Impact of Anode Doping on Avalanche in Vertical GaN Pin Diodes with Hybrid Edge Termination Design

Abstract

Introduction: Large area GaN power devices are seldom reported to avalanche and the theoretical studies of their edge termination still struggle to match experimental results especially for planar structures. We report for the first time a punch through avalanche on 1.2 kV vertical GaN diode with optimized hybrid edge termination design to fit the device structure. Furthermore, we provide a systematic study that pairs the Sentaurus TCAD simulation to the reverse characteristics of three different anode designs with three different remaining dose in the anode extension region. Device Fabrication: PiN diodes with three different anode doping were fabricated on a non-homogenous GaN substrate. 5×1017, 1×1018, and 2×1019 cm-3 Mg-doped layer were grown on a MOCVD lightly doped layer of 8um with doping concentration of the order of 1-2×1016 cm-3. The Edge termination process starts with etching a trench 1µm depth and about 140µm from anode edge, followed by isolation nitrogen implant using a box profile with three different energies and total depth of ~650 nm. Finally, the junction termination extension and guard rings (JTE/GR) hybrid is implemented by nitrogen implant to achieve a full planar device structure with final depth of 300 nm. Figure 1 depicts the device structure. The same JTE/GR hybrid implant depth into three different anode doping which results in different residual charge in the extension region thickness that is expected to impact the field management. Experimental Result: Three devices with the different anode doping were tested under identical forward and reverse biases conditions. The three devices display similar forward behavior with ideality factor ranging 2<n<2.2 as shown in Figure 2. Their reverse characteristics however, are varying proportional to the anode doping. The highest breakdown voltage and the sharpest breakdown characteristics was displayed by anode doping of 1×1018 cm-3. The other two anode doping devices fail short to achieve the expected breakdown voltage and both devices also exhibit higher leakage than the 1×1018 cm-3.. All devices are designed to withhold 1.2 kV, thus 2×1019 and 5×1017 cm-3 anode are ~ 500 V away from the targeted value. Considering the fact that these three wafers have the same drift region thickness and doping, and were processed together through the edge termination therefore, the drastic difference in the breakdown voltage can be only explained by the remaining dose in the extension region. To further understand the edge termination efficacy in these devices an avalanche test was conducted, reverse sweep was applied at 25,100,150, &200 °C for all three devices. Figure 3 depicts the avalanche results. Both devices with 1×1019 and 5×1017 are struggling to avalanche due to the high leakage and inconsistent trend with temperature. The PiN diode with 1×1018 cm-3 avalanched with an increase of breakdown voltage 10 V for every 50 °C. This trend of the breakdown with temperature is possible due to the increase in leakage current, consequently a temperature coefficient α= 3.1×10-4 K-1 was calculated using equation: . To better understand this difference a simulation was conducted using Sentaurus TCAD. The model is designed to observe the behavioral trends thus the absolute numbers in the model do not capture the devices non-idealities. The 2D distribution of the electric field is displayed in Figure 4 for all 3 doing levels. Conclusion: The reverse characteristic and TCAD simulation indicate an improvement in the field management with the decrease in the anode doping to 1×1018 cm-3 which subsequently means less charges in the edge termination region than the standard 2×1019 doping level . These results show that the anode dose is a function of anode doping level, implant depth and the edge termination region thickness. To optimize edge termination design one needs to manage the dose in the termination region which is a function of anode doping level, implant depth and the edge termination region thickness. Figure 1

Detecting defects that reduce breakdown voltage using machine learning and optical profilometry

“…2a. Some of the diodes that did not break down before 1000 V were measured at higher voltages in a vacuum system, and avalanching breakdown behavior was verified to match the theoretically expected breakdown voltage 25,26 .…”

“…1 for diagram): The p-layer was isolated by N implantation. A guard ring structure was fabricated using the design optimized in our previous research 11,25,26 . An outer trench isolation layer was produced using Cl 2 etching.…”

Using machine learning with optical profilometry for GaN wafer screening

“…The exact distribution devices and sizes can be determined from the x–y data in the supplemental materials (“Training Data.csv”). All versions of the diodes had a trench etched outside the devices using an Ar/Cl 2 plasma for isolation, a ~ 600 nm multi-energy nitrogen implant with a box profile for further isolation within the trench, and an implanted guard ring/JTE hybrid termination 26 approximately 300 nm deep also created with nitrogen implantation. Ohmic contacts were deposited using Pd/Pt/Au on the p-layer and Ti/Al/Ni/Au on the substrate.…”