International Journal of Heat and Mass Transfer, Vol.116, 496-506, 2018
Electrothermal studies of GaN-based high electron mobility transistors with improved thermal designs
In recent years, tremendous efforts have been dedicated to GaN high electron mobility transistors (HEMTs) for high-power and high-frequency applications. In general, the performance of these HEMTs is largely restricted by the significant overheating within the device, which would dramatically reduce the charge carrier mobility and thus lower the output current. To solve this problem, different thermal management strategies have been proposed and electrothermal simulations may play an important role here to save tremendous amounts of experimental efforts. However, existing electrothermal simulations are often oversimplified and do not include the details of electron and phonon transport. The large inaccuracies in temperature predictions can be misleading for the thermal improvement of these devices. In this aspect, coupled electron and phonon Monte Carlo (MC) simulations provide the most accurate temperature predictions of the transistor region. To further take into account the heat spreading across the whole sub-millimeter device, the phonon MC simulations can be coupled with conventional Fourier's law analysis for regions away from the transistor. This hybrid electrothermal simulation minimizes the heavy computational load of MC simulations but still incorporates the detailed energy transport processes at different length scales. In this work, this new hybrid simulation technique is used to re-evaluate one widely studied thermal management strategy that coats a high-thermal-conductivity layer on top of a device to spread out the heat. Defined as the maximum temperature rise divided by the total heating power, the device thermal resistance is also computed using the temperature rise of acoustic phonons. Difference is found from calculations based on the Fourier's law. The results provide important guidance for the future development of GaN HEMTs. (C) 2017 Elsevier Ltd. All rights reserved.