An Integrated Methodology for Accurate Extraction of S/D Series Resistance Components in Nanoscale MOSFETs Seong-Dong Kim, Shreesh Narasimha * , and Ken Rim * IBM Semiconductor Research and Development Center, Essex Junction, VT 05452, * Hopewell Junction, NY 12533, USA Phone: 1-802-769-2578 E-mail: sdkim@us.ibm.com Abstract A new integrated methodology for the accurate extraction of source/drain (S/D) series resistance components with emphasis on the spreading and contact resistance elements is presented. For the first time, detailed extractions of lateral extension doping abruptness and silicide specific contact resistance are made directly from 90nm-node SOI MOSFET characterization. The spreading resistance due to the lateral doping gradient is found to be a key component contributing to total parasitics, and the doping gradient engineering and scaling of specific contact resistance must be employed to overcome this parasitic limitation in future nanoscale CMOS performance roadmap. Introduction As CMOS technologies are scaled deeper into the nano-scale regime, the S/D extrinsic resistance (R ext ) will strongly limit MOSFET drive current. No FET-scaling roadmap can be completed without a detailed physics-based understanding of R ext and an accurate quantification of its multiple sub- components. However, due to the complexity in extracting many of these components, there has been no comprehensive treatment of R ext for deeply-scaled, state-of-the-art MOSFETs. This shortcoming is particularly evident for two of the primary R ext sub-components: 1) the extension (EXT) spreading resistance created by the lateral doping gradient and 2) the silicide-to-silicon diffusion contact resistance. In this paper, through a combination of detailed FET electrical characterization and physical modeling, we provide a comprehensive analysis of R ext and prescribe a new methodology for accurately quantifying the key R ext sub- components. Physical Resistance Model A physics-based equivalent-circuit R ext model is employed in this work [1]. The total R ext is divided into three key regions, or sub-components as illustrated in Fig. 1: 1) the resistance in the extension-to-gate overlap (R ov ), 2) the spreading resistance under the sidewall spacers (R spr ), and 3) the contact resistance at the silicide-to-silicon interface (R co ). The physical model is developed to accommodate different spacer geometries and dopant contours and takes important features of S/D parasitic structure into account, which include accurate calculation of accumulation channel potential, Gaussian function-based lateral and vertical doping gradient, analytical current spreading angle, and 1-dimensional (1-D) transmission line modeling (TLM) of distributed silicide contact resistance. The resultant key expressions of sub- resistance components are as follows: ∫ + - - = ov sp spr L L L s fb g ox ac ov ac x x V V C x dx R 1 )) ( ) ( ( ) ( , ψ μ (1) ∫ + + - + = 1 1 ) )( (tan ) ( , 1 , sp spr sp L L L c ov spr sp extx ov spr dx t x L L x R α ρ (2) ( ) dp sp ext s L c ov spr sp extx spr ext L L R dx t x L L x R sp - + + - + = ∫ 2 , 0 , 1 , 1 ) )( (tan ) ( α ρ (3) jdp dp dpx spr dp t L R ρ = , (4)         = t con t c co u L L L R coth , ρ (5) sili sw c co sw t R , , ρ = (6) where C ox is the oxide capacitance, μ ac is the accumulation channel mobility, V g , V fb and ψ s are the gate voltage, flatband voltage and the surface potential in overlap region, respectively, α is the current spreading angle (analytically α=1 radian), ρ extx and ρ dpx , are the resistivity of extension under spacer1 and the lateral diffusion of deep junction region, respectively, t c , t jdp and t sili are the thickness of channel, deep junction and silicide consumption, L sp1 , L sp2 , L dp , L con , and L spr,ov are the length of spacer1, spacer2, lateral deep junction diffusion, the effective silicide contact, and the spreading of overlap part, respectively, ρ c and ρ c,sw are the specific silicide contact resistivity in planer and sidewall contact, respectively, and L t is contact transfer length. Methodology The overall R ext calibration and component extraction strategy is illustrated in Fig. 2. First, the MOSFET R ext is determined by using the R on -L eff measurement technique of Fig. 3 described in [2,3]. Here, R on is defined as (V ds,lin /I dlin ) where I dlin is measured at a fixed overdrive (V gs -V tlin ) of 0.7V. Next, all of the relevant active doping concentrations and contours are determined from a combination of SIMS analysis, 1-D sheet resistance measurements, and TSUPREM4 process simulation. Fig. 4 shows the SIMS and calibrated active doping level for both S/D and EXT of N/P FET test structures. While this is a reasonable approach to calibrate 1-D active doping profiles, it doesn’t provide an accurate and sufficient method for determining the lateral doping gradient of S/D extension. Reasons for this shortcoming include the fact that in advanced CMOS technologies, both offset spacers (Sp1) and low gate-to- diffusion overlap are utilized to improve short-channel control. These design features have the potential to increase the lateral extension gradient, and to force more of the ‘spreading’ to take place in the low-doped extension tip 0-7803-9269-8/05/$20.00 (c) 2005 IEEE