I. Introduction
In recent years, GaN-based high-electron mobility transistors (HEMTs) on silicon substrates have received much attention in power electronics due to their excellent properties, such as low-channel resistance, high breakdown voltage , and high switching frequency. For conventional AlGaN/GaN HEMT structures, a large amount of 2-D electron gas (2DEG) exists because of the strong built-in polarization electric field in the Ga-face -plane epistructure [1]–[3]. The 2DEG in the channel cannot be easily depleted by the Schottky gate contact at zero gate bias. Thus, the inherent normally on behavior excludes GaN-based HEMTs from most of the power electronic applications. A more complex circuit design is required when using depletion-mode transistors. There have been many technologies proposed to raise the conduction band energy level underneath the gate contact so that a positive threshold voltage is obtained to ensure enhancement-mode (E-mode) operation. Examples include using the gate recess structure [4], [5], the fluorine treatment [6], [7], the p-type cap layer [8]–[13], the piezoneutralization layer [14], the nonpolar -plane channel [15], and the metal–oxide–semiconductor field-effect transistor structure [16]. Among the aforementioned methods, the literatures on HEMTs using a p-type cap layer have reported the threshold voltage ranging from 1 to 3 V with the applied gate voltage larger than 5 V [8]–[13]. The results show that the junction field-effect transistor-like (meaning with p-GaN and 2DEG channel structure) approach is promising for E-mode operation. Despite excellent performance reported, most of the works focused on demonstrating E-mode properties without comprehensive investigation on the correlation of the p-GaN layer structure with the electrical properties of an E-mode device, which is critical to ensure successful commercialization in the future.