Stress analysis of Si1−xGex embedded source/drain junctions

ARTICLE IN PRESS Materials Science in Semiconductor Processing 11 (2008) 285–290 Contents lists available at ScienceDirect Materials Science in Semiconductor Processing journal homepage: www.elsevier.com/locate/mssp Stress analysis of Si1xGex embedded source/drain junctions M. Bargallo Gonzalez a,b,, E. Simoen a, N. Naka c, Y. Okuno d, G. Eneman a,b,e, A. Hikavyy a, P. Verheyen a, R. Loo a, C. Claeys a,b, V. Machkaoutsan f, P. Tomasini g, S.G. Thomas g, J.P. Lu h, R. Wise h a IMEC, Kapeldreef 75, B-3001 Leuven, Belgium EE Department, KU Leuven, B-3001 Leuven, Belgium HORIBA Ltd., Miyanohigashi, Kisshoin, Minami-ku, 601 Kyoto, Japan d Panasonic (IMEC), Kapeldreef 75, B-3001 Leuven, Belgium e Research Foundation-Flanders (FWO), Belgium f ASM Belgium, Kapeldreef 75, B-3001 Leuven, Belgium g ASM America, 3440 East University Dr., Phoenix, AZ, USA h Texas Instruments, 13121 TI Blvd, Dallas, TX 75243, USA b c a r t i c l e in fo abstract Available online 17 November 2008 The purpose of this paper is to evaluate the impact of the geometry of embedded Si1xGex source/drain junctions on the stress field. Stress simulations were performed using TSUPREM4 2D software to further investigate the elastic strain relaxation as a function of Si1xGex alloy active size, in the regime where no plastic relaxation is present. Moreover, the role of the epilayer thickness and the Ge content on the stress levels is also discussed. The work is complemented with experimental Raman spectroscopy. & 2008 Elsevier Ltd. All rights reserved. Keywords: Strain engineering pMOSFETs Embedded source/drain regions Epitaxial deposition Process simulation Raman spectroscopy 1. Introduction The uniaxial compressive stress in the channel induced by Si1xGex in the source/drain (S/D) regions, which has been grown by selective epitaxial growth, is promising for strained channel pMOSFETs device design [1,2]. However, a major concern is the channel stress optimization with the device scaling planned by the ITRS roadmap [3]. This stress leads to a higher low-field mobility, which yields a drive current enhancement [2]. Furthermore, the use of embedded SiGe reduces the S/D resistance [2]. This work is focused on the impact of the Ge content, the epitaxial layer thickness and the active-area dimensions on the stress levels present in Si1xGex alloys, in the regime where no plastic strain relaxation is present [4]. The stress analysis is evaluated for fully strained Si1xGex/Si hetero-  Corresponding author at: IMEC, Kapeldreef 75, B-3001 Leuven, Belgium. Tel.: +32 16 281098; fax: +32 16 281706. E-mail address: [email protected] (M. Bargallo Gonzalez). 1369-8001/$ - see front matter & 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.mssp.2008.09.013 epitaxial device structures by two theoretical approaches and comparison with experiments. First, the misfit strain is calculated based on the standard elasticity theory model. Second, a stress simulation using TSUPREM4 2D [5] is performed in order to investigate the impact of the geometry on the stress field present in the Si1xGex alloys. Finally, a Raman spectroscopy analysis is added to the study. 2. Simulation setup The lattice constant of Si1xGex alloys is larger than the one for bulk Si. The difference in the lattice parameter creates a biaxial in-plane compressive stress in the Si1xGex epilayer and a tensile extension perpendicular to the interface when considering an ideal [1 0 0] surface. The geometrical impact on the sxx stress component in the [11 0] direction in the Si1xGex layer for Si1xGex/Si heterostructures is evaluated using the TSUPREM4 software [5] developed by SYNOPSYS. Two-dimensional (2D) ARTICLE IN PRESS 286 M. Bargallo Gonzalez et al. / Materials Science in Semiconductor Processing 11 (2008) 285–290 Fig. 1. Schematic cross-section of the simulated stress levels for a Si0.6Ge0.4/Si heterostructure with an active length w ¼ 1 mm and an epitaxial thickness t ¼ 120 nm, isolated with a 1 mm long and 300 nm deep shallow trench isolation box. The analysis of the geometrical impact on the simulated stress levels of Si1xGex/Si heterostructures is performed for four different Ge contents considering an active length between 300 nm and 10 mm and an epitaxial thickness between 40 and 160 nm. %Ge Content vs Mismatch Stress 4x109 σ (Pa) = 9.3*107 %Ge 3.5x109 3x109 2.5x109 σ (Pa) stress simulations have been performed based on a 1 mm wide n-doped Si substrate, with P and As n-well implantations, resulting in a donor doping density of 1017 cm3. The Si trenches were considered between 40 and 160 nm etch depth, and were refilled by selective epitaxial growth (SEG) of in-situ highly boron-doped SiGe epitaxial layers with a Ge content of 10%, 20%, 30% and 40%, and B levels of 1 1020 cm3. An active length between 300 nm and 10 mm is considered. The heterostructure is isolated with a 1 mm long and 300 nm deep shallow trench isolation (STI) box (see Fig. 1). Negative stress values refer to compressive stress, and positive stress values indicate tensile. 2x109 1.5x109 1x109 5x108 3. Experimental conditions Stress simulations are complemented with Raman spectroscopy experimental results. The spectroscopic analysis was performed on embedded Si1xGex S/D junctions fabricated on 300 mm n-type Czochralski silicon wafers. Active area regions were defined by STI followed by P and As n-well implantations. Before the epitaxial deposition an HF dip followed by an H2 bake at 850 1C for 2 min was carried out in order to remove native oxide. Subsequently 100 or 150 nm in-situ highly boron-doped Si0.75Ge0.25 epitaxial layers were selectively grown using an ASM Epsilons 3200 reactor, reaching B levels up to 5  1019 cm3 (evaluated by a spreading resistance probe). No implantation, nor anneal, was done after the epitaxial deposition. Another important observation is that no dislocations were detected by Nomarski optical microscopy, showing no evidence of plastic relaxation in the studied samples. The Raman measurements were done in backscattering geometry on a HORIBA Jovin Yvon LabRam HR-800 microRaman system using the 488 nm line of an argon ion laser (0.5 mW) as an excitation source. Laser plasma lines were used for peak shift correction. It should be noted that the penetration depth of 488.0 nm is 150–200 nm in SiGe. Hence, the obtained measurements enabled us to characterize the effective stress in both the SiGe alloy and the Si substrate. 0x100 0 5 10 15 20 25 %Ge content %Ge 10 20 30 40 GPa 0.93 1.86 2.79 3.72 30 35 40 Fig. 2. Predictions of the %Ge content dependence of the mismatch stress for fully strained Si1xGex/Si(1 0 0) heterostructures. 4. Results and discussion 4.1. Standard elasticity theory model For Si1xGex alloys the biaxial in-plane stress or lattice mismatch stress (s) is given by [4,6] s ¼ ½2Gð1 þ uÞ=ð1  uÞ (1) where G is the shear modulus and n the Poisson ratio of the Si1xGex alloy, and e the elastic strain, which can be expressed as a function of the lattice mismatch f, the dislocation density r and the effective component of the Burger’s vector beff as [7] jj ¼ f  rbeff (2) In the regime where no defect injection and plastic strain relaxation are present (r ¼ 0), no impact of the epilayer thickness on the mismatch stress levels is expected for t below the metastable critical thickness. In addition, in this study where the B concentration is less than 1%, it is assumed that boron doping does not introduce a significant variation of the biaxial compressive stress [8]. ARTICLE IN PRESS M. Bargallo Gonzalez et al. / Materials Science in Semiconductor Processing 11 (2008) 285–290 The misfit parameter associated with a Si1xGex epilayer on Si describes the elastic strain and can be calculated from the Ge content x in % and Vegard’s law [4,6,7]. A more accurate approach, also considering the thermal expansion at 300 K, is given as f(x)E0.041x/ 100 [6]. 287 Using standard values for Si1xGex alloys (G ¼ 64 GPa and u ¼ 0.28) [6], high mismatch stress levels are obtained for fully strained Si1xGex/Si(1 0 0) heterostructures (Fig. 2), which according to the model are independent of the film thickness. Fig. 3. Simulated stress profiles for the sxx stress component in the [11 0] direction across the pattern for 100 nm recess depth and (a) 500 nm and (b) 5 mm active size. The finite element simulations are based on Fig. 1. The stress values were taken at a depth of 50 nm. Fig. 4. Simulated stress profiles for the sxx stress component in the [11 0] direction as a function of epitaxial thickness for (a) 300 nm, (b) 500 nm, (c) 1 mm and (d) 5 mm active size. The finite element simulations are based on Fig. 1. The stress values were taken in the middle point of the two-dimensional simulated Si1xGex layer. ARTICLE IN PRESS 288 M. Bargallo Gonzalez et al. / Materials Science in Semiconductor Processing 11 (2008) 285–290 The degree of elastic relaxation as a function of the SiGe window size was calculated as 1sxx/s0xx, where sxx is the simulated stress value in the middle point of the 2D SiGe layer, and s0xx is the mismatch stress for fully strained Si1xGex/Si(1 0 0) heterostructures (Fig. 5). The 4.2. Stress simulations The geometrical impact on the stress levels of Si1xGex/ Si heterostructures is studied as the distribution of the sxx stress component in the [11 0] direction (Fig. 1). The simulated stress profiles for 100 nm recess depth along the 500 nm and 5 mm windows (Fig. 3) show a decrease of the compressive stress level at the edges of the SiGe stressors. This effect leads to a marked reduction of the stress levels for sub 5 mm wide Si1xGex windows. Simulation results for the sxx stress component as a function of the epitaxial thickness are shown in Fig. 4. A clear decrease of the stress levels with epitaxial thickness is observed for window sizes lesser than 5 mm. Furthermore, this trend is more pronounced for increased Ge content. For window sizes larger than 5 mm, the simulated stress becomes independent of the film thickness. For such windows, good quantitative agreement between elastic theory (Fig. 2) and simulated stress values is observed. It should be noted that in order to evaluate the stress field in the Si channel for embedded SiGe S/D technology, additional considerations need to be taken into account, such as the fact that an increase of lateral etching leads to a higher channel stress [3]. Fig. 6. Typical Raman spectrum for 100 and 150 nm in-situ highly borondoped Si0.75Ge0.25 epitaxial layers, reaching B levels 5  1019 cm3. 10% Ge :σ0XX = 0.93 GPa 20% Ge : σ0XX = 1.86 GPa Elastic Relaxation (%) 1 - (σXX/σ0XX) 1 1 40nm 60nm 80nm 100nm 120nm 140nm 160nm 0.8 0.6 0.4 0.6 0.4 0.2 0.2 0 1 Active Size (µm) 10 0 1 Active Size (µm) 30% Ge : σ0XX = 2.79 GPa Elastic Relaxation (%) 1 - (σXX/σ0XX) 10 40% Ge : σ0XX = 3.72 GPa 1 1 40nm 60nm 80nm 100nm 120nm 140nm 160nm 0.8 0.6 0.4 40nm 60nm 80nm 100nm 120nm 140nm 160nm 0.8 0.6 0.4 0.2 0 40nm 60nm 80nm 100nm 120nm 140nm 160nm 0.8 0.2 1 Active Size (µm) 10 0 1 Active Size (µm) 10 Fig. 5. Elastic relaxation as a function of active size for the stress component sxx in the [11 0] direction using TSUPREM4 2D simulations for (a) 10%, (b) 20%, (b) 30% and (d) 40% Ge content. The finite element simulations are based on Fig. 1. The stress values were taken in the middle point of the twodimensional simulated Si1xGex layer. ARTICLE IN PRESS M. Bargallo Gonzalez et al. / Materials Science in Semiconductor Processing 11 (2008) 285–290 obtained elastic relaxation profile shows a clear dependence on the active area size for windows smaller than 5 mm. This result is in excellent agreement with previously 289 published work [9], where HR-XRD analysis showed a window size dependence of the elastic relaxation for active areas smaller than 5 mm2. Fig. 7. Step scan measurements for the four different peaks shift at three different wafer positions, extracted from a 488 nm excitation line, for the studied 100 nm in-situ highly boron-doped Si0.75Ge0.25 epitaxial layers, across the horizontal component of a 4.5  31 mm2 pattern. Si-Si in Si Substrate Si-Si in SiGe 511.5 100 nm 150 nm 520.6 Peak Shift (cm-1) Peak Shift (cm-1) 520.8 520.4 520.2 520 519.8 100 nm 150 nm 511 510.5 510 509.5 509 508.5 508 519.6 3x3 30x30 50x50 70x70 100x100 3x3 Window Size (µm2) 50x50 70x70 100x100 Window Size (µm2) Ge-Ge in SiGe Si-Ge in SiGe 294 412 100 nm 150 nm 411.5 Peak Shift (cm-1) Peak Shift (cm-1) 30x30 411 410.5 410 409.5 409 100 nm 150 nm 293 292 291 290 289 288 408.5 3x3 30x30 50x50 70x70 100x100 Window Size (µm2) 3x3 30x30 50x50 70x70 100x100 Window Size (µm2) Fig. 8. Comparison of mean of peak shift in spot measurements as a function of window size, for the studied 100 and 150 nm in-situ highly boron-doped Si0.75Ge0.25 epitaxial layers. The measurements were done in the center of the window. ARTICLE IN PRESS 290 M. Bargallo Gonzalez et al. / Materials Science in Semiconductor Processing 11 (2008) 285–290 Another important observation is that the compressive biaxial in-plane stress in the Si1xGex epitaxial layer translates into reactive forces on the top of the underlying silicon substrate, leading to a biaxial tensile stress in the upper part of the silicon substrate (Fig. 1). 4.3. Raman spectroscopy analysis A typical Raman spectrum of in-situ boron-doped Si0.75Ge0.25 epitaxial layers is shown in Fig. 6. The three peaks in the proximity of 300, 400 and 500 cm1 wavenumbers correspond to the Ge–Ge, Si–Ge and Si–Si Raman modes, respectively [8,10,11]. It should be noted that the peak position of the Raman spectrum has a relation to stress or strain of Si and SiGe. If there is tensile stress/strain in Si or SiGe, the peak position shifts to lower wavenumber, while if there is compressive stress/strain, it shifts to higher wavenumber. Line scan measurements were performed across a Si0.75Ge0.25:B window, whose dimensions are 4.5  31 mm2 (Fig. 7). The measurements reveal higher elastic relaxation at the edges of the SiGe layer. This observation is in perfect agreement with the simulation results. Furthermore, a lower effective tensile stress at the edges of the pattern is also observed at the top layer of the silicon substrate, from the peak shift related to the stress-free Si–Si bond in the Si substrate (520.5 cm1). However, the effective tensile stress at 1 mm from the edges of the window shows a significant increase (Fig. 7). Therefore, it is expected that the stress level distribution in the Si substrate as a function of the active area size requires a different approach, which is not the purpose of the present paper. Finally, the size dependency of the effective stress levels in different geometries is observed in Fig. 8, where the peak shifts were obtained from spot measurements. The results show lower effective compressive stress in the SiGe alloy for the smallest windows of 3  3 mm2, no big impact of the size is observed for bigger windows. This observation is in good agreement with the simulation work and previously published data [9,12]. However, no strong impact of the epilayer thickness is observed in the studied samples from Raman analysis. Surprisingly, the effective tensile stress levels in the Si substrate show an opposite trend than the compressive stress in the SiGe overlayer, which points to a higher stress level for smaller window sizes. 5. Conclusion In summary, a decrease of the effective compressive stress levels at the edges of the Si1xGex alloys is revealed in this work. This effect leads to a marked impact on the stress levels and elastic relaxation for sub 5 mm wide Si1xGex active areas. Moreover, this dependence is substantially increased with the increase of the epitaxial thickness. No elastic strain relaxation was detected for window sizes bigger than 5 mm width. Finally, it is concluded that TSUPREM4 2D simulations are a useful technique to assess the stress levels with device scaling along the ITRS roadmap. Acknowledgement The European Commission is acknowledged for financial support within the frame of FP6 IP PULLNANO IST026828. References [1] Lee ML, Fitzgerald EA, Bulsara MT, Currie MT, Lochtefeld A. Strained Si, SiGe, and Ge channels for high-mobility metal-oxide-semiconductor field-effect transistors. J Appl Phys 2005;97(1):011101–27. [2] Washington L, Nouri F, Thirupapuliyur S, Eneman G, Verheyen P, Moroz V, et al. pMOSFET with 200% mobility enhancement induced by multiple stressors. IEEE Electron Device Lett 2006;27(6):511–3. [3] Eneman G, Verheyen P, Rooyackers R, Nouri F, Washington L, Schreutelkamp R, et al. Scalability of the Si1xGex source/drain technology for the 45-nm technology node and beyond. IEEE Trans Electron Devices 2006;53(7):1647–56. [4] Gonzalez MB, Simoen E, Vissouvanadin B, Thomas N, Taleb N, Verheyen P, et al. Impact of the Ge content and the epitaxial thickness on the bandgap shrinkage induced leakage current of recessed Si1xGex source/drain junctions. SAFE Proc 2007:496–500. [5] TSUPREM4 User Manual of Version A-2007.12, December 2007, Synopsys, Inc. [6] Hull R. Misfit strain accommodation in SiGe heterostructures. In: Hull R, Bean JC, editors. Germanium Silicon: Physics and Materials. Semiconductor and Semimetals Series, vol. 56. San Diego, CA: Academic Press; 1998. [7] Jain SC, Hayes W. Structure, properties and applications of GexSi1x strained layers and superlattices. Semicond Sci Technol 1991;6: 547–76. [8] Chopra S, Ozturk MC, Misra V, McGuire K, McNeil LE. Analysis of boron strain compensation in silicon-germanium alloys by Raman spectroscopy. Appl Phys Lett 2006;88(20). 202114-1/3. [9] Hikavyy A, Bhouri N, Loo R, Verheyen P, Clemente F, Hopkins J, et al., pMOS transistor with embedded SiGe: elastic and plastic relaxation issues. In: Fifth international conference on silicon epitaxy and heterostructures-ICSI-5 Proc 2007, p. 145–6. [10] Tsang JC, Mooney PM, Dacol F, Chu JO. Measurements of alloy composition and strain in thin Si1xGex layers. J Appl Phys 1994;75(12):8098–108. [11] Chen H, Li YK, Peng CS, Liu HF, Liu YL, Huang Q, et al. Crosshatching on a SiGe film grown on a Si(0 0 1) substrate studied by Raman mapping and atomic force microscopy. Phys Rev B 2002;65(23). 233303-1/4. [12] Usuda K, Irisawa T, Tezuka T, Moriyama Y, Hirashita N, Takagi S. Characterization of strain relaxation after mesa isolation with submm size for global strained substrates using Raman and NBD methods. In: Fifth international conference on silicon epitaxy and heterostructures-ICSI-5 Proc 2007, p. 20–1.