Chirality-Driven Analysis of CNFET 5T Spiking Neuron for Ultra-Low Energy Computing

Research into suitable alternatives to silicon has significantly expanded over the last decade due to the diminishing returns of Moore's law. Spiking Neural Networks (SNNs) offer the crucial advantage of low-power computing among Neural Networks (NNs). In this work, we propose an ultra-low energy, high-frequency spiking neuron circuit utilizing Carbon-Nanotube Field Effect Transistors (CNFETs). The key idea behind this work is to design a neuron with reduced energy consumption to enhance the fundamental advantage of SNNs, leveraging carbon instead of silicon, and to analyze performance by varying the chirality vector of CNFETs. Furthermore, we compare the computational outcomes of the circuit for CNFETs and CMOS. The CNFET neuron demonstrates 34.6 times faster performance than CMOS while consuming 5.25 times less energy. However, chirality vector variation is inversely proportional to the transistor's threshold voltage. Therefore, the peak spiking frequency is achieved at 80.32 GHz while consuming only 251.95 aJ per spike, using low-threshold voltage CNFETs with chirality number, <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\boldsymbol{m}=\mathbf{22}$</tex>. Overall, the design and performance evaluation across several CNFET characteristics, compared with CMOS, highlight the potential of carbon as a promising alternative to silicon for SNNs, aligning with their key advantages.