Abstract
Reverse-bias stress testing has been applied to a large
set of more than 50 AlGaN/GaN high electron mobility transistors,
which were fabricated using the same process but with different
values of the AlN mole fraction and the AlGaN barrier-layer
thickness, as well as different substrates (SiC and sapphire). Two
sets of devices having different defect types and densities, related
to the different growth conditions and the choice of nucleation
layer, were also compared. When subjected to gate–drain (or
gate-to-drain and source short-circuited) reverse-bias testing, all
devices presented the same time-dependent failure mode, consisting
of a significant increase in the gate leakage current. This failure
mechanism occurred abruptly during step-stress experiments
when a certain negative gate voltage, or “critical voltage,” was exceeded
or, during constant voltage tests, at a certain time, defined
as “time to breakdown.” Electroluminescence (EL) microscopy
was systematically used to identify localized damaged areas that
induced an increase of gate reverse current. This current increase
was correlated with the increase of EL intensity, and significant
EL emission during tests occurred only when the critical voltage
was exceeded. Focused-ion-beam milling produced cross-sectional
samples suitable for electron microscopy observation at the sites
of failure points previously identified by EL microscopy. In highdefectivity
devices, V-defects were identified that were associated
with initially high gate leakage current and corresponding to EL
spots already present in untreated devices. Conversely, identification
of defects induced by reverse-bias testing proved to be extremely
difficult, and only nanometer-size cracks or defect chains,
extending vertically from the gate edges through the AlGaN/GaN
heterojunction, were found. No signs of metal/semiconductor interdiffusion
or extended defective areas were visible.
