The physics behind the excitement lies in a deceptively simple scaling law. In rectifying power electronics, the Baliga figure of merit, which captures how well a semiconductor performs as a switch, is proportional to the bandgap raised to the power of 5.5. That steep exponent means that moving from silicon’s 1.1 electronvolts to gallium nitride’s 3.4 electronvolts already transforms what is possible in converters, chargers and grid infrastructure. Moving to a colossal bandgap above 6 electronvolts, the regime the Cornell-led team now enters, promises another leap in critical electric field and therefore in the voltage a device can block per unit thickness. Thinner devices switch faster, dissipate less heat and shrink the passive components around them, which is why the pursuit of ever-wider bandgaps has become one of the most competitive frontiers in materials science.

Yet widening the bandgap has historically come at a punishing cost: the loss of dopability. The defining trick of semiconductors is that their conductivity can be tuned over many orders of magnitude by deliberately introducing impurity atoms, a capability that underpins every transistor and diode ever made. As bandgaps grow, dopants tend to become so deeply bound that they fail to release their carriers at room temperature, and materials that were once dismissed as insulators, including gallium nitride, silicon carbide, aluminium nitride, gallium oxide and germanium oxide, only became useful semiconductors after years of work reducing defect densities and finding the right dopants. Diamond and cubic boron nitride, despite their superb intrinsic properties, remain locked behind enormous synthesis barriers: high pressures, tiny crystals, prohibitive costs and difficulty scaling. The new result breaks that trend by achieving record bandgap and record conductivity on one of the cheapest substrates in the industry.

The material at the heart of the study is the corundum-structured alloy α-(Al_xGa_1−x)₂O₃, in which gallium atoms in α-gallium oxide are progressively replaced by aluminium to widen the gap. The films were grown by suboxide molecular-beam epitaxy, a technique in which the metal sources are supplied as suboxides rather than pure elements, allowing adsorption-controlled growth of stoichiometric oxide films. The team grew the layers on sapphire, which is itself crystalline aluminium oxide, meaning the substrate is chemically identical to the aluminium-rich end of the alloy system. Sapphire is abundant, inexpensive, available in exceptional crystalline quality and produced at massive scale for everything from watch faces to LED wafers, with crystals as large as hundreds of kilograms grown by the Kyropoulos method. Building colossal-bandgap electronics on such a substrate is a decisive economic advantage over diamond and cubic boron nitride, which require high-pressure synthesis and yield only small, costly samples.

The headline number is conductivity. In the colossal-bandgap regime above 6 electronvolts, the silicon-doped α-(Al_xGa_1−x)₂O₃ films exhibit room-temperature conductivities more than 100 million times higher than all previous reports for materials in this range. Previous efforts to dope the corundum-structure alloy had achieved conductive films only up to bandgaps of about 6.22 electronvolts, and the conductivity collapsed rapidly as aluminium content rose. The new work shows that silicon, acting as a donor on the cation sublattice, can still ionize and release electrons even when the bandgap exceeds 7 electronvolts, provided the growth conditions are tuned correctly. Temperature-dependent Hall measurements on the films revealed carrier freeze-out as samples were cooled, allowing the researchers to extract donor densities and donor activation energies and to confirm that the conduction observed at room temperature arises from thermally activated, deliberately introduced silicon donors rather than accidental defects.

Characterizing such films demanded an unusually broad arsenal of techniques. High-resolution scanning transmission electron microscopy revealed atomically sharp, flat interfaces between the sapphire substrate, the buffer layers and the doped channel, with misfit dislocations confined to the earliest interface. Spectroscopic ellipsometry, performed at multiple angles of incidence and azimuthal orientations to account for the anisotropy of the corundum crystal structure, was used to determine the bandgap of each layer. Secondary ion mass spectrometry quantified the silicon incorporation down to a noise floor of roughly 1.2 times 10 to the 19 per cubic centimetre, while atom probe tomography mapped the three-dimensional distribution of silicon, aluminium and gallium atoms, showing that the alloy mixing follows a random, binomial distribution with no clustering. X-ray diffraction confirmed the crystal phase and composition of the doped overlayers through sharp reflections corresponding to aluminium fractions near 0.52.

Theory played an equal role in interpreting the results. Collaborators at the University of Texas at Austin, including Viet-Anh Ha, Nick Pant and Feliciano Giustino, calculated the theoretical electron mobilities of α-(Al_xGa_1−x)₂O₃ as a function of aluminium content and carrier density using density functional theory and electron-phonon coupling methods. Their calculations, which included phonon scattering, ionized impurity scattering and alloy scattering, and which accounted for the directional dependence of transport parallel and perpendicular to the basal plane, provided benchmarks against which the measured mobilities could be judged. Because the alloy is highly anisotropic, with bandgap energies and dielectric properties that depend on crystallographic direction, such first-principles guidance is essential for designing devices that exploit the material’s best transport directions.

Crucially, the team did not stop at films. They fabricated a Schottky barrier diode, dubbed the AlphaDiode, and a field-effect transistor, the AlphaFET, both with colossal-bandgap channels. The transistor structure stacked an aluminium-rich buffer on sapphire, a silicon-doped α-(Al_0.51Ga_0.49)₂O₃ channel, and a heavily doped regrown contact layer to achieve low-resistance ohmic access to the channel. The demonstration of both a rectifying diode and a transistor proves that the material can be patterned, contacted and switched, the minimum requirements for a practical electronic platform. These first devices are proof of concept rather than production-ready components, but they establish that doping, regrowth and contact formation, the traditional stumbling blocks for ultra-wide-bandgap materials, can all be executed on this oxide system.

The broader significance is strategic. Power electronics built on silicon carbide and gallium nitride are already reshaping electric vehicles, data centres and renewable energy systems, and gallium oxide in its beta phase has attracted intense interest because it can be grown from its own melt. The corundum-structure alloy reported here extends the reachable bandgap beyond anything doped before, and it does so on sapphire wafers that are already manufactured at scale, which the authors argue facilitates rapid development and adoption of colossal-bandgap electronics. The team has filed a provisional patent covering the growth procedure for conductive silicon-doped α-(Al,Ga)₂O₃ and the device structures that use it. Challenges certainly remain, including the modest mobilities typical of oxide alloys, deep donor levels that limit carrier activation, and the need for p-type doping, which remains unsolved. But the central barrier, the belief that semiconductors above 6 electronvolts cannot be made conductive, has now been decisively broken, and with it the assumption that the future of power electronics must be built on ever more expensive and difficult crystals.

Subject of Research: Silicon-doped α-(AlxGa1−x)2O3 as a record 7-eV bandgap semiconductor for colossal-bandgap power electronics

Article Title: A 7-eV bandgap semiconductor based on silicon-doped α-(AlxGa1−x)2O3

Article References: Steele, J., Bhattacharya, D., Nomoto, K., Pieczulewski, N. A., Sorensen, P., Ha, V.-A., Pant, N., Shukla, I., Das, S., Kilic, U., Ramesh, M., Giustino, F., Mazumder, B., Senevirathna, M. K. I., Williams, M. D., Schubert, M., Muller, D. A., Xing, H. G., Jena, D., & Schlom, D. G. (2026). A 7-eV bandgap semiconductor based on silicon-doped α-(AlxGa1−x)2O3. Nature. https://doi.org/10.1038/s41586-026-11085-3

Image Credits: AI Generated

DOI: 10.1038/s41586-026-11085-3

Keywords: colossal-bandgap semiconductor, alpha-gallium oxide, aluminum gallium oxide alloy, silicon doping, suboxide molecular-beam epitaxy, sapphire substrate, power electronics, Baliga figure of merit, Schottky diode, field-effect transistor, wide-bandgap materials, molecular-beam epitaxy

The device, which the authors call a Funnel Morph Antenna, gets its name from a distinctive hybrid geometry that combines a narrow tapered neck with a flared upper section, resembling the profile of an industrial funnel. That shape is not merely aesthetic. In millimeter-wave antenna design, one of the central challenges is impedance matching: the antenna must present a consistent 50-ohm load to the feeding circuitry across a wide range of frequencies, or else a large fraction of the captured signal simply reflects back instead of being delivered to the rectifier. The funnel profile solves this by providing a gradual electromagnetic transition from the narrow neck, which concentrates surface currents toward the radiating aperture, to the wide flared portion, which extends the effective radiating area. Because the current path changes smoothly rather than abruptly, discontinuities that would normally cause reflections are minimized, and the antenna maintains a reflection coefficient below minus 10 dB across the full operating band.

The physics behind the wideband behavior is a story of overlapping resonances. Simulations of the surface current distribution at 28, 33, and 38 GHz reveal that different parts of the structure dominate at different frequencies. At 28 GHz, current concentrates around the narrow neck and lower radiating sections, corresponding to a low-frequency resonant mode. By 33 GHz, the current spreads into the flared upper portion, indicating the formation of additional resonant modes, and at 38 GHz the maximum current density shifts to the upper flare and the cavity regions, where shorter current paths support the highest-frequency operation. The superposition of these multiple modes is what stretches the usable bandwidth to a full 14 GHz, an unusually wide span for a compact millimeter-wave element. A multilayered cavity embedded in the ground plane adds further control, confining the electromagnetic fields and stabilizing the directional radiation pattern.

Perhaps the most provocative engineering decision in the study is the choice of substrate. Millimeter-wave devices almost always use specialized low-loss laminates such as Rogers RT/duroid, because ordinary FR4, the fiberglass material found in countless consumer circuit boards, suffers from significant dielectric losses and dispersion at frequencies above 26 GHz. The team deliberately chose 1.6-millimeter FR4, with a dielectric constant of 4.3 and a loss tangent of 0.02, to test whether a genuinely low-cost design could still perform. Despite the higher insertion losses, careful optimization of the flare and feed geometry allowed the antenna to achieve wideband matching, and the geometry proved robust against the etching tolerances of standard printed circuit board fabrication, holding its performance even with dimensional deviations of plus or minus 0.05 millimeters. The single element measures only about 17.5 by 20.3 millimeters and delivers a simulated peak gain of roughly 5 dBi with radiation efficiency above 85 percent, figures that measurements closely confirmed.

A single antenna, however, captures only a modest amount of power, so the researchers replicated the element into a four-element array. To feed all four elements uniformly, they designed a 1-to-4 Wilkinson power divider, a classic microwave component built from three cascaded two-way dividers, each using quarter-wavelength transmission lines with a characteristic impedance of twice the system impedance and a 100-ohm isolation resistor between output ports. The divider performed impressively across the band, with insertion loss between roughly minus 6 and minus 6.2 dB, meaning each output port receives almost exactly one quarter of the input power, and port-to-port isolation exceeding minus 45 dB around 30 GHz, indicating negligible signal leakage between channels. The array elements were spaced 5 millimeters apart, half a wavelength at the operating frequency, to maximize gain while suppressing unwanted side lobes.

The assembled array, measuring 78.25 by 70.57 millimeters, delivered a measured peak gain of 10.25 dBi and an efficiency of about 93 percent at 28 GHz. The theoretical maximum gain for four ideal elements with a single-element gain of 4.8 dBi would be 10.82 dBi, so the measured value falls only 0.57 dB short, a gap the authors attribute to feed network losses, conductor losses in the microstrip lines, and the dielectric losses inherent to FR4. The aperture efficiency, a measure of how effectively the physical area of the array is used, remained a steady 78.5 percent despite mutual coupling between elements and phase errors across the wide band. The beam pattern is notably narrow, with a half-power beamwidth of about 9 degrees at 28 GHz and 10 degrees at 38 GHz in one plane, and side lobe levels between minus 7 and minus 9 dB, giving the array a focused, directional sensitivity well suited to harvesting energy from distant base stations.

Converting captured radio waves into usable electricity falls to the rectifier, and here the team navigated some practical constraints. They selected the 1SS351 Schottky diode from ON Semiconductor, chosen because its electrical characteristics closely match the widely used HSMS-2852 diode, which has recently become obsolete. Using Keysight Advanced Design System software, the researchers modeled the diode with harmonic balance and S-parameter simulations to extract its input impedance, finding a value of 47.4 minus j20 ohms near 28 GHz. Matching that impedance to the antenna required careful design work. The team compared half-wave and full-wave rectifier topologies and found that the full-wave bridge significantly raises the imaginary component of the input impedance, which would degrade efficiency without elaborate matching. A matched half-wave rectifier using microstrip lines emerged as the practical choice, especially after lumped-element matching proved infeasible because it required component values, such as a 53-ohm resistor and a 2.7-nanohenry inductor, that are difficult to source off the shelf.

The full system was validated experimentally in an anechoic chamber, with a broadband vector signal generator driving a standard gain horn antenna as the transmitter and the fabricated rectenna prototype as the receiver, separated by one meter, comfortably within the far-field region for these frequencies. Both antennas were carefully polarization-aligned, and the incident power density was calibrated with a broadband power sensor before characterization. When the array harvested 68.7 microwatts of incident RF power, the system produced an output voltage of 250 millivolts across a 2-kilohm load, corresponding to an overall RF-to-DC conversion efficiency of 45.5 percent. That figure is competitive with, and in several respects superior to, prior millimeter-wave rectennas, many of which required expensive substrates, waveguide structures, or far higher input power levels to reach comparable efficiency.

The comparison with earlier work underscores why the low-cost approach matters. Previous millimeter-wave rectennas have relied on Duroid substrates, substrate-integrated waveguides, liquid crystal polymer packaging, on-chip CMOS integration, and even air-filled waveguide Fabry-Perot resonators. Some achieved higher peak gains or higher conversion efficiencies, such as a waveguide-fed 35 GHz design reaching 68.5 percent, but typically with narrower bandwidth, more complex manufacturing, or substantially greater incident power. Others, like CMOS folded dipoles, achieved extreme miniaturization at the cost of negative gain. The new design occupies a distinctive middle ground: a full 26 to 40 GHz operating range, a respectable 10.25 dBi gain, a narrow 9-degree beamwidth, and 45.5 percent conversion efficiency, all on ordinary FR4 fabricated with standard PCB technology.

The implications reach beyond the laboratory. As 5G and future 6G networks densify urban radio environments, the ambient millimeter-wave energy available for harvesting will only grow, and devices that can scavenge it without batteries could power IoT sensors, wearables, and backscatter tags indefinitely. The authors suggest that future work will explore miniaturization techniques and adaptive beamforming to improve integration into compact real-world devices. If battery production costs, which are high relative to the energy those batteries actually store, continue to drive the push toward green electronics, a funnel-shaped patch of fiberglass that drinks from the 5G spectrum may prove to be one of the more elegant shortcuts to energy-autonomous wireless devices.

Subject of Research: A wideband funnel-geometry rectenna array for radio-frequency energy harvesting in the 26–40 GHz 5G millimeter-wave band

Article Title: Wideband funnel geometry rectenna array for RF energy harvesting in 26–40 GHz 5G networks

Article References: Megahed, A. A., Mousa, M. E., Al-Gburi, A., & Elabd, R. H. (2026). Wideband funnel geometry rectenna array for RF energy harvesting in 26–40 GHz 5G networks. Results in Optics, 25, Article 101178. https://doi.org/10.1016/j.rio.2026.101178

Image Credits: AI Generated

DOI: Not provided

Keywords: rectenna, RF energy harvesting, 5G, millimeter-wave, antenna array, Wilkinson power divider, Schottky diode, FR4 substrate, wireless power transfer, impedance matching, IoT, Results in Optics