Researchers at the University of California San Diego have demonstrated a way to control how an experimental electronic device forms and erases memory, using electrical currents to change the behavior of a material called vanadium dioxide.
The device, known as a neuristor, mimics certain electrical switching behaviors of biological neurons. The researchers found that they could influence whether its switching became more predictable or returned to a less predictable state.
The discovery could help scientists design brain-inspired computing hardware that stores information through changes in its physical structure. Such systems aim to perform some computing and memory functions more efficiently than conventional approaches that move data between separate processors and memory units.
Led by physics professor Oleg Shpyrko, the UC San Diego team used coherent X-ray diffraction to observe nanoscale changes inside the material while electrical current passed through the device. This allowed them to investigate how its internal structure affected its switching behavior.
Electrical pulses reshape memory
Vanadium dioxide can undergo a transition between electrically insulating and conducting states. This change is accompanied by structural changes in the material, making it useful for studying electronic devices that imitate some features of neurons.
In the experiments, researchers examined how repeated electrical stimulation affected the arrangement and movement of tiny regions within the material. These regions, called domains, can interact as the device switches between states.
The team identified a process known as jamming, in which the movement of these regions becomes increasingly restricted. Rather than behaving independently, their interactions begin to constrain how the material changes.
Repeated electrical stimulation above a threshold drove the material toward this jammed state. As a result, the device’s subsequent switching became more predictable, indicating that its earlier electrical activity had influenced its future behavior.
The researchers also found that lower-current stimulation could reverse this effect, restoring a less predictable switching state. This provided a way to control the device’s memory-like behavior through electrical inputs.
The experiments tracked structural changes over periods lasting thousands of seconds, allowing the researchers to connect the device’s electrical history with changes inside the material.
Controlling when memory fades
The findings offer a physical explanation for how a neuristor can retain information about previous electrical activity without requiring a separate conventional memory component.
In this case, learning does not mean that the device understands information or independently solves problems. It refers to a measurable change in switching behavior caused by previous stimulation. Forgetting describes the reversal of that change.
This distinction matters for neuromorphic computing, which seeks to build electronic systems inspired by how neurons and synapses process and retain information.
A device capable of adjusting its response based on earlier electrical signals could eventually contribute to circuits that combine memory and computation. However, the researchers have demonstrated the mechanism in an experimental material system, not a complete brain-inspired processor.
Further work would be needed to determine how reliably the behavior can be controlled across multiple devices, how quickly memory states can be changed and whether the approach delivers measurable energy savings in practical circuits.
The research also provides a method for studying how nanoscale structural changes influence electronic switching in materials with coupled structural and electrical transitions.
The findings were published in Nature Communications.