Power-delay-product optimal repeater design for horizontal and vertical multilayer graphene nanoribbon interconnects

Power-delay-product optimal design of repeater size and number design for both horizontal and vertical multilayer graphene nanoribbon (MLGNR) interconnects is implemented. Horizontal MLGNR (HMLGNR) interconnects, both top-contact and side-contact, and vertical MLGNR (VMLGNR) interconnects are considered. Four of the more common optimization algorithms, including ant colony optimization for continuous domains (ACOR), particle swarm optimization, artificial bee colony, and gray wolf optimization, are compared in terms of speed and accuracy. Accordingly, the best algorithm is chosen for optimizing both the size and number of repeaters in MLGNR interconnects. Edge backscattering probability, surface roughness amplitude, and doping concentration (parameters with a more significant effect on the performance of MLGNR interconnects) are considered as inputs to optimization algorithms. The ACOR optimization results are then utilized for training a back-propagation neural network that performs the optimal repeater design faster than optimization algorithms. Results show that VMLGNR or HMLGNR interconnects may require fewer repeaters depending on the edge backscattering probability, surface roughness amplitude, and doping concentration. It is also shown that regardless of the edge backscattering probability, surface roughness amplitude, and doping concentration, HMLGNR interconnects with top contacts always need smaller repeater size.