Korean Research Team Reveals Mechanism Behind Neutron Irradiation-Induced Reliability Degradation in Ferroelectric AI Semiconductors

Jeonbuk National University announced on the 20th that a research team led by Professor Bae Hak-yeol from the Department of Electronic Engineering at its College of Engineering, in collaboration with Professor Kim Tae-wan's team at Seoul National University and Q-Beam Solution, a startup affiliated with the Korea Atomic Energy Research Institute, conducted a joint study analyzing the causes of performance degradation in next-generation ferroelectric memory and neuromorphic semiconductor devices under neutron irradiation. The findings have been published in the latest issue of ACS Applied Electronic Materials.

Ferroelectric semiconductors, which can maintain their polarization state even after power is disconnected, are considered a key candidate for next-generation non-volatile memory and neuromorphic computing devices. Neuromorphic computing attempts to simultaneously implement both memory and computation within a single device, mimicking the working mechanism of human brain synapses, and is regarded as a promising approach to alleviating the “von Neumann bottleneck” of traditional computing architectures.

In this study, the research team focused on examining changes in two-dimensional ferroelectric semiconductor synaptic transistors after neutron irradiation. Experiments showed that neutron irradiation not only caused a decrease in current but also weakened charge transport through defect-related trap states and increased the uncertainty of synaptic weights. Energy-dispersive X-ray spectroscopy analysis revealed that the composition ratio of selenium to indium in the thin film changed after irradiation; electrical analysis further showed an increase in the density of electron trap defects, which in turn affected the charge transport process regulated by ferroelectric polarization.

Specifically, after neutron irradiation, the device's drain current decreased by approximately 88.95%, and the ratio of the memory window, which reflects storage characteristics, to the sweep range decreased by approximately 48.20%. However, the ferroelectric polarization switching characteristics of the device were maintained. Based on these results, the research team concluded that the performance degradation did not stem from complete loss of polarization, but was closely related to charge trapping and reduced transport caused by defect-related trap states induced by neutron irradiation.

The study also found that such degradation further affects neuromorphic synaptic operation. After neutron irradiation, the nonlinearity of synaptic weight updates increased and the dynamic range narrowed, leading to reduced inference accuracy in artificial neural networks. This indicates that radiation-induced defects not only affect the electrical performance of individual semiconductor devices but may also impact the reliability of AI computing at the system level.

Bae Hak-yeol stated that to ensure stable operation of next-generation AI semiconductors in extreme environments such as space, it is necessary not only to improve device performance but also to systematically understand the effects of radiation-induced semiconductor defects. This study, by mapping the neutron irradiation degradation pathways of two-dimensional ferroelectric semiconductor memory and neuromorphic devices, provides key physical insights for semiconductor device design and reliability assessment in radiation environments.