Physicists flip a single atom's nucleus with radio waves — and can't fully explain how

The nucleus of an atom is a terrible place to put information and a wonderful one, for the same reason. Its spin is feeble — a thousandth of the electron's magnetic strength, or less — but it is also shielded from the noisy world outside, which is precisely what a quantum memory needs. The problem has always been reaching it.

A group at Delft University of Technology now reports doing exactly that, on one atom at a time, and getting a result their own model says is impossible. The work, led by Sander Otte with first author Hester Vennema, is published in Nature Communications (DOI 10.1038/s41467-026-78801-5).

The tool is a scanning tunnelling microscope converted into an electron-spin resonance instrument: a metal tip a few atoms wide, held above a surface in ultra-high vacuum at cryogenic temperature, able to address a single atom electrically. In earlier work the group used the hyperfine interaction — the weak coupling that slightly entangles an electron spin with the nuclear spin sitting under it — to read and even to observe the nucleus switching back and forth.

This time they refined electron-nuclear double resonance for a single atom, sending two different radio-frequency signals at the atom simultaneously, one tuned to the electron and one to the nucleus. The guitar-string analogy in their release is apt: each allowed spin transition is a string with its own note, and you must play the right note to move it. The nucleus's notes are 100 to 1,000 times lower in frequency than the electron's.

With the two channels separated, the group could test the assumption directly. They found they could flip the nuclear spin even when the electron and nucleus were largely disentangled — that is, without using the hyperfine coupling as a handle.

"According to our simple understanding, this should not really happen," Otte said. "The nuclear spin should not be sensitive at all to the electrical signals from a scanning tunnelling microscope."

The honesty of that sentence is the most interesting part of the paper. Nuclear spins are attractive for quantum information precisely because they barely talk to their environment; a mechanism that lets a local electrical probe drive them anyway is either a new control channel or a measurement artefact, and the authors do not yet know which. "Back to the drawing board, then," Otte said. "That is really how science works."

What it is not: a device. This is one atom, on one surface, at millikelvin-adjacent temperatures, in a laboratory that has spent years learning to hold a tip still. Nothing here becomes a memory chip on a schedule. What it is: a demonstration that a widely assumed prohibition — you cannot drive a nuclear spin directly with a local probe — has an exception, and that the exception is reproducible enough to publish.

If the mechanism holds up, the payoff is architectural. Nuclear spins are the natural place to park a quantum state while the electron above them is busy being read out, and reading them back has been the bottleneck. A local, electrical handle on the nucleus would put both operations on the same tip. That is a long way from a computer, and close enough to be worth the confusion.