NAMBE 2024 Session NAMBE1-WeA: Heterogeneous Integration

Group III-Nitride materials have found applications in optoelectronic, photonic, and high power devices due to many factors, including the large variation in bandgap spanning from the infrared to the deep ultraviolet. Recent research has pursued the combination of this well-established material system with transition-metal nitrides for the creation of complex heterostructures which possess interesting optical, magnetic, or superconducting functionality. While GaN research has been primarily focused on the hexagonal allotrope, the metastable zincblende phase has a direct bandgap of 3.2 eV and providing an attractive option as a wide bandgap cubic material. Many transition metal nitrides have a stable cubic rocksalt structure. This includes ZrN, NbN, and TaN which are also well-known superconductors with relatively low lattice mismatch to GaN. This work will use molecular beam epitaxy for the epitaxial integration of these metal nitrides with hexagonal (c-plane) and cubic (001) GaN and will discuss the effect of growth parameters, including growth direction, on the structural and electrical quality of the metal nitrides. Reflection high energy electron diffraction, X-ray diffraction, and transmission electron microscopy reveal the epitaxial quality of single layer films and superlattices. Temperature dependent resistivity measurements show the superconducting critical temperature is strongly dependent on the growth conditions. These results open the door for new epitaxial superconductor-semiconductor systems and provide a platform for integration with other cubic materials to enable complex heterostructures. Such atomically precise hierarchical matter could be used for metamaterials or quantum science.

Development of GaN-based high electron mobility transistors (HEMTs) has led to significant performance improvements in solid-state power switching and RF electronics, owing to the wider bandgap and larger breakdown field strength of GaN compared to semiconductors such as Si and GaAs. Despite these enhanced material properties, GaN HEMT performance is limited by non-uniform peak electric fields that can cause premature breakdown. One field management strategy recently demonstrated leverages the high dielectric constant (κ) of a gate material such as BaTiO₃ (BTO) to both control the electrostatics and reduce the peak electric field in the gate-drain region of the HEMT [1]. These BTO dielectric layers are typically sputtered, leading to either poor crystal quality or polycrystalline films, which have lower dielectric constants than bulk or crystalline materials (κ > 1000 for bulk BTO).

In this work, we demonstrate epitaxial (111)-oriented BTO thin films grown on AlGaN/GaN HEMT heterostructures by oxide and nitride molecular beam epitaxy. Key in this approach is the use of a thin SrTiO₃ (STO)/TiO₂ bilayer in between the BTO and AlGaN to both grade the lattice mismatch between the oxide and nitride layers and to orient the BTO film. In previous work, we showed that a 1-nm TiO₂ buffer layer greatly improves crystallinity when cubic STO films are deposited on wurtzite AlGaN [2]. However, BTO has a tetragonal structure and a slightly larger unit cell volume than STO. Consequently, BTO films deposited on TiO₂-buffered AlGaN exhibit lower crystallinity compared to films with an additional STO buffer layer.

We determine the growth window of BTO/AlGaN films across substrate growth temperature, oxygen flow, and Ba/Ti flux, and characterize the crystal phase and structural properties with x-ray diffraction, atomic force microscopy, and reflection high-energy electron diffraction. We investigate the atomic microstructure with scanning transmission electron microscopy (STEM), which confirms the epitaxial relationship as (111)[1-10] BTO || (0001)[11-20] AlGaN, which is similar to the observed STO/AlGaN epitaxial relationship [2]. Additionally, the STEM imaging reveals that the BTO film is highly textured, with crystallites approximately 10 nm in size, and that the buffer layers are rougher than the BTO film. Finally, we show through van der Pauw Hall effect measurements that the electrical properties of the GaN channel are robustly maintained, with no appreciable degradation of sheet resistance, electron mobility, or charge density.

[1] N. K. Kalarickal et al., IEEE Trans. Electron Devices (2021)
[2] E. N. Jin et al., APL Mater. (2020)

While research on the direct growth of InAs quantum dot lasers on silicon has progressed rapidly in recent years, silicon photonic integrated circuits with heteroepitaxially-integrated lasers have not achieved the same level of success as heterogeneous wafer-bonded integration techniques. On-chip coupling between the silicon photonics passives and embedded heteroepitaxial III-V lasers has proven to be a significant challenge in the practical application of this technology with the best reported devices showing insertion losses of 7.35 dB. Heteroepitaxial integration has the potential to offer a significant manufacturing cost and throughput advantage over heterogeneous integration by allowing 300 mm wafer processing without the use of expensive III-V substrates. However, the coupling problem must be addressed for the technology to be competitive in device performance.

The large coupling losses in heteroepitaxially integrated lasers arise primarily from the air gap between the etched facets of the lasers and silicon photonics waveguides. Lasers are fabricated by first etching a pocket in a conventionally fabricated silicon photonics wafer, with the III-V laser material subsequently grown in the pocket to allow butt coupling between the active and passive regions. However, the growth at the edges of the pockets must be etched away to reduce the non-uniformity and to produce adequate facet mirrors for the III-V lasers. This results in a ~10 µm air gap between the laser facet and the waveguides that severely degrades coupling. In molecular beam epitaxy, these non-uniformities at the pocket edges are the result of polycrystalline III-V deposition on the silicon dioxide sidewalls. Selective area growth can reduce or eliminate this air gap by preventing the formation of polycrystalline III-V material on the sidewalls. However, selective area MBE growth is not practical for aluminum-containing alloys due to aluminum’s low volatility.

We report the growth and fabrication of III-V lasers heteroepitaxially integrated with silicon photonic integrated circuits using a partial selective area growth method to reduce the coupling gap. The use of a highly selective GaAs buffer layer and non-selective laser stack mitigates the formation of polycrystalline III-V material, reducing the effective coupling gap from 11 µm to 6 µm while maintaining previously reported growth conditions in the active region. Fabricated ridge lasers exhibit scattered light output powers as high as 20 mW continuous wave with thresholds as low as 205 mA. Coupling loss measurements are in progress and will be reported at the conference.

We investigate the dependence of the number of graphene layers on the strain and epitaxial orientation of GdAuGe films, grown on graphene / 6H-SiC (0001). Whereas GdAuGe films growth directly on SiC or on few layer epitaxial graphene on SiC forms a hexagon on hexagon epitaxial alignment (GdAuGe [10-10] // SiC [10-10]), GdAuGe films grown on buffer layer graphene on SiC are rotated 30 degrees with respect to the underlying substrate. These results cannot be explained by a “remote” epitaxy mechanism, but instead may result from the highly buckled structure of buffer graphene which has strong covalent bonding to the underlying SiC. We will present a detailed analysis of the structure and strain by cross sectional STEM and reciprocal space mapping, and the impacts on magnetism in GdAuGe.