A sulfur donor atom in indium arsenide donates an electron to the conduction band. In the textbook picture, the donor is either ionized or neutral — a static property determined by where its energy level sits relative to the Fermi level. Standard scanning tunneling microscopy (STM) measures the time-averaged tunneling current, which produces a clean spectral picture consistent with this static model.
Ortego Larrazabal et al. (arXiv:2603.13040) look at the noise instead of the signal. Using MHz-frequency noise spectroscopy at atomic resolution, they detect random telegraph noise from individual sulfur donors — the atom flickering between charged and neutral states on microsecond timescales. This flickering is invisible in time-averaged tunneling spectra. The standard measurement averages over thousands of charge-state transitions per data point, smoothing the dynamics into a static-looking spectrum.
The dynamics depend on bias voltage and tip position. As the STM tip voltage tunes the donor energy level through the Fermi level, the telegraph noise characteristics change — the switching rate, the duty cycle, the noise amplitude all shift in ways that a bias-dependent charge-state model quantitatively explains. The bulk semiconductor environment matters: surrounding donors, band bending, and the local Fermi level all influence the transition rates.
This reframes donor ionization as a non-equilibrium dynamical process rather than a static equilibrium property. For quantum devices that use individual dopant atoms — single-electron transistors, donor-based qubits — the flickering represents a noise source that doesn't appear in any equilibrium measurement. The atom looks stable in every standard diagnostic but is actually switching states millions of times per second. The dynamics are hidden by the very measurements designed to characterize the atom.