Electric/mechanical Regulation Of Nano-ferroelectric Domains And Its Application In Artificial Synapses

The current von Neumann architecture of general-purpose computers has the problem of"storage wall",which seriously hinders the further improvement of information storage and processing capabilities.There is an urgent need to develop a new computer architecture to meet the real needs of people’s production and life.Inspired by the human brain with parallel processing and massive storage capacity,researchers have proposed new brain-like computers and brain-like neural computing systems.Brain-like computers use a large number of artificial synapses to form a brain-like neural computing system to achieve learning and memory functions,which has lower power consumption and more efficient information processing capacity than computers with von Neumann architecture,and has been an important development direction for new computers in the future.The artificial synapse is the basic component unit of brain-like neural computing system,and it can realize information storage and processing at the same time.Now the high-performance artificial synapse is the key to establish brain-like neural computing system.The resistance of ferroelectric memristors depends largely on the polarization state of ferroelectric films,which can overcome the problem of randomness of conductive filament growth and defect migration in conventional memristors,and it is more promising for bionic applications in artificial synapses.In this thesis,we have investigated the growth and migration characteristics of nano-ferroelectric domains and their domain walls induced by the modulation of force and electrical.The metal probes,ferroelectric films and oxide bottom electrodes are used as presynaptic membrane,synaptic gap and postsynaptic membrane,respectively,to investigate the memory resistance properties of nano-ferroelectric domains and their applications in the simulation of artificial synaptic behavior.The main research and results are as follows:1.Artificial synaptic models have been constructed using Sr Ru O₃（SRO）as the postsynaptic membrane,（001）,（101）and（111）-Pb Zr_0.2Ti_0.8O₃（PZT）films as the synaptic gap,and piezoelectric force microscopy（PFM）nanoprobes as the presynaptic membrane.The structural characteristics of the domains in the synaptic gap and their electrical and force-induced switching properties have been studied under the stimulation of electrical and force signals.The results show that the ferroelectric domains in the synaptic gap can be electrically switched by 180°.When the applied force signal value is about 3000 n N,the three oriented PZT epitaxial films can be switched from polarization downward to polarization upward.Among the force-induced ferroelectric domain switchings,the ferroelectric domain switching in the（001）-PZT epitaxial film is isotropic,while the force-induced ferroelectric domain switchings in the（101）and（111）-PZT epitaxial films are anisotropic.These results indicate that the constructed artificial synaptic model is capable of responding to external electrical and force signals.2.Based on the above artificial synaptic model,the switching of nanoelectric domains in the synaptic gap modulated by spike parameters（pulse voltage amplitude and width）and their domain wall migration properties are also investigated.The results show that the domain wall migration rate along the stripe direction in（111）-PZT films is about twice as high as that in the vertical stripe direction,about twice as high as that in（101）-PZT films,and about 1.4 times as high as that in（001）-PZT films.The bubble domains in（001）-PZT and the nano-twinned domains in（111）-PZT with downward polarization will back-switch within 10 h.The bubble domains and nano-twinned domains with upward polarization can be maintained for at least 30 h.These results suggest that the PZT synaptic gap is capable of exhibiting different behavioral properties in response to different spike parameters.3.By replacing the nanoprobe presynaptic membrane with a micron-scale copper probe,the memory performance of the newly constructed series of artificial synaptic models under electrical stimulation and their synaptic behaviors are studied using a semiconductor analyzer.The results show that the constructed series of artificial synaptic models successfully simulate some plasticity behaviors similar to those of neurosynapses,such as synaptic short time course plasticity and long time course plasticity.In addition,when different forces are applied to the ferroelectric layer with different polarization states in artificial synaptic models,the reading of the resistive states of the artificial synaptic models can ben affected.