Spin-Orbit Torque-Driven Artificial Devices and Reconfigurable Logic gates by Anomalous Hall Effect in W/CoFeB/MgO Frame

Abstract

Spintronic devices employ various techniques for data storage, reading, and writing. Approaches encompass leveraging the PMA (Perpendicular Magnetic Anisotropy) property for recording magnetic domains, utilizing phenomena like TMR (Tunnel Magnetoresistance) and AHE (Anomalous Hall Effect) for reading resistance and Hall voltage, as well as employing methods such as STT (Spin-Transfer Torque) and SOT (Spin- Orbit Torque) for current-based writing. The phenomena and techniques are applied to create a diverse range of spintronic components, including memory with applications in neuromorphic computing, reconfigurable logic gates, multiple devices, adders, subtractors, and more. In this thesis, I demonstrate the implementation of various spintronic components by harnessing the SOT and domain wall motion phenomena to manipulate magnetic domains within a structure composed of W/CoFeB/MgO/Ta with PMA characteristics. The approach enables us to read Hall voltage using the AHE phenomenon. In chapter 2, I show synapse and neuron devices of neuromorphic computing. I finely adjust the up and down magnetization ratios within the Hall voltage detection area, using Stripe domain characteristics, to achieve perfectly linear and symmetric synaptic components. To control the parameters of the sigmoid function, I precisely adjust the x0 and k parameters by modulating the external magnetic field intensity and domain wall starting positions. Furthermore, I integrate and operate two synapses and one neuron, allowing the two synapses to learn and the neuron to integrate and fire, ultimately recognizing two different patterns. Finally, I achieve an accuracy of over 90% in MNIST simulation, demonstrating the robust recognition capabilities of our spintronic neuromorphic devices. In chapter 3, I focus on the development of 8 reconfigurable logic gates (NAND/AND/NOR/OR/Converse INH/Converse IMP/INH/IMP). I achieve the reconfigurable logic gates using a 2-channel crossbar array with integrated Hall voltage differences. In each channel, I induce different directions of magnetic domain switching to generate Boolean signals by voltage. The devices leverage the non-volatile property of magnetic domains, allowing for asynchronous operation. By reconfiguring the devices, initially by reversing the direction of the external magnetic field, and subsequently by utilizing inverting and non-inverting comparators, I produce distinct logic outputs. Finally, I configure the 8 reconfigurable logic gates (NAND/AND/NOR/OR/Converse INH/Converse IMP/INH/IMP) by modifying the direction of the read current. In chapter 4, I delve into the control of substrate roughness through ion-milling etching, resulting in alterations in PMA characteristics and subsequently impacting the switching current density of SOT. The changes enable the creation of 8 different states, which are further reflected in variations in Hall resistance. Leveraging the advancements, I develop a multi-level spin device and show linear synaptic characteristics. Finally, I demonstrate the control of the Hall voltage and Hall resistance resulting from the AHE by switching of magnetic domain walls induced by SOT current. Devices are applied to artificial intelligence, eight reconfigurable logic devices, and multi-state spin devices. The thesis laid the groundwork not only for demonstrating various possibilities of spintronics devices using SOT but also for providing a foundation that can be industrially realized through the explanation of the behavior of the free layer in future research on MTJ structures.