Process-induced Structural Variability-aware Performance Optimization for Advanced Nanoscale Technologies

Abstract

As the CMOS technologies reach the nanometer regime through aggressive scaling, integrated circuits (ICs) encounter scaling impediments such as short channel effects (SCE) caused by reduced ability of gate control on the channel and line-edge roughness (LER) caused by limits of the photolithography technologies, leading to serious device parameter fluctuations and makes the circuit analysis difficult. In order to overcome scaling issues, multi-gate structures are introduced from the planar MOSFET to increase the gate controllability. The goal of this dissertation is to analyze structural variations induced by manufacturing process in advanced nanoscale devices and to optimize its impacts in terms of the circuit performances. If the structural variability occurs, aside from the endeavor to reduce the variability, the impact must be taken into account at the design level. Current compact model does not have device structural variation model and cannot capture the impact on the performance/power of the circuit. In this research, the impacts of structural variation in advanced nanoscale technology on the circuit level parameters are evaluated and utilized to find the optimal device shape and structure through technology computer-aided-design (TCAD) simulations. The detail description of this dissertation is as follows: Structural variation for nanoscale CMOS devices is investigated to extend the analysis approach to multi-gate devices. Simple and accurate modeling that analyzes non-rectilinear gate (NRG) CMOS transistors with a simplified trapezoidal approximation method is proposed. The electrical characteristics of the NRG gate, caused by LER, are approximated by a trapezoidal shape. The approximation is acquired by the length of the longest slice, the length of the smallest slice, and the weighting factor, instead of taking the summation of all the slices into account. The accuracy can even be improved by adopting the width-location-dependent factor (Weff). The positive effect of diffusion rounding at the transistor source side of CMOS is then discussed. The proposed simple layout method provides boosting the driving strength of logic gates and also saving the leakage power with a minimal area overhead. The method provides up to 13% speed up and also saves up to 10% leakage current in an inverter simulation by exploiting the diffusion rounding phenomena in the transistors. The performance impacts of the trapezoidal fin shape of a double-gate FinFET are then discussed. The impacts are analyzed with TCAD simulations and optimal trapezoidal angle range is proposed. Several performance metrics are evaluated to investigate the impact of the trapezoidal fin shape on the circuit operation. The simulations show that the driving capability improves, and the gate capacitance increases as the bottom fin width of the trapezoidal fin increases. The fan-out 4 (FO4) inverter and ring-oscillator (RO) delay results indicate that careful optimization of the trapezoidal angle can increase the speed of the circuit because the ratios of the current and capacitance have different impacts depending on the trapezoidal angle. Last but not least, the electrical characteristics of a double-gate-all-around (DGAA) transistor with an asymmetric channel width using device simulations are also investigated in this work. The DGAA FET, a kind of nanotube field-effect transistor (NTFET), can solve the problem of loss of gate controllability of the channel and provide improved short-channel behavior. Simulation results reveal that, according to the carrier types, the location of the asymmetry has a different effect on the electrical properties of the devices. Thus, this work proposes the n/p DGAA FET structure with an asymmetric channel width to form the optimal inverter. Various electrical metrics are analyzed to investigate the benefits of the optimal inverter structure over the conventional GAA inverter structure. In the optimum structure, 27% propagation delay and 15% leakage power improvement can be achieved. Analysis and optimization for device-level variability are critical in integrated circuit designs of advanced technology nodes. Thus, the proposed methods in this dissertation will be helpful for understanding the relationship between device variability and circuit performance. The research for advanced nanoscale technologies through intensive TCAD simulations, such as FinFET and GAA, suggests the optimal device shape and structure. The results provide a possible solution to design high performance and low power circuits with minimal design overhead.