PNNL tests AI method for radiation damage in satellite power electronics

(SpaceEnergy) US researchers have demonstrated an artificial intelligence-assisted method for analysing radiation damage in gallium nitride semiconductors used in satellite systems and power electronics, potentially accelerating qualification of spacecraft components.

The Pacific Northwest National Laboratory (PNNL) announced the research on 26 October, describing experiments combining ion-beam irradiation with automated analysis of semiconductor materials.

Researchers used data from experiments at the University of Tennessee in which ion beams simulated radiation damage that spacecraft electronics could accumulate over years in orbit.

The investigation examined semiconductor structures incorporating aluminium gallium nitride and gallium nitride, materials used in satellite electronics, communications equipment, power systems and radar.

PNNL deployed its SciLink artificial intelligence platform to coordinate specialised agents responsible for experimental analysis, planning and computational simulation.

The agents combined information from materials-characterisation techniques, including atom probe tomography and scanning transmission electron microscopy.

According to PNNL, the system completed analyses in hours rather than the days or weeks normally required.

Its assessment assigned 90% confidence to the conclusion that ions dominated the observed damage, compared with 60% confidence for its findings concerning proton damage.

Researchers independently examined the results, checked calculations and assessed potential errors.

The work could support development of radiation-resistant spacecraft power electronics by improving understanding of semiconductor degradation.

PNNL did not report qualification of a flight-ready component, demonstrate improved radiation resistance or identify a spacecraft mission incorporating the technology.

The laboratory also did not disclose the precise dates of the experiments.