PERFORMANCE OPTIMIZATION OF 14nm FINFETS WITH DIFFERENT FIN SHAPES by Srishti
Abstract
With the development in semiconductor industry, the size of FETs has scaled down from deep sub-micron to nanometer domain. But literature shows that beyond 22 nm, it is almost impossible to scale the dimensions of MOSFET with acceptable SCEs. As a solution, field effect transistors such as FinFETs have emerged as novel devices having superior controls over short channel effects than the conventional MOS transistor devices. It is also observed that fin shape has considerable impact on SCEs. In the view of above issues, this thesis explores FinFET devices at 14 nm technology. In this work, the FinFETs with different fin shapes have been designed using TCAD. Literature shows that the performance of FinFET is influenced by parameters like fin width, height, angle etc. So, to examine the clear effect of top fin width on performance of FinFET; the fin height & other parameters are kept constant. And optimum top fin width with higher drive current & lesser possible leakage current is achieved. Trapezoidal fin FinFETs with different dimensions of top fin width were designed and simulated. The impact of fin shape has been evaluated in terms of V-I characteristics, performance parameters like Ion, Ioff, Ion/Ioff and SCEs. The performance improvement was observed in terms of Ioff, Ion/Ioff, Subthreshold Swing, Drain Induced Barrier Lowering and power dissipation with the reduction in top fin width. But reduction in top fin width slightly degraded the performance in terms of Ion. To verify these simulation results, the FinFETs with similar trapezoidal fins were simulated at 20nm technology and compared with the available literature. This comparison depicted the improvement in the present work. Then using the artificial neural network training and particle swarm optimization algorithm, optimum top fin width was obtained. Further, the FinFET with achieved optimum top fin width was simulated on TCAD to check the accuracy of the results. The optimization results after ANN & PSO and TCAD simulation results of optimum top fin width are quite close with an error of 0.48%.
