Interactions and Fast Dynamics in Magnetic Nanoparticles and Magnetic Tunnel Junction Based Spintronics

Nanometer-sized magnets have been widely used in the semiconductor industry and medical treatment. The delivery and performance for such applications need thorough knowledge about the material properties, device fabrication and underlying physics dictating the evolution of the system. I worked on two kinds of nanomagnets, the chemically synthesized magnetic nanoparticles (NPs) and the thin film-based spintronics made by nanofabrication. I used Electron Microscopes to learn the geometry and structure of the system and improvements were made for the application purposes. I also used magnetometries to study the macroscopic picture of the magnetic response, at a relatively long measurement timescale (1 − 10² 𝑠). To measure the small timescale (𝑛𝑠 − 𝜇𝑠 ) response, I did scanning probe measurements to get the real time electric signal, which can directly measure the time evolution of a single device. I then applied the theory of nanomagnet interactions and the algorithms for computing to make the nanomagnets into functional assemblies. 

In Chapter 1, I studied the spin configuration for the NP assemblies using polarization analyzed small angle neutron scattering (PASANS). I synthesized the manganese ferrite (𝑀𝑛𝐹𝑒₂𝑂₄) NPs using a modified hot injection method. The 𝑀𝑛𝐹𝑒₂𝑂₄ NPs had a uniform 𝑀𝑛²⁺ and 𝐹𝑒³⁺ ion distribution, comparing to the traditional synthesis method where a core-shell structure would be induced by the decomposition temperature difference of the precursors. The NPs were made into FCC packed nanocrystals and put into the neutron beamline to collect the scattering data. I developed a numerical method based on the 3-D Fourier transformation method to simulate the single particle form factor and the structure factor of the FCC packaging. This method could be used to select the possible spin configurations in the NPs to fit the experimental measured data. The (111) Bragg peak of the Cobalt Ferrite (𝐶𝑜𝐹𝑒₂𝑂₄) and Magnetite (𝐹𝑒₃𝑂₄) NPs were studied. 

In Chapter 2, I synthesized the 𝐹𝑒₃𝑂₄ nanoparticles using a two-step growth method and the 𝐹𝑒₃𝑂₄ NPs were diluted in the eicosane to measure the susceptibility, under a field consisting of both AC and DC components. This gives information about the heat delivery of the NPs when and AC magnetic field is used to drive the NP assembly. The system was analyzed based on the picture of an interacting assembly that has a distribution of energy barriers. 

In Chapter 3, I fabricated magnetic tunnel junctions (MTJs) with shape anisotropy and measured the tunnel magnetoresistance (TMR) using conductive atomic force microscope (cAFM) in a field with both hard axis and easy axis component. The devices were hardwired to prevent side wall oxidation. The Stoner-Wohlfarth model was applied to the analysis of the field dependent hysteresis loops to estimate the shape anisotropy energy barrier. It was found that a field along the hard axis direction can reduce the energy barrier and make the device telegraph. Assisted with the external field, the MTJs could be driven to an unstable state and work as a probabilistic bit to carry out probabilistic computing algorithms. 

In Chapter 4, I built a CMOS circuit to transform the TMR from the parallel and antiparallel states in a MTJ to the computable voltage levels in a circuit. The circuit has a voltage controlled probabilistic output. The output of the circuit is fluctuating between 0 (voltage = 0 𝑉) and 1 (voltage = 5 𝑉) and the fraction of the 0’s and 1’s is controlled through a sigmoidal function. Spin transfer torque is used as the underlying physics to drive the device from preferring one state to another. I then built the stochastic logic gates based on the theory of the Boltzmann machine. Feedback was introduced from the outputs of the blocks of the CMOS circuits to their inputs, which selected the states of the system to fit the truth table of the logic gates. The stochastic NOT gate was demonstrated. The feedback mechanism and the invertibility of the stochastic AND gate were then discussed. 

In Chapter 5, I used MuMax3 software package to simulate the simplest case in a cellular automata (CA), a closely packed dot chain. The signal propagation, represented by the cascade flipping of the dots, was simulated while an excitation was injected from one side of the chain through a spin nucleation. I first presented the logic to design a system that could be used to carry out logic operation. The coercivities and stray fields for the single dot building blocks were simulated and the properties for the dots were matched for their own functions in the dot chain. I then simulated the dynamics for a dot chain containing 30 dots with thermal fluctuation to understand the interaction between the dots and the energy diagram landscape for a closely packed nanomagnet system. This provides guidelines for the design of the spin ice based computing devices. 

In Chapter 6, I proposed a method to detect the skyrmion in the racetrack. While skyrmions can be used as 0’s and 1’s in the racetrack for computing purposes, it is important to detect them with an electrical read out. I installed a magnetic detector on the racetrack to read out the TMR and determine the core direction and the chirality of a skyrmion. I first used COMSOL Multiphysics to address several concerns in the strategy, the current spreading which could wash out the TMR signal, the disturbance of the current used to drive the skyrmion to move and the working voltages for such a racetrack device. After validating the current distribution, I simulated the spin dynamics using Mumax3 software package and calculated the TMR value based on a finite element method. Despite the observed small trajectory distortion from the stray field and the thermal fluctuation, the core direction can be detected using a perpendicular magnetized circular dot and the chirality can be detected using a rectangular synthetic antiferromagnet (SAF) detector.