Hafnia films below 10 nanometers open path to denser memory

Hafnia, the insulating material already sitting inside every silicon-based transistor, keeps its useful electrical behavior at thicknesses below 10 nanometers — a result that could let chipmakers build denser memory without adding new components. A team led by Rohan Mishra, associate professor of mechanical engineering and materials science at Washington University in St. Louis, published the finding in Science on 24 September 2026.

The work addresses a structural limit. A human hair is roughly 80,000 nanometers wide, and most ferroelectric materials lose their switchable polarization once they are made thinner than 10 nanometers. Researchers in Germany previously found that hafnia (HfO2) breaks that rule: the material, which is not ferroelectric in bulk, becomes ferroelectric when grown as an extremely thin film. "It was a puzzle we are trying to understand," Mishra said.

How the team cracked the hafnia puzzle

Mishra's group attacked the problem from two directions. Xiaoshan Xu's team at the University of Nebraska-Lincoln grew well-ordered, single-crystal films of lanthanum-doped hafnia on a substrate. Guodong Ren, who completed his doctorate under Mishra and is now a postdoctoral researcher at the University of Washington, then confirmed the single-crystal structure with atomic-resolution microscopy that resolves individual atomic columns.

Electron microscopy measurements of the crystals' phase and behavior matched the team's quantum-mechanical models. That agreement between prediction and experiment is the study's core contribution.

The practical takeaway, Mishra said, is that engineers should stop trying to force hafnia into a conventional ferroelectric phase. "Instead of fighting nature, we can grow high-quality antiferroelectric films and we can then switch them during operation," he said. Antiferroelectrics also store large amounts of energy, which points toward use in high-energy-density capacitors.

What did the zirconia study find?

A second study, co-led by Mishra with Asif Khan at Georgia Tech and Jun Hee Lee at UNIST in South Korea, appeared in Advanced Materials. Ren is a co-first author. The team observed ferroelectric domains in zirconia (ZrO2), a close chemical analogue of hafnia, with footprints smaller than one square nanometer.

Inside each domain, electrical polarization points in a particular direction. Extra oxygen atoms balance the electrical charge at the boundaries between domains. The group's calculations predicted that these domain walls can conduct electricity and move under an electric bias. That result suggests atomically thin walls could one day serve as memory-storage components in their own right.

Why modeling and microscopy travel together

Mishra's group pairs atomic-scale modeling — tracking the stability and properties of ensembles of atoms — with microscopy of real synthesized samples. "When you synthesize real materials, they always have defects, so our work was to understand how to minimize these defects so we can grow a high-quality film," Mishra said.

The pairing matters because microscopy alone shows only a flat projection. "Microscopy is a two-dimensional projection of a three-dimensional structure, so it helps us retrieve the real structure of the materials we study using a combination of microscopy and quantum mechanical calculations," Mishra said.

Ren's ability to bridge both sides is rare. "Guodong is an expert in combining theory and microscopy, one of the very few people who can do it at that level," Mishra said. Ren described the group's advantage in similar terms: "We have a unique chance to combine theoretical studies and also microscopy experiments at the atomic level so we can gain a better understanding about different material systems."

What this means for chip manufacturing

The National Science Foundation funded the work at WashU under awards DMR-214579 and DMR-2122070, with additional collaboration from Georgia Tech, the University of Nebraska-Lincoln, UNIST and other institutions. Collaborators spanned synthesis, modeling and characterization.

The commercial appeal is straightforward. Hafnia and zirconia are already present in silicon chips in production today, so manufacturers would not need to introduce new materials or components to exploit the effect. The demonstrated control at sub-10-nanometer thickness and sub-square-nanometer domain sizes suggests a route to higher memory density within existing process flows. Both studies remain laboratory demonstrations on single-crystal films, and any transfer to production wafers will depend on whether defect-controlled growth of these phases scales beyond the research setting.