EPFL · Wetenschap

New transistor brings high voltage to microchip scale

EPFL engineers have developed a new type of gallium nitride (GaN) transistor, called the intrinsic polarization superjunction (iPSJ), capable of handling very high voltages with minimal energy loss. This innovation addresses a key limitation in current GaN technology, where high voltages can cause premature device failure.

The iPSJ transistor, built on a silicon base, can withstand nearly 4 kilovolts (kV) before breaking down, a record for this technology, while maintaining low resistance. This combination is crucial for efficient power conversion in applications like AI data centers, electric vehicles, and solar power systems. The breakthrough exploits a natural polarization effect unique to GaN.

Gallium nitride's crystal structure creates internal electric forces that gather mobile electrons into current-carrying sheets. In conventional GaN transistors, when the device turns off, this can lead to a concentration of voltage at a single point, causing breakdown. The POWERlab's iPSJ design engineers the GaN layers so that a second sheet of positive charge forms alongside the electron sheet, balancing the charges and allowing voltage to spread evenly.

This balanced design allows the iPSJ transistor to withstand over five times the voltage of commercial GaN power devices. Its ability to hold high voltage across a wide temperature range makes it suitable for demanding environments like electric vehicles and industrial power systems. Furthermore, the doping-free design enhances robustness by avoiding temperature sensitivity issues associated with conventional doping methods.

The POWERlab is also exploring methods to add multiple conduction channels to power electronics components to distribute current, reduce resistance, and prevent overheating. The next objective is to combine the iPSJ's high-voltage capability with this multi-channel approach to tackle the challenges of next-generation power electronics, aiming for safe high-voltage handling and minimized energy loss.

AI-samenvatting op basis van de bron.