Equivalent Thermal Conductance Network (ETCN) Model for Domain Decomposition and Efficient Thermal Simulation of GaN HEMTs

Abstract

Accurate and efficient simulation is essential for the thermal management of gallium nitride (GaN) high-electron-mobility transistors (HEMTs). However, conventional numerical approaches are usually time-consuming when dealing with transient thermal simulations with temperature-dependent parameters. To conquer this problem, we present a novel method based on the equivalent thermal conductance network (ETCN) model for efficient thermal simulation of GaN HEMTs. First, according to the temperature-dependent characteristics of the materials of the device, the entire structure is divided into region of variation (ROV) and region of fixity (ROF). Then, we decompose the transient response of ROF into a zero-input (ZI) and a zero-state (ZS) response based on the intrinsic property of linear time-invariant systems. After that, a novel ETCN model is developed for efficient transient simulation of GaN HEMTs. The principle of the ETCN model is to transform the impacts of the ROF on the ROV in the form of the equivalent thermal boundary conditions. By this means, we only need to focus on the ROV in the nonlinear iteration, enabling a significant reduction of degrees of freedom in solving the nonlinear equation and thus significantly improving the computational efficiency. In addition, the ETCN model is also available to handle the steady-state thermal problems. Several numerical examples are provided to validate the accuracy and efficiency of the proposed method. Compared with the conventional finite volume method, a speed-up of 105x is achieved by the ETCN model in simulating a typical multifinger GaN HEMT with microchannel cooling.