Silicon-Doped Oxide on Sapphire Passes 7 eV, a New Bandgap Record

A Cornell University-led team has made conductive semiconductor films with bandgaps above 7 electronvolts. That is wider than any semiconductor known before. According to the team's paper in Nature, the silicon-doped α-(Al_xGa_1−x)₂O₃ films exceed 7.0 eV and surpass cubic boron nitride, the next-widest-bandgap semiconductor known. The larger result is electrical rather than optical. In what the authors call the "colossal-bandgap" regime above 6 eV, room-temperature conductivities are over 100 million times higher than all previous reports.

A wide gap alone is not the achievement. Many materials with gaps that wide are insulators. The work shows that one can be doped into a useful conductor, and that it can be grown on a cheap, commercially available substrate.

01Why bandgap matters for power devices

The bandgap is the energy an electron needs to move from the valence band, where it is bound, to the conduction band, where it carries current. A wider gap lets a material withstand a stronger electric field before it breaks down. For power switches, that means a thinner drift region can block a given voltage, which allows smaller devices with lower conduction losses. Gallium nitride and silicon carbide already exploit this in fast phone chargers and electric vehicles, according to the Phys.org summary of the work.

Aluminum gallium oxide pushes the idea further. Changing the aluminum-to-gallium ratio tunes the bandgap anywhere from 5.4 to 8.6 eV, the widest range of any known semiconductor. The group's 2025 APL Materials paper notes that the alloy's breakdown field is anticipated to reach about 30 MV/cm at the top of that range. That is a projection, not a measured device result.

02Growing the film on sapphire

The α-phase of gallium oxide has the same corundum crystal structure as sapphire (α-Al₂O₃). Pure gallium oxide prefers the more stable β-phase, which has a narrower gap. A 2021 Science Advances paper from the same Cornell groups showed that growing on m-plane sapphire stabilizes the α-phase across the full alloy range. In that work, the c-plane of the crystal catalyzes the transition to the narrower-bandgap phase, and the m-plane orientation avoids that route.

The Nature films are grown by suboxide molecular-beam epitaxy (S-MBE). In ordinary MBE of gallium oxide, the rate-limiting step is oxidizing the gallium on the surface. S-MBE avoids it. According to the group's 2024 paper, 99.98% of the gallium beam arrives already oxidized as Ga₂O. That allowed growth rates above 1 µm/h at a relatively low substrate temperature of 605 ± 15 °C. Aluminum content, and therefore bandgap, follows the flux ratio closely: the 2024 paper reports that aluminum incorporation is linear with the relative aluminum-to-Ga₂O flux ratio.

Crystal quality in that earlier work was good. The films showed a rocking-curve width of about 12 arc seconds, rms roughness below 1 nm, and full commensuration for x ≥ 0.5 at 20–50 nm thickness. The 2024 paper covered compositions up to x = 0.95. Even so, the authors reported a problem there: their silicon-doping attempts produced highly insulating films.

03The doping problem

That result was the obstacle. As aluminum content rises, supplying free electrons gets harder. According to Phys.org, aluminum-rich films had conducted several orders of magnitude worse than gallium-rich ones, which made them impractical.

The general physics of ultra-wide-bandgap materials explains why. In a wide-gap host, a donor's electron is more strongly bound, and the donor level can sit deeper below the conduction band. Compensating defects can also trap the electrons the donor supplies. Silicon, a donor on a metal site in these oxides, therefore gives diminishing returns as the gap grows. This is general semiconductor physics, not a finding of the paper. The abstract I could access does not describe the specific mechanism by which the team overcame it, so I cannot say which growth-condition changes mattered. The route did run through a step-by-step campaign. Silicon-doped α-Ga₂O₃ was shown to be conductive in October 2025, and the Nature paper extends that to aluminum-rich alloys.

The paper's framing is historical. It notes that GaN, SiC, AlN, Ga₂O₃ and GeO₂ became useful semiconductors after defect densities fell and suitable dopants were found. This work is presented as the same pattern repeated at a higher energy.

04Devices and what they show

The team built two devices. One is a Schottky diode. The other is a field-effect transistor the authors call the AlphaFET, with colossal-bandgap channels. A secondary summary reports that the transistor channel is α-(Al_0.51Ga_0.49)₂O₃. I could not confirm that from the primary text, and I could not find breakdown voltage, on-resistance or mobility figures in the material I accessed. Those are the numbers that will decide whether the material competes with GaN and SiC. They should be checked against the paper itself.

05Substrate economics and open questions

The commercial argument rests on the substrate. According to the paper, the films and devices use sapphire, an abundant, inexpensive substrate of excellent quality produced at massive scale. The authors say this breaks the trend of rising synthesis difficulty, cost and small size in ever-wider-gap materials such as diamond and cubic boron nitride.

Several things remain unproven:

• Reliability and yield. One secondary summary lists reliability, contact resistance, thermal management and manufacturing yield as still to be shown. That fits the proof-of-concept stage of the devices.
• Thermal behavior. Oxides are generally poor heat conductors, and this is a known concern for gallium-oxide power devices. A high-voltage device is only compact if it can shed its heat. Whether thin films on sapphire manage this is not addressed in the sources I reviewed.
• Scale of growth. MBE is a research-grade tool, and the sapphire substrate can be large, but the film growth is not yet shown at production volume or throughput. Some commentary treats the substrate's low cost as implying easy commercialization. The sources do not demonstrate it.
• Metastability. The α-phase is not the thermodynamically stable form of gallium oxide. Its long-term and thermal stability in processed devices needs data.

The authors hope the films could lead to power electronics that waste far less energy as heat when converting and controlling electricity. That is a reasonable aim. It is also a long way from demonstrations of a diode and a transistor. What the paper does establish is narrower: doping control can survive past 6 eV, and it can do so on a commodity substrate.