A research team from King Abdullah University of Science and Technology in Saudi Arabia has developed an artificial neuron that can use optoelectronics to perform neural morphology calculations. New technologies mimic the function of synapses or neurons, adapting and reconfiguring their response to light to complete calculations. This breakthrough progress was published in the latest issue of the journal "Light: Science and Applications". The team used two-dimensional material hafnium diselenide to design and manufacture metal oxide semiconductor capacitors (MOSCap). This device adopts a vertical stacking structure, in which hafnium diselenide is sandwiched between two layers of aluminum oxide and placed on a p-type silicon substrate. The top is covered with a transparent layer of indium tin oxide, allowing light to enter from above. When hafnium diselenide nanosheets are integrated into charge capture storage devices, optical data sensing and retention functions can be achieved, allowing them to reconfigure to sense light or store optical data even after the light source is removed. Experiments have shown that the charge capture and capacitance of MOSCap vary with changes in lighting conditions, making it a smart memory that can be trained and responded to using optical signals. For example, exposure to blue light with a wavelength of 465 nanometers can enhance the response to red light with a wavelength of 635 nanometers, a behavior known as associative learning. In neuromorphic computing, MOSCap is like an artificial synapse that can exhibit both long-term enhancement (increasing synaptic response) and long-term inhibition (weakening synaptic response) capabilities. The team further explored how these artificial neurons respond and adapt to light stimuli of different intensities, durations, and wavelengths. The prediction of simulation experiments shows that the capacitive synaptic array circuit based on these devices can achieve an accuracy of 96% in recognizing handwritten digits in industry standard databases. In addition, these devices also demonstrate the potential for detecting exoplanets, with an accuracy rate of up to 90% by identifying transient changes in stellar intensity. Research shows that this kind of device with memory optical sensing function is very suitable for edge computing applications, especially in the field of artificial intelligence that requires rapid processing and storage of large amounts of optical data. Its potential applications are more extensive, including autonomous vehicle, virtual reality and Internet of Things systems, paving the way for more adaptive and energy efficient solutions in the future. (New Society)