Fig. 2: Ultra-broadband ultrafast spectroscopy of the photoinduced phase transition in VO₂.

a Transient reflectivity of a 45 nm thick VO₂ thin film on sapphire when excited through the phase transition by a few-fs-duration pulse centred at 610 nm and a flucence of 29 mJ/cm². A variety of non-trivial dynamics are seen at all regions of the spectrum. b Lineouts (coloured lines) and rise time fits (dashed lines) of the dynamics averaged over wavelength in the three spectral bands indicated in (a). Transition times as low as 5 fs (temporal resolution limited) are observed. c Fit of the transient differential reflectivity using the metallic phase dielectric function. d The dielectric function of the rutile metallic phase (dashed line), as well as the constituent resonances (solid). Only the Drude term and the two indicated optical transitions (coloured lines) are allowed to vary in fitting the experimental data in panel a. The \({{{{\rm{O}}}}_{2p}\to d}_{\parallel }\) resonance is modelled as a Gaussian while the \({{{\rm{O}}}}_{2p}\to {\pi }^{*}\) is modelled as a Tauc-Lorentz resonator. e–j Variation of the fitting parameters from (d) with pump-probe delay. All parameters of the Drude and \({{{{\rm{O}}}}_{2p}\to d}_{\parallel }\) transitions are allowed to vary freely, while for the \({{{\rm{O}}}}_{2p}\to {\pi }^{*}\) resonance the width and amplitude are fixed. This set of parameters represents the minimal best fit to the data; further details on the fitting procedure and model in Methods. Shaded areas around the lines show the error on the fit parameters as given by the square root of the diagonal of the covariance matrix.