Lithium niobium oxide multifunctional materials and applications in neuromorphic computing

Abstract

This work explores the growth fundamentals and multifunctional applications of materials in the Li-Nb-O family with specific focus on the memristive applications of LiNbO2 as a synaptic analogue in neuromorphic computing architectures. Initial studies include the development of a flux versus temperature growth phase diagram for lithium niobium oxides using molecular beam epitaxy at high substrate temperatures. Using this growth understanding, various multifunctional materials were epitaxially grown and characterized by structural, chemical, and morphological methods. The optical and electrical characteristics of Li1-xNbO2 were also investigated as a function of lithium stoichiometry. LiNbO2 samples were then fabricated into devices for use in neuromorphic computing, specifically memristors and batteries. Among other important studies, the use of Li-alloying contacts is explored as a method to induce non-volatile behavior in natively volatile LiNbO2 memristors, a feature critical for neuromorphic behavior. The findings suggest that LiNbO2 can exhibit memristive resistance changes in an analog manner which show tunable timescale ranges appropriate for biologically realistic synaptic behavior. The current and future state of memristors in neuromorphic computing is discussed, focusing on the role of volatile decay and short term effects in biological systems.