Researchers develop ultrathin magnetic material to reduce AI chip energy consumption

Researchers led by Brown University have developed a method to preserve the electrical and magnetic properties of bismuth ferrite (BFO) at extremely thin layers. This breakthrough addresses the dead-layer effect, a phenomenon where material properties typically vanish when reduced to the thicknesses required for modern microelectronic scaling.

By growing the bismuth ferrite on a dysprosium scandate substrate, the team used compressive strain to maintain the necessary atomic arrangement. Additionally, a metallic strontium ruthenate layer was used to provide electrical shielding against disruptive internal fields. This approach allows the material to retain its coupled electrical and magnetic behavior even at thicknesses below three nanometers.

This advancement could lead to the development of lower-energy memory and computing components. Because the material’s electrical and magnetic behaviors are linked, magnetic information could potentially be controlled using small voltages rather than high-energy electric currents, offering a solution to the growing energy demands of AI processing.