Yale-led team simulates the Kondo effect with better accuracy than state of the art methods

[PrimeDirective27]

Researchers led by Professor Tianyu Zhu at Yale and Dr Linqing Peng at Princeton, working with Professor Garnet Chan at Caltech, have published a new way of simulating one of the classic puzzles of quantum materials. The work appears in Science and focuses on the Kondo effect. It is a good example of how much quantum physics can be done with clever classical computing. Science is about as prestigious as journals get

The Kondo effect is a strange one. Normally metals like copper conduct better as they get colder, because the atoms vibrate less and get in the electrons' way less often. But if you add a tiny amount of a magnetic impurity like iron, the resistance starts going back up at very low temperatures. The magnetic moment of the impurity gets screened by a cloud of surrounding electrons, and that increases scattering. It puzzled physicists for decades until Jun Kondo explained it in the 1960s

Simulating it properly has always been difficult. Most approaches simplify the material into a basic model, which loses important detail. Zhu's team instead used a first principles approach that includes the real atomic and electronic structure. They borrowed techniques from quantum chemistry and treated the magnetic impurity a bit like a molecule, while keeping the detailed interactions between many electrons

They tested seven different transition metal atoms embedded in copper and got better accuracy than standard model based calculations. Zhu said they had demonstrated that the method can simulate a challenging quantum materials problem with better accuracy than state of the art approaches. The team is now applying it to two dimensional magnetic materials for memory and computing, and to single atom alloy catalysts. The same ideas could eventually help with high temperature superconductors and quantum magnets

What I like about this is that it is not a quantum computer story at all, yet it tackles exactly the kind of problem quantum computers are meant to help with. Classical methods keep getting better, which raises the bar for quantum advantage. Does that make you more or less optimistic about quantum chemistry on real quantum hardware?

New computational strategy modeling electron interactions advances quantum materials researchchem.yale.edu