Perspective—Opportunities and Future Directions for Ga₂O₃

Bulk, insulating or conducting β-Ga₂O₃ crystals can be grown by edge-defined film-fed (EFG) growth using Ir crucibles, by Czochralski or by float zone and are commercially available.^2–4 Each has advantages and limitations, but to date, each has proven capable of producing high quality crystals. EFG substrates with excellent transparency and uniformity are commercially available in up to 2" sizes,² with demonstrations at 4" and 6" plates in development, as shown in Figure 2.² The 10 × 15 mm² substrates are available in three orientations, namely (⁻201), (010) and (001) while the larger circular substrates are available only in the (⁻201) orientation. The wafers contain no twin boundaries.²

Zoom In Zoom Out Reset image size

There has also been progress in development of epi growth by a number of techniques, including MOCVD, HVPE and MBE with controlled n-type doping over the range 10¹⁵–10¹⁹ cm⁻³ using Sn or Si shallow donors.^1,5–8 HVPE is probably the most developed at this point, but requires subsequent polishing to provide an acceptable morphology. The dislocation density of current bulk wafers is of order 10³ cm⁻²_, a key result for making large area power rectifiers and the n-type doping is stable against annealing.² Progress in bulk and n-type epi films has been impressive, with future need for development of coherent β-(Al_xGa_1x)₂O₃/Ga₂O₃ and β-(In_xGa_1−x)₂O₃ heterostructures.⁹ The solubility limit of Al₂O₃ in β-Ga₂O₃ is ∼65% in the range 850–1950 °C and for MBE growth temperatures (600–800 °C), the solubility limit of Al₂O₃ in β-Ga₂O₃ is drastically reduced due to the formation of a AlGaO₃ intermediate compound.⁹ This necessitates use of high oxygen fluxes during MBE to achieve high growth rate. Similar limitations will be encountered in MOCVD and HVPE growth and exploration of optimized growth conditions and the role of strain and extended defects is an important area of future research.

For p-type doping, there have been no reports on achieving conductive material with acceptors,¹⁰ due to their predicted large ionization energies, presence of common n-type impurities and native defects and the resulting n-type background conductivity that must be overcome. This will require a systematic measurement of the ionization energies of candidate acceptors (by optical methods initially) and a better understanding of the compensation by native point defects and their complexes plays an important role in influencing the efficiency of doping in Ga₂O₃ and related alloys. These compensating centers may be native defects or complexes from either sublattice or simply impurities. At high enough levels, the doping itself might induce self-compensation. This is a common phenomenon in other compound semiconductors where the formation energy of acceptor-like vacancies in n-type material decreases as the Fermi level rises closer to the conduction band edge. The understanding of native defects in Ga₂O₃ is far from complete and has been largely driven by first principal calculations using different approaches.¹¹ The formation energy of these defects is generally dependent on the Fermi level.

Many very promising Ga₂O₃-based power rectifiers, MESFETs, MOSFETs and even finFETs have been reported,^1,10–20 along with various types of solar blind photodetectors.^21,22 Another limitation is the lack of high quality patterning, doping and contacting processes that exist for the more mature semiconductors.²³ High quality ohmic contacts are a prerequisite for any device and should provide low contact resistance at moderate anneal temperatures. Additional contact resistance leads to slower device switching speeds as well as reliability issues due to local contact heating during current flow during device operation. Etching is needed for intra-device isolation or for exposing layers for subsequent contacting. Ion implantation can be used for channel/contact region doping and device isolation where planarity requirements preclude the use of mesa etching. The other major issue is the limited and anisotropic thermal conductivity of Ga₂O₃ (0.27 W.cm⁻¹ K⁻¹ for [010] and 0.11 W.cm⁻¹ K⁻¹ for [100]),¹⁰ which might be addressed using heatsinks like diamond or microfluidic approaches or transfer to a metal substrate, all currently being developed for GaN electronics. However, the lower thermal conductivity of Ga₂O₃ relative to other wide bandgap materials means these techniques must be even more effective. For this reason, Ga₂O₃ is projected for high voltage, lower current applications than GaN.

The usual approaches involve surface etching or cleaning to reduce barrier height or increase of carrier concentration of the surface through preferential loss of oxygen.^{10,12,23–25} To date, contact schemes involving IZO or dry etching in BCl₃/Ar to enhance the surface n-type conductivity, followed by Ti/Au annealed at 500°C have been common. Specific contact resistances of ∼4.6–8 × 10⁻⁶ Ω ⋅ cm⁻² were reported for Ti/Au contacts on n-Ga₂O₃ epitaxial layers in which Si was implanted and annealed at 925°C, followed by dry etching, metal deposition and annealing at 470°C.^23,24 Many published I-V characteristics are only quasi-linear at low current and shows the need for improved contact approaches. Cr has a low work function of 4.5 eV and is currently one of the best choices for ohmic contacts on n-Ga₂O₃.

For an n-type semiconductor, to achieve an ohmic contact means that the work function of the metal must be close to or smaller than the electron affinity of the semiconductor (affinity of β-Ga₂O₃ is ∼4.00 ± 0.05 eV),²⁵ and thus potential choices if the surface is unpinned include Hf (work function 3.9 eV), Sc and La (both 3.5 eV) and Gd (2.9 eV). Bilayers of these with Au should reduce the contact sheet resistance. The transport mechanisms in the contacts can be determined by temperature-dependent TLM measurements. A traditional ohmic contact follows a thermionic field effect (TFE) or field-effect (FE) depending on the semiconductor effective doping and/or temperature. No significant dependence of the contact resistance on measurement temperature would indicate the dominant transport mechanism is field emission.

Wet etchants have been reported for Ga₂O₃, including HNO₃/HCl, H₂SO₄, H₃PO₄ and HF-based solutions,^26–28 but little is known about its dry etching characteristics and the associated mechanisms and effects on the properties of the material. Plasma chemistries of Cl₂/BCl₃ for reactive ion etching and BCl₃, BCl₃/SF₆, and CF₄/O₂ for inductively coupled plasma conditions showed generally low etch rates.^29,30 Plasma-induced damage in Ga₂O₃ is found to increase the conductivity of the near surface and lead to improved n-type ohmic contact resistivities, as was the case in the early days of GaN technology. Other options include methane/hydrogen (CH₄/H₂) and mixed Cl₂/H₂ or BCl₃/Cl₃/H₂ chemistries. It needs to be established whether they are ion-driven or have a significant chemical etch component. In etching processes occurring by ion-enhanced sputtering in a collision-cascade process, the etch rate will be proportional to E^0.5-E_TH^0.5, where E is the ion energy and E_TH is the threshold energy.

Implant isolation is a method of using ion implantation of either electrically inactive or deep level impurities to produce electrically insulating regions that can be used for inter-device isolation. It has a major advantage over mesa isolation of maintaining a planar surface. It has not yet been investigated in Ga₂O₃.

Similarly, little is known about the electrical properties of implanted dopants in Ga₂O₃, especially for p-type doping. For n-type doping, Si, Ge and Sn are donors if incorporated on the Ga site.^21,23 Conversely, the column V elements should be acceptors if incorporated on the O site. Of course, their energy level in the gap determines whether they are shallow or deep and therefore whether they are significantly ionized at room temperature even if they are soluble on the appropriate lattice site. The systematics of damage recovery and substitutionality of the implanted dopant still need much work in Ga₂O₃.³¹

There are three main criteria that a gate dielectric for Ga₂O₃ must possess. First, it must be thermodynamically stable with the semiconductor and not react during processing. Secondly, it should have a high quality interface with low defect and trap density to ensure high carrier mobility. Finally, it must have sufficient band offsets (typically >1 eV for both conduction and valence band offset) to act as barriers to both electrons and holes. The device performance depends strongly on the type of band alignment, and on the band offsets. The wide bandgap of Ga₂O₃ limits the available choices of gate dielectrics to those with bandgaps >6 eV (eg. SiO₂, Al₂O₃, HfSiO₄, Y₂O₃, La₂O₃, LaAlO₃) and it will also be necessary to establish the type of band alignments: type I, type II staggered, type III broken gap. If the offsets are not large enough, the dielectric may still be used as a surface passivation layer to protect the surface of Ga₂O₃ from the instabilities induced by exposure to atmosphere. Measurements of band offsets are in their early stages.³² To date, Al₂O₃, SiO₂ and HfO₂ have been used in metal-oxide-semiconductor (MOS) structures on Ga₂O₃. The process and etch conditions for patterning the dielectrics also need to be established

There is interest in the properties of hydrogen in Ga₂O₃, because of the predictions from density functional theory and total energy calculations that it should be a shallow donor.³³ The generally observed n-type conductivity, therefore, may at least in fact be explained by the presence of residual hydrogen from the growth ambient, rather than to native defects such as Ga interstitials or O vacancies. In some related oxides, oxygen vacancies may even be deep donors. There is some experimental support for the fact that hydrogen may be a shallow donor in Ga₂O₃ from experiments on its muonium counterpart and from electron paramagnetic resonance of single-crystal samples.³⁴

In terms of projected devices utilizing Ga₂O₃, the high fields and potentially high temperatures in some of the intended applications are beyond the capabilities of current device models. Device design simulators should include high fields and high temperatures as early as possible so that they can provide accurate guidance. The monoclinic crystal structure has different symmetry than the GaN hexagonal or GaAs zincblende lattices most commonly simulated. The Ga₂O₃ has lattice constants of a = 12.23 Å, b = 3.04 Å, c = 5.80 Å and two types of two kinds of coordination for Ga³⁺ ions in the structure, namely distorted tetrahedral and octahedral, meaning the oxide ions are in a distorted cubic closest packing arrangement, and the gallium (III) ions in the two types of sites have Ga-O bond distances of 1.83 and 2.00 Å, respectively.