4A-2 Strained Si Channel with Embedded Formed by Solid Phase Epitaxy Yaocheng Liu*, Oleg Gluschenkov, Jinghong Li, Anita Madan, Ahmet Ozcan, Byeong Kim, Tom Dyer, Ashima Chakravarti, Kevin Chan#, Christian Lavoie#, Irene Popova, Teresa Pinto, Nivo Rovedo, Zhijiong Luo, Rainer Loesing, William Henson, and Ken Rim IBM Semiconductor Research and Development Center (SRDC), Hopewell Junction, NY 12533, USA #IBM T. J. Watson Research Center, Yorktown Heights, NY 10598, USA *e-mail: [email protected] phone: +1-845-894-3564 Abstract control wafers went through exactly the same process as described Current drive enhancement is demonstrated in sub-40 nm NFETs above except the C implantation. Splits with combinations of e-Si:C with strained silicon carbon (Si:C) source and drain using a novel and neutral liner (NL) or tensile stress liner (TL) were designed to solid-phase epitaxy (SPE) technique for the first time. The very clearly highlight the impact of e-Si:C stressor as well as to simple process uses no recess etch or epi deposition steps, adds investigate the interaction between e-Si:C and stress-inducing liner. minimal process cost, and can be easily integrated into a standard CMOS process. With a record high 1.65 at% substitutional C Results and Discussion concentration in source and drain, 615 MPa uniaxial tensile stress Fig. 3 (a) shows the cross-sectional STEM image of an NFET was introduced in the channel, leading to a 35% improvement in with e-Si:C. Convergent beam electron diffraction (CBED) and 6% and 15% current drive increase in sub-40 technique was used to measure the stress in the channel of this and 200 nm channel length devices respectively. (Keywords: silicon device. Fig. 3 (b) shows the measured uniaxial tensile stress as a carbon, MOSFET, strained Si, mobility and solid phase epitaxy) function of depth at the center of the channel. The stress increases as the location gets closer to the top surface of the channel, with the Introduction maximum stress of 615 MPa, roughly equivalent to the channel Strained Si channel has become an essential component of modern stress induced by 100 nm-thick nitride film with 2 GPa tensile stress. high-performance CMOS technology. Materials with same crystal The Vt roll-off (Fig. 4) and Ioff-Vtsat (Fig. 5) plots indicate that the structure but different lattice constants are good candidates for strain Vt and short channel effects of the e-Si:C devices and the control engineering. SiGe has been successfully incorporated in the source devices are matched very well. Fig. 6 shows the Ron vs. LGATE plots and drain of PFETs to strain the channel compressively and increase of the e-Si:C NFET and the control devices. The electron mobility the hole mobility [1, 2]. With the lattice constant that is smaller than was extracted using �n=1/[W·Qinv·(dRon/dLGATE)] [7] and the results Si, silicon carbon alloy (Si:C) embedded in the source and drain can showed a 35% �n enhancement for the e-Si:C NFETs compared to induce tensile strain to the channel and improve the electron mobility the control devices. However, series resistance of the e-Si:C NFET is for NFETs. A relatively small substitutional C concentration clearly higher than the control device, and offsets the performance ([C]sub=1~2 at%) in Si is sufficient to generate a high strain level. gain obtained from the strain effect. The series resistance increase is A straightforward approach to form embedded Si:C (e-Si:C) is to due to the lower activation in Si:C compared to that in Si, as recess source and drain with an etch and then deposit Si:C using a shown by the Rs-[C]tot relationship in Fig. 7. selective epi process [3]. However, due to the extremely low solid In spite of the higher series resistance, Ion-Ioff comparison in Fig. 8 solubility of C in Si, it is difficult to grow Si:C with [C]sub >1 at% indicates a 6% Ion improvement due to the enhanced mobility for e- selectively, and in many cases such approach can result in more than Si:C NFETs at Ioff=100 nA/�m. When the e-Si:C stressor is half of the total carbon content ([C]tot) at interstitial sites [4]. integrated with tensile stress liner, the performance enhancements Furthermore, thermal processes after the Si:C epi can easily from both stress elements are retained and additive to each other. precipitate C atoms out from the substitutional sites, causing the loss Longer channel devices are less influenced by series resistance. Fig. of stress and affecting the junction and transport properties. 9 shows the typical I-V characteristics of an e-Si:C NFET and a We report a novel technique that combines carbon implant and control device at LGATE=200 nm with matched Vt. There is a 15% solid-phase epitaxy (SPE) to form high quality e-Si:C in the source drive current improvement with e-Si:C NFETs. and drain of NFETs. Compared to the conventional recess etch- Fig. 10 shows the reverse diode leakage current of the e-Si:C selective epi approach, the SPE method has significant advantages: NMOS devices compared to the typical leakage obtained without 1) high substitutional C concentration can be achieved, 2) dopants in incorporating e-Si:C. There is some increase in the junction leakage source and drain are activated during the SPE and there is no need but the increase is within an acceptable range that can be optimized for high-temperature process after the Si:C growth, and 3) the simple with junction design. process for NFETs can readily be integrated with embedded SiGe for With the implantation-SPE process for e-Si:C, a lithography step PFETs without adding additional masking and epi deposition steps. can be easily added to block C implant in PFETs. For certain PFETs, we intentionally integrated e-Si:C in the source and drain to monitor Process and Device Fabrication the effect of tensile strain as the hole mobility is expected to As shown in Fig. 1, the process for e-Si:C device fabrication with decrease with the uniaxial tensile strain. The R -L plots (Fig. SPE is very simple compared to the recess etch-selective epi process. on GATE 11) shows a 60% hole mobility degradation, confirming the Epitaxial Si:C film was formed by implanting C into an amorphized successful introduction of tensile strain by e-Si:C. Si layer and regrowing it with SPE. HRXRD measurement (Fig. 2) and calculation with Berti and Kelires’s methodology [5, 6] showed Conclusion [C]sub=1.65 at%, the highest substitutional C concentration for Si:C Embedded Si:C in source and drain was formed for the first time that is incorporated in the source and drain of NFETs to date. The by very simple implantation and SPE regrowth process to enhance well defined thickness fringes in XRD rocking curve indicate high NFET performance. Tensile stress of 615 MPa was generated in the crystal quality and smooth surface. SIMS analysis confirms that channel with a record high 1.65 at.% substitutional C concentration [C]tot=[C]sub within the measurement accuracy, i.e. C is completely in the source and drain regions, resulting in 35% electron mobility substitutional in the SPE Si:C film. improvement and 6% and 15% drive current improvement for sub-40 The device fabrication steps followed most part of an industry- and 200 nm channel length NFETs, respectively. standard 65 nm bulk CMOS process. C was implanted together with the source/drain doping. A protective layer was used to block carbon References implant into the poly gate. Si:C SPE and doping activation were [1] T. Ghani et al., IEDM Tech. Dig., pp. 978, 2003. [2] W. –H. Lee et al., completed at one anneal. The wafers then went through standard IEDM Tech. Dig., pp. 61, 2005. [3] K. W. Ang et al., IEDM Tech. Dig., pp. NiSi formation and BEOL. No high temperature process was needed 1069, 2004. [4] A. C. Mocuta et al., J. Appl. Phys., pp. 1240, 1999. [5] M. Berti et al., Appl. Phys. Lett. 85, pp. 1602, 1998. [6] P. C. Kelires, Phys. Rev. after Si:C SPE, so there were no C precipitation issues. The baseline B 55, pp. 8785, 1997. [7] K. Rim et al., IEDM Tech. Dig., pp. 43, 2002.

44 978-4-900784-03-1 2007 Symposium on VLSI Technology Digest of Technical Papers Doping implant, (a) Recess etch Selective Si:C epi activation anneal

Si:C Si:C Si:C Si:C STI Si STI STI Si STI STI Si STI STI Si STI

(b) PAI, doping I/I, C I/I, SPE anneal Fig. 1 Process flow for NFET with e-Si:C using (a) the conventional recess etch and selective epi approach, (b) the novel Si:C Si:C implantation and solid-phase epitaxy technique. STI Si STI STI Si STI

1E+6 �xx (MPa)

) Si peak 1E+5 800 PS

C 1E+4 600 ( Si:C peak with

y 0 x

it 1E+3 400 [C]sub=1.65 at% ns

te 1E+2 e-Si:C e-Si:C 200 n I 1E+1 0 0 100 200 300 400 1E+0 z z (nm) -1000 -500 0 500 1000 1500 2000 2500 3000 �-2� scan (sec) (a) (b) Fig. 2 HRXRD rocking curve of the SPE Si:C film Fig. 3 (a) Cross-sectional STEM image of an NFET with e-Si:C before � ([C]sub=1.65 at.%) grown on (100) Si substrate. The silicidation. (b) The uniaxial tensile stress ( xx) measured by CBED as a well defined Si:C peak and fringes indicate that the function of depth (z) at the center of the channel (x=0). The stress at z=25 nm Si:C film is high-quality single crystal. is 615MPa. 240 0.5 Slope change caused by Vtlin 1E-5 220 35% mobility enhancement 0.4 m) Control NFET � 200 (V) m) 1E-6

t e-Si:C NFET 0.3 hm· �

V 180 (O

0.2 (A/ 1E-7 f on 160

V of tsat R I Control NFET 0.1 Control NFET 1E-8 140 e-S:C NFET e-Si:C NFET 0 1E-9 120 0.02 0.03 0.04 0.05 0.06 0 0.1 0.2 0.3 0.4 0 0.01 0.02 0.03 0.04 0.05 0.06 LGATE (�m) Vtsat (V) � Fig. 4 NFET V as a function of gate LGATE ( m) t Fig. 5 NFET I as a function of length. The e-Si:C devices match with off Fig. 6 NFET resistance as a function of gate length saturation threshold voltage. the control devices very well. showing a 35% electron mobility improvement but higher series resistance with e-Si:C. 250 1E-5

q) 200

s Vdd=1V / 13% 6% Fig. 8 Ion-Ioff performance m 150 comparison for NFETs with 1E-6 6% (Oh 100 m)

s different stress elements (NL: 40% increase in Rs � / R 50 with 1.65 at% C neutral liner, TL: tensile liner). 0 (A 1E-7 Drive current is improved by off 0 1 2 3 4 I 6% due to the e-Si:C tensile NL only Total C concentration (at%) stressor. When e-Si:C is 1E-8 NL + e-Si:C Fig. 7 Sheet resistance of As doped integrated with TL, both the SPE Si:C as a function of total C 13% TL only TL + e-Si:C improvement by e-Si:C (6%) content. The As activation is degraded 1E-9 and the improvement by TL as more C is implanted. 7.0E-4 8.0E-4 9.0E-4 1.0E-3 1.1E-3 1.2E-3 (13%) are retained and they

Ion (A/�m) are additive to each other. 6E-4 1400 Control NFET VG=1.2V Control PFET .U.) m) 5E-4 e-S:C NFET 100 1200 m) A � ( / e-Si:C PFET �

V =1.0V · 4E-4 G e (A 10 1000 t hm en

r 3E-4 r VG=0.8V 1 (O 800 on

n cu n 2E-4 R 600 ai V =0.6V 0.1 r G Leakag tion 60% steeper c D 1E-4 400 Jun 0E+0 Control e-Si:C 0 0.02 0.04 0.06 0 0.2 0.4 0.6 0.8 1 1.2 NFET NFET LGATE (�m) Drain Voltage (V) Fig. 10 NFET junction leakage Fig. 11 Ron vs LGATE plots for PFETs. Si:C Fig. 9 Typical I-V characteristics for NFETs comparison between e-Si:C and was incorporated in the source and drain of with LGATE=0.2 �m. The drive current (at control devices. There is a slight certain PFET devices. The 60% slope VD=VG=1V) of e-Si:C NFET is 15% higher leakage increase due to e-Si:C but increase due to �p reduction as expected than the control device with matched Vt. it is within the acceptable range. confirms the tensile strain induced by e-Si:C.

2007 Symposium on VLSI Technology Digest of Technical Papers 45