Modeling Of Conductive Filament Memristor And Study On Synaptic Characteristics

As an important basic electronic component,memristor is regarded as the best candidate for constructing neuromorphic network and simulating human brain in the future,and has become the focus of attention in the field of microelectronics.Compared with other electronic devices,it has obvious advantages such as simple structure,low power consumption,good size miniaturization and fast switching speed.The diverse synaptic functions of memristor can make the artificial neural network more similar to the function of human brain.However,memristor-based synaptic biomimetic research is still in its infancy,due to limited physical models,unclear physical mechanisms,and single simulated synaptic function,which limit its further development.Although researchers have proposed corresponding physical models for different structures and material systems of memristors,many details are still controversial.Therefore,it is important to establish some more reasonable and effective memristor models,which will help us to predict the mechanism of resistance switching and simulate more complex and diverse synaptic functions.Because metal and oxygen vacancy conductive filament memristors have balanced performance in all aspects and are closer to practical application,they have been widely concerned.In this paper,a series of new models are established for the two typical filament memristors,and synaptic characteristics and influencing factors are studied here,the details are as follows:(1)A novel conductive filament model of metal clusters is established based on ion drift and spontaneous decay of clusters(Cluster-decay model,CDM model).In some conductive filament devices composed of metal clusters,the conductance decay is caused by the spontaneous contraction of the clusters in the filaments.This paper will establish a new metal conductive filament model based on this new phenomenon in such devices.The ion drift under the action of electric field causes the dynamic changes of the conductive channel in both vertical and horizontal directions.The spontaneous decay of metal clusters gives the model forgetting property.Through the verification of memristor characteristics and paired-pulse facilitation synaptic function,it can be seen that the basic characteristics of the model are in good agreement with the actual device.The effects of pulse interval,duration and amplitude on synaptic weight are studied.The forgetting properties of the model and the relationship between atomic transition distance,surface tension coefficient and forgetting properties are discussed.In addition,the effect of ion transition distance on the resistance transition and memristor properties,and the effect of metal ion/atom activation energy on the properties are also studied.The model provides a theoretical basis for the study of the filament memristor with cluster decay and its application.(2)Based on cluster residual effect,a metal cluster conductive filament model with memory retention characteristic is established(Cluster-residual model,CRM model).On the basis of CDM model,CRM model is established by introducing the residual effect of cluster decay.This model not only has the forgetting characteristic but also has the memory retention characteristic,so it is closer to the human forgetting law.Through the verification and analysis of CRM model,it can be seen that the model can be well matched with the actual device.The forgetting characteristic of CRM model is further discussed and compared with CDM model.The empirical learning functions of the two models are studied.Through comparison the empirical learning efficiency of CRM model with memory retention characteristic is higher than that of CDM model.In addition,by adding a window function,the boundary effect similar to that of the actual device can appear in the model,so as to avoid the situation that the size of the conductive channel exceeds the maximum physical limit,and realize the hard-switching mode of the model.The introduction of CRM model enables the cluster filament model to simulate the forgetting effect with memory retention characteristic,which can be better used in the application and development of memristor.(3)Based on ion drift,Fick and Soret diffusion mechanism,and the coupling relationship between the length of conductive channel and the electric field,an improved oxygen vacancy conductive filament model is proposed.The horizontal and vertical dynamic changes of conductive channel are described by two variables: the cross-sectional area and the length of conductive region,respectively.Two diffusion mechanisms can make the model have forgetting and memory retention characteristics.The coupling between the length of the conductive channel and the electric field can correct the deviation between the previous model and the actual device on memristor characteristics,which makes the model more reasonable and comprehensive.The boundary problem of this model is solved by using a window function.Moreover,the improved model is verified and analyzed,and compared with other models.The improved model can better capture the characteristics of the actual device.In addition,the model can match different memristor devices by adjusting fitting parameters.The work provide a solid theoretical basis for the performance prediction and optimization of oxygen vacancy filament devices and synaptic bionics.(4)The basic functions of synapses,including short-term plasticity,long-term plasticity,habituation and dishabituation,are simulated based on the improved oxygen vacancy conducting filament model.The realization of these functions shows that the model has multiple synaptic capabilities at the same time,which provides a reliable choice and theoretical basis for the construction of artificial neural networks in the future.In addition,the influence of temperature on the synaptic characteristics of the improved model is further discussed.The influence of temperature change on synaptic weight based on Joule thermal effect and thermal coupling between the inner and outer regions of the conductive channel is mainly studied.Moreover,the potential applications of the model are briefly introduced.