Abstract
In recent years, GaN-based HEMTs have great potentials for high power and RF applications. The candidate of substrate for HEMTs was usually Sapphire or SiC material in previous day, however, Si has gradually been the choice of substrate for HEMTs recently. The GaN-based HEMT on silicon substrate, due to low cost and the integration to CMOS in the future work, has become a popular research topic. In this thesis, we focus on the design of AlGaN/GaN HEMTs on silicon substrate and a small signal equivalent circuit modeling for AlGaN/GaN HEMTs. And by the results of the equivalent circuit model analysis, we propose some methods which can extremely improve the RF characteristics of AlGaN/GaN HEMTs. First part of this thesis, for the purpose of improving the RF device characteristics, we scale down the device size and fabricate T-shaped gate structure. Then we investigate how the device layout affects the RF performance of AlGaN/GaN HEMTs, resulting in the design of optimized device layout. Second part of this thesis, we investigate how the short channel effect and the gate length (Lg) affects the RF performance. In this study, we have fabricated 2×(0.2×25) m AlGaN/GaN HEMTs grown on (111) high resistivity Si substrate with f_T more than 35 GHz and f_MAX more than 100 GHz (g_m=225"mS" ⁄"mm" ，I_d=822"mA" ⁄"mm" ). Furthermore, by the analysis of the equivalent circuit model for different device layout, we figure out how the intrinsic parameters like capacitors and resistors affect the RF characteristics of devices. Final part of this thesis, we use the small signal equivalent circuit model to simulate the Si substrate parasitic effect, which largely affects the RF performance of the devices. Due to the GaN/Si structure, there will have high impact ionization and high electron density layer at the Si interface indicating a parasitic leakage path which decreases power at the high frequencies. We figure out that the parasitic parameters such as Rsub and Csub provided by Si substrate largely affect devices’ output power for high frequency operations. By the parameters extracted by this small signal extraction modeling method, we can evaluate what kind of the substrate structure or the material suits for RF devices. In the future, we will use this extraction method as a basis for deign layer structure or device layout, in order to improve the RF performance of the devices.