Shrink some of the most promising memory materials down to the thickness a microchip needs, and they quietly stop working. The layer that matters goes inert, losing the electrical and magnetic behavior that made the material worth studying. A team led by Lucas Caretta, an assistant professor of engineering at Brown University, reports that it has gotten rid of that “dead layer” in thin films of bismuth ferrite, and Brown says the approach could ease the energy appetite of AI hardware.
Read next on Morning Overview: Cadillac cut the price of its cheapest EV, which goes up to 317 miles on a charge
The claim is about materials, not about finished chips, and no device has been built.
The dead layer in ultrathin multiferroics
Bismuth ferrite is a multiferroic, a material that carries both an electrical polarization and a magnetic order. Brown’s Oct. 28 release explains why thin versions fail: the atomic lattice loses its functional shape and chemistry, and internal electric fields overwhelm the polarization. Caretta said the group had established that the dead layer can be removed. The result appeared in Nature Materials, and trade coverage such as the electronics magazine EFY picked it up in early November.
Ångström-scale thickness is the point. An ångström is a hundred-millionth of a centimeter, and Basic Thinking reports the team stabilized bismuth ferrite at about one ångström. Polarization matters because it can be flipped with small voltage pulses to encode binary data, while the magnetism offers a way to read that data back out.
Strain from dysprosium scandate and a metal layer that soaks up stray fields
The recipe has two parts. The bismuth ferrite is grown on a substrate of dysprosium scandate, whose narrower crystal structure squeezes the film as it forms; that compressive strain helps the film hold on to its functional lattice. A second, metallic layer in the stack absorbs the stray electric fields that would otherwise swamp the polarization. Placing the film on that metal layer is the step that keeps the polarization alive at low thickness. With both in place, the film kept its multiferroic properties and showed enhanced magnetism, and Caretta says it can be made down to a few angstroms thick while staying ferroelectric and magnetic.
The layered structure also turned the bismuth ferrite into an altermagnet, a hybrid of ferromagnet and antiferromagnet. Caretta describes altermagnets as the best of both worlds, because they combine fast switching, dense packing and a strong electrical readout signal. A German-language report on the work, at Basic Thinking, frames altermagnetism as a way to pack memory cells tightly without magnetic interference, and treats the energy and AI-efficiency benefits as projections from the researchers’ perspective rather than measured results for a device.
How much energy for AI chips is actually on the table
Here the careful reading matters. The caveats are plain. Basic Thinking calls it basic research with considerable development time ahead, and no market-ready processor or memory chip exists. Brown’s release says multiferroics could run at much lower power than current devices and could ease AI’s energy demands, but it gives no quantified energy figure. Nobody has published a measured saving for a chip built from these films. A separate team at Chalmers University of Technology, the Indian Institute of Technology Ropar and Uppsala University has claimed that a different two-dimensional magnetic material could cut memory power use by a factor of ten, which Hackster reports, yet that is a different material and a researcher’s estimate, not a result for Brown’s films. The Chalmers-led group’s argument is that dropping the need for power-hungry external magnetic fields is what produces the saving, a mechanism with no counterpart in the Brown paper as summarized by the release. That group also argues, per Hackster, that a single material with several magnetic behaviors avoids interface problems in multilayer stacks and is far easier to manufacture, a pointed contrast with Brown’s approach, which depends on a three-layer stack of substrate, bismuth ferrite and metal. Whether a layered film or a single crystal wins on cost and yield is a question the two groups have not tested against each other.
The Brown work was funded by the U.S. Army Research Office (grant W911NF-21-2-0162), the National Science Foundation (DMR-2039380 and OIA-2327352) and the U.S. Department of Energy (DE-AC02-05CH11231), according to the release. The electronics magazine EFY describes the possible payoff in similar conditional terms, adding that altermagnets may offer faster operation and reduced stray magnetic fields while still producing a detectable readout signal: multiferroics could enable memory and logic devices that consume less power, with polarization states storing bits and magnetic properties handling readout, including for AI computing.
What the dead-layer fix supplies is a stable, ultrathin film, the raw ingredient for memory or logic elements that could switch on small voltage pulses. Turning that film into working memory cells, measuring what they draw per operation and showing they can be manufactured at scale are the steps still ahead, and they will decide whether “far less energy” becomes a number.
This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.
More from Morning Overview
- 11 engines built to run well past 200,000 miles
- Two Kentucky residents died when a small plane crashed at Missouri’s Lake of the Ozarks, and the NTSB is investigating
- A handful of car transmissions are so tough that mechanics say they almost never die
- Melting permafrost is setting off a newly identified kind of catastrophic landslide, scientists say
