Spacer Optimization using a Neuro-Pso Approach for Improving FinFET Repeater Performance in On-Chip Global MLGNR Interconnects

Spacers are often used in fin-shaped field effect transistors (FinFETs) to increase the ON current at the cost of higher parasitic capacitances. In this paper, a neural network assisted particle swarm optimization technique called neuro-PSO is developed to optimize the geometrical and material parameters of spacers in order to enhance the performance of FinFET based repeaters in on-chip global MLGNR interconnects. The proposed neuro-PSO approach is able to achieve a much smaller power delay product of MLGNR interconnects while requiring a smaller repeater area than what is possible when no spacers are used.