A computationally efficient hybrid approach based on artificial neural networks and the wavelet transform for quantum simulations of graphene nanoribbon FETs

• Published in: Journal of computational electronics 2019-09, Vol.18 (3), p.813-825
• Main Authors: Tamersit, Khalil, Djeffal, Fayçal
• Format: Article
• Language: English
• Subjects: Accuracy; Artificial neural networks; Charge density; Computational efficiency; Electrical Engineering; Engineering; Field effect transistors; Flexibility; Flight simulators; Graphene; Green's functions; Mathematical and Computational Engineering; Mathematical and Computational Physics; Mathematical models; Mechanical Engineering; Methods; Nanoelectronics; Nanoribbons; Nanotechnology devices; Neural networks; Numerical models; Onsite; Optical and Electronic Materials; Poisson equation; Semiconductor devices; Simulation; Theoretical; Transistors; Wavelet transforms
• ISSN: 1569-8025; 1572-8137
• DOI: 10.1007/s10825-019-01350-2

Numerical modeling of graphene nanoribbon field-effect transistors (GNRFETs) using quantum-mechanical approaches is often associated with a heavy computational burden, indicating the urgent need for new methods to resolve this issue. A computationally efficient hybrid approach for quantum simulations of ballistic GNRFETs is developed herein. The proposed simulation method is based on solving the two-dimensional (2D) Poisson equation using both wavelet-based adaptive meshing and wavelet-based matrix compression techniques, self-consistently, with well-trained neural network models that imitate the mode-space (MS) nonequilibrium Green’s function (NEGF) solver in generating the charge density. The results obtained by applying the hybrid approach show good agreement with NEGF simulations. Numerical experiments reveal that the developed simulation method can offer a speed-up of about one order of magnitude over the conventional MS NEGF approach. The encouraging results obtained in this work indicate that the developed numerical model is particularly appropriate for incorporation into nanoelectronic device simulators to investigate future GNRFET-based circuits.