Trapping Phenomena in GaN HEMTs with Fe- and C-doped Buffer

GaN-based high electron mobility transistors (HEMTs) are a promising technology for high-frequency and high-power applications due to their high breakdown strength, superior electron transport characteristics, and their ability to support a large polarization-induced electron concentration. However, reliability issues in GaN HEMTs, such as trap-induced degradation, have drawn considerable attention in both academia and industry. Studies have been carried out on reducing the effect of traps via the optimization of the epitaxial structure of the HEMT. In this work, we focus on extracting and further analyzing the properties of traps in an AlGaN/GaN HEMT, shown in Fig. 1, with a doped GaN buffer. This device is fabricated and characterized at Mitsubishi Electric Corporation (Japan) and additional details regarding experimental methods will be presented elsewhere. The doping profile achieved during the epitaxial growth is shown in Fig. 2. Fe and C doping in the GaN buffer are typically employed to enhance the confinement of the two-dimensional electron gas (2DEG) in the channel and thus reduce buffer leakage [1]. The doping process also introduces traps in the buffer, which are found to be responsible for current collapse (CC) in GaN HEMTs. We analyze the trap characteristics in fabricated AlGaN/GaN HEMTs as a function of C doping in the buffer, while Fe doping concentration is fixed. The activation energy $(E_{A})$ and cross section $(\sigma)$ of traps in the fabricated devices are extracted from the Arrhenius plot of the drain current transient (DCT) measurements. Similar to previous works, we find that the C doping in GaN layer is mainly responsible for the acceptor-like trapping states with activation around 0.5 eV. We also conduct time-domain simulations of the HEMT using Sentaurus from Synopsys to understand the impact of trap characteristics on the transient response of this device. We conclude that for acceptor-like trapping states with large $E_{A}$ and small $\sigma$, the current takes longer to recover from CC, while the trap concentration affects the degree of collapse.