Fast unveiling of Tmax in GaN HEMT devices via the electrical measurement-assisted two-heat source model
Date
2022-04-07Source Title
IEEE Transactions on Electron Devices
Print ISSN
0018-9383
Electronic ISSN
1557-9646
Publisher
IEEE
Volume
29
Issue
5
Pages
2319 - 2324
Language
English
Type
ArticleItem Usage Stats
13
views
views
5
downloads
downloads
Abstract
Gallium nitride (GaN) high-electron-mobility transistor (HEMT) devices, which have wide application potential from power amplifiers to satellite, need to be thoroughly examined in terms of reliability in order to benefit the superior intrinsic properties of the device. The most critical parameter in the device reliability is the hotspot, or Tmax , which occurs somewhere on the subsurface and along the channel of the GaN HEMT, which is optically inaccessible due to optical path disability. Therefore, the Tmax value is underestimated in optical measurements, such as the thermographic IR and Raman methods. With 3-D electrothermal simulations, Tmax is obtained close to reality, but it requires a huge computation load and the complex modeling of semiconductor device physics. In 2-D or 3-D thermal simulations that do not use electrothermal simulations, since the self-heating is mostly modeled with a single heat source, neither the correct Tmax value is obtained nor the effect of bias conditions is considered. To address the aforementioned shortcomings, a hybrid method is demonstrated, which exploits the electrical measurements of GaN HEMT, which RF and reliability engineers often and easily do. It is demonstrated that Tmax can be determined quickly and close to the electrothermal simulations in a GaN HEMT device with a two-heat source method and finite element analysis (FEA) hybrid interaction with respect to various bias conditions. Moreover, the impact of the knee voltage is investigated with different knee-detection techniques. The proposed method provides GaN HEMT reliability engineers with an easy-to-implement alternative to reveal the hotspot location and the value.