br Acknowledgments This work was funded by the EU ERDF

Acknowledgments This work was funded by the EU ERDF and the Spanish Ministry of Economy and Competitiveness (MINECO) under Project TIN2013-41576-R, and the Spanish Ministry of Education, Culture and Sport (MECD) under a FPU grant (AP2010-3259), and the Andalusian Regional Government (Spain) under Project P10-TIC-6114. This work was also supported by the CEiA3 and CEIMAR consortiums.
Introduction Few studies address phenomena generated by heavy ions in the COTS power GaN transistors. Bazzoli et al. [1] show that COTS GaN transistors (RT240PD, 70 V) are not sensitive to Single Event Burnout (SEB) under neutron and heavy Cynomolgus Erythropoietin / EPO Protein (His Tag) irradiation tests; however, a phenomenon similar to Single Event Gate Rupture (SEGR) was observed in these devices although no oxide was under the gate. They supposed that the defects created in AlGaN layers by incident particles could be the origin of the gate insulation. Although, there is no insulator between the gate and the AlGaN layer in this type of component, but the radiation can create defects in the gate interface (pGaN)/AlGaN which can play an insulating role. They confirm that GaN transistors are less sensitive to Single Event Burnout than MOSFETs. Recently, different generations of this GaN technology (EPC) have been tested under heavy ion irradiation. Scheick assumes that the critical region seems to be near the edge of the gate on the drain side [2]. The lot-to-lot variance that has been taken into account appears to be a very significant parameter [2]. Other results consider that these devices are not as robust as expected and show that normally-off power GaN HEMTs are affected by a significant charge amplification mechanism [3,4].
Test vehicles Firstly, reverse engineering was carried out in order to define the technological parameters of these devices. Fig. 1 gives an example of cross-section view of the GaN Systems GS66508P HEMT that was tested. The structure consists of a substrate (silicon), a nucleation region comprising different AlGaN layers with aluminum rates gradually decreasing, an unintentional doped (UID) GaN buffer layer, an AlGaN barrier layer, a p-doped GaN layer and a SiN passivation layer.
Test circuit The complete circuit of the test board is provided in Fig. 2 and Fig. 3 provides a typical view of the test board. The proposed methodology consists in testing the device under heavy ion beam and monitoring the current level in the component with a discharge RC network (C24, R24 connected to DrainMon net) connected to the drain of the component. On HV2 net, R2L resistor ensures that the current is limited in the drain path to limit the extension of ion induced failures. The gate of the component is maintained at a regulated voltage of 0.00 V which ensures that the component is maintained in a blocking state (off-state). This setup is based upon similar nondestructive techniques as those described in [5,6] and [7].
High penetration beam The penetration of each ion has been compared to the front side device stack of material. To ensure that their range (see Table 1) allows generating charges in the active surface of the GaN/AlGaN structure, each ion track was analyzed with SRIM Software. Three ions were especially considered for this campaign: 582 MeV Nickel (described as Ni onward), 769 MeV Krypton (described as Kr onward) and 995 MeV Xenon (described as Xe onward). Their Energy loss profiles are given in Fig. 4 for the GaN material. A slight difference can be observed compared to LETs in Silicon. Nickel ion's initial LET is 16 MeV/mg/cm2, Krypton is 26 MeV/mg/cm2 and Xenon is 50 MeV/mg/cm2. All the three ions have been compared to atmospheric environment secondary ions [16]. The nickel ions which have the lowest LET of the three ions remain more severe than the majority of secondary ions that could be produced in atmospheric environment by neutrons and protons. Consequently, the sensitivity under nickel ion provides a good figure of merit for atmospheric applications.