2.6.1 AlGaN/GaN HBTs

In the end of the 1990s III-Nitride technology advanced rapidly. The interest was mainly driven by the progress in lasers and LEDs but also in HEMTs. The latter soon reached record microwave performance and thus prompted the development of heterojunction bipolar transistors (HBTs). They offer traditional advantages over field effect transistors, such as linearity, threshold uniformity, and current handling. One of the first working samples was demonstrated in [78], but was plagued by poor contacts and the quality of the material. Subsequently a current gain ($\beta$) of only 3 was reached with a very high V $_\ensuremath{\mathrm{CE}}$ offset of 5 V. Other groups [79] faced similar problems. However, while the technology was still lacking, various theoretical studies pointed out the large potential of the devices. Using a two-dimensional model and based on material properties reported in literature, [80] predicted a maximum $\beta$ of 1130, collector saturation current of 3.5 kA/cm$^2$, breakdown voltage of 55 V, and $f_\ensuremath {\mathrm {t}}$ of 18 GHz. Even higher theoretical values (reaching $\beta$=2000 and $f_\ensuremath {\mathrm {t}}$=30 GHz) were reported by employing compact models [81]. Other studies focused on a comparison between npn- and pnp-structures [82] and the worse high-frequency characteristics of the latter. Using models and material parameters verified by modeling experimental device characteristics, an optimization was performed in [83] and a theoretical $f_\ensuremath {\mathrm {t}}$ of 44 GHz was predicted. Based on improved fabrication processes developed previously for other compound systems, HBTs with a similar performance ( $\beta \approx$3 at room temperature) were fabricated on MBE and MOCVD grown material [84]. The device performance was again limited by the base resistance. Better results were achieved in [85] by using MOCVD grown material, with $\beta$=80, but high forward resistance at the base-emitter junction due to possible diffusion of Magnesium into the emitter. The same group also introduced selective area growth resulting into high crystalline quality, but a still leaky base-emitter diode [86]. The same technique was used by McCarthy et al. [87], while they also employed emitter mesa regrowth to avoid etch damage, and material grown using lateral epitaxial overgrowth technique to achieve low dislocation density. Through optimization of the width and grading of the base, operation at 70 V with a $\beta$=6 was possible.

In the following years work on AlGaN/GaN based HBTs continued [88,89,90,91,92,93,94] (Fig. 2.9). A possible issue with the measured values of the extrinsic current gain, was however pointed out [92]. Due to the low quality ohmic contacts and the leaky base-collector junction, the anomalous current gain at low current levels can be erroneously attributed to the intrinsic device performance. This was also observed by Hsueh et al. [95], who proposed that a common-emitter I-V under high current bias is the best way to evaluate the transistor performance, instead of the Gummel plot. Some of the experimental studies focused on high-temperature performance [90,96]. As the hole concentration increases by thermal activation, the current gain of pnp structures is enhanced [90]. Device operation at temperatures up to 400$^\circ$ C was demonstrated, although the device performance shows a degradation after prolonged operation at this temperature [96]. Despite the progress in the last years the technology faces still several major problems:

• low base conductivity,
• deep acceptor levels of Mg,
• and emitter-collector leakage currents.

Figure 2.9: Current gain (measured and simulated) of GaN HBTs over time.