Atomera has announced a GaN-on-silicon development intended to reduce parasitic RF losses and support wider use of gallium nitride devices on lower-cost silicon substrates.
The semiconductor materials and technology licensing company says its Mears Silicon Technology, or MST, can address a key performance barrier that has limited GaN-on-silicon adoption in mainstream radio-frequency applications. The company is targeting 5G, future 6G, satellite communications, and other high-frequency RF devices where GaN performance is attractive but substrate cost and scalability remain difficult.
High-performance RF GaN devices are typically built on silicon carbide substrates. SiC provides strong thermal and electrical performance, but it is more expensive and harder to scale than silicon. Silicon offers larger wafer potential, better compatibility with established semiconductor manufacturing infrastructure, and a lower-cost route if performance barriers can be overcome.
The problem has been parasitic channel loss. GaN-on-silicon devices can suffer from losses at high frequencies that reduce efficiency and weaken performance compared with GaN-on-SiC. Atomera’s MST approach introduces a thin, oxygen-modified layer near the surface of the silicon wafer. The controlled layer is designed to modify the silicon lattice, block dopant diffusion, improve crystal quality at the GaN/silicon interface, and reduce mechanisms that contribute to RF power loss.
Independent RF characterisation work on early MST-enabled samples has shown a substantial reduction in parasitic interface charge and a reduction in RF losses. Atomera has also highlighted large-signal results, including improved linearity and power handling, with the strongest samples approaching performance territory more commonly associated with advanced RF silicon-on-insulator technologies.
Linearity is a critical parameter in RF design. Communications infrastructure, satellite links, radar, and high-frequency industrial systems need devices that preserve signal quality while handling power efficiently. Poor linearity can increase distortion, complicate amplifier design, and reduce system performance. If GaN-on-silicon can deliver stronger linearity while retaining silicon’s cost and manufacturing advantages, it opens a broader design window.
The development sits alongside wider adoption of wide-bandgap devices, including QPT’s 1MHz GaN motor-drive platform, where high-frequency switching is being pushed into application-level engineering. GaN is spreading through compact power conversion, RF systems, motor drives, and emerging infrastructure where conventional silicon devices face efficiency, speed, or thermal limits.
Atomera’s work sits on the RF side of that same shift. Demand for more bandwidth, higher frequencies, lower losses, smaller systems, and better power efficiency is increasing. Mobile networks are moving towards denser infrastructure and future 6G research. Satellite communications are expanding into electronically steered antennas and high-throughput systems. Defence and aerospace applications need compact, efficient RF front ends. Industrial sensing and imaging are moving into higher-frequency regimes.
Substrate economics will shape how widely these technologies can be deployed. GaN-on-SiC will remain important where top-end performance is required, but a credible GaN-on-silicon route could support higher-volume or cost-sensitive applications. It may also improve integration with silicon-based manufacturing flows, although device qualification, reliability, process transfer, and foundry adoption will determine how quickly the approach can move from test data to production designs.
The announcement also reflects a broader materials trend in semiconductor engineering. Performance improvements are no longer driven only by smaller geometries. Interface control, epitaxy, substrate engineering, packaging, thermal paths, and process integration are all becoming central to device competitiveness. In wide-bandgap devices, the interface between materials can determine whether theoretical advantages survive in manufactured components.
Customer evaluation and manufacturing validation now become decisive. RF device makers will want repeatable data across wafers, operating conditions, frequencies, temperatures, and ageing profiles. Foundries will need to assess process compatibility, yield, and whether MST can be implemented without disrupting established manufacturing economics. System designers will look for device availability, models, application data, and qualification evidence.
Atomera’s proposition targets a known bottleneck rather than a vague performance claim. If parasitic losses in GaN-on-silicon can be reduced consistently while preserving silicon’s cost and scale benefits, RF GaN could move into a wider set of industrial, communications, and aerospace systems. Production adoption still has to follow, but high-frequency electronics are increasingly becoming a materials engineering contest as much as a circuit design challenge.
