The influence of Ti capping layers on CoSi2 formation

The influence of Ti capping layers on CoSi2 formation

Microelectronic Engineering 50 (2000) 125–132 www.elsevier.nl / locate / mee The influence of Ti capping layers on CoSi 2 formation a, a a b b,c C. D...

161KB Sizes 1 Downloads 6 Views

Microelectronic Engineering 50 (2000) 125–132 www.elsevier.nl / locate / mee

The influence of Ti capping layers on CoSi 2 formation a, a a b b,c C. Detavernier *, R.L. Van Meirhaeghe , F. Cardon , R.A. Donaton , K. Maex a

Vakgroep Vaste Stof-Wetenschappen, Laboratorium voor Kristallografie en Studie van de Vaste Stof, Universiteit Gent, Krijgslaan 281 /S1, B-9000 Gent, Belgium b IMEC, Kapeldreef 75, B-3001 Leuven, Belgium c INSYS, Katholieke Universiteit Leuven, B-3001 Leuven, Belgium

Abstract Cobalt silicide formation is very sensitive to the presence of oxygen. Oxygen contamination may originate from different sources: impurities in the annealing ambient, oxygen incorporated within the deposited Co layer and interfacial oxide at the Co / Si interface. In this work, it is shown that the cause of the sensitivity towards oxygen contamination is the formation of a SiO x diffusion barrier between CoSi and the unreacted Co. This causes an increase in the activation energy for CoSi formation. Furthermore, we will show that a titanium capping layer eliminates the sensitivity of CoSi 2 formation for oxygen contamination, thus improving the formation of CoSi 2 layers.  2000 Elsevier Science B.V. All rights reserved. Keywords: Cobalt silicide; Titanium capping layer; Oxygen contamination; Silicidation

1. Introduction At present, silicides are commonly used in VLSI technology [1]. Currently, mainly TiSi 2 is applied as contact and interconnect material. However, further downscaling of the TiSi 2 process is difficult due to problems with the nucleation of the low resistive C54 phase in narrow lines. Due to its low resistivity, CoSi 2 is a possible alternative to TiSi 2 . The main problem for the use of CoSi 2 is the sensitivity of the silicide formation for oxygen [2,3]. Oxygen may delay or even inhibit silicide formation. Other technologically important silicides, like PtSi and NiSi, suffer from this same problem [1]. In the case of TiSi 2 , the problem of oxygen contamination is less critical due to the reducing nature of Ti. Possible solutions consist of using cluster tools or the use of capping layers. The use of a TiN capping layer has been proven useful. TiN is able to protect the forming silicide from contamination from the annealing ambient, but due to its non-reactive nature, it is not able to getter oxygen that was incorporated in the deposited metal or to reduce interfacial oxide. In previous work [4], it has been shown that the use of a Ti capping layer improves the formation of CoSi 2 in narrow lines. During RTP, the reactive Ti capping layer getters water vapour that is adsorbed onto the field *Corresponding author. Fax: 1 32-9-264-4996. E-mail address: [email protected] (C. Detavernier) 0167-9317 / 00 / $ – see front matter PII: S0167-9317( 99 )00272-5

 2000 Elsevier Science B.V. All rights reserved.

126

C. Detavernier et al. / Microelectronic Engineering 50 (2000) 125 – 132

oxide, and thus protects the forming silicide from oxygen contamination. In this work, we will show that the physical background for the sensitivity towards oxygen contamination is the formation of a SiO x diffusion barrier between CoSi and the unreacted Co. Furthermore, we will show that a titanium capping layer eliminates the sensitivity of CoSi 2 formation for oxygen contamination, thus improving the formation of CoSi 2 layers. Other authors [5,6] have already pointed at certain advantages of Ti capping layers, among them an increased uniformity and thermal stability of the CoSi 2 layer. These advantages were, however, not in the context of oxygen contamination.

2. Experimental The p-Sik100l substrates were cleaned using a standard RCA cleaning, followed by a dip in diluted HF. On some samples, a thin oxide was intentionally grown by immersion in a boiling HCl / H 2 O 2 / H 2 O solution (3 / 1 / 1) for 5 min (Shiraki cleaning [7]). Co layers were deposited using e-gun evaporation (99.9% pure Co source) in a vacuum of 10 26 mbar. In order to reduce the oxygen partial pressure in the vacuum chamber, Ti was evaporated (while the samples were protected by a shutter) just prior to Co evaporation. This simulated the effect of a titanium getter pump. Without breaking the vacuum, Ti capping layers of various thicknesses were deposited on top of the Co. Layer thickness was monitored in situ by a quartz crystal balance and checked ex situ by profilometry and X-ray reflectivity. To study the silicidation reaction, isochronal RTP-annealing (30 s, nitrogen ambient) was done at various temperatures. RTP annealing was done in an AST Superheat 1000 system. A thermocouple was used to measure temperatures lower than 6008C, for higher temperatures a pyrometer was used. The sheet resistance of the samples was measured by a four-point probe. XRD was done using a Siemens D5000 diffractometer and XPS by means of a Phi 5500 system (Al Ka source, hemisperical analyser). Oxygen contamination during silicidation may originate from different sources: (1) the annealing ambient; (2) oxygen incorporated within the deposited Co layer; and finally (3) interfacial oxide at the Co / Si interface. In order to separate these different sources: (1) we have tried annealing in pure oxygen; (2) we compared Co layers evaporated under high vacuum and poor vacuum; and (3) we used a thin, chemically grown oxide as a prototype system for a thin interfacial oxide.

3. Results In a first series of experiments, we studied cobalt silicide formation by annealing in oxygen ambient. After annealing, the samples were selectively etched in a boiling H 2 SO 4 / H 2 O 2 (3 / 1) solution during 10 min. After etching, the sheet resistance of the samples was measured using a four-point probe. The sheet resistance as a function of annealing temperature and for different thicknesses of the Ti capping layer is given in Fig. 1 (20-nm thick Co layer). Annealing of the uncapped system in pure oxygen ambient did not result in the formation of a continuous CoSi 2 layer. The decrease in sheet resistance between 550 and 7008C is probably due to the formation of a very thin, irregular and, therefore, poorly conductive layer. This decrease in sheet resistance was less pronounced for 10 nm of Co, probably because in case of thick Co layers the upper part of the Co layer acts as a kind of capping layer, permitting limited silicide formation at the Co / Si interface at the

C. Detavernier et al. / Microelectronic Engineering 50 (2000) 125 – 132

127

Fig. 1. Sheet resistance as a function of annealing temperature (30 s RTP) for oxygen annealing of 20-nm Co layers; uncapped (j) and capped with 3 nm (d), 7 nm (m), 10 nm (.), 15 nm (♦) and 20 nm ( 1 ) of Ti. The sheet resistance was measured after selective etching of the unreacted metal.

beginning of the annealing treatment. However, the silicide layer thus formed is clearly of poor quality. If the Co is capped with Ti, annealing in pure oxygen ambient resulted in normal silicide formation as evidenced by the measured sheet resistance. If the capping layer is very thin (3 nm), it improves CoSi 2 formation, but the oxygen ambient still influences CoSi formation. If the capping layer is thick enough (7–15 nm), both CoSi and CoSi 2 formation proceed as in N 2 ambient (CoSi formation is initiated at about 4008C, CoSi 2 formation at about 5508C). If the capping layer is too thick, CoSi and CoSi 2 still form, but the sheet resistance of the silicide layer is slightly increased, and its thermal stability is decreased. This may be explained by the fact that due to intermixing of Co and Ti, a certain amount of Co cannot be used for the silicidation reaction, resulting in a thinner CoSi 2 layer. Moreover, Ti is known to be unstable on top of CoSi 2 at high temperatures ( . 8508C), resulting in the formation of ternary Co x Ti y Si z phases [8]. The phase formation was also studied by grazing incidence XRD. For uncapped Co (20 nm) annealed at 7008C, evidence was found for the formation of cobalt oxide, and a small amount of CoSi 2 grains (Fig. 2). For Ti-capped Co annealed under the same conditions, grazing incidence XRD showed the presence of CoSi 2 and TiO 2 phases. XPS measurements also showed that the Ti capping was transformed into TiO 2 . This TiO 2 proved to be a very efficient diffusion barrier, protecting the unreacted Co from oxygen contamination. The results obtained for annealing in pure oxygen ambient demonstrate that oxygen in the annealing ambient is very detrimental for CoSi 2 formation, since the oxygen is able to penetrate the unreacted Co layer, and inhibit silicide formation. Titanium capping is very efficient in protecting the unreacted Co from oxygen contamination. The results obtained for annealing in pure oxygen atmosphere clearly illustrate that the Ti capping process will protect the unreacted Co from small traces of oxygen contamination in the normally used nitrogen ambient. This oxygen contamination in the RTP ambient may originate, e.g., from a bad purge management or even from outgassing of the wafer itself [9].

128

C. Detavernier et al. / Microelectronic Engineering 50 (2000) 125 – 132

Fig. 2. XRD for Ti(20 nm) / Co(20 nm) / Si (a) and Co(20 nm) / Si (b) annealed for 30 s at 7008C in oxygen ambient. XRD peaks could be identified as CoSi 2 (h), TiO 2 (앳) and cobalt oxide (,).

In order to investigate the effect of the cleanliness of the as-deposited Co layer on the silicidation reaction, we compared the silicidation reaction for ‘clean’ Co layers that were evaporated rapidly (2 ˚ / s) under high vacuum (10 26 mbar) to ‘contaminated’ Co layers that were evaporated slowly (0.5 A ˚ / s) at 10 24 mbar. XPS depth profiles of the as-deposited layers indicated 2–3% of oxygen in the A contaminated Co, while the oxygen level in the clean Co was below the detection limit of XPS. In a first series of experiments, the sheet resistance was measured as a function of annealing temperature (30 s nitrogen annealing). In order to eliminate effects of contamination from the annealing ambient, all samples were annealed together (in the same nitrogen ambient). Using XRD, we only detected Co peaks after annealing the contaminated Co at 4008C (no Co 2 Si peaks were detected). From the sheet resistance measurements (Fig. 3), it is clear that for the contaminated Co, the formation of CoSi is delayed to higher temperature, while CoSi 2 formation occurs at the same temperature (5508C) as compared to clean Co. In the presence of a Ti cap, the formation of CoSi occurs at the same temperature as for the uncapped, clean Co, and the formation of CoSi 2 is delayed to higher temperature (6258C). XPS depth profiling during CoSi formation for the contaminated layer indicates the presence of a SiO x layer between the already formed CoSi and the unreacted Co. In order to continue CoSi formation, Co has to diffuse through this SiO x diffusion barrier. This explains the increase in reaction temperature. For the Ti-capped contaminated Co, XPS depth profiling did not show any oxygen between CoSi and the unreacted Co. However, the XPS peak of Ti in the capping layer contained a TiO component, indicating that oxygen contamination from the Co layer was gettered by the Ti cap. In order to study the kinetics of the CoSi formation in detail, RTP annealing was done for different durations at different temperatures. By measuring the reaction rate R at different temperatures, the activation energy Ea can be determined. The reaction proved to be diffusion controlled for all samples. However, the activation energy for CoSi formation for the uncapped, contaminated Co (3.77 eV) was

C. Detavernier et al. / Microelectronic Engineering 50 (2000) 125 – 132

129

Fig. 3. Sheet resistance as a function of annealing temperature (30s RTP in N 2 ) for uncapped Co (10 nm) evaporated in high (j) and poor (♦) vacuum and for Ti (5 nm)-capped Co (10 nm) evaporated in high ( 1 ) and poor ( 3 ) vacuum.

significantly larger than the activation energy for the Ti-capped, contaminated Co (2.24 eV). This is caused by the fact that cobalt has to diffuse through the oxide before it is available for the silicidation reaction. For the Ti-capped Co, no diffusion barrier is present, resulting in a lower activation energy for CoSi formation. In order to study the influence of a thin interfacial oxide on CoSi 2 formation, we have deposited 10 nm of Co followed by various thicknesses of Ti on samples that were chemically oxidised. To study the silicidation reaction, isochronal annealing (30 s) was done at various temperatures. In the absence of a Ti cap, ex situ annealing of thick (20 nm) Co layers resulted in laterally inhomogeneous and discontinuous CoSi 2 formation. Annealing in vacuum, however, resulted in reproducible silicidation. In the presence of a Ti cap, silicidation always occurred in a reproducible way. The formation temperature of the CoSi and CoSi 2 phases is, however, strongly dependent on the thickness of the Ti capping layer. For thin Ti capping layers (1–2 nm), silicidation always occurred, but the silicidation reaction is delayed up to 700–8008C. When the thickness of the capping layer is increased, the formation temperature for CoSi is lowered, almost down to its normal value for the direct Co / Si reaction without interfacial oxide. XPS measurements provided proof that Ti from the capping layer diffuses through the Co layer. Once enough Ti has reached the Co / SiO 2 interface, the interfacial oxide is reduced to probably some form of Co x Ti y O z . At first, SiO 2 acts as a diffusion barrier. Once enough Ti from the cap has reached the Co / SiO 2 interface, Ti transforms the SiO 2 diffusion barrier into a Co x Ti y O z diffusion membrane, enabling faster Co diffusion towards the Si substrate, thus initiating CoSi formation. This mechanism is not restricted to chemically grown oxides, but also works for thin, thermally grown oxide layers. XRD measurements provided proof for different preferential orientation of the CoSi 2 layer formed by the different processes (Fig. 4): uncapped Co on clean Si results in polycrystalline CoSi 2 , Ti-capped Co on clean Si results in a preferential k220l orientation for the CoSi 2 , while Ti-capped Co on oxidised Si results in preferential k400l oriented CoSi 2 . By optimising the Ti capping thickness,

130

C. Detavernier et al. / Microelectronic Engineering 50 (2000) 125 – 132

Fig. 4. Main XRD peaks for CoSi 2 . The CoSi 2 was formed by RTP (11008C, 30 s, N 2 ) from an uncapped Co (20 nm) layer ( 3 ), a Ti (10 nm) / Co (20 nm) / Si bilayer ( 1 ) and a Ti (10 nm) / Co (20 nm) / SiO 2 / Si structure (s).

good quality epitaxial CoSi 2 may be formed. This epitaxial growth may be explained by a combination of oxygen-mediated epitaxy (OME [10]) and Ti interlayer-mediated epitaxy (TIME [11]), and will be treated in more detail in a forthcoming publication.

4. Technological applications The Ti capping process holds much promise for technological applications: it leads to a process in which CoSi 2 formation is independent of oxygen contamination. This oxygen contamination may originate from various sources: the annealing ambient, impurities in the Co layer itself, water vapour adsorbed on the field oxide or interfacial oxide that was not completely removed by the HF dip, or which has regrown before the wafers were introduced into the deposition chamber. Other authors [4] have shown that the resulting CoSi 2 lines that are formed using this process have a resistance that is linewidth independent, opening the possibility of further downscaling linewidth down to at least 0.08 mm [12]. One negative aspect is the possible interaction of Ti from the capping layer with the field and spacer oxide. However, in device manufacturing silicidation is always carried out in a two-step process in order to avoid lateral overgrowth and to limit the amount of Si consumption. A first annealing step is done at low temperature (typically 5008C), followed by a selective etch and a second annealing step at higher temperature (about 7508C). The selective etch removes the Ti capping layer, together with unreacted Co. Thus, the danger for Ti oxide interaction is only present during the first, low-temperature annealing step. During this step, some Ti can reach the interface, enough to reduce

C. Detavernier et al. / Microelectronic Engineering 50 (2000) 125 – 132

131

thin layers of interfacial oxide there present, but the effect on the thick field oxide will be negligible. Moreover, by tuning the thickness of the Ti cap, one may further optimise the process in order to minimise interaction between Ti and the field oxide. One can observe (Fig. 3) that Ti capping broadens the process window for CoSi formation during the first annealing step as compared to uncapped Co layers. Furthermore, a titanium cap may be very beneficial if regions of the wafers are covered with a thin interfacial oxide (e.g., due to insufficient cleaning or re-oxidation): without Ti cap, no CoSi would be formed in these regions during the first annealing step, resulting in a loss of yield. Using a thick enough Ti cap, CoSi can also be formed in these regions. In summary, the Ti capping method reduces the need for tight process control (cleanliness of the deposited Co film, RTP temperature, contamination of the RTP ambient) during CoSi 2 formation, rendering the silicide formation a more robust and reliable process. 5. Conclusion In conclusion, we have shown that the use of a reactive Ti cap eliminates the sensitivity of CoSi 2 formation for oxygen contamination, thus resulting in a robust process to form CoSi 2 . Firstly, the Ti cap protects the unreacted cobalt from oxidation and indiffusion of oxygen from the annealing ambient. Secondly, it can getter oxygen that was incorporated into the cobalt film during deposition. Thirdly, Ti from the cap can reduce interfacial oxide layers that are present at the cobalt–silicon interface. There is no apparent reason why the Ti capping method could not be applied to other silicides, whose formation is also strongly influenced by oxygen contamination. We have started experiments on PtSi and NiSi / NiSi 2 to verify this. Acknowledgements We would-like to thank Ing. L. Van Meirhaeghe for technical assistance and Ing. U. Demeter for XPS measurements. C. Detavernier thanks the ‘Fonds voor Wetenschappelijk Onderzoek–Vlaanderen’ (FWO) for a scholarship. K. Maex is a research director for the FWO.

References [1] J.P. Gambino, E.G. Colgan, Silicides and ohmic contacts, Mater. Chem. Phys. 52 (1998) 99–146. [2] W.J. Freitas, J.W. Swart, The influence of impurities on cobalt silicide formation, J. Electrochem. Soc. 138 (1991) 3067–3070. [3] K. Maex, Mater. Sci. Eng. 11 (1993) 53–153. [4] P.R. Besser, A. Lauwers, N. Roelandts, K. Maex, W. Blum, R. Alvis, M. Stucchi, M. De Potter, The influence of capping layer type on cobalt salicide formation in films and narrow lines, in: MRS Symp. Proc., 1998. [5] R.T. Tung, F. Schrey, Increased uniformity and thermal stability of CoSi 2 thin films by Ti capping, Appl. Phys. Lett. 67 (1995) 2164–2166. [6] D.K. Sohn, J.S. Park, B.H. Lee, J.U. Bae, J.S. Byun, J.J. Kim, Formation of CoTi barrier and increased thermal stability of CoSi 2 film in Ti-capped Co / Si(100) system, Appl. Phys. Lett. 73 (1998) 2302–2304. [7] A. Ishizaka, Y. Shiraki, Low temperature surface cleaning of silicon and its application to silicon MBE, J. Electrochem. Soc. 133 (1986) 666–671.

132

C. Detavernier et al. / Microelectronic Engineering 50 (2000) 125 – 132

[8] M. Setton, J. van der Spiegel, Silicide formation for Co / Ti / Si structures processed by RTP under vacuum, Appl. Surf. Sci. 38 (1998) 62–71. [9] K. Maex, E. Kondoh, A. Lauwers, A. Steegen, M. De Potter, P. Besser, J. Proost, in: MRS Symp. Proc., 1998. [10] R.T. Tung, Oxide mediated epitaxy of CoSi 2 in silicon, Appl. Phys. Lett. 68 (24) (1996) 3461–3463. [11] M. Lawrence, A. Dass, D.B. Fraser, C.S. Wei, Growth of epitaxial CoSi 2 on (100)Si, Appl. Phys. Lett. 58 (1991) 1308–1310. [12] Q.F. Wang, K. Maex, S. Kubicek, R. Jonckheere, B. Kerkwijk, R. Verbeeck, S. Biesemans, K. De Meyer, New CoSi 2 salicide technology for 0.1-mm process and below, in: Symp. on VLSI Tech., 1995.