Graphene Nanoribbon-FET for Higher Drive Current using Machine Learning-Enhanced First Principles Analysis

This work presents a novel and innovative design approach for Graphene Nano-Ribbon Field Effect Transistors (GNR FETs), uniquely employing Zigzag Graphene Nano-Ribbons (ZGNRs) as electrodes and Armchair Graphene Nano-Ribbons (AGNRs) as the channel region. To deeply understand device performance, rigorous first-principles modeling was conducted, leveraging Extended-Hückel formalism alongside Landauer-Buttiker transport theory. Extensive Technology Computer-Aided Design (TCAD) simulations systematically explored the impact of critical parameters such as doping concentration (ND), gate voltage (Vg), and drain voltage (Vd) on transistor behavior. However, the computational intensity associated with such comprehensive analyses necessitated the introduction of an advanced Machine Learning (ML)-assisted methodology, specifically employing a Conventional Artificial Neural Network (C-ANN). Remarkably, this ML-driven strategy achieved highly accurate results within significantly reduced computational times of just 80–90 seconds, underscoring its practicality and efficiency. Furthermore, the intrinsic 2.71 eV band gap of the pristine AGNR channel was effectively modulated in a broad range (0.013–1.6 eV) through controlled doping and engineered defects. An N-passivated AGNR FET demonstrated an extraordinary 157 times enhancement in drive current, although its negligible band gap raised concerns regarding leakage currents. Alternatively, the N-doped Stone-Wales AGNR FET provided a well-balanced performance with a 33.21 nA drive current and a suitable 0.58 eV band gap, substantially reducing leakage risks, enhancing thermal stability, and improving peak inverse voltage robustness. This pioneering ML-assisted C-ANN approach highlights significant potential for accelerating accurate and reliable nano-transistor analyses.