Phase separation, similar to how oil and water separate, works in tandem with oxygen diffusion to assist memristors—electronic components that maintain information through electrical resistance—retain data after power is turned off, according to a new study.
Phase separation, akin to the way oil and water don’t mix, collaborates with oxygen diffusion to enable memristors—devices that use electrical resistance to store information—to hold onto data even when the power is cut off, as indicated by a study led by the University of Michigan and published in Matter.
Until now, the exact mechanism behind how memristors retain information without needing power (known as nonvolatile memory) has not been completely understood, as observed models and experimental results have often not correlated.
“Experiments show that these devices can preserve information for over a decade, yet existing models suggest that they can only hold data for a few hours,” stated Jingxian Li, a doctoral graduate in materials science and engineering from U-M and the study’s primary author.
To enhance comprehension of the nonvolatile memristor memory phenomenon, the researchers examined a type of device called resistive random access memory (RRAM), which serves as an alternative to the volatile RAM used in traditional computing and shows great promise for energy-efficient artificial intelligence applications.
The specific RRAM analyzed was a filament-type valence change memory (VCM), which consists of an insulating layer of tantalum oxide sandwiched between two platinum electrodes. By applying a certain voltage to the platinum electrodes, a conductive filament is created, forming a bridge of tantalum ions that allows electrical flow, placing the system in a low-resistance state that represents a “1” in binary. Conversely, when a different voltage is applied, the filament dissolves as returning oxygen atoms react with the tantalum ions, effectively “rusting” the conductive bridge and reverting it to a high-resistance state, symbolizing a binary “0.”
Initially, it was believed that RRAM retained information over time because oxygen diffusion was too slow. However, a new series of experiments uncovered the importance of phase separation, which had been overlooked in prior models.
“In these devices, oxygen ions tend to avoid the filament and will not diffuse back, even after an extended time. This behavior is comparable to how oil and water never mix, regardless of how much time passes, due to lower energy levels in a separated state,” explained Yiyang Li, an assistant professor of materials science and engineering at U-M and the study’s senior author.
To investigate the retention time, researchers accelerated the experiments by raising the temperature. For instance, an hour at 250°C mimics nearly 100 years at 85°C—the standard operating temperature for computer chips.
Utilizing high-resolution imaging techniques through atomic force microscopy, the researchers captured images of filaments that are just about five nanometers wide, or about 20 atoms thick, within the one micron wide RRAM device.
“Finding the filament in the device was surprising; it was akin to locating a needle in a haystack,” Li remarked.
The team discovered that filaments of varying sizes exhibited different retention characteristics. Filaments smaller than approximately 5 nanometers decayed over time, while those larger than 5 nanometers became more robust. This size-related variation cannot be solely attributed to diffusion.
When combined, the experimental outcomes and models that included thermodynamic principles indicated that the formation and stability of the conductive filaments are influenced by phase separation.
The research team utilized phase separation to enhance memory retention from a single day to over 10 years in a radiation-resistant memory chip—a device designed to endure radiation exposure, suitable for space missions.
Potential applications include in-memory computing for more energy-efficient AI solutions and memory devices for electronic skin—a flexible electronic layer designed to replicate human skin’s sensory functions. This e-skin could potentially offer sensory feedback for prosthetic limbs, contribute to new wearable fitness trackers, or assist robots in developing tactile sensing for delicate tasks.
“We aspire for our discoveries to inspire innovative uses of phase separation in crafting information storage devices,” Li added.
Collaborators on this study included researchers from Ford Research, Oak Ridge National Laboratory, University at Albany, NY CREATES, Sandia National Laboratories, and Arizona State University. The device was fabricated in the Lurie Nanofabrication Facility and analyzed at the Michigan Center for Materials Characterization. The University of Michigan’s work was primarily funded by the National Science Foundation (ECCS-2106225).
