Modelling of Terahertz IMPATT Diodes Based on GaN Using Artificial Neural Networks

This paper explores the application of artificial neural networks (ANNs) to predict static and dynamic performance metrics of GaN impact avalanche transit time (IMPATT) diodes operating in the terahertz (THz) frequency range. Using datasets derived from self-consistent quantum drift-diffusion (SCQDD) simulations, the ANN models capture key parameters like breakdown voltage (V<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">B</inf>), RF power output (PRF), and DC-to-RF conversion efficiency (η<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">L</inf>), with significantly reduced computational time. The results show a reduction of computational time by 80%, while achieving accuracy within 10<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-4</sup> to 10<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-6</sup> in prediction errors.