Light-induced negative differential resistance and neural oscillations in neuromorphic photonic semiconductor micropillar sensory neurons.

Neuromorphic systems, inspired by nature, are sought to efficiently process analogue inputs in real and complex environments. This could lead to ultralow-power in-sensor intelligent edge computers. Here, we present an artificial sensory oscillator neuron consisting of a III-V semiconductor micropillar quantum resonant tunnelling diode (RTD) with GaAs photosensitive absorption layers. The oscillatory optical neuron encodes incoming analogue optical data into spatiotemporal oscillatory signals. We demonstrate that near-infrared light within a certain intensity range activates a region of negative differential resistance, and subsequently, large-amplitude voltage oscillations. As a result, optic analogue information is encoded into electrical oscillations resulting in amplification of sensory light inputs. Under pulse-modulated light, excitation and inhibition of burst firing patterns can be controlled within a single oscillatory neuron, simulating neural activity in networks in the form of breather-type oscillatory phenomena. Such spatiotemporal oscillatory patterns (burst firing) form the basis for the combined sensing, pre-processing, and encoding abilities of the vision-nervous system found in biological organisms. This work paves the way for future artificial visual systems using III-V semiconductor nano-optoelectronic circuits in applications for light-driven neurorobotics, bioinspired optoelectronics, and in-sensor neuromorphic computing systems for real-time processing of sensory data.