Have you ever wondered how your brain effortlessly recognizes faces, understands language, makes quick decisions, or learns from past experiences? These seemingly simple tasks pose significant challenges for conventional computers, which consume considerable energy handling precise numerical calculations step-by-step. The brain’s remarkable efficiency arises from its ability to perform numerous parallel computations simultaneously, often with lower precision. This observation raises an intriguing question: Can artificial systems replicate the brain’s impressive efficiency?
Neuromorphic computing tackles exactly this challenge. This emerging technology aims to mimic the brain’s structure and function, specifically its essential components—neurons and synapses. Neurons are tiny processing units interconnected by synapses that transmit electrical signals known as spikes. Remarkably, the human brain accomplishes all these tasks while consuming only about 20 watts of energy—comparable to a dim light bulb—demonstrating exceptional efficiency. By replicating these biological interactions, neuromorphic computing promises substantial improvements in computational power and energy efficiency, potentially leading to powerful yet sustainable solutions.
One promising technology for neuromorphic computing is spintronics, which utilizes the magnetic spins of electrons to build innovative devices. There are several ways to use spintronic devices for neuromorphic computing. I will talk about my favorite: magnetic skyrmions! Imagine a skyrmion as a tiny magnetic vortex, where electron spins align into stable, swirling patterns resembling miniature whirlpools or galaxies, but in a magnetic material. This intricate configuration gives skyrmions remarkable stability, allowing them to be reliably manipulated by electrical currents at nanoscale dimensions (see figure for a visual example).
Biological neurons can be effectively modeled using RC circuits—basic electronic components combining resistance (R) and capacitance (C) to control electrical signals. These circuits capture the essential dynamics of neurons through the Leaky-Integrate-Fire (LIF) model, which is widely adopted in neuroscience to describe neuron behavior. Remarkably, skyrmions can replicate these neuronal dynamics at a nanoscale. By placing a skyrmion within a specially engineered magnetic environment where the perpendicular magnetic anisotropy (PMA)—which defines how strongly spins prefer aligning perpendicular to the surface—varies quadratically, the skyrmion moves predictably under electrical stimulation. The applied current induces a force that competes with the force due to the gradient, causing the skyrmion to shift its position, similar to how voltage accumulates in a capacitor during charging. Once the current ceases, the skyrmion naturally returns to its original position, effectively simulating the discharge phase of a capacitor. This motion directly mirrors the LIF neuron’s behavior, which shows that skyrmions have the potential to efficiently emulate biological neuron dynamics.
Scaling these skyrmion-based neuromorphic devices involves overcoming challenges in precise fabrication, maintaining stability, and integrating them into existing technologies. Despite these challenges, skyrmions might have a crucial role in neuromorphic computing. So, can we emulate the brain? Well, at least NOT YET :((. However, scientists have been able to emulate individual neurons with skyrmions (and other spintronic devices). Given that the brain consists of billions upon billions of interconnected neurons, I guess we could say, “Let’s take one neuron at a time… one neuron at a time.”
