Kinetic energy dependence of TiSi2 film growth from low energy Ti+ ion beams

Kinetic energy dependence of TiSi2 film growth from low energy Ti+ ion beams

Nuclear Instruments and Methods in Physics Research B 157 (1999) 220±225 www.elsevier.nl/locate/nimb Kinetic energy dependence of TiSi2 ®lm growth f...

168KB Sizes 0 Downloads 53 Views

Nuclear Instruments and Methods in Physics Research B 157 (1999) 220±225

www.elsevier.nl/locate/nimb

Kinetic energy dependence of TiSi2 ®lm growth from low energy Ti‡ ion beams S.M. Lee, E.T. Ada, H. Lee, D. Marton, J.W. Rabalais

*

Department of Chemistry, University of Houston, Houston, TX 77204-5641, USA

Abstract The growth of TiSi2 thin ®lms by using low energy (10±500 eV) Ti‡ ion beams on Si(1 1 1) has been studied by in situ Auger electron spectroscopy (AES) and re¯ection high energy electron di€raction (RHEED) as well as ex situ depth pro®le X-ray photoelectron spectroscopy (d-XPS). Stoichiometric TiSi2 ®lms were grown with Ti‡ energies of 50 eV at a temperature of 600°C. The in situ RHEED and AES data reveal the existence of some elemental Si in the surface region, whose concentration is dependent on the kinetic energy of the incoming Ti‡ ions. Resolution of the chemical shift between the Si(2p) lines of the silicide and elemental silicon allows the ex situ d-XPS pro®les to reveal the ®lm composition and thickness. The e€ect of energetic ions in silicide ®lm formation is discussed accordingly. The data show consistent trends that demonstrate that irradiation with Ti‡ ions suppresses the di€usion of Si as the ion kinetic energy increases. It is proposed that this radiation inhibited di€usion process is due to void formation resulting from the high density of defects in the subsurface region which inhibits di€usion of Si atoms through the silicide ®lm to react with incoming Ti atoms. Ó 1999 Elsevier Science B.V. All rights reserved. PACS: 81.15; 68.55; 81.15.A; 81.15.H Keywords: Ti in sapphire (a-Al2 O3 ); Thin ®lms; Deposition methods; Ion beam deposition

1. Introduction Refractory metal silicides have been intensively studied due to their low resistivity and high thermal stability. Titanium silicide is considered to be an optimum choice for applications as contacts and interconnects on Si metal oxide semiconductor (MOS) devices because it exhibits very low resis-

* Corresponding author. Tel.:+1-713-743-3282; fax: +1-713743-2709; e-mail: [email protected]

tivity [1±4]. In most of the studies reported to date, silicide ®lms were formed by physical vapor deposition on silicon surfaces at room temperature followed by high temperature treatment. Four di€erent crystalline phases (Ti5 Si3 , TiSi, C49± TiSi2 , and C54±TiSi2 ) and an amorphous phase have been reported to form in these thin ®lms [5,6]. When the ®lm thickness is approximately 200±600  C54±TiSi2 is reported to be the thermodynamA, ically favorable phase; this is the only phase existing at temperatures above 700°C. C49±TiSi2 is the kinetically favorable metastable structure

0168-583X/99/$ ± see front matter Ó 1999 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 8 - 5 8 3 X ( 9 9 ) 0 0 4 1 6 - 4

S.M. Lee et al. / Nucl. Instr. and Meth. in Phys. Res. B 157 (1999) 220±225

which forms at the lower temperatures of 500°± 600°C in this thickness region [7]. While C54±TiSi2 is the most stable state with the lowest resistance, it has several drawbacks, such as rough surface morphology, incomplete surface coverage, and high fabrication temperature [2,8] which are incompatible with the fabrication of shallow junctions. Various ®lm fabrication methods, such as codeposition of Si and Ti and Ti deposition on amorphous Si substrates, have been attempted in order to overcome these shortcomings. Synergism between ion beam induced e€ects and substrate temperature has been observed in various systems. These include, for example, the case of Si‡ ion beam homoepitaxy, where a combination of elevated substrate temperature (160°C) and low ion energy (20 ‹ 10 eV) can signi®cantly enhance low temperature epitaxial ®lm growth beyond the molecular beam epitaxy (MBE) process [9]. Such e€ects have also been observed in the case of preparation of clean, smooth and well ordered silicon surfaces at a low processing temperature of 500°C by simultaneous irradiation with a 100 eV Ar‡ ion beam [10]. In this paper, we present a study of the formation of TiSi2 thin ®lms by low energy Ti‡ deposition on a Si(1 1 1) substrate at elevated temperature. A substrate temperature of 600°C was chosen because only the TiSi2 phase exists at this temperature. By comparing ®lms grown from Ti‡ beams of di€erent kinetic energies (10, 50, 100 and 500 eV), we show how the energy of the impinging Ti‡ ions a€ect the structure and composition of the ®lm surface, the bulk composition of the ®lm and the sharpness of the interface. 2. Experimental methods Titanium silicide ®lms were grown by direct impact of 48 Ti‡ ion beams on clean Si(1 1 1) surfaces. The ion beam instrument provides low energy (5 eV±10 keV), mass-selected ion beams that impinge on a target in an UHV chamber [11]. The spread of the ion energy is typically ‹3 eV. The base pressure in the deposition chamber is 2 ´ 10ÿ8 Pa and during the deposition it rises to 1 ´ 10ÿ7 Pa, the pressure increase being due to incoming Ti‡

221

ions into the main chamber. The sample temperature is monitored by using an infrared pyrometer which is calibrated with a thermocouple. The instrument is equipped with in situ Auger electron spectroscopy (AES), X-ray photoelectron spectroscopy (XPS) and re¯ection high energy electron di€raction (RHEED). AES spectra were collected with a 3 keV electron gun using a VSW Scienti®c hemispherical electron analyzer. The surface structures of the titanium silicide ®lms were examined using RHEED at 20 kV acceleration voltage with an incident angle of 1.5° with respect to the sample surface. Ex situ XPS spectra were collected on a Physical Electronics Model 5700 instrument using standard Al Ka X-rays. XPS depth pro®les (d-XPS) were obtained on this instrument by 5 keV Ar‡ sputtering. Si(1 1 1) samples (CZ, p-type, B-doped, 0.3 X cm) measuring approximately 7 ´ 7 mm2 were ultrasonically cleaned in methanol and loaded into the UHV chamber. Before deposition, the Si(1 1 1) substrate was annealed at 1200°C for 1 min. for cleaning. Its cleanness and crystallinity were then checked by in situ AES and RHEED. Ti‡ beams of four di€erent kinetic energies, i.e. 10, 50, 100, and 500 eV, were used for deposition while the substrate was maintained at 600°C. The Ti‡ ion current to the sample was 15 nA/mm2 and the deposition spot size was 8 ´ 10 mm2 . Depositions were carried out for 3 h for a total Ti‡ ion dose of 1 ´ 1017 ions/cm2 for each di€erent kinetic energy. If the Ti‡ ions have unit sticking probability, this dose is equivalent to 45 nm thickness of deposited titanium. After deposition, in situ AES and RHEED were used to study the surface composition and structure. The samples were then removed from the deposition chamber for d-XPS studies. 3. Results 3.1. Depth pro®le of a deposited ®lm An example of an XPS depth pro®le of a TiSi2 ®lm which was deposited from a 50 eV Ti‡ beam at 600°C is shown in Fig. 1. A small chemical shift (0.4 eV) between elemental Si and Si in the tita-

222

S.M. Lee et al. / Nucl. Instr. and Meth. in Phys. Res. B 157 (1999) 220±225

Fig. 1. Ex situ XPS depth pro®le of a TiSi2 ®lm grown by means of a 50 eV Ti‡ beam on a Si(1 1 1) substrate at 600°C. The depth pro®le used 5 keV Ar‡ ions for sputtering. Clear resolution of the TiSi2 ®lm from the substrate Si and a sharp interface region are observed. The insert shows the chemical shift between the 2p peaks of the silicide Si and the elemental Si in the bulk.

nium silicide ®lm was observed in the XPS analysis (inset of Fig. 1). The silicide layer could be resolved from the substrate Si by deconvolution of the Si peaks. Fig. 1 shows a clearly resolved titanium silicide ®lm layer in which elemental Si is below the detection level. Note that the relative atomic concentrations of Si and Ti in the silicide ®lm have a nearly perfect stoichiometric composition of TiSi2 . The d-XPS plot shows that there are three regions of interest in the pro®le: (1) the outermost surface layers where concentrations are changing rapidly; (2) the silicide ®lm region where the concentrations are constant as a function of sputtering time; (3) the interface region between the silicide ®lm and the silicon substrate where concentrations again change rapidly.

20 eV, only a faint trace of these spots can still be observed (Fig. 2(b)). These faint spots totally disappear into a di€use pattern for kinetic energies of 30 eV (Fig. 2(c)). These images show that crystalline Si is in the surface region of the 10 eV deposited sample and that this crystallinity is loss for slightly higher ion energies. In situ AES analysis of the [Si]/[Ti] signal ratios in the samples after deposition in Fig. 3 show that this ratio is 12 for 10 eV deposition and then decreases rapidly with increasing energy. This large surplus of elemental Si near the surface when low energy Ti‡ is used for deposition is in agreement with the observation of elemental Si in Fig. 2(a). The [Si]/[Ti] ratio decreases exponentially as the Ti‡ kinetic energy increases.

3.2. Outermost surface layers

3.3. Titanium silicide ®lm

In situ RHEED images taken immediately after ®lm deposition exhibit spotty patterns when the ®lm is formed by 10 eV Ti‡ beam deposition as shown in Fig. 2(a). The lattice constant calculated from this image corresponds exactly to that of Si(1 1 1). When the Ti‡ kinetic energy increases to

The relative atomic concentrations of Si and Ti in the silicide ®lms have a nearly perfect stoichiometric composition of TiSi2 as shown in Fig. 1. The individual elemental concentrations are relatively ¯at over a large sputtering time, indicating that the composition is uniform over the width of

S.M. Lee et al. / Nucl. Instr. and Meth. in Phys. Res. B 157 (1999) 220±225

223

Fig. 3. Dependence of the [Si]/[Ti] concentration ratio on the kinetic energy of the impinging Ti‡ ions. (a) [Si]/[Ti] ratio in surface region by in situ AES; [Si] represents the total AES Si signal, i.e silicide plus elemental Si, (b) [Si]/[Ti] ratio in TiSi2 ®lm region by ex situ d-XPS; [Si] represents the total XPS Si signal, however only silicide Si is observed in the ®lm region (Fig. 1).

the ratio [Si]/[Ti] is 2 for E < 100 eV and drops to 1 for E ˆ 500 eV, indicating that stoichiometric TiSi2 ®lms are formed with Ti‡ energies <100 eV. These ratios also exhibit an exponential dependence, however they are considerably lower than those obtained from in situ AES, particularly at low kinetic energy. The reason for this is that there is no elemental Si in the ®lm region. 3.4. Interface region

Fig. 2. RHEED images of the ®lm surface formed by di€erent deposition energies of the Ti‡ ions. (a) 10 eV, (b) 20 eV, (c) 30 eV.

the ®lm. This stoichiometry persists for Ti‡ deposition energies of less than 100 eV. This is exempli®ed in the ex situ XPS measurements of the [Si]/[Ti] ratio in Fig. 3, which were taken in the ¯at region of the d-XPS pro®les of Fig. 1. Note that

In the interface region of Fig. 1 between the titanium silicide ®lm and the Si(1 1 1) substrate, the concentration gradients of the species are di€erent for each ®lm made by Ti‡ deposition at di€erent kinetic energies. In order to compare the widths of this interface region for ®lms deposited at di€erent kinetic energies, the ®rst-order derivatives of each species in the XPS depth pro®les in the interface region were calculated. The full-width-at-halfmaximum (FWHM) of these derivatives is a measure of the width of the interface between the silicide layer and the underlying Si substrate. These FWHM values are plotted in Fig. 4 as a function of Ti‡ deposition energy. The widths of the interfaces for all three species, i.e., Ti, Si in TiSi2 ,

224

S.M. Lee et al. / Nucl. Instr. and Meth. in Phys. Res. B 157 (1999) 220±225

Fig. 4. Dependence of the FWHM of the ®rst-order derivatives of each species in the interface region of the d-XPS pro®les of Fig. 1.

and elemental Si in the substrate, decrease exponentially as the kinetic energy of the impinging Ti‡ ions increases.

4. Discussion The penetration depth of 500 eV Ti‡ ions into silicon is estimated from TRIM [12] simulations to be 22 eV. These simulations are not accurate at the low energy end of the range used here, i.e., 10 eV. It is expected that 10 eV will be below the penetration threshold, resulting in deposition of elemental Ti on the surface. Thus, for the very low energy depositions, it is expected that excess titanium will be observed in the surface region. This is corroborated by the observation of a metallic Ti layer on the Si surface when 10 eV Ti‡ is deposited at room temperature where di€usion is negligible [13]. Therefore the surplus elemental Si observed in the surface region in Fig. 3(a) indicates that Si is di€using through the silicide layer to the surface at the temperature of 600°C. From the results of an RBS study [14], the di€usion coecient of Si into Ti is known to be much larger than that of Ti into Si. Therefore, Si is the dominant di€using species. Also, since the surface energy of elemental Si (1.01 Jmÿ2 ) is much smaller than that of metallic Ti (2.22 Jmÿ2 ) at 600°C, a Si terminated surface is more stable than a Ti terminated surface [15].

These facts explain why the surface has a surplus of elemental silicon. Consider the various processes that occur simultaneously during energetic Ti‡ deposition at 600°C. (1) Ti atoms are being deposited on the surface and in the subsurface region. (2) These Ti atoms react with elemental Si to form a silicide. (3) Si from the substrate di€uses through this silicide region to the surface region where it reacts with deposited Ti atoms. (4) The low surface energy of elemental Si compared to metallic Ti favors elemental Si in the surface layer. As a result, the relative ratio of Si to Ti in the surface region includes information about the di€usion process in this system. During prolonged Ti deposition, the surface region is under a steady state condition where the ¯ux density of Si atoms diffusing to the surface is equivalent to the ¯ux density of Si which is being consumed in forming TiSi2 . This di€usion process also in¯uences the TiSi2 ®lm-substrate interface width as shown in Fig. 4, where an exponential decrease of the width of the interface is observed as a function of ion energy. It is well-known that high energy particles such as electrons and ions are ecient in creating vacancy-interstitial pairs in solids. At certain temperatures where these defects are mobile and hence can anneal out, the balance of these two rates, i.e., formation vs. annihilation, typically leads to a steady state excess concentration of defects. Since the di€usion rates are normally directly proportional to the defect concentrations, an enhancement of di€usional processes results [16] for radiation damaged solids. This is called radiation enhanced di€usion. Considering the current experiment, even for the low kinetic energy of the impinging Ti‡ employed herein, increasing kinetic energy should alter the surface region by creating more defects. Indeed, the RHEED images of Fig. 2 demonstrate that increasing the Ti‡ energy changes the surface structure from crystalline to disordered at very low energies. However, despite this increased defect density caused by the increased kinetic energy, the di€usion rate is decreased rather than increased. This is radiation inhibited di€usion. Why does the higher energy deposition tend to suppress the di€usion process?

S.M. Lee et al. / Nucl. Instr. and Meth. in Phys. Res. B 157 (1999) 220±225

The defect density in the surface region where the silicide ®lm is being grown is expected to be high as the kinetic energy increases above 100 eV due to the relatively shallow penetration depth of the Ti‡ ions. As a result, there is a high probability for defects to encounter each other and form divacancies. If the defect densities are high enough, these divacancies can coalesce together to form vacancy clusters. Such vacancy clusters can eventually form voids which are nearly immobile and can contribute to blocking the ¯ow of di€using elemental Si to the surface. This radiation inhibited di€usion process minimizes the amount of elemental Si in the surface layers. Therefore the growth of pure stoichiometric TiSi2 ®lms on silicon using energetic Ti‡ ions requires a careful choice of the ion energy. The bene®cial e€ect for high energies (E > 100 eV) is that radiation inhibited di€usion limits the amount of excess elemental Si in the surface region (Fig. 3) and that a sharp ®lmsubstrate interface is formed (Fig. 4). The detrimental e€ect is that at higher energies (E  500 eV) this is exaggerated such that the silicide ®lm is nonstoichiometric due to the Si de®ciency resulting from the inhibited di€usion of Si into the surface region (Fig. 3) as well as a high concentration of defects. The bene®cial e€ect for low energies (E  50 eV) is the growth of a stoichiometric silicide ®lm (Fig. 3), although at the expense of having some elemental silicon in the outer layers. A balance between these e€ects can be best accomplished by using Ti‡ ion energies in the range of 50±100 eV. 5. Conclusions Ion beam kinetic energy e€ects in low energy Ti‡ ion beam growth of titanium silicide on a Si(1 1 1) surface have been studied by RHEED, AES and d-XPS. Stoichiometric TiSi2 ®lms can be grown with Ti‡ energies of 50 eV at a temperature of 600°C. The elemental ratios of [Si]/[Ti] in the ®lm surface and bulk as well as the FWHM of the interface derivative, all show consistent trends that

225

demonstrate that irradiation with Ti‡ ions suppresses the di€usion of Si as the ion kinetic energy increases. It is proposed that this radiation inhibited di€usion process is due to void formation resulting from the high density of defects in the subsurface region which inhibits the di€usion of Si atoms through the silicide ®lm to react with incoming Ti atoms. Acknowledgements This work was supported by the National Science Foundation under award number DMR9616440. References [1] K. Maex, L. Van Den Hove, R.F. Keersmaecker, Thin Solid Films 140 (1986). [2] M.H. Wang, L.J. Chen, J. Appl. Phys. 66 (1992) 5918. [3] C.A. Sukow, R.J. Nemanich, J. Mater. Res. 9 (1994) 1214. [4] R.T. Tung, Appl. Phys. Lett. 68 (1996) 1933. [5] R. Butz, G.W. Rublo€, T.Y. Tan, P.S. Ho, Phys. Rev. B 30 (1984) 5421. [6] J. Vahakangas, Y.U. Idzerda, E.D. Williams, R.L. Park, Phys. Rev. B 33 (1986) 8716. [7] H. Jeon, C.A. Sukow, J.W. Honeycutt, G.A. Rozgonyi, R.J. Nemanich, J. Appl. Phys. 71 (1992) 4269. [8] S.B. Herner, V. Krishnamoorthy, A. Naman, K.S. Jones, H.J. Gossmann, R.J. Tung, Thin Solid Films 302 (1997) 127. [9] J.W. Rabalais, A.H. Al-Bayati, K.J. Boyd, D. Marton, J. Kulik, Z. Zhang, W.K. Chu, Phys. Rev. B 53 (1996) 10781. [10] S.M. Lee, C.J. Fell, D. Marton, J.W. Rabalais, J. Appl. Phys. 83 (1998) 5217. [11] A.H. Al-Bayati, D. Marton, S.S. Todorov, K.J. Boyd, J.W. Rabalais, D.G. Armour, J.S. Gordon, G. Duller, Rev. Sci. Instrum. 65 (1994) 2680. [12] J.F. Zeigler, J.P. Biersack, U. Littmark, in: J.F. Zeigler (Ed.), The Stopping and Range of Ions in Solids, Pergamon, NY, 1985. [13] S.M. Lee, E.T. Ada, H. Lee, J.W. Rabalais, to be published. [14] W.K. Chu, S.S. Lau, J.W. Mayer, H. Muller, K.N. Tu, Thin Solid Films 25 (1975) 393. [15] L.Z. Mezey, J. Giber, Jpn. J. Appl. Phys. 21 (1982) 1569. [16] R.J. Borg, G.J. Dienes, An Introduction to Solid State Di€usion, Academic Press, New York, 1988, p. 247.