Drexel Engineers and Collaborators Uncover a Material That Could Cut the Energy Cost of AI

Image of a row of computer servers in a data center

The artificial intelligence boom has a power problem. The data centers that train and run AI models are consuming electricity at a rate that is straining grids and drawing comparisons to the energy appetite of entire countries. A meaningful part of that consumption comes down to a basic fact about how computers work: moving and storing information using electrical charge generates heat, and heat is wasted energy. One path toward more efficient computing runs through a different approach entirely, using electron spin rather than charge to carry and store information. The obstacle is that no single material has been able to bridge the electric and magnetic worlds reliably, at room temperature, in a way that could be practically built upon.

A study just published in Proceedings of the National Academy of Sciences reports that a new phase of a well-known material may have broken that pattern. Jonathan E. Spanier, Hess Family Endowed Chair Professor and department head in mechanical engineering and mechanics and affiliated faculty in physics at Drexel University, was part of the international research team, which included Lane W. Martin, Robert A. Welch Professor of Materials Science and Nanoengineering at Rice University, and Ilya Grinberg, associate professor of chemistry at Bar-Ilan University. Together with colleagues at MIT, the University of Pennsylvania, Northeastern University, the U.S. Naval Research Laboratory, the University of California at Berkeley, and Lawrence Berkeley National Laboratory, the team set out to unravel the findings.

The material is a solid solution of bismuth ferrite and barium titanate. When combined in a specific ratio and grown as an ultrathin film under precisely controlled conditions, it takes on a crystal structure that researchers had not previously observed in bismuth ferrite-based systems.

In that new structure, the film simultaneously exhibits strong, spontaneous, remanent, and re-orientable electrical polarization and magnetization at room temperature, and those two properties are tightly coupled. That coupling is the key for potential applications: it means an applied electric (magnetic) field can be used to change the magnetic (electric) dipolar state, without the energy losses that come from engineering magneto-electric interactions across the boundaries between different materials.

“This material and its intriguing new mechanism brings the possibility of using a magnetic field to control electronic charge, and a voltage to manipulate electron spin, and realized through the exquisite quality films attained by Professor Martin and postdoctoral scientist Dr. Taeyeon Kim,” Spanier said.

“The original goals of this research were very different,” Spanier recalled. “Our initial team conceived and launched an investigation of the high-frequency dynamics of how domain walls, or boundaries between different regions of dipolar ordering, fluctuate in a material containing magnetic ions, based on our earlier work. As is typically the case, we started by making the bulk ceramic material needed for the thin-film growth and characterization.”

But as different research groups brought their measurements and theory together, something unexpected came into focus.

“This work unfolded somewhat serendipitously,” Spanier said, “which is one of the things I find enthralling about doing science.”

The Drexel team’s contribution to characterizing the new material included inelastic light scattering spectroscopy, which uses laser light to probe the collective dynamics of excitations in a material, including spins. Those measurements helped the collaboration track how the internal magnetic ordering shifts as the film’s composition changes, providing evidence that something fundamentally different was happening in the new structural phase.

First-principles calculations by collaborators at Bar-Ilan University and the University of Pennsylvania identified a proposed mechanism: the unusually large magnetic response appears to arise from a specific intermediate-range ordering of, paradoxically, the nonmagnetic titanium atoms within the crystal, substituted for the magnetic iron, producing an imbalance between opposing magnetic spins that generates a net moment far larger than the parent material.

What makes breakthroughs such as these possible is long-term investment in multi-institution research. “Funding for students and postdoctoral scholars in science and engineering research is crucial,” he said. “Our team is grateful to the Army Research Office and its program managers Dr. James Harvey and Dr. Joe Qiu for their support, along with Dr. Brendan M. Hanrahan of the Army Research Laboratory.”

That support has already opened a new chapter. The work is the foundation for a new five-year, $1 million grant from the Office of Naval Research, on which Spanier serves as principal investigator with Martin as a key collaborator. That project, titled “Energy-Efficient Single-Phase Multiferroic Magnetoelectric Thin Films,” extends the research into new families of materials and will work to deepen understanding of why and how the coupling between electrical and magnetic order arises, and how strong it can be made.

For Spanier, the arc from an unexpected experimental result to the research team’s publication in one of science’s most prestigious journals and a new grant reflect something important about how basic research works at its best. The ceramic his lab synthesized set something larger in motion, and following where that led produced results that none of the collaborators had mapped out at the start. The need now is to better understand the coupling between dielectric polarization and magnetism and what future applications it may enable.


In This Article