By using indium aluminum nitride (InAlN) rather than the more usual aluminum gallium nitride (AlGaN) as the barrier layer, one can lattice match (In0.17Al0.83N) with the underlying gallium nitride (GaN), removing strain effects. It is found that spontaneous polarization in the nitride semiconductor materials is sufficient to create the two-dimensional electron gas (2DEG) channel at the barrier/substrate interface that is necessary for HEMT operation. In other words, one does not need strain-induced (piezoelectric) polarization from a mismatched barrier to produce GaN HEMTs.
While strain can produce desirable effects in some semiconductors, such as increased mobility in certain directions, it can also lead to unstable structures that can fail under electrical and thermal stress. Among the achievements for InAlN/GaN HEMTs have been drain current densities exceeding 3A/mm and HEMT operation at 1000°C without permanent damage.
Figure 1: Comparisons of normalized drain current (a), intrinsic channel resistance (b), and threshold voltage (c) after negative gate bias (NGB), off, and semi-on stresses at the same drain-gate potential VDG. VGS=–3V during semi-on stresses. Corresponding bias conditions are VDG =–VG=VDS – VGS.
The researchers from the Technische Universität Wien (TU Vienna), Institute of Electrical Engineering Slovak Academy of Sciences and École Polytechnique Fédérale de Lausanne (EPFL) aimed to fill the gap in the analysis of possible degradation mechanisms in InAlN/GaN HEMTs at various device working points and electrical stressing conditions in a similar way to other groups studying AlGaN/GaN devices. The team also hoped to confirm its expectation that the absence of strain in InAlN can reduce some device degradation processes.
The epitaxial layers (10nm InAlN/1nm AlN/1μm GaN/150nm AlN) for the tested devices were constructed using metal-organic chemical vapor deposition (MOCVD). Source/drain contacts consisted of titanium, aluminum, nickel and gold, while the Schottky barrier for the gate consisted of nickel and gold. No passivation procedure was used.
Stressing experiments were carried out under negative gate bias (NGB), off and semi-on conditions. NGB stresses are found to damage AlGaN/GaN devices because inverse piezoelectric effects (strain produced from electric fields) generate defects. The InAlN/GaN devices were also subjected to testing under different temperature conditions up to 250°C. Five or six devices were subjected to each test and it was found that there were no qualitative differences in behavior.
NGB tests revealed less variability in parameters when compared with published results for AlGaN/GaN HEMTs. Although gate leakage is initially a little higher for InAlN/GaN with gate-source voltage (VGS) less than –26V, it increases by less than 80% in going to –50V (catastrophic breakdown), in contrast to the four-order-of-magnitude increase for AlGaN/GaN. Further, when degradation does occur, it is reversible, unlike the traditional AlGaN/GaN HEMT set-up. While some parameters need several hours to recover their initial values, the drain and gate leakage currents need only about 100 minutes.
Irreversible damage was seen in off and semi-on tests when the drain-gate voltage (VDG) exceeds 38V. In the off-state the intrinsic channel resistance (Rch) and drain current are most affected. The resistance increased by one order of magnitude and the drain current decreased by 70% after off-state bias tests to a VDG of 50V. The researchers believe that these degradation effects are related to hot-carrier injection into the GaN buffer layer, creating defects and ionizing existing states.
Improvements may be sought using double-heterostructure channels, field plates or recessed gates, with the aim of reducing the hot-carrier injection. Surface passivation is another possible route to more reliable InAlN/GaN HEMTs.
Fuente:
Gerald Soto, CRF 2010-1.
No hay comentarios:
Publicar un comentario