Year: 2023
Type: dissertation
Abstract: Inherently nonlinear, two-terminal circuit elements with state-dependent electrical properties are collectively termed “memelements” and include the memristor, the memcapacitor, and the meminductor. The first intentional memristor and memcapacitor were physically realized in 2008 and 2019 respectively, and in this work, I report a two-terminal passive element bearing the three fingerprints of a meminductor thus providing the first physical evidence of meminductance. I discuss the role of series resistance as a parasitic component which obscures potential meminductive behavior and present a technique to extract meminductance in resistance dominated systems by removing parasitic resistance as “resistive flux”. \nMemelements are capable of being biased at “locally active” steady states and may therefore generate persistent dynamical response to a constant excitation when passively coupled to an appropriate passive network. Reports on mathematical proof of the existence of such locally active steady states in memristors and experimental evidence of resultant persistent oscillations can be found in the literature. However, similar results for a meminductor are currently missing due to the absence of its physical realization until recently and an associated unavailability of realistic compact models. In this work, I tackle this challenge by fabricating a volatile meminductor whose small signal model mathematically identifies locally active bias points where the meminductor operates at the “edge-of-chaos (EOC)”. I present experimental evidence of persistent oscillatory behavior when this EOC meminductor is coupled to a passive network and explain the rationale behind the choice of the coupling elements. I show that local activity/passivity of each bias point depends on the meminductor's material properties and geometry and quantify this dependence through the parameters of a “circle of inversion (COI)”. Further, I show the emergence of the COI to be a consequence of state-dependent inductance, thus highlighting that local activity- and thereby, complexity- cannot emerge without “mem”-behavior. Finally, I present results demonstrating that the coupled meminductor system emulates second order neuronal behavior in its dynamical response thus experimentally affirming the potential of meminductors in neuromorphic computing applications.
Author: Dinavahi, Abhiram
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7002042127