- Email: [email protected]
Gas immersion laser doping (GILD) for ultra-shallow junction formation
Thin Solid Films 453 – 454 (2004) 106–109
Gas immersion laser doping (GILD) for ultra-shallow junction formation a ´ G. Kerriena, T. Sarneta,*, D. Debarre , J. Boulmera, M. Hernandezb, C. Lavironc, M.-N. Semeriac a
ˆ 220, Universite´ Paris-Sud, 91405 Orsay Cedex, France IEF, Bat. SOPRA, 26, rue Pierre Joigneaux, 92270 Bois Colombes, France c CEA-LETI, 17, avenue des Martyrs, 38054 Grenoble Cedex 9, France b
Abstract Gas immersion laser doping (GILD) is a very attractive technique to realize the ultra-shallow and highly doped junctions required by the International Technology Roadmap for Semiconductors (ITRS) for future CMOS technologies. In the present work, gaseous dopant precursors (BCl3) are chemisorbed on the Si surface, and partially incorporated during the meltingy recrystallisation of the Si top layer induced by an UV laser pulse (ls308 nm, pulse duration f25 ns). The resulting thickness and dopant concentration of the doped layer depend on the laser energy density and the number of chemisorptionylaser-induced incorporation cycles (up to 200). GILD processed junctions are box-like and exhibit depths ranging from 14 nm to 65 nm, with sheet resistances ranging from f110 to 20 Vyh (respectively), dopant concentrations well above the B solubility limit in Si (up to 3=1021 atycm3) at local thermodynamic equilibrium (LTE) and abruptness of 5–2 nmydecade. Moreover, in situ optical characterization shows the GILD technique capabilities to realize the sub-10 nm thick shallow junctions needed for the sub-40 nm node ITRS predictions. 䊚 2003 Elsevier B.V. All rights reserved. Keywords: Laser doping; Ultra-shallow junction; Excimer laser annealing
1. Introduction The future CMOS generations will require doping techniques capable of forming ultra-shallow, highlydoped junctions with abrupt profiles. Such specifications will be hardly met by present technologies such as rapid thermal annealing (RTP) or spike annealing. Recent experiments have shown the potential capabilities of laser processing of USJ for CMOS technologies. According to the ITRS w1x, two laser processes are able to reach ultimate predictions: laser thermal processing (LTP) and gas laser immersion doping (GILD) w2–7x. LTP and GILD are based on rapid meltingysolidification of the substrate. During solidification, the liquid silicon, which contains the dopants, is formed epitaxially from the underlying crystalline silicon. In the case of LTP, dopants are implanted before laser processing. GILD skips the ion-implantation step: in this case dopants are chemisorbed on the Si surface before the laser shot. The dopants are then incorporated and activated during the meltingysolidification cycle. Activation is limited to the *Corresponding author.
liquid layer and this chemisorptionylaser-shot cycle can be repeated until the desired concentration is reached. 2. Experimental procedure GILD is performed in a high vacuum chamber (10y7 mbar) on Si and SOI wafers, using a homogenized XeCl excimer laser (308 nm, 30 ns, 200 mJ per pulse, 1–25 Hz). After cleaning and removing native oxide the substrate is introduced in the chamber. The dopant precursor gas (BCl3) is injected and chemisorbed on the substrate before each laser pulse w8,9x. Samples have been characterized using four-point probe method, SIMS, IR spectrometry and in-situ optical characterization at 675 nm w10x. 3. Experimental results 3.1. Results on Si bulk Dopant profiles have been measured after GILD laser implantation by SIMS w11x. Fig. 1 presents boron concentration profiles obtained with 50, 100 and 200 laser shots at 670 mJycm2. Box-like and abrupt profiles have
0040-6090/04/$ - see front matter 䊚 2003 Elsevier B.V. All rights reserved. doi:10.1016/j.tsf.2003.11.151
G. Kerrien et al. / Thin Solid Films 453 – 454 (2004) 106–109
Fig. 1. SIMS depth profiles of B doped layers realized by GILD with 50, 100 and 200 laser shots at 670 mJycm2.
107
been obtained, with high dopant concentration and very low sheet resistance. For the same laser energy density, the junction thickness increases with the number of shots due to boron incorporation which changes the thermodynamic parameters and the optical absorption at 308 nm. Thanks to the optical characterization, we have been able to follow the junction evolution during the GILD process. Fig. 2a presents the linear evolution of the junction thickness as a function of the number of laser shots for three different laser energy densities. Boron incorporation is constant during the process, for the same laser energy density. The linear evolution of the boron concentration (atycm2) vs. the number of laser shots at different energy densities is shown in Fig. 2b. We have measured the transmission coefficient of the doped layer with an IR spectrometer. By using the Drude theory for doped layer index measurement we have been able to find the ratio of the activated dose over the effective mass for each sample. We have chosen
Fig. 2. Evolution during GILD process of the junction depth (a) and the boron concentration (number of B atomsycm2 ) (b) as a function of the number of laser shots for different laser energy densities.
108
G. Kerrien et al. / Thin Solid Films 453 – 454 (2004) 106–109
Fig. 3. Boron dose incorporation (number of B atomsyshotsycm2), comparison on silicon bulk and SOI 200y1000 substrates.
an effective mass of 0.37 to find the best fit between the SIMS and FTIR measurements. The evolution of boron concentration incorporated by shots on silicon substrate vs. the laser energy density is shown in Fig. 3. 3.2. Results on SOI Additional GILD experiments have been carried out on SOI (200y1000) substrate. Due to the 1000 nm of silicon oxide the heat is confined to the 200 nm of the c-Si layer, decreasing the melting threshold to 260 mJy cm2. An amorphization threshold also appears above 610 mJycm2 because the liquid phase reaches the oxide layer. Sheet resistances evaluated from FTIR (effective mass of 0.37) are in good agreement with those measured by the four-point probe. We have observed that junction depth increases with the number of laser shots, as in the GILD experiment on c-Si. In Fig. 3 we compare the evolution of boron concentration incorporated by shots as measured by FTIR, vs. the laser energy density on SOI and c-Si substrates. For SOI, the boron dose is much higher than that on silicon bulk and the boron incorporation can reach up to 30% of the total chemisorbed sites on the silicon surface. We explain this fact by the important difference during the heating of the substrate before melting. The boron incorporation efficiency increases with the temperature gradient before melting. As a consequence, the measured sheet resistance on SOI substrate is lower than that on Si substrate, but the evolution of sheet resistance with the laser shots is the same. Fig. 4 shows the evolution of the sheet resistance vs. the laser energy density for different laser shots on SOI 200y1000 nm and Si substrate.
low energy implantation associated with a rapid or spike annealing for activation of dopants. However, channel implantation effects and annealing dopant diffusion are problems inherent to these techniques. Laser processing can solve these technological problems. Thanks to optical characterization of laser processing, we have been able to determine the evolution of junction thickness and boron incorporation as a function of the laser energy density on c-Si bulk. Therefore, we can extrapolate a theoretical limit of GILD process for USJ. In Fig. 5, we present these limitations and characteristics of junctions made by GILD as measured by four-point probes and SIMS. We compare the universal Rs yXj trade-off of implanted and thermally annealed boron with recent conventional techniques w2x, and with laser thermal processing (LTP) studies from our previous work and from Felch et al. w12x. The measurements show that GILD and LTP meet the ITRS 2002 specifications for the 45 nm technology node and laser processing is more efficient than any other usual technique. In case of LTP, transient enhanced diffusion (TED) does not occur but the implanted profile determines the doping profile. Because GILD skips the implantation step, problems coming from this implantation step can be avoided. Junctions can reach depths as thin as 6 nm, and because the activation rate is very high, the sheet resistance can decrease down to 400 Vy h with only 10 laser shots and with an energy density just above the melting threshold. The corresponding abruptness is then 2 nmydecade w8x. 5. Conclusion In this work we present the GILD process which could be an alternative to implantation and rapid annealing for making ultra shallow junctions. By studying GILD on Si and SOI 200y1000 nm substrates, we have seen that the number of boron atoms incorporated by shots depends on the laser energy density and tempera-
4. ITRS comparison results According to the ITRS predictions, no conventional techniques for USJ formation can meet the CMOS sub0.1 mm specifications. Usual technologies are based on
Fig. 4. Evolution of sheet resistance Vs laser energy density for different laser shots on SOI 200y1000 nm and Si substrates.
G. Kerrien et al. / Thin Solid Films 453 – 454 (2004) 106–109
109
Fig. 5. Sheet resistance vs. junction depth according to the ITRS 2002 specifications. Comparison of GILD results with conventional techniques and laser annealing.
ture gradient. Thanks to optical characterizations, we have been able to estimate the theoretical limit curve for the Rs yXj on Si bulk. Finally, we have been able to determine the laser energy density and the number of laser shots to make an ultra thin junction with excellent activation rate, which can meet the final ITRS specifications. Acknowledgments This work has been supported by the French Ministry of Research and Technology in the framework of the RMNT DOLAMI project. References w1x The International Technology Roadmap for Semiconductors, 2002, http:yypublic.itrs.net. w2x D. Lenoble, A. Halimaoui, O. Kermarrec, Y. Campidelli, D. Bensahel, J. Bonnouvrier, O. Menut, E. Robilliart, E. Perrin, F. Arnaud, F. Bœuf, T. Skotnicki, A. Grouillet, G. Bignell, 2nd INTL Workshop on Junction Technology, Tokyo, Japan, 2001, p. 29.
w3x T. Gebel, M. Voelskow, W. Skorupa, G. Mannino, V. Privitera, F. Priolo, E. Napolitani, A. Carnera, Nucl. Instrum. Methods Phys. Res. B 186 (2002) 287. w4x C. Laviron, M.N. Semeria, D. Zahorski, M. Stehle, ´ M. Hernan´ dez, G. Kerrien, D. Debarre, J. Boulmer, 2nd INTL Workshop on Junction Technology, Tokyo, Japan, 2001, p. 91. w5x W. Luo, S. Yang, P. Clancy, M.O. Thompson, J. Appl. Phys. 90 (5) (2001) 2262. w6x S. Earles, M. Law, K. Jones, R. Brindos, S. Talwar, Mater. Res. Soc. 610 (2000) B10.5.1. w7x L. Mariucci, G. Fortunato, S. Whelan, V. Privitera, G. Mannino, ESSDERC, 2002, p. 595. w8x G. Kerrien, J. Boulmer, D. Debarre, ´ D. Bouchier, A. Grouillet, D. Lenoble, Appl. Surf. Sci. 186 (2002) 45. w9x D. Debarre, ´ G. Kerrien, T. Noguchi, J. Boulmer, IEICE Trans. Elect. E85-C (5) (2002) 1098. w10x G. Kerrien, M. Hernandez, C. Laviron, T. Sarnet, D. Debarre, ´ ´ ´ T. Noguchi, D. Zahorski, J. Venturini, M.N. Semeria, J. Boulmer, Appl. Surf, Sci. 208y209 (2003) 277. w11x N. Baboux, J.C. Dupuy, G. Prudon, P. Holliger, F. Laugier, A.M. Papon, J.M. Hartmann, J. Cryst. Growth 245 (2002) 1. w12x S.B. Felch, D.F. Downey, E.A. Arevalo, S. Talwar, C. Gelatos, Y. Wang, 2000 Int. Conf. on Ion Implantation Technology, IEEE, Alpbach, Austria, 2000, p. 167.
Gas immersion laser doping (GILD) for ultra-shallow junction formation a ´ G. Kerriena, T. Sarneta,*, D. Debarre , J. Boulmera, M. Hernandezb, C. Lavironc, M.-N. Semeriac a
ˆ 220, Universite´ Paris-Sud, 91405 Orsay Cedex, France IEF, Bat. SOPRA, 26, rue Pierre Joigneaux, 92270 Bois Colombes, France c CEA-LETI, 17, avenue des Martyrs, 38054 Grenoble Cedex 9, France b
Abstract Gas immersion laser doping (GILD) is a very attractive technique to realize the ultra-shallow and highly doped junctions required by the International Technology Roadmap for Semiconductors (ITRS) for future CMOS technologies. In the present work, gaseous dopant precursors (BCl3) are chemisorbed on the Si surface, and partially incorporated during the meltingy recrystallisation of the Si top layer induced by an UV laser pulse (ls308 nm, pulse duration f25 ns). The resulting thickness and dopant concentration of the doped layer depend on the laser energy density and the number of chemisorptionylaser-induced incorporation cycles (up to 200). GILD processed junctions are box-like and exhibit depths ranging from 14 nm to 65 nm, with sheet resistances ranging from f110 to 20 Vyh (respectively), dopant concentrations well above the B solubility limit in Si (up to 3=1021 atycm3) at local thermodynamic equilibrium (LTE) and abruptness of 5–2 nmydecade. Moreover, in situ optical characterization shows the GILD technique capabilities to realize the sub-10 nm thick shallow junctions needed for the sub-40 nm node ITRS predictions. 䊚 2003 Elsevier B.V. All rights reserved. Keywords: Laser doping; Ultra-shallow junction; Excimer laser annealing
1. Introduction The future CMOS generations will require doping techniques capable of forming ultra-shallow, highlydoped junctions with abrupt profiles. Such specifications will be hardly met by present technologies such as rapid thermal annealing (RTP) or spike annealing. Recent experiments have shown the potential capabilities of laser processing of USJ for CMOS technologies. According to the ITRS w1x, two laser processes are able to reach ultimate predictions: laser thermal processing (LTP) and gas laser immersion doping (GILD) w2–7x. LTP and GILD are based on rapid meltingysolidification of the substrate. During solidification, the liquid silicon, which contains the dopants, is formed epitaxially from the underlying crystalline silicon. In the case of LTP, dopants are implanted before laser processing. GILD skips the ion-implantation step: in this case dopants are chemisorbed on the Si surface before the laser shot. The dopants are then incorporated and activated during the meltingysolidification cycle. Activation is limited to the *Corresponding author.
liquid layer and this chemisorptionylaser-shot cycle can be repeated until the desired concentration is reached. 2. Experimental procedure GILD is performed in a high vacuum chamber (10y7 mbar) on Si and SOI wafers, using a homogenized XeCl excimer laser (308 nm, 30 ns, 200 mJ per pulse, 1–25 Hz). After cleaning and removing native oxide the substrate is introduced in the chamber. The dopant precursor gas (BCl3) is injected and chemisorbed on the substrate before each laser pulse w8,9x. Samples have been characterized using four-point probe method, SIMS, IR spectrometry and in-situ optical characterization at 675 nm w10x. 3. Experimental results 3.1. Results on Si bulk Dopant profiles have been measured after GILD laser implantation by SIMS w11x. Fig. 1 presents boron concentration profiles obtained with 50, 100 and 200 laser shots at 670 mJycm2. Box-like and abrupt profiles have
0040-6090/04/$ - see front matter 䊚 2003 Elsevier B.V. All rights reserved. doi:10.1016/j.tsf.2003.11.151
G. Kerrien et al. / Thin Solid Films 453 – 454 (2004) 106–109
Fig. 1. SIMS depth profiles of B doped layers realized by GILD with 50, 100 and 200 laser shots at 670 mJycm2.
107
been obtained, with high dopant concentration and very low sheet resistance. For the same laser energy density, the junction thickness increases with the number of shots due to boron incorporation which changes the thermodynamic parameters and the optical absorption at 308 nm. Thanks to the optical characterization, we have been able to follow the junction evolution during the GILD process. Fig. 2a presents the linear evolution of the junction thickness as a function of the number of laser shots for three different laser energy densities. Boron incorporation is constant during the process, for the same laser energy density. The linear evolution of the boron concentration (atycm2) vs. the number of laser shots at different energy densities is shown in Fig. 2b. We have measured the transmission coefficient of the doped layer with an IR spectrometer. By using the Drude theory for doped layer index measurement we have been able to find the ratio of the activated dose over the effective mass for each sample. We have chosen
Fig. 2. Evolution during GILD process of the junction depth (a) and the boron concentration (number of B atomsycm2 ) (b) as a function of the number of laser shots for different laser energy densities.
108
G. Kerrien et al. / Thin Solid Films 453 – 454 (2004) 106–109
Fig. 3. Boron dose incorporation (number of B atomsyshotsycm2), comparison on silicon bulk and SOI 200y1000 substrates.
an effective mass of 0.37 to find the best fit between the SIMS and FTIR measurements. The evolution of boron concentration incorporated by shots on silicon substrate vs. the laser energy density is shown in Fig. 3. 3.2. Results on SOI Additional GILD experiments have been carried out on SOI (200y1000) substrate. Due to the 1000 nm of silicon oxide the heat is confined to the 200 nm of the c-Si layer, decreasing the melting threshold to 260 mJy cm2. An amorphization threshold also appears above 610 mJycm2 because the liquid phase reaches the oxide layer. Sheet resistances evaluated from FTIR (effective mass of 0.37) are in good agreement with those measured by the four-point probe. We have observed that junction depth increases with the number of laser shots, as in the GILD experiment on c-Si. In Fig. 3 we compare the evolution of boron concentration incorporated by shots as measured by FTIR, vs. the laser energy density on SOI and c-Si substrates. For SOI, the boron dose is much higher than that on silicon bulk and the boron incorporation can reach up to 30% of the total chemisorbed sites on the silicon surface. We explain this fact by the important difference during the heating of the substrate before melting. The boron incorporation efficiency increases with the temperature gradient before melting. As a consequence, the measured sheet resistance on SOI substrate is lower than that on Si substrate, but the evolution of sheet resistance with the laser shots is the same. Fig. 4 shows the evolution of the sheet resistance vs. the laser energy density for different laser shots on SOI 200y1000 nm and Si substrate.
low energy implantation associated with a rapid or spike annealing for activation of dopants. However, channel implantation effects and annealing dopant diffusion are problems inherent to these techniques. Laser processing can solve these technological problems. Thanks to optical characterization of laser processing, we have been able to determine the evolution of junction thickness and boron incorporation as a function of the laser energy density on c-Si bulk. Therefore, we can extrapolate a theoretical limit of GILD process for USJ. In Fig. 5, we present these limitations and characteristics of junctions made by GILD as measured by four-point probes and SIMS. We compare the universal Rs yXj trade-off of implanted and thermally annealed boron with recent conventional techniques w2x, and with laser thermal processing (LTP) studies from our previous work and from Felch et al. w12x. The measurements show that GILD and LTP meet the ITRS 2002 specifications for the 45 nm technology node and laser processing is more efficient than any other usual technique. In case of LTP, transient enhanced diffusion (TED) does not occur but the implanted profile determines the doping profile. Because GILD skips the implantation step, problems coming from this implantation step can be avoided. Junctions can reach depths as thin as 6 nm, and because the activation rate is very high, the sheet resistance can decrease down to 400 Vy h with only 10 laser shots and with an energy density just above the melting threshold. The corresponding abruptness is then 2 nmydecade w8x. 5. Conclusion In this work we present the GILD process which could be an alternative to implantation and rapid annealing for making ultra shallow junctions. By studying GILD on Si and SOI 200y1000 nm substrates, we have seen that the number of boron atoms incorporated by shots depends on the laser energy density and tempera-
4. ITRS comparison results According to the ITRS predictions, no conventional techniques for USJ formation can meet the CMOS sub0.1 mm specifications. Usual technologies are based on
Fig. 4. Evolution of sheet resistance Vs laser energy density for different laser shots on SOI 200y1000 nm and Si substrates.
G. Kerrien et al. / Thin Solid Films 453 – 454 (2004) 106–109
109
Fig. 5. Sheet resistance vs. junction depth according to the ITRS 2002 specifications. Comparison of GILD results with conventional techniques and laser annealing.
ture gradient. Thanks to optical characterizations, we have been able to estimate the theoretical limit curve for the Rs yXj on Si bulk. Finally, we have been able to determine the laser energy density and the number of laser shots to make an ultra thin junction with excellent activation rate, which can meet the final ITRS specifications. Acknowledgments This work has been supported by the French Ministry of Research and Technology in the framework of the RMNT DOLAMI project. References w1x The International Technology Roadmap for Semiconductors, 2002, http:yypublic.itrs.net. w2x D. Lenoble, A. Halimaoui, O. Kermarrec, Y. Campidelli, D. Bensahel, J. Bonnouvrier, O. Menut, E. Robilliart, E. Perrin, F. Arnaud, F. Bœuf, T. Skotnicki, A. Grouillet, G. Bignell, 2nd INTL Workshop on Junction Technology, Tokyo, Japan, 2001, p. 29.
w3x T. Gebel, M. Voelskow, W. Skorupa, G. Mannino, V. Privitera, F. Priolo, E. Napolitani, A. Carnera, Nucl. Instrum. Methods Phys. Res. B 186 (2002) 287. w4x C. Laviron, M.N. Semeria, D. Zahorski, M. Stehle, ´ M. Hernan´ dez, G. Kerrien, D. Debarre, J. Boulmer, 2nd INTL Workshop on Junction Technology, Tokyo, Japan, 2001, p. 91. w5x W. Luo, S. Yang, P. Clancy, M.O. Thompson, J. Appl. Phys. 90 (5) (2001) 2262. w6x S. Earles, M. Law, K. Jones, R. Brindos, S. Talwar, Mater. Res. Soc. 610 (2000) B10.5.1. w7x L. Mariucci, G. Fortunato, S. Whelan, V. Privitera, G. Mannino, ESSDERC, 2002, p. 595. w8x G. Kerrien, J. Boulmer, D. Debarre, ´ D. Bouchier, A. Grouillet, D. Lenoble, Appl. Surf. Sci. 186 (2002) 45. w9x D. Debarre, ´ G. Kerrien, T. Noguchi, J. Boulmer, IEICE Trans. Elect. E85-C (5) (2002) 1098. w10x G. Kerrien, M. Hernandez, C. Laviron, T. Sarnet, D. Debarre, ´ ´ ´ T. Noguchi, D. Zahorski, J. Venturini, M.N. Semeria, J. Boulmer, Appl. Surf, Sci. 208y209 (2003) 277. w11x N. Baboux, J.C. Dupuy, G. Prudon, P. Holliger, F. Laugier, A.M. Papon, J.M. Hartmann, J. Cryst. Growth 245 (2002) 1. w12x S.B. Felch, D.F. Downey, E.A. Arevalo, S. Talwar, C. Gelatos, Y. Wang, 2000 Int. Conf. on Ion Implantation Technology, IEEE, Alpbach, Austria, 2000, p. 167.