New ferroelectric material sandwiched between GaN gates, makes the transistor switch faster and more efficient, through negative capacitance.
Researchers from the University of California, Berkeley, along with Stanford collaborators, have successfully used a ferroelectric material to overcome a longstanding limitation in gallium nitride (GaN) transistors. Their findings, published in Science, show that integrating a material exhibiting negative capacitance into GaN devices helps improve performance without sacrificing energy efficiency.
GaN-based transistors are critical components in 5G base stations and compact power supplies. However, scaling them for higher power and frequency has always involved trade-offs. A key challenge lies in maintaining high current when the device is on, while also reducing energy leakage when it’s off. This is typically constrained by the Schottky limit, a trade-off dictated by the thickness of insulating layers in the transistor.
The research team addressed this issue by applying a 1.8-nanometre-thick bilayer made of hafnium oxide and zirconia, known as HZO. It has crystal structure, allowing it to maintain an internal electric field, without external voltage applied.
Unlike conventional insulators, HZO is a ferroelectric material that supports negative capacitance. This phenomenon enhances gate control and increases on-state current flow, while still limiting leakage when the transistor is off.
Normally, increasing the dielectric thickness weakens control over the transistor. But with negative capacitance, the new design defies that logic. The internal field of the HZO layer interacts with the applied voltage in a way that boosts charge accumulation at the gate. This directly translates to better switching behaviour and greater efficiency in GaN transistors.
Although the experimental devices are still relatively large, the team plans to apply this approach to more advanced, miniaturised radio-frequency transistors. The research opens new paths for extending negative capacitance applications beyond silicon into GaN, and possibly into other high-power semiconductors like silicon carbide and diamond. If proven scalable, this innovation could dramatically enhance the performance of future electronic and telecom devices.