Time resolved CoSi2 reaction in presence of Ti and TiN cap layers

Time resolved CoSi2 reaction in presence of Ti and TiN cap layers

Materials Science and Engineering B 114–115 (2004) 232–235 Time resolved CoSi2 reaction in presence of Ti and TiN cap layers A. Albertia,∗ , R. Front...

130KB Sizes 1 Downloads 8 Views

Materials Science and Engineering B 114–115 (2004) 232–235

Time resolved CoSi2 reaction in presence of Ti and TiN cap layers A. Albertia,∗ , R. Fronterr´ea , F. La Viaa , E. Riminia,b b

a CNR-IMM Sezione Catania, Stradale Primosole 50, 95121 Catania, Italy Dipartimento di Fisica e Astronomia dell’Universit`a di Catania, Via S. Sofia 64, 95123 Catania, Italy

Abstract In this work we study the phase transition of Co/poly-Si layers capped with Ti or TiN films. Silicide reaction has been performed by isothermal annealing in the temperature range between 420 and 510 ◦ C and studied by measuring the sheet resistance during time. The time interval associated to the CoSi–CoSi2 phase transition has been extracted as a function of the cap layer and the annealing temperature. The presence of the Ti cap systematically reduces the rate of CoSi2 formation with respect to the sample with TiN. It has been found that the cap type has an impact on the pre-exponential factor of the growth time but it does not affect the activation energy. As an effect, the silicide capped with Ti has a flat interface with the substrate. © 2004 Elsevier B.V. All rights reserved. Keywords: TiN films; MOS; X-ray spectroscopy

1. Introduction

2. Experimental

CoSi2 thin layers are nowadays used on polycrystalline silicon in place of TiSi2 as gate metallization of metal oxide semiconductor (MOS) devices. In order to preserve the integrity of the device from metallic (e.g. aluminium) or external (e.g. oxygen) contamination, Ti or TiN layers have been used as barrier or cap layers, and therefore they have often been placed on cobalt before silicidation. Silicide reaction in presence of capping layers was of particular interest in the past years since this process has opened the possibility of growing epitaxial CoSi2 layers on silicon [1,2] or improving silicide thermal stability [3]. It is known that a Ti cap, rather than TiN, is able to guarantee a good interface between silicide and silicon, therefore preserving the integrity of gate oxides and shallow junctions. In this work we present the results on the phase transition of Co/poly-Si layers capped with Ti or TiN films in order to study the effect of the cap on the nucleation and diffusion processes and to correlate these to the structural properties of the silicide.

Cobalt layers, 15 nm thick, have been deposited by sputtering on polycrystalline silicon substrates and then capped by 10 nm Ti or TiN layers. Samples have also been left without cap to be used as a reference. Silicidation processes have been performed by furnace annealing at a base pressure of 2–5 × 10−7 mbar in the temperature range between 420 and 510 ◦ C, and studied by in situ sheet resistance, transmission electron microscopy (TEM) and energy dispersive X-ray spectroscopy (EDS) analyses.



Corresponding author. E-mail address: [email protected] (A. Alberti).

0921-5107/$ – see front matter © 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.mseb.2004.07.021

3. Results and discussions A typical transition curve is shown in Fig. 1. After a step of heating up to the desired temperature, during which the resistance of the sample increases as an effect of the phonon scattering, the system is maintained at a fixed temperature, 455 ◦ C in this case, for 4 h. During annealing cobalt atoms diffuse downwards and react with the substrate by sequentially forming Co2 Si and CoSi phases [4]. The formation of these high resistivity layers (≥150 ␮ohm × cm) increases the sheet resistance of the sample up to 45 ohm/sq. By prolonging the annealing, the CoSi phase is transformed into CoSi2

A. Alberti et al. / Materials Science and Engineering B 114–115 (2004) 232–235

Fig. 1. Sheet resistance vs. time during annealing at 455 ◦ C. In the part of the curve labeled with I the nucleation of the CoSi2 phase takes place, followed by the growth of the silicide layer toward the surface of the sample (region II).

233

annealing time, a continuous CoSi2 layer forms in contact with the substrate and grows towards the surface of the sample by consuming the CoSi layer (region II in Fig. 1). The time interval between the maximum and the flexus has been associated to the nucleation process of the CoSi2 phase (nucleation time), while the time from the flexus and the plateau to the growth of the silicide layer (growth time). These characteristic times have been measured for a silicidation process performed at 455 ◦ C with Ti and TiN cap layers and compared to the reference case (Fig. 2). The reaction done without cap does not significantly differ from the case of using a TiN cap. On the contrary, CoSi2 reaction in presence of a Ti cap is characterized by having longer specific times both for the nucleation and the growth processes. CoSi2 nucleation takes about 45 min, 35 min more than the other two cases, and the growth process about 130 min, 110 min more than the references (Fig. 2).

and the resistance progressively reduces to a plateau value of 12.5 ohm/sq. The portion of the resistance curve where CoSi2 reaction takes place is characterized by having a flexus whose presence indicates the occurrence of two different reaction processes. In the time interval between the maximum and the flexus of the curve (region I in Fig. 1), grains of the low resitivity CoSi2 phase sporadically nucleate at the triple grain boundaries of the CoSi layer near the interface with the substrate (TEM not shown). By further prolonging the

Fig. 2. Nucleation and transition times measured at 455 ◦ C.

Fig. 3. Arrhenius plot of nucleation and transition times. The activation energies for silicide nucleation are 1.9 and 2.3 eV for TiN and Ti capped silicide layer, respectively. The energy for the layer growth is independent of the cap type (2 eV) whilst the pre-exponential factors introduce a delay time in the process.

234

A. Alberti et al. / Materials Science and Engineering B 114–115 (2004) 232–235

Silicidation has additionally been performed in the temperature range between 420 and 510 ◦ C, and the specific times for the nucleation and growth of the CoSi2 phase measured as a function of the annealing temperature and the cap type. These time parameters are compared in the Arrhenius plot of Fig. 3. The activation energy found for nucleation is 1.9 ± 0.2 eV for the TiN capped silicide, and 2.3 ± 0.2 eVfor the Ti capped silicide. The fact that in presence of a Ti cap the barrier for CoSi2 nucleation becomes higher than in the reference case is in agreement with other works in the literature [5]. The effect of Ti capping layers in delaying cobalt silicide phase transition was observed as a shift of the transition temperature and the phenomenon was attributed to a change of the nucleation barrier [5]. The growth process of the CoSi2 layer is characterized by an activation energy of about 2 eV that is independent of the cap nature. The two curves are shifted to each other by the pre-exponential factor, which is higher in the Ti capped reaction. Tough the equivalency of the energy, it is to note that the silicide growth in presence of Ti is systematically slowed down with respect of the TiN case. A possible explanation of the phenomenon involves the diffusivity of cobalt through the forming layer. From the growth times, it can be estimated that cobalt diffusivity with the Ti cap is lower, by at least one order of magnitude, with respect to that of the TiN capped samples and the literature values measured for thin films [6]. The diffusion coefficients found in Ti capped silicides is also about two-order of magnitude higher than in bulk systems [7], and this leads to the conclusion that bound-

Fig. 5. EDS analyses of the Ti capped silicide annealed at 455 ◦ C. Ti atoms diffuses downwards into the sample and participates to the reaction processes.

ary diffusion drives CoSi2 layer growth. The reduction of cobalt diffusivity using Ti cannot be attributed to structural differences between Ti and TiN capped silicides since the two-silicide layers have grains of comparable size (260 nm). As a consequence, it is argued that the role of titanium atoms consists of reducing cobalt diffusivity by occupying silicide grain boundaries. Slowing down cobalt diffusion from CoSi through the forming CoSi2 layer has the straightforward effect of smoothing the interfacial roughness with the substrate. This result is shown in Fig. 4a where the reacted silicide has a uniform thickness of 50 nm and a very good interface with silicon. On the opposite, the silicide reacted in presence of TiN (Fig. 4b) has a rough interface with the substrate. Titanium distribution inside the sample has been detected by performing EDS analyses at the TEM through the cross section of the sample (Fig. 5). The surface layer is titanium oxide (3), the intermediate layer (2) is a ternary compound containing Co Ti Si and the inner film is the silicide. With this technique, no a clear evidence of the presence of titanium at the interface has been found probably due to the sensitivity of the technique (∼1019 atmos/cm2 ). Nevertheless, it is to note that, with respect to the TiN behaviour, Ti atoms have a role in the reaction process since they diffuse inside the forming layer and participate to the reaction.

4. Conclusions

Fig. 4. TEM analyses of the silicide layers reacted at 455 ◦ C. The interface between silicide and silicon is very flat when using titanium.

We have studied the phase transition of Co/poly-Si layers capped with Ti or TiN by measuring the sheet resistance as a function of time in the range between 420 and 510 ◦ C. From the resistance curve the time intervals, during which CoSi2 nucleation and growth take place, have been extracted as a function of the annealing temperature. It has been found that silicide nucleation is influenced by

A. Alberti et al. / Materials Science and Engineering B 114–115 (2004) 232–235

the cap with activation energy of 1.9 and 2.3 eV, respectively for TiN and Ti capped silicidation processes. We have additionally found that titanium atoms have an effect on the growth rate. It has activation energy of 2 eV, which is independent of the cap. As a possible explanation of the phenomenon, the diffusion of Ti atoms inside the forming layer slows down Co motion and results in systematically longer time of growth. In the range of temperature investigated, nucleation is a relatively fast process whilst diffusion is the slowest process coming into play. As an important effect, the silicide capped with Ti has a flat interface with the substrate.

235

References [1] S.-L. Zhang, J. Cardenas, F.M. d’Heurle, B.G. Svensson, C.S. Petersson, Appl. Phys. Lett. 66 (1) (1995) 58. [2] G.B. Kim, J.S. Kwak, H.K. Baik, S.M. Lee, S.H. Oh, C.G. Park, Appl. Phys. Lett. 77 (10) (2000) 1443. [3] D.K. Sohn, J.-S. Park, B.H. Lee, J.-U. Bae, J.S. Byun, J.J. Kim, Appl. Phys. Lett. 73 (16) (1998) 2302. [4] C. Lavoie, F.M. d’Heurle, C. Detavernier, C. Cabral Jr., Microelectron. Eng. 70 (2003) 144. [5] C. Detavernier, R.L. Van Meirhaeghe, F. Cardon, K. Maex, W. Vandervorst, B. Brijs, Appl. Phys. Lett. 77 (20) (2000) 3170. [6] C.-D. Lien, M.-A. Nicolet, S.S. Lau, Appl. Phys. A 34 (1984) 249. [7] T. Barge, P. Gas, F.M. d’Heurle, J. Mater. Res. 10 (5) (1995) 1134.