Investigation On The Spiking Dynamic Characteristics In Cascaded VCSEL-SA Photonic Neurons

Neural network,as the core of artificial intelligence,has been extensively applied in these fields such as information processing,biological medicine and intelligence controlling.However,with the increase of information requirements and emerging of more complex tasks,exploring the neural network with more functions and faster proceeding speed becomes necessary.Photonic spiking neural networks(SNNs)can break through the bottle neck including of bandwidth,power consumption and information exchange capacity in the traditional SNNs based on electrical method.Moreover,photonic SNNs can provide ultrafast spiking dynamics up to 8 orders of magnitude faster than biological neurons.Therefore,the related research on photonic SNNs has become a hot issue.In recent years,different photonic neural models based on semiconductor optical amplifiers and semiconductor lasers have been proposed,and the super-fast dynamic characteristics of spiking have been provided.Amongst of these photonic neural models,Vertical-cavity surface-emitting lasers with saturable absorber(VCSEL-SA),as a typical leaky integrate-and-fire(LIF)neuron model,possesses some unique advantages such as easy to integration and scale,low cost and shorter sub-ns spike output,have become an idea neuron-like device.Obviously,research on the spiking dynamics of VCSEL-SA and its application has significant academic and application value.In this paper,the spiking dynamic characteristics of the photonic neuron of the vertical-cavity surface-emitting lasers with an embedded saturable absorber(VCSEL-SA)under external perturbations are investigated numerically.The input strength and the temporal duration of the input optical pulse effect on the spiking patterns,and the refractory period of VCSEL-SA neurons under the two continuous perturbations are discussed.The spikes can be fired when the input strength and the temporal duration exceeds a certain threshold.And with the increase of injection intensity and temporal duration,the neurons may produce more spike signals in the temporal duration.In the case of the two continuous perturbations,the spike signal will not be triggered by the neuron during the refractory period when the time interval between the two disturbances is shorter than the refractory period of the VCSEL-SA neurons respond to the spike signal.On that basis,we proposes a photonic neural system composed of three cascaded VCSEL-SAs and numerically investigate the encoding,propagation and storage characteristics of the spiking patterns in this system.The results show that,with suitable perturbation strength,the first VCSEL-SA(VCSEL-SA1)can convert the regular and irregular stimulus into spike signals.Ideally,the conversion rate from binary data to spike(BTS)can reach 1Gbps.To some extent,increasing both the perturbation strength and the bias current is beneficial to improve the conversion rate.Moreover,the spiking patterns generated by VCSEL-SA1 can be stably propagated into another two VCSEL-SAs(VCSEL-SA2 and VCSEL-SA3)with a certain delay through adjusting the coupling weight.Finally,after introducing a feedback into VCSEL-SA1,the fired spiking patterns can be successfully stored in this proposed system.The relatively larger feedback weight and feedback delay are helpful to store the spiking patterns.The obtained results can offer great some theoretical supports for the application of VCSEL-SAs neurons in the future fields of brain-inspired ultrafast neuromorphic computing system and brain-like computing.