Characterization of Sb-Doped Fully-Silicided NiSi/SiO 2 /Si MOS Structure T. Hosoi 1 , K. Sano 1,2 , M. Hino 1,2 , A. Ohta 2 , K. Makihara 2 , H. Kaku 2 , S. Miyazaki 2 , and K. Shibahara 1,2 1 Research Center for Nanodevices and Systems, Hiroshima University 2 Graduate School of Advanced Sciences of Matter, Hiroshima University 1-4-2 Kagamiyama, Higashi-Hiroshima 739-8527, Japan Phone: +81-82-424-6265, Fax: +81-82-424-3499, E-mail: [email protected] Abstract X-ray photoelectron spectroscopy (XPS) measurement of Sb-doped fully-silicided (FUSI) NiSi/SiO 2 interface has been carried out to evaluate location of Sb pileup and to discuss its role for workfunction shift. The XPS result revealed Sb encroachment into SiO 2 . Workfunction characterization by XPS implied that NiSi workfunction was identical to its original value without Sb pileup located inside the gate oxide. The impact of predoping on silicidation reaction was also investigated. Introduction Single-metal tunable-workfunction gate is required for the next generation devices. Fully- silicided (FUSI) NiSi is one of the most promising candidates, because NiSi workfunction that is near Si midgap can be modulated by impurity pileup formation at the NiSi/gate oxide interface [1-5] by snowplow effect [6]. The reported largest workfunction shift toward Si conduction band was -0.40 eV obtained with Sb predoping [3]. However, a lot of issues to be solved remain for growing this technology to a production level. For examples, difficulty in NiSi phase control has been recently pointed out [7]. This problem becomes more severe accompanying defect formation by doping of poly- Si prior to silicidation in order to modulate NiSi workfunction [4, 5, 8]. Moreover, the physical origin of the FUSI NiSi gate workfunction modulation is still unclear. In this paper, precise evaluation of Sb location by x-ray photoelectron spectroscopy (XPS) is described to discuss the role of Sb for the NiSi workfunction shift. FUSI NiSi MOS Structure Fabrication Fabrication process flow of FUSI NiSi gate MOS diodes is shown in Fig. 1. The gate poly-Si was doped with 30 keV Sb ion implantation at a dose of 5×10 15 cm -2 . As reported previously, silicidation temperature is a key parameter for workfunction modulation [5, 8]. Fast silicidation at 500 ºC resulted in no function shift, and by lowering silicidation temperature to at 450 ºC, workfunction reduction was obtained as shown in Fig. 2. Slower silicidation increased Sb pileup concentration at the NiSi/SiO 2 interface. However, Sb introduction gave rise to various changes in the silicide film. Figure 3 shows Raman scattering spectra for various NiSi films. The spectra were obtained from NiSi Surface. Growth of both Si-Si and NiSi 2 peaks in Sb or B predoped samples indicated that the silicidation reaction was retarded by the existence of impurities. In addition, void formation, shown in Fig. 4, attributable to anomalous Si diffusion toward the surface was observed in predoped NiSi gate. Thus, the silicidation was strongly affected by predoping. XPS Measurements Change due to predoping was also found in the SiO 2 film. As illustrated in Fig. 5, the Sb predoped NiSi/SiO 2 /Si MOS structure could be cleaved into upper and lower parts. Cleaved face was located 2 nm away from the NiSi/SiO 2 interface. The cleaving technique was not applicable to an undoped NiSi MOS structure. RMS values obtained by AFM for upper and lower parts were 0.17 and 0.29 nm, respectively. The NiSi/SiO 2 interface was evaluated by XPS utilizing these specimens, as shown in Figs. 6-8. Sb 3d XPS spectra in Fig. 6 shows that Sb bonded with O is located in both upper and lower parts. In other words, Sb atoms, driven by snowplow effect during silicidation, encroached into SiO 2 . As shown in Fig. 7, Ni atoms also encroached. However, its encroachment depth was much shallower than Sb, because no Ni-O signal was observed for the lower part specimen. Sb-O and Ni-O peaks for 450 ºC silicidation were higher than those for 500 ºC. NiSi workfunction at the SiO 2 interface was determined by measuring the low-energy threshold in XPS measurements [9], as shown in Fig. 9. The obtained workfunction was about 4.6 eV regardless of silicidation temperature. This value is also identical to the undoped NiSi workfunction obtained by C-V measurement shown in Fig. 1. Specimens for workfunction measurement were lightly sputtered by Ar + ion in order to avoid influence of C contamination. By this sputtering, Sb-O signal that exists in spectra in Fig. 6(a) disappeared. However, Si-O bond signals still remained after the sputtering. These results indicate that Sb pileup observed by back-side SIMS [5] is located mainly inside the gate oxide as illustrated in Fig. 8, and the piled-up Sb atoms in SiO 2 are essential to workfunction shift. Summary It has been found that the predoped Sb atoms encroach into the gate oxide during silicidation, and their pileup inside the SiO 2 is considered to be related to NiSi gate workfunction shift. The existence of impurities also causes the retardation of silicidation reaction and anomalous Si diffusion toward the surface during silicidation, resulting in nonuniform silicide phase and the void formation at NiSi/SiO 2 interface. Acknowledgement Part of this work was supported by a Grant-in- Aid for the 21st Century COE program “Nanoelec- tronics for Tera-bit Information Processing” and STARC. 2005 International Semiconductor Device Research Symposium Proceedings WP4-05-1