Synaptic Phototransistors Devices Based On Silicon Nanocrystals

Neuromorphic computing that mimics the functionalities of a brain is now one of the most important technologies beyond the conventional von Neumann computing.In a brain intelligent activities like self-learning and memory are implemented through the modification of synaptic weight between biological synapses.This signifies that the development of artificial synapses(i.e.,synaptic devices)is crucial for the realization of neuromorphic computing.In the past few years synaptic devices with the structures of both memristors and field-effect transistors(FETs)have been actively investigated.For a synaptic FET augmentative functionalities may be conveniently incorporated to enhance plasticity and parallelism,facilitating the development of robust neuromorphic circuitry with drastically reduced complexity.Hence,significant efforts have been devoted to the investigation of high-performance synaptic FETs.Up to now,most of synaptic FETs have been demonstrated with electrical stimulations.The development of optogenectics in neuroscience,however,has recently inspired the study on optically-stimulated synaptic FETs.The incorporation of light into the operation of synaptic FETs may not only significantly broaden the bandwidth of synaptic devices,but also effectively mitigate the interconnect issues of synaptic devices.Optically-stimulated synaptic FETs have been recently fabricated by using two-dimensional(2D)materials such as graphene and MoS2.Although these devices showed excellent synaptic plasticity,their optical stimulations only efficiently took place in a narrow ultraviolet(UV)-to-visible spectral range.The near-infrared(NIR)optical stimulation with the advantageous compatibility with optical communication remains elusive.In addition,the optical absorption of 2D materials is usually weak.Hence,large optical power is required for the stimulation of the devices based on 2D materials.This is in fact a serious issue for the development of neuromorphic computing with ultralow energy consumption.Please note that optically-stimulated synaptic FETs can also work with electrical stimulations.However,the synergy of optical and electrical stimulations has not been well explored to enable high-performance multi-functional synaptic devices.In this work,we have fabricated optically-stimulated synaptic FETs by using boron-doped silicon nanocrystals(Si NCs),which have strong optical absorption in the broad UV-to-NIR spectral range.These devices are essentially Si-NC phototransistors.They exhibit a series of important synaptic functionalities with the stimulation of UV-to-NIR light,which leads to the photogating of the devices.The synaptic Si-NC phototransistors can be also readily stimulated by electrical spikes at the backgates,rendering the parallelism of the devices.We have taken advantage of the synergy of the optical and electrical stimulations to not only realize the metaplasticity and logic functions of the synaptic Si-NC phototransistors,but also mimic the neurological behavior of aversion learning.The current synaptic Si-NC phototransistors may have rather low energy consumption,indicating a great promise for the development of highly energy-efficient neuromorphic computing.The main reserch contants and results are as follows:(1)Using non-thermal plasma method to prepare boron-doped silicon nanocrystals(Si NCs)with strong optical absorption in the broad UV-to-NIR spectral range.The TEM was performed to analyze its structure and size.The UV-vis-NIR absorption spectra was employed to demonstrates its optical absorption.The synaptic FET devices were fabricated by using boron-doped silicon nanocrystals(Si NCs).By employing electrical spikes at the backgates,we successfully simulate the main synaptic plasticity of EPSC and PPF,which show excellent synaptic plasticity.(2)The synaptic FETs exhibit a series of important synaptic functionalities with the stimulation of UV-to-NIR light.The incorporation of light into the operation of synaptic FETs helps to realize the STDP,behavior of aversion learning,metaplasticity and logic functions.We analyze the write noise,variability,non-linearity and energy consumption of our synaptic phototransistors devices based on silicon nanocrystals.We simulated a spiking neural network(SNN)for the modified national institute of standards and technology(MNIST)database of handwritten digits based on the synaptic Si-NC phototransistors and obtained a 93.6%classification accuracy on the test set.