Gallium Oxide Trumps Traditional Wide Bandgap Semiconductors
Two separate developments are needed to secure stable energy supplies in the near future: Widespread adoption of revolutionary technologies that can replace burning of fossil fuels; and the introduction of a multitude of products that consume less energy, a step that will also trim greenhouse gas emissions.
One way to fulfil this latter goal is to turn to more efficient power electronics. This is possible by replacing silicon devices with those made from wide bandgap semiconductors, which have superior material properties, such as higher breakdown voltages and lower switching losses.
The two most promising, highly developed wide bandgap materials for power electronics are GaN and SiC. However, both of them have massive weaknesses when it comes to mass production. In an ideal world, compound semiconductor devices are built on an affordable native substrate, but it is impossible to make GaN and SiC crystals with simple, low-cost methods. Current methods to make GaN lead to sale prices that are prohibitively high for power electronics, and while SiC substrates don’t suffer from the same fate, they are still costly and their material quality is far from perfect.
A promising wide bandgap alternative that has been overlooked up until now is Ga2O3. Thanks to a bandgap that is significantly larger than that of SiC and GaN, this oxide promises to enable the production of devices with higher breakdown voltages and higher efficiencies than those stemming from its wide bandgap rivals. What’s more, Ga2O3 power devices could be manufactured at low cost in high volume, because it is possible to produce single-crystal native substrates from a melt using the same method employed for manufacturing sapphire substrates. This platform for making GaN LEDs is now being manufactured in a low-cost commercial process in numbers rivaling those for silicon substrates; there is no major obstacle to prevent Ga2O3 substrates from treading the same path.
To try and convert the promise of Ga2O3 power devices into a reality, since 2010 I have been working in collaboration with Kyoto University, the Tokyo Institute of Technology, and Tamura and Koha Corporations. Our team has already developed several elemental technologies and hit several key milestones, including the world’s first demonstration of Ga2O3 transistors.
The character of gallium oxide
Ga2O3 has many forms, and to date there have been confirmed reports of five different polytypes, which are denoted α, β, γ, δ and ε. The β-polytype is the most stable, and the four other polytypes are classed as quasi-stable. Since efforts on Ga2O3 are in their infancy, there have only been a handful of reports on crystal growth and material properties of this oxide, and they have focused on the β-phase. This polytype has several attractive attributes, including an incredibly wide bandgap of 4.8-4.9 eV and an n-type conductivity that can be controlled from 1016-1019 cm-3 through doping with tin or silicon. However, so far there are no reports of clear hole conduction in p-type Ga2O3 layers.
Thanks to its wider bandgap than SiC and GaN, the β-polytype of Ga2O3 promises to enable fabrication of devices with excellent characteristics, including high breakdown voltages, high power capacity and high efficiency. (See Table 1 for a comparison of the important material properties for power device applications for various popular semiconductors, plus β-Ga2O3).
Table 1. In terms of break down field and Baliga’s figure of merit, Ga2O3 is superior to all the popular compound semiconductor materials
The great material properties of β-Ga2O3 indicate that this wide bandgap semiconductor should have a breakdown electric field of 8 MV/cm, three times those of either SiC or GaN. This very high value for the electric field strength is a trump card for Ga2O3 because Baliga’s figure of merit – the basic parameter to show how suitable a material is for power devices – is far more dependent on breakdown field than mobility. Values for this figure of merit are proportional to the cube of the breakdown field, but only linearly proportional to mobility.
It is possible to use values for the material properties of semiconductors to calculate theoretical limits of on-resistance at a range of breakdown voltages (see Figure 1). Such efforts indicate that the on-resistance of Ga2O3 devices can be an order of magnitude lower than those for SiC and GaN devices at the same breakdown voltage.
Figure 1. The β-poltype of Ga2O3 promises to excel in breakdown voltage and on-resistance
Preparing a platform
To fabricate Ga2O3 devices that can fulfill their potential, one must begin with a native substrate. The good news is that large, single-crystal Ga2O3 substrates can be fabricated from the melt at low cost using very little energy. This is in stark contrast to the expensive, energy-consuming methods employed for creating GaN and SiC bulk crystals and substrates: Sublimation, vapor phase epitaxy, and high-pressure synthesis.
Two of the members of our team, the Tamura and Koha companies, have already succeeded in developing 2-inch-diameter single-crystal β-Ga2O3 wafers via a melt growth technology known as edge-defined film-fed growth (EFG). This growth technology has a good track record for producing large sapphire substrates, and the same system configuration can be used for making Ga2O3 substrates.
This is not the only reason why melt growth is attractive: In principle, melt growth of Ga2O3 should never generate micropipe defects, which have historically plagued SiC substrates produced by the most popular method, sublimation. In addition, melt growth does not need a high-pressure environment, which is required for the production of SiC substrates. And on top of these benefits, the size of substrates produced by the EFG method is just determined by the dimensions of the equipment – scaling substrate size is simply a matter of increasing the size of the growth tool.
It should be possible to make Ga2O3 substrates very cheaply. We estimate that a mass-production system could churn out high-quality, 6-inch-diameter Ga2O3 substrates at a unit cost of $120. Thanks to a relatively efficient process, we estimate that the power dissipated per-unit-area of substrate at the time of production is just one-third of that associated with SiC sublimation, due to a lower growth temperature and a higher growth rate.
We recently succeeded in fabricating the first field-effect transistors based on a single-crystal Ga2O3 channel grown on a β-Ga2O3 (010) substrate. These devices were metal-semiconductor field-effect transistors (MESFETs), simple structures that are suitable for demonstrating transistor action. Our devices are certainly not the first oxide transistors ever produced. For several years, oxide semiconductors such as InGaZnO4 (IGZO) and ZnO have been attracting attention as new transparent transistor materials for display applications. However, unlike Ga2O3, these crystal structures are amorphous and/or polycrystalline, and are consequently unsuitable for the fabrication of high-power devices. To produce our Ga2O3 transistor, we deposited a tin-doped n-Ga2O3 channel layer on a native substrate by MBE. Conventional Knudsen cells provided the gallium and tin sources, and the oxygen source comprised 5 percent ozone and 95 percent oxygen.
Figure 2. A melt-based process,which is similar to that employed for sapphire production, can yield 2- inch Ga2O3substrates
Device isolation techniques have not been developed, so we employed a circular FET pattern (see Figure 3). To form ohmic contacts, we used a mixture of BCl3 and argon to perform reactive-ion etching, followed by evaporation of Ti/Au. We have found that this etching process substantially reduces contact resistance. After this, we fabricated Schottky gates by Pt/Ti/Au deposition and lift off. MESFETs that did not feature surface dielectric passivation were created that had a 4 μm gate length and a source-drain spacing of 20 μm. The diameter of the inner circular electrodes for the drain was 200 μm.
Figure 3. A crosssectional schematic illustration of the n-Ga2O3 MESFET structure (a) and an optical micrograph of the fabricated device (b)
DC output characteristics for one of our circular Ga2O3 MESFETs include a maximum drain current density of 25 mA/mm and perfect pinch-off (see Figure 4). A destructive measurement of the three-terminal breakdown voltage in the off-state, which resulted in burned gate electrodes, produced a value of more than 250 V. Transconductance peaked at 2.3 mS/mm for a 40 V drain bias.
Figure 4. DC characteristics of the n-Ga2O3 MESFET
Other promising characteristics of our MESFET include an off-state drain leakage current of just 5 μA/mm and an on/off drain current ratio that could hit 10,000. A high proportion of this leakage current will be associated with the large gate pad (see Figure 3(b)), and the leakage from the gate finger will be at least one order of magnitude below this value. The off-state drain current is comparable to the gate leakage current, indicating that there is a negligibly small leakage through the semi-insulating Ga2O3 substrate. It is possible to drive down the off-state current even further by simply adjusting the device configuration.
A benchmark to judge these devices by is that of the performance of the GaN MESFETs of the early 1990s: Our oxide transistors deliver comparable or better performance. It is no surprise that our devices are inferior to the far more mature, state-of-the-art SiC and GaN devices of today, but their combination of high breakdown voltage and low leakage current shows that they have great potential as power devices.
To improve the performance of our fledgling devices, we must address a slew of technological challenges. These include the production of substrates with diameters of more than 4-inches, device processing, and epitaxial growth that includes doping. So that our devices can find practical application fast, we want to develop normally off transistors that can serve in switching equipment. The best structures for this purpose are MOSFETs, which are likely to use either Al2O3 or SiO2 for the gate dielectric film. There is good reason to believe that it will be possible to produce such structures with a high-quality, low-defect-density interface, because both materials are oxides. We believe that our recent development of Ga2O3 transistors could herald a new era for high-performance power electronics. Such devices will not only help to reduce global energy use; they will also trim energy consumption in the semiconductor industry.
This work is partially supported by the New Energy and Industrial Technology Development Organization (NEDO) and the Japan Science and Technology Agency (JST), Japan.