Modeling, optimization and performance benchmarking of multilayer graphene nanoribbons for on-chip interconnect applications

Abstract

Using current techniques, GNR layers suffer from edge roughness during fabrication. In addition, MLGNRs turn into graphite as number of GNR layers increase that further degrades the performance of MLGNRs. This thesis aims toward higher performance building in MLGNRs as compared to traditional and industry preferred interconnect material i.e. Cu. Closed-form analytical models for transient analysis of MLGNRs are proposed in this work. It is seen that accurate extraction of circuit parameters of MLGNRs is extremely important. In that context, the analytical model for computation of equivalent capacitance in MLGNRs interconnects is proposed by considering interlayer coupling. In our analysis, it is observed that Fermi level, mean free path in GNR layers and inter-layer resistivity between GNR layers significantly affect the performance of MLGNRs. This thesis proposes two approaches to improve the performance of MLGNRs. First is the intercalation of metal and compounds in MLGNRs that leads to increase in Fermi level in GNR layers and lowering in the inter-layer resistivity between GNR layers. Here, AsF5 and Li have been used as intercalants. The thickness of intercalated MLGNR stack is optimized to obtain best performance for interconnect applications. Based on our analysis, we proposed optimized Li-intercalated MLGNR structures as potential candidates to substitute Cu at local, intermediate and global level interconnect applications. The optimized top-contact Li-intercalated MLGNRs with edge roughness exhibit significantly lower energy-delay product and higher bandwidth density as compared to Cu at all interconnect applications. Our second approach is the insertion of high-k dielectric between GNR layers that leads to higher mean free path in each GNR layers. This technique prevents the MLGNRs from converting into graphite. An analytical model is proposed to compute scattering rate, mean free path and mobility where dielectric is inserted between GNR layers. It is seen that the transport properties of GNRs are strongly depend on the surrounding dielectric environment and the quality of dielectric sample. Using cleaner samples, mean free path higher than 1 𝜇𝑚 can be obtained.