Analysis of high power gallium nitride based high electron mobility transistors for next generation electronics applications

• Main Author: Vompolu, Ganesh Sampath
• Other Authors: Ng Geok Ing
• Format: Theses and Dissertations
• Language: English
• Published: 2019
• Subjects: Engineering::Electrical and electronic engineering::Microelectronics
• Online Access: http://hdl.handle.net/10356/79010
• Institution: Nanyang Technological University

AlGaN/GaN high electron mobility transistors (HEMTs) are being widely investigated and increasingly considered as a promising device for the next generation of power electronics and microwave applications because of their key characteristics like high breakdown voltage, high power density and high temperature operability. It is also widely recognized that, the semiconductor devices which can function at ambient temperatures beyond 150°C without any external cooling mechanisms can very well help a lot of future electronics applications in the fields of aerospace, automotive, energy production, etc. Issues like self-heating at high temperature and bias levels are known to result in reduced mobility and increased leakages in HEMTs. These effects also impact their reliability. Hence, to improve and optimize the performance of HEMTs, it is important to study and understand their thermal behaviour through accurate, compact and reliable models. In this work, the commercially available modelling software AtlasTM from Silvaco was used to develop a 2D physical simulation of AlGaN/GaN HEMT on silicon substrate to understand the thermal behaviour of HEMTs. Emphasis has been to understand the impact of temperature and bias conditions on the performance of AlGaN/GaN HEMTs on silicon substrate. The DC device characteristics and lattice temperature profiles of the AlGaN/GaN/Si HEMT have been simulated and analysed to understand the role of self-heating effects on their electrical characteristics. Simulation results of lattice temperature profiles have confirmed the existence of a hotspot at the gate edge of gate-drain region. The junction temperature was determined from the lattice temperature profiles. The results exhibited the impact of self-heating effects on the drain current at large bias conditions. The need and demand for having accurate, compact and reliable models for design and development of next generation electronic applications motivates us to explore other advanced nonlinear device modelling techniques to model HEMTs. Lately, Deep Learning tools such as artificial neural networks (ANNs) have been gaining a lot of attention in various fields because of their remarkable information crunching capabilities. By leveraging these advantages, behaviour models of AlGaN/GaN/Si HEMTs have been developed in this work using ANNs to model drain current (Id), gate leakage current (Ig), transconductance (gm) and channel temperature for a wide range of operating temperature and bias conditions. The detailed modelling and optimization approach has been discussed. The models were evaluated by validating their ability to interpolate the outputs beyond training dataset. Good accuracy with few hidden layer neurons have been obtained for all the four models. The key advantage and limitations of using ANN based modelling approach instead of using convolutional modelling techniques for next generation electronic applications were also discussed.