Extended Data Fig. 1: I/W(110) and W(110) measurements.

a, Main panel, XUV I/W(110) photoelectron spectra taken at 105 eV central photon energy for different adsorbate surface coverages (key in top left inset). Three effects are observable with decreasing iodine surface coverage: a decrease of the I4d peak intensity, an increase of the W4f peak intensity due to reduced inelastic scattering, and a shape change of the valence-electron peak towards a clean tungsten spectrum. The iodine surface density was calibrated by taking a full monolayer I/W(110) photoelectron spectrum as a reference before thermal desorption and comparing the I4d photoelectron flux before and after the thermal desorption of iodine. Because the iodine surface coverage saturates, this allows for a reliable coverage calibration. Top centre inset, illustration of the employed Shirley background (BG) subtraction scheme; top right inset, the magnified valence photoelectron spectrum. b, Relative photoemission delays for I/W(110). Both W4f to I4d (blue) and valence to I4d (red) delays are shown. All individual measurements are depicted by crosses, averages for individual coverages are depicted by circles. Vertical error bars mark 95% confidence assuming a Student’s t-distribution and horizontal error bars mark maximum errors. The blue line represents a linear regression to the W4f to I4d delay (blue line), the shaded area represents the 95% confidence interval of this model. c, Attosecond streaking delay measurements on a pristine W(110) surface. The W4f−CB emission delay (blue circles and histogram) as a function of the time of measurement after the preparation of a pristine surface (yellow) reveals a small deviation of the centre (red, dashed line) of the normal distribution (red, solid line) fit to all measurements from the extrapolation to an instantaneous measurement due to surface contamination. The large number of measurements allows the extraction of the photoemission timing of the clean surface by extrapolation.