The von Neumann structure, which has been previously used for data storage and processing, sequentially executes input arithmetic commands, and thus is inefficient due to a delay in the memory device. To solve this problem, interest in neuromorphic computing that can achieve high speed and low power by emulating the neural network structure of the brain is growing. Many researchers are actively studying the neuromorphic properties of next-generation memories, such as resistive switching memory, phase change memory, ferroelectric memory, and magnetoresistive memory. Among them, the resistance switching memory has advantages such as a fast operation speed, high integration, low power operation. Meanwhile, in order to emulate the brain that stores and processes information in a parallel way, devices that act as synapses and neurons are needed.
In this thesis, a method to control the local volatility in one ECM device by a laser annealing method is demonstrated for application to future one-chip neuromorphic system. In addition, neurons and synapses were simulated simultaneously. As a result of irradiating the excimer laser to the silver filament-based volatile ECM device, the data retention time increased from 0.6 ms to 8250 s, indicating local modulation to the non-volatile memory. The redistribution of silver nanoparticles in SiO$_2$ matrix, which is the cause of the volatility change, was analyzed through TEM, XPS, and ToF-SIMS. Both synapses and neurons were simulated by observing Integrate-and-fire and Spiking-timing-dependent plasticity phenomena with volatile and non-volatile devices, respectively. Furthermore, concomitant plasticity, a phenomenon that occurs in a neural network where neurons and synapses coexist, was confirmed through local volatility control in one device, showing the potential for implementing simple structured memristive neural networks with volatile tunable ECM memristors.