Studying the potential of memristors

By Caitlin Gittins, Editor, Electronic Specifier

Juston is an undergraduate at Embry-Riddle Aeronautical University’s Prescott, Arizona, campus. He admitted when I spoke to him that he didn’t set out to be an aerospace engineer, per se: “I went to Embry-Riddle originally to become a pilot, but halfway through I switched to engineering and took all [of] the basic engineering classes."

His memory of being taught by Dr. John Sevic, Associate Professor of Electrical Engineering at the university led him to enquire about ongoing research projects that Dr. Sevic was running. This led him down the path to memristor technology.

“I had no idea what memristors even were,” he confessed. “Before I started this project, I was supposed to be using [Multiphysics] phase-field modelling to model snowflakes and/or melting ice.”

A memristor is a two-terminal electronic component that regulates the flow of electrical current within a circuit and remembers the amount of charge that has passed through it. As a result, it functions as non-volatile memory, whereby it preserves information without needing power. It functions with two resistance states: a ‘1’ and a ‘0’, a low resistance state and a high resistance state.

“What is very important about a memristor is that pattern is repeatable, and it is path dependent,” explained Juston. “That’s why we say it has hysteresis: that means if you were to apply voltage in an increasing manner, it goes in a straight line or a path. But as you take it [voltage] out, it won’t go back, it will go a different path downwards.”

Their research focused on exploring the result of using conducting filaments (CFs) with memristor technology, as they found that without CFs, memristors are high resistance; but with CFs, it becomes conductive. “It’s like resistive switching,” added Juston.

Wider implications

The implication of using memristors for memory storage is that it can store more memory in a smaller package.

“That has been the limitation of current architecture, because if you create a smaller package, at some point you experience quantum effects and it’s not as reliable,” said Juston. “More memory is good everywhere.”

This is good news not just for AI applications, but also for computing capabilities in general.

To test their theory, Juston used computer simulations, but he acknowledged that an existing limitation has been that they haven’t been able to directly validate results using physical memristors – as manufacturing memristors is costly and is a capability Embry-Riddle hopes to add in the near future, funding dependent.

“The benefits of [using] computer simulations are it’s easy to do, but the most useful thing is that you can test lots of things quickly, and you don’t have the cost of manufacturing that memristor,” he said. “But as a computer, it’s only as good as what you tell it to do. We can change the numbers here and there, but we have to validate everything … Most of the research I did was helping [with] validation.”

Besides potentially enabling the creation of more memory in a smaller package, Juston’s research also has implications for Moore’s Law: a prediction revised in the 1970s by Gordon Moore who said every two years, the number of transistors on a microchip doubled, translating to computational power doubling.

However, advancements in the size of transistors shrinking in the 21st century has led some to speculate whether Moore’s Law has reached its constraints.

“Transistors are small to the point where quantum efforts are a huge constraining factor, and we can’t make them smaller, we’re just stacking them up for now,” shared Juston. “The idea of this research is creating next-generation memory where we can make them smaller so we can get around those quantum effects that would restrain us in traditional architecture, so we can keep doubling [the number of] transistors.”

Looking to the future, Juston said that he expected their research to move on from validating their work, to summarising it. “We’re going to use our application as a method to simulate memristors, that’s what we’re working on currently,” he said. “The key thing I appreciate about working with Ambroise is that on one hand we’re doing semiconductor physics here in my group, and on the other hand this is at an aeronautical university, and he has bridged the gap very well,” said Dr. Sevic. “I think this process he’s learned scales to just about anything he wants to do for the rest of his life.”

This article originally appeared in the January'25 magazine issue of Electronic Specifier Design – see ES's Magazine Archives for more featured publications.