Technical Digest of 1996 IEEE International Electron Devices Meeting Low Resistive Ultra Shallow Junction for Sub 0.1µm MOSFETs Formed by Sb Implantation K. Shibahara, M. Mifuji*, K. Kawabata*, T. Kugimiya**, H. Furumoto, M. Tsuno, S. Yokoyama, M. Nagata*, S. Miyazaki*, and M. Hirose* Research Center for Nanodevices and Systems/*Faculty of Engineering, Hiroshima University, **Electronics Research Laboratories, Kobe Steel Corporation 1-4-2, Kagamiyama, Higashi-Hiroshima, 739 Japan TEL: +81-824-24-6267, FAX: +81-824-22-7185 Abstract Nineteen-nm depth ultra shallow and 1.7 kΩ/sq. low resistive junctions were fabricated by Sb implantation. The shallowness of the junction is attributed to the low diffusive nature of Sb. The junction was applied to 0.15 µm MOSFETs, and excellent suppression of short channel effect and Gm improvement were confirmed. Introduction Shallow junction formation is one of the most important issues in MOSFET scaling. Junction depth (Xj) of source and drain (S/D) extensions should be reduced as gate length becomes shorter to suppress short channel effects. 40 nm gate length nMOSFETs [1] featured 10 nm depth ultra shallow junction formed by solid phase diffusion (SPD) of P from PSG. However, from the view points of CMOS integration, conventional ion implantation is more attractive than SPD because of its adaptable nature for fabrication processes [2]. Especially in the case of nMOSFETs, dopants with low diffusivity and heavy mass which are necessary features for ultra shallow junction formation are applicable. sub 0.1 µm gate length n-MOSFETs were successfully fabricated with low energy As ion implantation [2, 3]. However, fundamental characteristics of As implanted layers, such as Xj and sheet resistance (Rs), were not evaluated sufficiently in these cases. Since Sb is heavier and less diffusive than As, Sb implantation has been focused on recently p-MOSFET channel doping [4, 5] instead of As implantation. It has not been sufficiently investigated in terms of the S/D extensions except for a few reports on 40 nm depth junction and application for MOSFETs [6]. For sub 0.1 µm era, the potentialities of Sb implantation for shallower junctions should be investigated. Well Formation and Isolation 5 nm Cap Oxide Formation Sb+ or As+ Implantation 5~20keV, 1x1014~1x1015cm-2 Cap oxide Removal Profile Evaluation by SIMS Horizontal Furnace Annealing in N2 850°C, 30 min (Typically ~2 nm Oxide is Formed) Xj Evaluation by Selective Wet Etching or SIMS Contact Formation and Metallization Fig. 1 Flow chart of junction formation and evaluation. Under the background described above, ultra shallow junction formation by Sb implantation was investigated with the aim of usage for S/D extensions of sub 0.1 µm MOSFETs and comparison with As implantation. Sheet resistance is discussed in conjunction with Xj in this paper because the shallow extension with high Rs is an obstacle to current drive performance improvement by device scaling. Effectiveness of the Sb S/D extensions is demonstrated by 0.15 µm gate length MOSFET fabrication. Fabrication and Evaluation Fig. 1 shows a fabrication process sequence of junction formation. Sb or As were implanted into Si through oxide 5 nm thick. After the oxide was removed, annealing was carried out at 850 °C for 30 min in nitrogen ambient with a horizontal furnace. During the annealing, oxide 2 nm thick grew typically. Junction depth was evaluated with selective wet etching of n+ portion by mixed solution of HF, HNO3, and CH3COOH. As shown in Fig.2, the etching rate decreases as carrier concentra- 21.7.1 IEDM 96-579 Etched Depth ( nm ) Sb+ 10keV 1x1014cm-2 40 Xj 20 0 0 100 200 Time ( s ) 300 Fig.2 Variation of etched n+-layer depth against etching time. The n+-layer is selectively etched by mixed acid solution. Etching stops at the junction because of carrier reduction. 10 P-SPD(Ref. 1 and 7) 1x101 4 cm - 2 1 S b+ 10 keV A s + 10 keV 0 0 10 1x101 5 cm - 2 20 30 40 Junction Depth, Xj ( nm ) Fig.3 Relationship between junction depth and sheet resistance. The combination of ultra shallow junction and low sheet resistance can be achieved by Sb implantation. 14 12 10 8 6 A s + 10keV 4 S b+ 10keV 2 0 1 01 4 Implantation Dose ( 1 01 5 cm- 2 ) Fig.4 Dif ference of junction depth (∆Xj) between before and after annealing is plotted against implantation dose. Sb shows smaller ∆Xj due to lower diffusivity. ∆Xj for a square plot was obtained from SIMS different from others. 21.7.2 580-IEDM 96 HF : HNO3 : CH3COOH = 1: 3: 8 Sb+ 20keV 1x1014cm-2 Sheet Resistance, Rs ( kΩ/sq. ) Characteristics of Sb and As Implanted Layers Fig. 3 shows relationship between Xj and Rs. Sb leads to shallower junction than As dose as expected. When Sb implantation dose is 1x1014 cm-2 and energy is 10 keV, Xj is 19 nm and Rs is 1.7 kΩ/sq. Although P SPD from PSG can achieve shallower junction [1, 7], it resulted in much higher Rs close to 10 kΩ/sq., as plotted in Fig. 3. Thus Sb is the most suitable to achieve the combination of ultra shallow junction and low sheet resistance when Xj is around 20 nm. The difference of junction depth (∆Xj) between before and after annealing is compared in Fig. 4. Xj for before and after annealing was obtained using process simulator and the selective wet etching method, respectively. The result obtained by this method agrees with a square plot in Fig. 4 which was obtained from SIMS. ∆Xj for Sb is much smaller than that for As. This is a clear evidence for smaller diffusivity of Sb. When implanted dose is 1x1014 cm-2, Rs for Sb is smaller than that for As, as shown in Fig. 3. This result is attributed to the difference of diffusivity and a segregation problem. The segregation is observed in SIMS depth profile (Fig. 5) as a pile up peak. The peak disappears after HF treatment (Fig. 6), which is an evidence that the Sb pileup does not exist in Si. The similar segregation was observed by RBS for both Sb and As [8]. In the report it was concluded that Sb and As piled up not in Si or SiO2 but at the Si-SiO2 interface. Since As has larger diffusivity, much more As atoms can arrive at the interface and segregate during annealing. Since the segregated atoms are electrically inactive, Rs for the As case is higher. The dose of 1x1015 cm-2 results in higher Rs for Sb, on the contrary to the lower dose case. This result is considered to be due to bulk precipitation caused by small solid solubility of Sb. The thermal equilibrium solid solubility at 850 °C was reported to be 3.0x1019 cm- 3 [9]. Activation rate of As decreases to about 50% by clustering [10] when total As concentration is 2x1020 cm- 3 and heat treatment temperature is 850 °C. Though the activation rate for Sb is not evaluated, it seems to be smaller than the As case. 60 ∆ Xj ( nm ) tion reduces, and the etching nearly stops at the junction. The junction depth was defined against the bulk acceptor concentration of 5x1017 cm-3. The junction depth was measured with SIMS as well. Sheet resistance was extracted with the two terminal type test structures with various length implanted layers. Sheet Resistance, R ( kΩ /sq. ) as Implanted Annealed 850°C, 30min Before HF Treatment 10 20 10 19 Sb + 10keV 1x10 14 cm -2 10 18 10 17 0 5 10 15 20 25 30 + Sb + As 8 6 14 S b cm Dose 1x10 -2 5nm Oxide Through 4 850°C 30min 2 0 0 5 10 15 20 25 Energy ( keV ) Fig.7 Dependence of sheet resistance on implantation ener gy. Depth ( nm ) Fig.5 SIMS depth profile of implanted Sb. The surface of an annealed specimen is covered by 2 nm oxide (cf. Fig.1). Surface peak in the annealed specimen is due to segregation of Sb at the interface between the oxide and Si. 1 02 1 Simulated S i O2 S i Concentration ( cm- 3 ) Ion Count ( Arb. Unit ) 10 s Concentration ( cm -3 ) 10 21 100 Annealed 850°C, 30 min After HF Treatment 10 Sb+ 10keV 1x1014cm-2 S b+ 1x101 4 cm- 2 1 02 0 1 01 9 5keV 10keV 25% 75% 1 01 8 1 01 7 -5 1 0 5 10 15 20 25 Depth ( nm ) 0 2 4 6 Fig.8 Simulated as-implanted depth profile of Sb. In the case of 5 keV , a quarter of implanted Sb cannot arrive in Si. Depth ( Arb. Unit ) Fig.6 SIMS depth profile, taken after HF treatment, for the annealed specimen in Fig.5. The segregated Sb disappears as a result of the treatment. Fig. 7 shows the dependence of Rs on implantation energy. Shallower Xj is obtained in general by reducing the energy. However, when the energy is reduced down to 5 keV, Rs rapidly increases. There are two considerable reasons for the increase. One is that the energy is too small for the dopant atoms to penetrate into Si through 5 nm oxide. A quarter of implanted atoms cannot reach to Si as shown in the depth profile of asimplanted Sb calculated with Montecarlo model process simulator [11] (Fig. 8). The other is the segregation. The peak of the implanted atoms is so close to the interface that major part of dopant is considered to be lost as segregation. MOSFETs with Sb Doped S/D Extensions MOSFETs with Sb or As doped extensions were fabricated for the comparison of these two dopants. The principal features of the fabricated MOSFETs are listed in Table 1. Implantation Table1 Features of fabricated MOSFET s. Channel Doping B: 7x1017 cm-3 Uniformly Doped Gate Oxide 3.5 nm Side Wall 150 nm S/D Extenstion Sb+ or As+ 10 keV, 1x1014 cm-2 5 nm Oxide Through S/D As+ 80 keV, 5x1015 cm-2 Annealing 850 °C, 30 min (b) poly-Si AA AAA AAAAA SiN n+ n+ Extensions Fig.9 (a) Cross-sectional FIB microscope photograph and (b) schematic crosssection of a fabricated 0.15 µm gate length MOSFET. 21.7.3 IEDM 96-581 dose and energy were 1x1014 cm-2 and 10 keV, respectively. Under this doping condition, Sb leads to shallower Xj and lower Rs than As dose, as described before. Fig. 9 shows a crosssectional focused ion beam (FIB) microscope photograph of the 0.15 µm gate length MOSFET with the Sb doped extensions. Its ID-VD characteristics are shown in Fig. 10. As shown in Fig. 11, short channel effect is well suppressed by the ultra shallow junction with Sb. Fig.12 shows relationship between effective gate length (Leff) and Gm. Gm is improved compared with the As case by the low resistive extensions with Sb. Sb+ 10 keV 1x1014 cm-2 4.0 ID ( mA ) 1.6 V 2.4 582-IEDM 96 1.2 V 1.6 0.8 0.8 V 0.4 0.8 1.2 1.6 2.0 VD ( V ) Fig.10 I D-VD characteristics of the fabricated MOSFET with Sb doped S/D extensions. 0.8 Vt h ( V ) 0.6 Sb Xj = 19 nm 0.4 As Xj = 26 nm 0.2 0 0.0 Acknowledgments The authors would like to thank Prof. K. Terada, Hiroshima City University for his useful discussion. 0.2 0.4 0.6 0.8 1.0 L g ( µm ) Fig.1 1 Vth dependence on Lg. Short channel effect is well suppressed by the ultra shallow junction fabricated by Sb implantation. 1000 Sb 10keV 1x101 4 cm- 2 Rs 1.7 kΩ/sq. G m ( mS/mm ) References [1] M. Ono et al., Tech. Dig. IEDM 1993, pp. 119-122. [2] K. Takeuchi et al., VLSI. Tech. Symp. Dig. 1995, pp. 9-10. [3] A. Hori et al., Tech. Dig. IEDM 1994, pp. 485-488. [4] G. G. Shahidi et al., VLSI. Tech. Symp. Dig. 1993, pp. 9394. [5] L. Su et al., VLSI Tech. Symp. Dig. 1996, pp. 12-13. [6] Y. Taur et al., Tech. Dig. IEDM 1993, pp. 127-130. [7] H. S. Momose et al., Tech. Dig. IEDM 1994, pp. 593-596. [8] G. A. Sai-Halasz et al., IEEE EDL Vol. EDL-6 (1985), pp. 285-287. [9] F. A. Trumbore, Bell System Tech. J. (1960) pp. 205-233. [10] R. O. Schwenker, J. Appl. Phys. Vol. 42 (1971), pp.31953200. [11] “ATHENA User’s Manual”, SILVACO International, 1995. VG = 2.0 V 3.2 0.0 0.0 Conclusion Shallow junction formation was investigated with Sb ion implantation. Nineteen-nm depth ultra shallow and 1.7 kΩ/sq. low resistive junctions were fabricated by Sb implantation. Junction depth for Sb was shallower than that for As because of low diffusivity of Sb. The junctions were applied to S/D extensions for 0.15 µm MOSFETs, and improvements in short channel effect and Gm were achieved in consequence of shallow and low resistive Sb doped junctions. Thus Sb implantation is concluded to be promising for S/D extensions of sub 0.1 µm MOSFETs. Lg = 0.15 µm 100 10 0.1 As 10keV 1x101 4 cm- 2 Rs 2.2 kΩ/sq. 1.0 L eff ( µm ) Fig.12 Relationship between Lef f and Gm. Sheet resistance of S/D extensions must be low to improve Gm by scaling. 21.7.4