Inspired by the architecture and efficiency of the human brain, neuromorphic systems aim to integrate memory and processing within the same physical location, enabling low-power and energy-efficient computation. In this context, memristive devices capable of storing and updating analog conductance states are seen as key building blocks for implementing artificial synapses and neurons in hardware. Recently, the potential of the TiN/La₂NiO_4+δ/Pt memristive devices for artificial synapse applications was demonstrated. La₂NiO_4+δ (L2NO4) exhibits oxygen over-stoichiometry and can accommodate a wide range of excess oxygen content (δ). The presence of highly mobile interstitial oxygen ions in L2NO4, combined with the TiN electrode, facilitates the formation of a TiNxOy interlayer at the metal/oxide interface, which plays a key role in the device operation. This thesis presents a comprehensive study of TiN/L2NO4/Pt memristive devices, focusing on their application as artificial synapses in spiking neural networks (SNN). The L2NO4 thin films are grown by pulsed injection metal-organic chemical vapor deposition (PI-MOCVD) at 600 °C, followed by the microfabrication of the memristive devices and extensive structural characterization. Furthermore, electrical characterization is performed to evaluate the key metrics of resistive switching behavior: retention, endurance, and variability. Importantly, the potential of the devices for analog computing is assessed through controlled multilevel switching and emulation of spike-timing-dependent plasticity (STDP). In addition, the ability to tune analog properties through control of the oxygen content in the L2NO4 layer is analyzed. In the next stage, to explore integration scalability, the devices are fabricated in both shared bottom electrode and cross-point architectures, with lateral dimensions scaled down to 5×5 µm². The influence of fabrication strategies, particularly continuous and etched oxide layers, on device variability, forming voltage, switching kinetics, and analog performance is systematically studied. Finite element modeling is used to simulate electric potential and Joule heating distributions across the device stack, shedding light on the thermoelectric environment that governs the switching behavior. Furthermore, a detailed investigation of the resistive switching mechanisms of the TiN/L2NO4/Pt devices is carried out. A number of advanced

Membres du jury/ Jury members :