Simulation And Study Of Electronic Synaptic Plasticity Based On PCNE

The 21 st century has an urgent need for artificial intelligence,which is driving the rapid development of artificial neural network system.For human brain,the synaptic plasticity is considered to be the cytological basis of learning and memory,therefore,the electronic synaptic plasticity has also become a prospective research hotspot in the field of human brain biomimetic,especially artificial neural networks.So it is very important to realize electronic synapses with ability of synaptic weight and biological synaptic density by using the existing technology.The phase change memory which has both storage function and nanometer-integrated characteristic can solve this problem to the greatest extent,considering that,the study of electronic synaptic plasticity for the phase change nanometer element(PCNE)is of great significance.Based on MATLAB simulation platform and the storage principle of PCNE,the article establishes the physical and mathematical models of the T-type three-dimensional structural PCNE with the finite element method.And the volta ge pulse with adjustable amplitude and width is applied to implement the electrical performance simulation,thermal performance simulation and the phase transition process simulation,based on that,we emulate the change of the resistance and the temperature/phase distribution,validate the feasibility of PCNE as an electronic synapse,and simulate the performance and working mechanism of the electronic synaptic plasticity especially STDP learning rules.Firstly,the voltage pulse is applied on one end of the PCNE electrode,by controlling the amplitude and width of the pulse,PCNE can complete the correction and storage of the synaptic weight well,it owns the abilities of nonlinear transmission and leaning including long-term depression(LTD)and long-term potentiation(LTP).Like the biological synapse,PCNE also exhibits characteristic of spiking-frequency dependent plasticity,and the bigger pulse frequency is,the bigger the synapse weight change is.Secondly,based on the basic STDP learning rule of biological synapse,we design the pulse scheme of the presynaptic pulse and the postsynaptic pulse flexibly.In addition,the SET or RESET threshold is estimated according to the R-V characteristic curve of the PCNE with a specific initial state,which will help to determine the parameter of the presynaptic and postsynaptic pulses.Ultimately,we achieve both the symmetrical and asymmetric STDP learning rules whose time windows are in nanoseconds level,and also,the energy consumption in the learning process is only in the order of picojoule,reach the energy consumption level of biological synapses,to some extent,the above results confirm that the electronic synapses based on PCNE have the characteristics of both low energy consumption and high speed.