Researchers have manipulated the nuclear spin of a single atom by applying radio waves to it. A team from Delft University of Technology and CIC nanoGUNE-BRTA in Spain measured the electron spin resonance of a titanium-47 atom placed on a surface and observed two of its six peaks become equal in height. A closer analysis showed that the nuclear spin can be driven between two specific levels even under conditions where the electron and nuclear spin states hardly mix.
Manipulating the spin of a single nucleus has been demonstrated before in experiments using diamond and silicon. In earlier scanning tunneling microscopy (STM) experiments on copper and titanium atoms on surfaces, however, nuclear spin resonance could be observed only under conditions where the electron and nuclear states were strongly mixed. The new study goes beyond that limitation. The paper, by co-first authors Hester G. Vennema, Cristina Mier and colleagues, was published in Nature Communications on October 24, 2026. Delft University of Technology also announced the results on October 27.
Six resonance peaks from a single titanium atom
The team studied titanium atoms adsorbed on oxygen sites of a two-layer magnesium oxide film grown on a silver substrate. The sample was cooled to 0.4 K (about −272.75 °C), and a magnetic field was applied perpendicular to the surface. Measurements used a tungsten tip made magnetic by attaching 10 to 20 iron atoms. When the tip is brought close to the titanium atom, the tunneling current flowing between the tip and the atom changes depending on the direction of the electron spin, which makes it possible to read the magnetic state of a single atom.
The isotope studied, titanium-47 (⁴⁷Ti), has an electron spin of 1/2 and a nuclear spin of 5/2. The electron spin has two possible components along the field direction, while the nuclear spin's magnetic quantum number can take six values: −5/2, −3/2, −1/2, +1/2, +3/2 and +5/2. The electron and nuclear spins are coupled through the hyperfine interaction, so the radio frequency needed to flip the electron spin shifts slightly depending on the nuclear spin state.
When the researchers measured the tunneling current while sweeping the radio frequency, the electron spin resonance spectrum of a single titanium-47 atom showed six peaks. A titanium atom with zero nuclear spin, measured for comparison, showed only one peak. Each of the six peaks of titanium-47 corresponds to a different nuclear spin state, and the relative peak heights allow the fraction of time the nucleus spends in each state to be estimated.
The conventional approach of applying a single radio wave near the nuclear spin resonance frequency and directly detecting the change in tunneling current has limits, however. When the electron and nuclear states are strongly mixed, a nuclear spin transition also affects the electron side and can be seen as a change in current. But when a stronger magnetic field weakens the mixing, the change appearing on the electron side also becomes small. In fact, in the single-frequency measurements here, the signal corresponding to nuclear spin resonance disappeared when the field was raised to 200 mT. That result alone could not distinguish whether the nuclear spin had stopped responding to the radio wave or was responding but could no longer be detected.
Confirming nuclear spin transitions with a 49.3 MHz radio wave
To resolve this, the team used electron-nuclear double resonance (ENDOR), which applies radio waves of two different frequencies at the same time. A radio wave in the MHz range induces nuclear spin transitions, while one in the GHz range probes the electron spin resonance. Separating the driving of the nuclear spin from the electron-spin-based readout makes it possible to detect the nuclear response even when the electron and nuclear states are not strongly mixed.
In the experiment, the magnetic field was fixed at 450 mT, the tip setpoint current was 4 pA, the bias voltage was 80 mV, and the voltage amplitude of each of the two radio waves was 20 mV. When the nuclear-spin radio wave was set to 30.2 MHz, off resonance, the electron spin resonance peaks differed in height. When the nuclear-spin frequency was changed to 49.3 MHz, the first two peaks became equal in height.
These two peaks correspond to the nuclear spin states with magnetic quantum numbers −5/2 and −3/2. This can be interpreted as the resonant radio wave driving the nucleus between these two levels until it reached saturation, where their average occupations are equal. In other words, rather than observing the nuclear state directly, the researchers confirmed from the electron spin resonance peak heights that the nuclear spin had responded to the radio wave.
The team also ran measurements in which the frequency of the electron-spin readout wave was fixed at a particular peak while the nuclear-spin frequency was swept. They found that the transition between the same pair of nuclear levels has resonance frequencies that split into about 50 MHz and about 85 MHz depending on the direction of the electron spin. Measuring the electron resonance associated with one nuclear level produced an increase in signal, and the other produced a decrease.
In a control experiment, when the electron-spin frequency was moved off the resonance peak, the signal stayed constant as the nuclear-spin frequency was swept. In addition, a titanium atom with zero nuclear spin in the same field of view showed no signal corresponding to nuclear spin resonance in this frequency range. These results support the conclusion that the observed changes originate from titanium-47 nuclear spin transitions.
The team could further identify which pair of neighboring nuclear levels was involved in each transition. This is because the nuclear quadrupole interaction, in which the nucleus interacts with the inhomogeneous electric field around it, makes the spacings between nuclear spin levels slightly different. In addition to the transition between −5/2 and −3/2, the researchers separated the resonance lines for the transition between −3/2 and −1/2 according to the electron spin direction. In this way they identified at least four nuclear spin transitions.
Resonance persists as electron-nuclear coupling weakens
The most important aspect of the study is that nuclear spin transitions were observed even under conditions where the electron and nuclear spin states barely mix. The team varied the magnetic field perpendicular to the surface from 200 mT to 1400 mT and examined how the resonance frequencies changed. The resonance frequencies shifted with the field, but the pair of nuclear spin transitions identified near about 50 MHz and about 85 MHz was detected at every field condition examined with double resonance.
Calculations using the model in the paper indicate that, across this field range, a stronger field weakens the mixing of the electron and nuclear spin states. That is, the decrease in mixing is a result derived from the theoretical model, while the observation of nuclear spin resonance under such conditions was confirmed by experiment. Double resonance made it possible to read the nuclear spin response even under conditions where the conventional single-frequency measurement lost the signal.
However, what drives the nuclear spin transitions when mixing with the electron is weak has not been identified. The previously considered mechanism, in which the electron and nuclear spins flip simultaneously in opposite directions, has difficulty explaining both the observed frequencies and the weak mixing. The nuclear quadrupole interaction can explain the differences in level spacing, but if fluctuations of that interaction were the main cause of the transitions, this would be hard to reconcile with the transition selection rules.
The authors raise the possibility that an electric field modulates part of the hyperfine interaction. An oscillating magnetic field generated by the tip acting directly on the nuclear magnetic moment would also be consistent with the observed selection rules. But the coupling between the nuclear spin and a magnetic field is weak, and the strength of the driving field that actually acted on the nucleus beneath the tip is not known quantitatively.
The measurement setup also has limitations. Above 500 mT, the radio-frequency transmission characteristics of the apparatus make it difficult to keep the amplitude of the electron-spin wave constant. As a result, although nuclear spin resonance was observed over a wide range of fields, it cannot be concluded that the efficiency of driving the nuclear spin was constant across fields.
The main measurements in the paper used the same tip, apart from Figure 1d, which examined the field dependence with a single frequency. The authors have made the measurement data and the analysis and simulation code available on Zenodo. The paper does not, however, state the total number of independently measured samples or the experimental success rate.
Next challenges on the path to quantum memory
The spin of an atomic nucleus is relatively well isolated from its surroundings and may be able to hold a quantum state for a long time. This property has made nuclear spins candidates for quantum memory that stores quantum information. At the same time, weak interaction with the outside world makes nuclear spin states harder to manipulate and read out.
The study showed that a single nuclear spin transition can be induced, and detected through the electron spin, even under conditions of weak mixing with the electron. What was measured, however, was the average occupation of nuclear spin states under continuous radio waves, and the coherent control needed to store quantum information has not been demonstrated.
Writing quantum information into a nucleus requires not only changing the occupations of two levels but also manipulating their superposition while preserving its phase. The next step will be to vary the length of short radio-wave pulses and see whether Rabi oscillations, in which the nuclear spin state changes periodically, can be observed. If demonstrated, this would provide a more concrete foundation for using the spin of a single atomic nucleus to store and manipulate quantum information.
