1 / 2020-10-03 11:11:50

Characterization of the SEE Performance of eGaN HEMTs at Various LET using an Alternative Test Approach

SEE,GaN,HEMT,LET

GaN HEMTs have gathered a lot of momentum for high-reliability (Hi-Rel) applications because of their excellent radiation hardness and exceptional circuit performance. Until recently, single event effects (SEE) testing on GaN has been carried out predominantly with heavy ions bombarding from the bump side. This is owing to the fact that the typical Si substrate thickness on the backside exceeds the nominal range of common heavy ion beams. Efficient Power Conversion Corp. (EPC) developed an alternative SEE test method, allowing incident ions to directly access the GaN layer with minimal energy attenuation. This is enabled by completely removing the Si substrate using EPC’s proprietary etching technology. We have demonstrated that this substrate removal process does not have any adverse effects on the electrical characteristics of the device. In fact, devices with no substrate still retained superior long-term electrical performance under extreme hard switching conditions.

The majority of GaN SEE testing has followed traditional Si MOSFET Hi-Rel testing, where 85 LET (Si) is commonly used to determine the SEE performance in GaN. This is invalid for the following reasons: 1) GaN intrinsically has lower LET than Si at a given ion beam energy, primarily due to its higher density; 2) the Si substrate is internally shorted to the source node of the device, and therefore has no electric field and should not be sensitive to ionization effects in SEE; 3) the two-dimensional electron gas (2DEG) in GaN provides the carriers for conduction in eGaN HEMTs, so the volume within the GaN in the vicinity of the 2DEG is expected to be the most sensitive to ion radiation.

SRIM simulations demonstrate that the backside approach can achieve identical LET in the GaN layer as compared to the conventional bump side method. In addition, this new approach can achieve the equivalent LET with significantly lower ion beam energy. To validate the simulation results, we performed bump side and backside SEE testing on EPC’s 300V developmental eGaN HEMTs using ~3GeV Au ion beam. Both approaches yielded similar results. Simulations also show that using this new testing approach, we can easily modulate the LET in the GaN layer by adjusting the air gap distance between the ion beam source and device backside surface. We conducted SEE testing in Texas A&M University at various LET in GaN, where incident ions directly access the GaN layer. EPC’s eGaN HEMTs showed that the SEE response tracked with the simulated LET.