Fig. 5: Ultrafast Coulomb blockade detection using THz pump – THz probe time-domain sampling.

a STM topography and STS spectrum of Vac_Se in 1 ML WSe₂. The gray spectrum references pristine WSe₂. The illustration shows the concept of pump-probe measurements, where a THz pump pulse (red) charges the system while a time-delayed THz probe pulse (blue) measures the transient charge occupation. b The rectified charge of the THz probe pulse as a function of THz–THz delay. The black line is an exponential fit to the relaxation of the charge state. Orbital images at selected delays, illustrating the change of rectified charge ΔQ_LW with respect to an equilibrium reference at 10 ps. c Dependence of the charge-state lifetime extracted via exponential fitting in panel b as a function of ΔV = V_LUMO − V_dc. The horizontal line (gray) marks the intrinsic charge-state lifetime from dc measurements. Error bars correspond to the fit uncertainty, and lifetime values below 0.5 ps are shorter than the time resolution of the experiment. d–f Same analysis for Vac_Se in 2 ML. Due to significant THz reflections, apparent as dips in Q_LW(Δt), the extraction of a charge-state lifetime from the pump-probe measurement in panel (e) must be matched with a simulation as detailed in the Methods. The data was obtained on different Vac_Se characterized in Fig. S1: a–c at \({{{{\rm{Vac}}}}}_{{{{\rm{Se}}}}}^{1.2}\), d–e at \({{{{\rm{Vac}}}}}_{{{{\rm{Se}}}}}^{2.3}\), and f at \({{{{\rm{Vac}}}}}_{{{{\rm{Se}}}}}^{2.4}\). Integration time per data point is 100 ms for LW-STS and 30 ms (1 ML) or 15 ms (2 ML) for LW-STM, corresponding to 4.1 M and 1.2 M or 0.6 M THz pulses at 41 MHz repetition rate.