(Invited) Recent Advances and Challenges of MOCVD-Grown AlScN/GaN HEMTs

Abstract

The allocated frequency bands of next-generation wireless communication networks will be shifting further up in the spectrum to be able to meet the stringent data rate requirements. Doing so, demands higher bandwidth capabilities from the electronic devices in the RF chain. Numerous cutting-edge systems are using AlGaN/GaN-HEMT based components, in particular in the transmit chain, due to their favorable combination of output power and bandwidth capability. Still, this semiconductor technology is approaching an upper bound in this regard, known as the Bode-Fano limit. Devices based on an AlScN/GaN heterointerface have been identified as candidates to push these boundaries and enable components that deliver higher power over a broader frequency range. We report on the status of MOCVD-grown AlScN/GaN HEMTs, their current capabilities but also on challenges that still hamper the exploitation of the material system's full potential.The first reported AlScN/GaN HEMTs were fabricated on epitaxial stacks grown by MBE. However, the majority of commercial GaN-based HEMT technologies is relying on MOCVD-grown epitaxy due to higher growth rate and lower cost of operation. We managed to grow AlScN/GaN heterostructures showing sheet carrier densities (nS) of up to 3·1013cm-2 and mobilities of 1100 cm2/Vs. These results were achieved with a MOCVD reactor that had been modified to cope with the low vapor pressure of the used precursors Cp3Sc and (MCp)2ScCl. Depending on the temperature, the obtained growth rates were as high as 0.015 nm/s. Currently, further precursors with allegedly higher vapor pressure are evaluated that should allow for even higher growth rates, further improving the quality of the fabricated heterostructures. Furthermore, HEMT devices were fabricated on top of the epitaxial wafers that showed power densities (Pout) of up to 8.4 W/mm and power-added efficiencies (PAE) of more than 48 % at a frequency of 30 GHz, even surpassing the current state-of-the-art in terms of combined Pout and PAE of Ga-polar devices.However, there are still a number of unsolved issues visible in most reported AlScN/GaN HEMT devices, regardless of the used growth method. One is concerning the level of leakage currents and the consequential premature soft breakdown. The root cause of this phenomena is not yet fully understood but is believed to originate from poor dielectric/isolating properties of the Sc-containing barrier. This hypothesis is supported by a comparably high level of impurites that is found in the AlScN layers. Minimizing theses residuals by purification of the precursors is, therefore, necessary to further improve the electrical device characteristics. Secondly, transport properties of AlScN/GaN heterointerfaces are not as well studied as more established junctions, e.g., AlGaN/GaN and AlN/GaN. An important parameter that has not been experimentally identified for 2DEGs induced by AlScN/GaN heterojunctions yet, is the saturation drift velocity (vsat). It is of particular importance since the Bode-Fano limit, among other, depends on it. Since the extension of the latter is one of the great promises of AlScN/GaN devices, it is clear that the value of vsat is essential to know in order to make quantitative predictions of the level of improvement that is achievable in terms of bandwidth capability but also with regard to power density.We will give an overview of recent advances of MOCVD-grown AlScN/GaN HEMTs with respect to their capability to extend on the achievable performance level of present GaN-based devices, in particular in the millimeter-wave frequency range. Further, we will benchmark the results against other competing technologies, i.e. GaAs pHEMT and Si-CMOS, to give an outlook of where optimized AlScN/GaN devices can takes us and where they cannot.