Two-Terminal SWIR Photosynaptic Memory for Neuromorphic Vision
When and Where
Nov 30, -0001
00:00 - 00:00
Presenter(s)
Chan So (Postech)
Co-Author(s)
Daesung Chung (Postech)
Abstract
Short-wave infrared (SWIR) photosynaptic memory is a promising technology for artificial vision systems because it can combine optical sensing, memory, and computation within a single device. However, many existing photonic memory devices require complex structures, such as additional gate electrodes or charge-storage layers, which limit scalability and high-density integration.
This work demonstrates a simple two-terminal photosynaptic diode that stores SWIR optical information through an internal photogating effect. In this device, light-induced charges remain inside the active layer and serve as an internal memory node. As a result, the device can retain previous optical input and convert it into an electrical response during readout, without relying on an external gate structure. Importantly, this operating principle enables non-destructive readout, meaning that the stored optical memory can be measured without being immediately erased.
The device also shows gradual conductance modulation, allowing repeated or weak SWIR signals to be accumulated as analog memory states. This behavior is particularly meaningful for neuromorphic vision, where sensory information should be processed in a history-dependent and energy-efficient manner. By translating optical input history into stable synaptic states, the device provides a compact hardware platform for infrared sensing and in-sensor computing.
Overall, this work highlights a practical strategy for simplifying SWIR photosynaptic memory devices while preserving key neuromorphic functions. The proposed internal photogating approach offers a pathway toward low-power, high-density artificial visual systems capable of sensing, storing, and processing infrared information directly at the sensor level.