Review of carbon nanotube on-chip interconnects in RLC prediction using artificial intelligence

Carbon nanotube based on-chip interconnects hold great promise for resolving the scalability problems associated with conventional metal interconnects in integrated circuits. They are a desirable option because of their unique mechanical, electrical, and thermal properties for creating dependable, low-power, high-speed interconnects in upcoming nanoelectronic systems. As a result of their complex structure and behavior, it is still very difficult to estimate the resistance, inductance, and capacitance of interconnects or the RLC parameters collectively. In the RLC parameter prediction for carbon nanotube on-chip interconnects, this review paper provides an extensive overview of recent advances and challenges. The present modeling approaches, such as deep learning, machine learning, and other analytical models, are first highlighted, which have helped in determining the key structural and electrical characteristics of carbon nanotubes that influence RLC prediction. Furthermore, the influences of several parameters on accuracy of RLC prediction, including contact resistance, defects, quantum effects, frequency dependence etc., are investigated. In conclusion, the paper recommends future directions to push the limits of RLC prediction for carbon nanotube on-chip interconnects and overcome current constraints. The analysis is conducted on a 22 nm technology node, and the estimated values indicate that compared to a traditional copper interconnect, the carbon nanotube on-chip interconnects had the lowest delay and power dissipation, measuring 46.902 ns and 0.314 µW by mixed-wall carbon nanotubes. Therefore, by overcoming these challenges, carbon nanotube interconnects have the potential to fundamentally change the design and optimization of next-generation nanoelectronic systems, enabling the development of integrated circuits that are both extremely effective and energy-efficient circuits. The integration of advanced AI approaches, such as deep learning and Bayesian optimization, has significantly advanced the modeling of CNT interconnects. AI-enabled techniques facilitate accurate data-driven prediction of RLC parameters, reduce computational complexity, and enable predictive optimization of interconnect architectures. Unlike conventional analytical or numerical methods that are often limited by scalability and modeling assumptions, AI-based methodologies offer a robust and efficient framework for designing high-performance, reliable, and scalable nano electronic interconnect systems.