Ultraviolet Optoelectronic Synapse Based on AlScN/p-i-n GaN Heterojunction for Advanced Artificial Vision Systems.

Abstract

Ferroelectric materials represent a frontier in semiconductor research, offering the potential for novel optoelectronics. AlScN material is a kind of outstanding ferroelectric semiconductor with strong residual polarization, high Curie temperature, and mainstream semiconductor fabrication compatibility. However, it is challenging to realize multi-state optical responders due to their limited light sensitivity. Here, a two-terminal AlScN/p-i-n GaN heterojunction ultraviolet optoelectronic synapse is fabricated, overcoming this limitation by leveraging hole capture at the AlScN/p-GaN hetero-interface for multi-state modulation. The novel structure maintains excellent memristor characteristics based on the ferroelectric of AlScN, realizing an on/off ratio of 9.36 × 10⁵. More importantly, the device can mimic synaptic characteristics essential for artificial vision systems, achieving an image recognition accuracy of 93.7% with a weight evolution nonlinearity of 0.26. This approach not only extends the applications of AlScN in optoelectronics but also paves the way for advanced artificial vision systems with image preprocessing and recognition capabilities. The findings provide a step forward in the development of non-volatile memories with potential for on-chip sensing and computing.