A team at Caltech and Yale used a chemistry-inspired computational approach to predict the Kondo effect in real materials without reducing electronic structure to simplified models.

Funding for the work came from the Air Force Office of Scientific Research, the US Department of Energy and its Center for Molecular Magnetic Quantum Materials, and the National Science Foundation.

In tests on seven magnetic transition-metal impurities embedded in copper, the new method improved prediction accuracy for most elements by up to two orders of magnitude compared to conventional model-based calculations.

The Kondo effect, a well-understood yet quantitatively challenging phenomenon, involves a magnetic impurity in a metal causing resistance to first decrease with cooling and then increase after a characteristic Kondo temperature due to complex electron interactions.

The work demonstrates a prototype approach for predicting material-specific Kondo behavior and other strongly correlated phenomena, potentially aiding the design of complex quantum materials like high-temperature superconductors and quantum magnets.

Lead authors Linqing Peng of Yale and Tianyu Zhu, both in Garnet Chan’s Caltech lab, led the project with results published in Science.

The study is described as a baby step toward forecasting the properties of challenging quantum materials from first principles, without experimental input, signaling progress toward computational material design for correlated electron systems.

Researchers treated magnetic impurities more like molecules, using quantum-chemistry tools to preserve the impurity’s full electronic structure rather than relying on reduced orbital models.