started · updated
Brown University researchers develop energy-efficient multiferroic material for AI chips
Researchers at Brown University have developed a technique to overcome the “dead layer” effect, a major obstacle in using multiferroic materials for modern computer chips. Previously, when these materials were made thin enough for semiconductor integration, they lost the essential electrical and magnetic properties required for efficient computing.
The team engineered ultra-thin films of bismuth ferrite by depositing them onto a dysprosium scandate foundation. This method uses atomic mismatch to compress the film, preserving its crystal structure, while an additional metallic layer neutralizes internal electric fields that would otherwise erode electrical polarization.
This breakthrough allows the material to remain ferroelectric and magnetic even at thicknesses of just a few angstroms. The resulting films exhibit altermagnetism, which combines the rapid operation of antiferromagnets with the strong signals of ferromagnets. This development could lead to hardware that consumes significantly less energy, specifically addressing the high power demands of artificial intelligence workloads. The findings were published in the journal Nature Materials.