Zimmerman, Jeramy D.Warren, Emily L.Schneble, Olivia D.2024-07-032024-07-032023https://hdl.handle.net/11124/179107Includes bibliographical references.2023 Fall.Neuromorphic computing aims to mimic the principles or functions of biological brains for improved speed and energy efficiency in computing. The popularity of bio-inspired neural network algorithms has already begun to show the potential improvements in speed and accuracy when processing large sets of data or performing abstract tasks, but improvements in energy efficiency only come from biomimetic hardware. With increasing demand for computing resources, increasing energy efficiency is a crucial goal. Developing neuromorphic hardware that approaches the spatial density and energy efficiency needed will require investment in new materials and fabrication techniques that will supplement or replace existing complementary metal-oxide-semiconductor technologies. These efforts must also consider scalability and repeatability for eventual large artificial neural network implementations. This work first presents a novel technique for the growth of III-V semiconductors on silicon that would enable their use for optoelectronic neuromorphic devices. Templated liquid-phase epitaxy is a scalable and low-cost technique that bypasses the need for organometallic precursors and can significantly improve material utilization. We focus on heteroepitaxial growth, as control of the crystallographic orientation is critical for device fabrication. These improvements make III-V semiconductors more applicable to low-cost device applications. The rest of this work will describe artificial neuron devices implemented in insulator-metal transition (IMT) materials. First, we use circuit simulations to evaluate the impact of thermal and electronic properties on the behavior of volatile neuron-like devices. By independently varying material parameters, we identify control of the IMT as a fundamental task for device engineering. Second, we discuss the growth of rare-earth nickelates and their application to artificial neuron devices. We show that these devices can be driven electrically via resistive heating consistent with the thermally-driven behavior of the film, and highlight some challenges in the implementation of neuronal devices.born digitaldoctoral dissertationsengCopyright of the original work is retained by the author.Text2024-06-25