Simulation Of Memristors And Bionic Research Of Neural Synapses

The human brain is not as accurate or fast as a traditional computer,but it is significantly better than traditional computer in dealing with complex tasks.The reason is that there are a lot of neurons inside the brain and a lot of connections,known as synapses,are established between neurons.The strength of connections between neurons is represented by synaptic weights,which are modulated in neuronal activity.In general,high frequency action potentials increase the weight of synapses,while low frequency action potentials decrease the weight of synapses.The conductance of memristors can be continuously adjusted and is non-volatile,making them ideal for simulating synapses.At present,it is necessary to change the direction of voltage to adjust the conductance of memristors when simulating synapses with memristors.It is impossible to adjust the conductance of the memristor by changing the frequency of the input pulse,as is the case with synapses.Here in this work,we modeled the memristor in COMSOL.The capacitor-memristor circuit(CMC)were simulated and analyzed.The concept of action pulse was proposed.The action pulses were input into the CMC,and synaptic plasticity was successfully simulated by changing the frequency of the input pulses.By connecting two memristors to the same capacitor,the Pavlov associative memory function was realized.The main contents and conclusions of this paper are as follows:(1)Pt/Zn O/Ag and ITO/Ag Ge Te/Ag memristors were prepared.It is found by fitting the I-V curve of the device that the device obeys Ohm’s law in the low-resistance state and follows the space-charge-limited conduction mechanism in the high-resistance state.(2)The memristor was modeled in COMSOL.The effects of RESET voltage,compliance current,conductive filament radius,electrode material and hopping distance of dielectric layer material on the resistive switching behavior of memristor were explored.In addition,two ways(CRS,1S1R)to suppress the sneak current were simulated.(3)A capacitor-memristor circuit was built.It is found that when a rectangular pulse is applied to it,a positive/negative voltage spike would appear across the memristor at the rising/falling edge of the input pulse,indicating the charging and discharging of the capacitor,respectively.It causes the resistance of memristor to change if the amplitude and pulse width of the voltage spike are large enough.(4)The concept of action pulse was proposed.It is composed of a positive rectangular pulse and a negative rectangular pulse.Applying low frequency action pulses to the CMC decreases the conductance value of the memristor,and applying high frequency action pulses increases the memristor conductance value.This result is consistent with long-term depression(LTD)and long-term potentiation(LTP)at synapses.(5)When simulating the long-term plasticity of synapses,too small capacitance in the CMC causes a rapid drop in positive voltage spikes during charging and increases the magnitude of negative voltage spikes during discharge,making LTD difficult to achieve.Too large capacitance increase the critical pulse count,making LTP difficult to achieve.(6)The Pavlov associative memory experiment was simulated by connecting two memristors to the same capacitor.It is found that the timing of the input pulses affect the change of the conductance of the memristors if the action pulses are input to two memristors at the same time.