Optimal segmentation of Cu–CNT interconnects

Hybrid copper–single-walled carbon nanotube (Cu–SWCNT) interconnects are gaining attention recently due to their inherent advantages of better conductivity than Cu and ease of fabrication. This work focuses on optimizing propagation delay and power delay product by optimizing the number of segments using particle swarm optimization (PSO) and machine learning (ML) techniques, for the first time. First, a PSO algorithm is deployed to find the optimal number of segments to minimize the delay and power delay product (PDP). Then, a random forest ML (RFML) algorithm is implemented to find the optimal number of segments for interconnect lengths 1000–6000 µm where the computational time is 9.34 ms which is found to be less compared with PSO algorithm which took 796 ms to run. The data obtained from PSO algorithm are used to train the RFML algorithm and its time consumption is validated. It has also been observed that the number of segments is less for Cu–CNT interconnects compared to Cu interconnects.