The epitaxial layers were grown by MOCVD including a ~3900-nm buffer layer, a 325-nm GaN channel layer, a 1-nm AlN spacer layer, a 21.5-nm Al0.25Ga0.75N barrier, and a 4.1-nm in-situ SiN cap layer. First, device isolation was implemented by ICP etching down to the buffer layer. The ohmic contact was made by the deposition of Ti/Al/Ni/Au (25/125/45/55 nm) without annealing. Using ICP to perform gate recess, and then annealed by RTA at 875 °C in an N2 circumstance for 40 s. The gate dielectric SiN was deposited at 300 °C by PECVD, then annealed by RTA at 450 °C for 1 min in an N2 circumstance. Ni/Ti/Al/Ti/Au (30/25/250/25/200 nm) metal stack was used to form gate metal contact. Finally, the devices were passivated by a 200-nm SiN deposited by PECVD.
The C-V and I-V characteristics of fabricated devices were measured by the Agilent B1500A and B1505A. The leakage current was reduced in the devices with proposed annealing. Double sweep C-V characteristics at 100-kHz of devices with proposed annealing step show a smaller hysteresis and lower flat band voltage shift than devices with traditional ohmic annealing. Moreover, the frequency dependent C-V measurements show that device with proposed annealing step effectively reduces the dispersion phenomenon. Conductivity method was applied to extract interface trap density as a commonly used analytical method for the III-V materials [3]. The calculated Dit of the device with proposed annealing step shows lower interface trap density than the device with traditional ohmic annealing. The calculated Dit was reduced from 7.1E12 to 4.8E12 eV-1cm-2. As a result, annealing after gate recess can really improve the interface quality and reduce the damage from dry etching, and form good ohmic contact without the extra annealing step.
