Texas A&M Discovers New Circuit Element: The Meminductor
The physical realization of the third “mem” component gives designers access to new elements for the next generation of electronics.
Texas A&M researchers recently demonstrated "meminductance," forming the foundation of a new circuit element: the meminductor. Previous researchers have demonstrated both memresistance and memcapacitance, making the contribution from Texas A&M a major leap in the scientific community.
Memristors and memcapacitors are well-researched "mem" variations of basic circuit elements. Now, Texas A&M has created the meminductor.
The “mem” variations of basic circuit elements, while they may not be common on hobbyist breadboards, have demonstrated considerable utility in computing and AI/ML applications. The memristor is arguably the most common today due to its earlier discovery (2008 versus 2019 for memcapacitors). But as more is understood about the components, their utility may grow exponentially.
In this article, we'll highlight the Texas A&M research to show designers how the meminductive behavior was discovered. We'll also discuss how the completed trinity of “mem” devices could prove useful in the future.
Purely Passive Memory
The “mem” prefix indicates that a circuit element incorporates some form of memory. And while the memory isn’t of the random-access or read-only variety, it does allow for unique properties to be leveraged in emerging applications.
Memristors, for example, have seen a myriad of uses in recent research. In one such example, memristors were used for image processing inspired by mammalian pattern recognition. Memristive devices have also been used to realize compute-in-memory architectures, where a central processor no longer performs calculations thanks to the variability in the device’s resistance.
This image of various inductors highlights the distinguishing features of meminductors, showing how the inductance can change over time and the impact that it has on flux. The “figure-8” shown in (c-d) is one of the fingerprints of a meminductor. Image used courtesy of Nature Scientific Reports
In simple terms, “mem” devices demonstrate characteristics (resistance, capacitance, inductance) that can change depending on their previous state. In this way, the element has “memory,” setting it apart from non-mem devices that are independent of previous states.
Hiding Behind Resistance
While the meminductor has been theorized for some time, definitive proof of a true two-terminal device has yet to be observed prior to the Texas A&M group’s recent findings. This is because series resistance effectively obscures the meminductive properties, especially at low frequencies when the desired effect is strongest.
To negate the effects of series resistance, the group adopted a clever technique that effectively subtracted the effects of series resistance on the device’s operation in order to isolate the effects of meminductance. Since resistance can be readily measured and ideally doesn’t change with frequency, this makes the job of homing in on meminductance nearly trivial.
The simulated and experimental setups for the meminductor demonstrated the ability to remove the effects of resistance and provided experimental evidence of meminductance. Image used courtesy of Nature Scientific Reports
To create an experimental meminductor, the Texas A&M group needed a mechanism that could passively modify inductance in relation to the applied current. The team placed an air-wound coil on a rod that partially contained a ferromagnetic material between two magnets. In this setup [GIF linked], as the current through the coil changed, the coil moved in relation to the ferromagnetic rod, changing the inductance.
The results from the experimental setup (shown under Supplementary Information in the paper) demonstrate that by subtracting the well-known effect of series resistance, the meminductive properties can be observed, providing experimental proof of passive, two-terminal meminductance.
Completed Table of Circuit Elements
The new circuit element may consume less power in reactive devices and provide more efficient computing. Using this new component, AI and ML fields may benefit from improved neuromorphic computing, allowing for hardware-enabled performance increases. The high-performance computing field may also benefit from the programmable properties of the meminductor, enabling complex or high-density calculations directly in memory without requiring a massive computational load on the CPU.
The contributions of the Texas A&M group complete Chua’s “Periodic Table of Two-Terminal Circuit Elements” with the creation of the meminductor. Image used courtesy of Nature Scientific Reports
The physical realization of the meminductor has given each of the three basic circuit elements its mem-counterpart. This research development comes at an opportune time as Moore's law is stretched to its limits. As engineers extend the momentum that has propelled the developments of the last century, the meminductor could help play a role in continued innovation.
An important to note that “discovered” does not equal “first theorized.” Memristors were first theorized to exist in 1971 by Leon Chua, but it wasn’t until 2008 (as the article says) that they were “discovered” or physically realized (or implemented). Meminductors were theorized to exist in 2009 (also by Chua, alongside predictions of the memcapacitor), and evidence of their existence demonstrated in 2014 in Pt/YIG crossbars (spin currents, etc). The basic idea of the version of the meminductor demonstrated by this group isn’t new either, I might add (First paragraph of Sec. V in https://journals.aps.org/prresearch/abstract/10.1103/PhysRevResearch.3.013135 mentions several possible mechanisms of meminductance - the paper should be open access, though the rest of the paper is irrelevant to the nature of meminductors). The Texas group DID have the novel idea to make it so the coil itself wasn’t fixed, so that’s neat.