning, no impurities were visible in the Auger spectrum.

220

M. Draxler et al. / Nucl. Instr. and Meth. in Phys. Res. B 203 (2003) 218–224

With the set-up described above, TOF-LEIS spectra were measured for Heþ ions at 1, 3, 5, 7 and 9 keV (see Fig. 1). These measurements were performed with post-acceleration, thus the ion spectra are separated from the spectra of the neutrals. The specific experimental geometries (angles of incidence and exit, a and b, respectively) are shown in the insets. In Fig. 1, all spectra show a sharp onset at kE0 . A prominent feature of the TOF spectra of the

neutrals (neutral spectra) shown in Fig. 1 is the existence of a more or less pronounced peak at kE0 , the so-called surface peak. The relative height of the surface peak is largest at 1 keV, even if it will be damped due to limited experimental resolution; it is weaker at 3 keV and is hardly visible at 9 keV. In order to be able to compare the spectra to computer simulations, we convert our TOF spectra to energy spectra by standard procedures [13]. 2000

600

400 300

3 keV He+ → Cu α =0°, β= 51°, θ = 129°

100 80

1500 60 40

Counts

500

Counts

1 keV He+ → Cu α= 20°, β =31°, θ =129°

80 70 60 50 40 30 20 10 0 10600 10700 10800 10900

20

1000

0 6200

6300

6400

6500

6600

200

500 100 0 10000

0 10500

11000

11500

12000

12500

13000

13500

14000

t TOF [ns]

(a)

5000

5 keV He+ → Cu α =0°, β =51°, θ = 129°

30

500 20

7000

7500

8000

8500

9000

t TOF [ns]

7 keV He+ → Cu α =0°, β =51°, θ =129°

30 20

400

10 0 4700 4800 4900 5000 5100

200

300

10 0 3900 4000 4100 4200 4300

200

100

100

0 4000

(c)

6500

500

Counts

Counts

300

6000

600

600

400

5500

(b)

4500

5000

5500

6000

6500

7000

(d)

t TOF [ns]

0 3000

3500

4000

4500

5000

5500

6000

t TOF [ns]

Counts

6500 6000

500

5500

400

5000

300

4500

200

4000 3500 3000

9 keV He+→ Cu α =54°, β =3°, θ =129°

100 0 3500

3600

3700

2500 2000 1500 1000 500 0 3000

(e)

3500

4000

4500

5000

time of flight t TOF [ns]

Fig. 1. TOF-LEIS spectrum, obtained for (a) 1, (b) 3, (c) 5, (d) 7 and (e) 9 keV Heþ projectiles and a Cu target.

M. Draxler et al. / Nucl. Instr. and Meth. in Phys. Res. B 203 (2003) 218–224

3. Discussion

energy spectrum N(Ef) (arb.u.)

In Fig. 2(a)–(c), the energy converted spectra of Fig. 1 are compared to the corresponding MARLOWE spectra. The experimental spectra shown in Fig. 2 are corrected for the energy dependent detector efficiency [15], assuming identical efficiency for ions and neutrals [16]. For 9 keV, the experimental energy spectrum is very well reproduced by the MARLOWE spectrum in the whole simulated energy range, i.e. from 4500 to 8000 eV (see Fig. 2(a)). In the sim-

(a)

500 + 9 keV He → Cu

400 300 200 100

corr. experimental data MARLOWE result MARLOWE*Gaussian

kE0

0 4500 5000 5500 6000 6500 7000 7500 8000

final energy E f [eV]

ulation, there is a small indication of a surface peak with a width of 350 eV, which is hardly visible since it exceeds the spectrum plateau by only three standard deviations. In order to facilitate a comparison of simulated and experimental spectra, the simulated spectrum was convoluted with a Gaussian with a width of 265 eV matching the experimental resolution. In the convoluted spectrum, the intensity at kE0 still exceeds that of the experimental spectrum. For E0 ¼ 3 keV, the simulation shows a pronounced surface peak in a very narrow energy interval with a width of <50 eV around kE0 and a relative height that is much larger than in the experiment. The reduced width of the simulated peak may partly be due to the experimental resolution (70 eV), and partly due to the local electronic energy loss model used in the simulation, which yields the same final energy for projectiles after binary scattering from an atom in the surface layer and in the second layer, in contrast to what is expected from a nonlocal energy loss model. Similarly as for 9 keV, the simulated spectrum was convoluted with a Gaussian (see Fig. 2(b)) of 60 eV width, such that the peak height agrees with experiment. In energy spectrum N(Ef) (arb.u.)

The simulations were performed for 1, 3 and 9 keV Heþ ions and Cu as target, by using the computer code MARLOWE [14], which treats scattering as a series of binary collisions within classical mechanics. The polycrystalline target model was chosen, using a Cu(1 0 0) single crystal as a target that is rotated after each collision by a random angle. All particles were registered that have been scattered into the interval 127:5° 6 h 6 130:5°. In order to obtain acceptable statistics, 2

106 trajectories were calculated for each spectrum.

221

1000 3 keV He+ →Cu

800 600 400

corr.experimental data MARLOWE results MARLOWE*Gaussian

200

kE0

0 1600

1800

2000

2200

2400

2600

final energy Ef [eV]

(b)

energy spectrum N(Ef) (arb.u.)

2500 1 keV He+ → Cu

2000 1500 1000 500 0 500

(c)

corr.experimental data MARLOWE results MARLOWE*Gaussian 600

700

kE0

800

900

final energy Ef [eV]

Fig. 2. Energy converted TOF-LEIS spectrum of (a) 9, (b) 3 and (c) 1 keV Heþ as projectiles and Cu as a target (as in Fig. 1), in comparison with MARLOWE calculations.

222

M. Draxler et al. / Nucl. Instr. and Meth. in Phys. Res. B 203 (2003) 218–224

the convoluted MARLOWE spectrum, the width of the surface peak and therefore also its intensity is larger than in the experiment. This indicates some need for refinement of the simulation model, e.g. in the electronic energy loss model. For the plateau of the spectrum, very good agreement between the MARLOWE simulation and the measured spectrum is obtained (see Fig. 2(b)). For 1 keV He, the comparison between the MARLOWE simulation and the measured data yield similar qualitative results as for 3 keV. At kE0 , the simulation exhibits a surface peak with a width of less than 10 eV that exceeds the rest of the spectrum by a factor of two in height. The experimental surface peak is again less pronounced than the simulated one, as in Fig. 2(b). For the MARLOWE spectrum convoluted with a Gaussian of 50 eV width, the surface peak is in good agreement with the measured one. Also for the plateau of the spectrum, the agreement is satisfactory for energies above 700 eV (see Fig. 2(c)). At energies below 700 eV, the simulated spectrum is higher than the experimental data, by up to 40%. This may be due to

different reasons: either the model assumptions are not realistic (scattering potential, electronic energy loss), or the detection efficiency correction of the experimental data is inadequate at low energies. In Fig. 2, the increasing importance of ms with decreasing primary energy is visible in the experimental spectra as well as in the simulations: at energies E > kE0 , there is intensity (high-energy background) due to multiple collisions by large angles, and the relative intensity in this highenergy background with respect to the spectrum height increases with decreasing energy. Let us now discuss the surface peak. As stated before, the simulated surface peaks have an energy width of 350, 50 and 10 eV at E0 ¼ 9, 3 and 1 keV. These widths would correspond to a scattering depth of 12, 4 and 1.5 atomic layers, respectively, in the single scattering approximation [17]. Since in the simulation kinematic broadening due to ms is considerable, these numbers represent a safe upper limit for the number of target layers contributing to this peak. In order to look for the physical origin of this peak, we present in Fig. 3(a)–(c) the energy 300

60

MARLOWE simulation 9 keV He+→ Cu

50 1st layer 2nd layer 3rd layer 4th layer 5th layer 6th layer 7th layer 8th layer

40 30 20 10 0 6800

(a)

7000

7200

7400

scattered yield (arb.u.)

scattered yield (arb.u.)

70

250 200

1st layer 2nd layer 3rd layer 4th layer 5th layer 6th layer 7th layer 8th layer

150 100 50 0 2100

7600

final energy E f [eV]

MARLOWE simulation 3 keV He+→ Cu

2200

2300

2400

2500

2600

final energy E f [eV]

(b)

scattered yield (arb.u.)

1400 1200 1000 800 600 400 200 0 500

(c)

MARLOWE simulation 1 keV He+→ Cu 1st layer 2nd layer 3rd layer 4th layer 5th layer 6th layer 7th layer 8th layer

600

700

800

900

final energy E f [eV]

Fig. 3. Contributions of the outermost atomic layers of a polycrystalline Cu target to the energy spectrum of (a) 9, (b) 3 and (c) 1 keV Heþ projectiles, obtained by MARLOWE simulations (for the sum spectrum see Fig. 2).

M. Draxler et al. / Nucl. Instr. and Meth. in Phys. Res. B 203 (2003) 218–224

distributions of ions scattered in the eight outermost atomic layers, for E0 ¼ 9, 3 and 1 keV. Common to all three plots is that scattering by the outermost atomic layer results in a very sharp binary collision peak, with an energy width limited just by the collisional broadening due to the finite acceptance angle (127.5–130.5°), proving the strict validity of the ss model in this case. Note that due to the statistical nature of the electronic energy loss process, there is an energy loss broadening, which is neglected in the simulation. At high energies (9 keV) the ss model is a reasonable assumption also for the other layers close to the surface, so that a surface peak does not appear. However, at low energies (1 keV), the intensity due to scattering from atoms in the second layer is only partly due to a single binary collision, because of the large scattering cross section. As a consequence, at 1 keV already for second layer atoms ms is significant yielding additional scattered intensity with a broad energy distribution, and for scattering in the third and in deeper layers ms dominates. This is supported by the fact that at 1 keV the number of projectiles backscattered per atomic layer corresponds to the value expected from the ss model only for the outermost atomic layer, while already for the second layer the intensity is larger by 80%, due to ms. For 3 keV, this increase in backscattered intensity from the second layer is 50%, while it is negligible for 9 keV. Note that channeling rather would act in the opposite way, such that deeper layers would contribute less to the scattered intensity. The relevance of ms at different energies may also be seen from comparison of the electronic stopping cross section ee , to the nuclear stopping cross section, en , which describes the energy loss of the projectile due to collisions with target atoms. Even if en does not include the increase of the mean path length due to ms, the comparison of ee and en will yield a good qualitative measure of the importance of ms for given experimental conditions. From Table 1 it is obvious, that for protons, nuclear stopping is of very minor relevance at all energies, and the same holds true for ms. For He ions, ee and en are of similar magnitude at 1–3 keV, and for Ne ions en ee holds in the whole range 1–10 keV. Under these conditions, the ss model

223

Table 1 Electronic and nuclear stopping powers, ee and en , respectively, in units of 1015 eV cm2 , from [5] and the ratio ee =en , for Hþ , Heþ , Neþ and Arþ ions in Cu for various energies ee

en

ee =en

1 keV Hþ Heþ Neþ

3.8 1.9 4.3

0.4 2.2 17.2

9.5 0.86 0.25

3 keV Hþ Heþ Neþ

6.2 3.3 7.5

0.3 2.6 29.0

21.0 1.3 0.26

5 keV Hþ Heþ Neþ

7.8 4.6 9.7

0.3 2.6 34.9

26.0 1.8 0.28

10 keV Hþ Heþ Neþ

10.7 7.8 13.7

0.2 2.3 41.7

53.5 3.4 0.33

will be of limited validity (except for the outermost two atomic layers as discussed above), and the relative importance of the surface peak should be large there, as observed for He ions. As a consequence, projectiles scattered by atoms in the outermost surface and to a lesser extent from the second layer, experience a lower stopping power than those backscattered from deeper layers, so that their energy spectrum has a narrow width, and consequently a large height, which in the spectrum appears as surface peak. In order to further check the importance of the first two atomic layers for the surface peak, we performed TOF-LEIS measurements on a Cu sample, for which the AES analysis had indicated a surface contamination by two monolayers of hydrocarbons (see Fig. 4(a)), and found that the surface peak was quenched by the surface contamination. After sputtering, the surface peak was obtained again (see Fig. 4(a)). To conclude this study, we also performed MARLOWE simulations for Cu with a monolayer of C and of H, respectively, adsorbed on a clean Cu(1 0 0) surface (see Fig. 4(b)). The qualitative result of these calculations is a reduction of the surface peak intensity by the adsorbed layers, without changing the rest of the spectrum noticeably.

time-of-flight spectrum N(ttof) (arb.u.)

224

M. Draxler et al. / Nucl. Instr. and Meth. in Phys. Res. B 203 (2003) 218–224

Acknowledgements 0. 016

3 keVHe →Cu

This project was partially supported by the Austrian Science Fund (FWF) under project number P12471-NAW. Inspiring discussions with H.H. Brongersma and W. Heiland are gratefully acknowledged.

0. 014 0. 012 0. 010 0. 008 0. 006 0. 004

Cuclean Cu +2 layer of hydrocarbons

0. 002 0. 000 500

520

540

560

580

600

References 620

t-ttof (arb.u.)

(a)

250

3 keV He+ →Cu(100)with adsorbates

N(Ef) (arb.u.)

200 150 100 50 0 1600

(b)

1800

2000

2 200

2400

2600

final energy E f [eV]

Fig. 4. (a) TOF-LEIS spectrum, for a clean Cu target and for a Cu target, covered with two monolayers of hydrocarbons, measured at a ¼ 0°. (b) MARLOWE simulation, obtained for a ¼ 0°, for clean Cu(1 0 0) and with one monolayer of H and of C, respectively (see insert).

4. Conclusions We have shown that in the LEIS regime neutral spectra exhibit a surface peak, due to a reduced energy width of the contribution of the surface layer. Even in the regime of strong ms, scattering from surface atoms occurs practically exclusively by binary collisions, while in deeper layers ms sets in. As a consequence, in the regime of strong ms, a pronounced surface peak will appear, fading away at higher energies where ms is of minor importance. Experiment as well as simulation show this behavior, so that other possible reasons for the appearance of a surface peak like channeling can safely be ruled out.

[1] E. Taglauer, in: J.C. Vickerman (Ed.), Surface Analysis – The Principal Techniques, Wiley-VCH, Weinheim, 1997. [2] P. Bauer, in: H. Bubert, H. Jenett (Eds.), Surface and Thin Film Analysis, Wiley-VCH, Weinheim, 2002, p. 150. [3] L. Palmetshofer, in: H. Bubert, H. Jenett (Eds.), Surface and Thin Film Analysis, Wiley-VCH, Weinheim, 2002, p. 141. [4] J.A. Leavitt, L.C. McIntire, M.R. Weller, in: J.R. Tesmer, M. Nastasi (Eds.), Handbook of Modern Ion Beam Materials Analysis, Materials Research Society, Pittsburgh, 1995, p. 37. [5] J.F. Ziegler, J.P. Biersack, U. Littmark, The Stopping and Range of Ions in Solids, vol. 1, Pergamon, New York, 1985. [6] E. Rauhala, in: J.R. Tesmer, M. Nastasi (Eds.), Handbook of Modern Ion Beam Materials Analysis, Materials Research Society, Pittsburgh, 1995, p. 3. [7] P. Sigmund, B. Winterbon, Nucl. Instr. and Meth. 119 (1974) 541; 125 (1975) 491. [8] J.A. Davies, W.N. Lennard, I.V. Mitchell, in: J.R. Tesmer, M. Nastasi (Eds.), Handbook of Modern Ion Beam Materials Analysis, Materials Research Society, Pittsburgh, 1995, p. 343. [9] Z. Smit, Phys. Rev. A 48 (1993) 2070. [10] R. Souda, N. Aono, Nucl. Instr. and Meth. B 15 (1986) 114. [11] T. Buck, G.H. Wheatley, D.P. Jackson, A.L. Boers, S. Luitjens, E. Van Loenen, A.J. Algra, W. Eckstein, H. Verbeek, Nucl. Instr. and Meth. 194 (1982) 649. [12] M. Draxler, S.N. Ermolov, K. Schmid, C. Hesch, A. Poschacher, R. Gruber, M. Bergsmann, P. Bauer, in: Proceedings of the 15th International Conference on Ion– Surface Interections (ISI-2001), Zvenigorod, Moscow district, Russia, 27–31 August 2001, to be published. [13] T. Buck, G.H. Wheatley, G.L. Miller, D.A.H. Robinson, Y.-S. Chen, Nucl. Instr. and Meth. 149 (1978) 591. [14] M.T. Robinson, I.M. Torrens, Phys. Rev. B 9 (1974) 5008. [15] R. Cortenraad, A.W. Denier van der Gon, H.H. Brongersma, Surf. Interface Anal. 29 (2000) 524. [16] M. Barat, J.A. Fayeton, Y.J. Picard, Rev. Sci. Instr. 71 (2000) 2050. [17] E. Steinbauer, Program RBS, 1993, private communication.