Synaptic silicon-nanocrystal phototransistors for neuromorphic computing

The incorporation of augmentative functionalities into a single synaptic device is greatly desired to enhance the performance of neuromorphic computing, which has brain-like high intelligence and low energy consumption. This encourages the development of multi-functional synaptic devices with architectures that are capable of achieving demanded synaptic plasticity. Here we take advantage of the remarkable optical absorption of boron (B)-doped silicon nanocrystals (Si NCs) to make synaptic phototransistors, which can be stimulated by both optical and electrical spikes. The optical and electrical stimulations enable a series of important synaptic functionalities for the synaptic Si-NC phototransistors, well mimicking biological synapses. It is interesting that the synergy of the photogating and electrical gating of the synaptic Si-NC phototransistors leads to the implementation of aversion learning and logic functions. We show that a spiking neural network based on the synaptic Si-NC phototransistors may be trained for the recognition of handwritten digits in the modified national institute of standards and technology (MNIST) database with a recognition accuracy around 94%. The energy consumption of the synaptic Si-NC phototransistors may be rather low, which should help advance energy-efficient neuromorphic computing.