β-Ga2O3: The Next Chapter
Speeding MOCVD growth sets the stage for device development by Ross Miller, Fikadu Alema and Andrei Osinsky from Agnitron Technology
Gallium oxide is a very promising material for making power and optoelectronic devices. It has a great set of intrinsic properties, and preliminary device testing indicates that it could have the upperhand over SiC and GaN power devices in high-voltage switching. What’s more, it has the potential to produce highperformance solar-blind UV photodetectors.
This ultra-wide band gap material comes in five different flavours. The most promising is the monoclinic polymorph β-Ga2O3. It has the unique combination of a very large bandgap energy – it is 4.9 eV – and a conductivity that can be varied over a vast range, due to doping control that extends from 1 x 1015 cm-3 to 1 x 1020 cm-3. Armed with these attributes, β-Ga2O3 can withstand electric fields that could cause failure in SiC and GaN.
Figure 1. (a) Crosssectional view of the photodiodes based on a β-Ga2O3:Ge homoepitaxial film grown by MBE on the bulk n+ β-Ga2O3 substrate. (b) Optical image showing the top view of the vertical Schottky photodiode.
Holding back progress with β-Ga2O3-based devices are weaknesses associated with the growth of its epilayers. Drawbacks with MOCVD, MBE and HVPE are preventing the rapid deposition of thick films of β-Ga2O3 and its related alloys with properties needed to underpin the commercialisation of power devices and photodetectors.
But these issues are not insurmountable, according to our work at Agnitron Technology of Eden Prairie, MN. Our efforts have shown that with the right approach, MOCVD has been demonstrated to support growth rates up to 10 μm/hr, which is more than an order of magnitude higher than previously; far higher electron mobilities than before; and complementary AlGaO alloys with an aluminium content of more than 40 percent.
Another appeal of β-Ga2O3-based technology is that native substrates are available and can enable growth of high-quality epilayers without special techniques required when epilayers and substrates are crystal lattice mismatched. This foundation can be produced by several techniques, including edge-defined film-fed growth and the Czochralski technique.
One tremendous advantage of being able to produce native substrates is that it enables the growth of lowdefect-density epitaxial films. These are a prerequisite for producing power electronics devices with a high critical electric field strength as defects are a primary cause for device breakdown at lower than expected fields.
Table 1. Comparison of various semiconductor material properties as well as several key figures of merit (taken from the last two papers listed in Further reading). Commercial development of β-Ga2O3 devices will have to consider thermal management and packaging R&D, due to the relatively low thermal conductivity of this oxide.
In addition, melt techniques – and specifically the Czochralski technique that is used to make countless silicon substrates – have already realised low material costs by scaling up production. Assuming sufficient demand, one should expect that costs for β-Ga2O3 substrates can asymptotically approach cost parity with those for silicon and sapphire.
Note that the road to commercialisation for β-Ga2O3 is markedly different to that for SiC and GaN. Both those materials have far better material properties for high-voltage operation than those of silicon, but have faced manufacturing challenges, due to various issues related to substrates and processes. This leaves the door ajar for alternative material systems, including Ga2O3, which is not hampered by those weaknesses and may offer differentiated performance for some applications.
Our initial involvement with β-Ga2O3 was not associated with power devices, but with UVC photodetectors. Working in partnership with researchers at the University of California, Santa Barbara (UCSB), and the University of Central Florida (UCF), we fabricated Schottky photodiodes with this material system.
Despite a lack of optimisation, these detectors exhibited a high photoresponsivity and a high rejection ratio (see Figures 1 and 2). We believe that further effort, enabling a tuning of the energy gap and cut-off wavelength by bandgap engineering via compositional tuning of β-(AlGaIn)2O3 alloys, could lead to the commercialisation of true solar-blind, full UV spectrum detectors.
Figure 2. (a) Current-time characteristics of the vertical Schottky device based on β-Ga2O3: Ge epitaxial film from Figure 1 under dark and illuminated conditions. (b) Spectral response of the Pt-β-Ga2O3:Ge vertical Schottky photodiode. (c) Comparison of the photoresponsivity of the non-optimized β-Ga2O3: Ge photodiode (magenta) with commercial devices based on GaN (black), SiC (red) and AlGaN (blue) wide bandgap semiconductors. Estimated quantum efficiency is represented by the dashed lines.
There is no doubt, however, that β-Ga2O3 is attracting the greatest interest as a material for making power electronics. Offsetting its many strengths is that, like other oxide semiconductors, realising a usable p-type doping in this material is elusive – and this might always be so.
Glow from oxide heater assembly during burn-in reflects out of the multiwafer MOCVD chamber while open to atmosphere at Agnitron’s MOCVD Lab.
However, due to its ultra-wide bandgap and controllable n-type doping over a vast range, there is much interest in developing unipolar devices with vertical and lateral conducting topologies for switching and RF device applications. Progress to date for demonstration of over 1 kV vertical devices includes Schottky diodes, reported in 2017 by a team from NICT in Japan as well as FETs recently by Cornell University – and far greater success could follow, as the great intrinsic properties of β-Ga2O3 put realisation of vertical geometry devices with switching voltages exceeding 10 kV on the horizon. Recent demonstration of high mobility two-dimensional electron gas (2DEG) in β-Ga2O3 based heterostructures by Ohio State University and Air Force Research Laboratory also indicates a path to devices which may exceed the voltage performance of GaN HEMTs.