PUB - Publikationen an der Universität Bielefeld

Photoemission dynamics in solid-state materials can be observed in real-time via attosecond time-resolved photoelectron spectroscopy [1]. In this work *Reconstruction of Attosecond Beating By Interference of Two-photon Transitions* (RABBITT) [2] is applied to Bi_*2*Te_*3* and Bi_*2*Se_*3* crystals. Here, the first experimental observation of photoemission delays between the two finestructure components of a core level band in a solid-state system is reported. In Bi_*2*Te_*3* spin-orbit delays between photoemission from d_*5/2* and d_*3/2* initial states of Bi 5d and Te 4d core level bands of t_*LS*(Bi 5d) = 30 +/- 13 as and t_*LS*(Te 4d) = -39 +/- 18 as were measured at a central XUV photon energy of 92.2 +/- 0.4 eV. Corresponding spin-orbit delays of Bi 5d and Se 3d in Bi_*2*Se_*3* are t_*LS*(Bi5d) = 31 +/- 10 as and t_*LS*(Se3d) = -93 +/- 83 as. It is shown by solving the time-dependent Schrödinger equation for photoemission in Jellium-type model including electron-hole interaction and IR field penetration that photoelectron transport cannot account for the experimentally observed delays. Instead, calculations indicate that RABBITT can probe large Eisenbud-Wigner-Smith (EWS) delays in the vicinity of Cooper minima. A hydrogenic atom model is presented that yields intra-atomic spin-orbit delays in the same order of magnitude as the experimental observations. However, *ab initio* single-configuration Dirac-Fock calculations do not predict a Cooper minimum in isolated Bi atoms at the aforementioned photon energy. Since the crystal environment has a strong impact on photoemission final states in this energy regime, additional Cooper minima might appear in the photoemission transition matrices of Bi_*2*Te_*3* and Bi_*2*Se_*3*. Moreover, the first experimental comparison between RABBITT and attosecond streaking spectroscopy on a solid target is reported. Streaking experiments on Bi_*2*Te_*3* reveal a photoemission delay between Bi 5d and Te 4d bands of t_*streaking* = 7 +/- 17 as [3]. If spin-orbit delays are not resolved, an effective RABBITT delay can be extracted. The effective RABBITT delay between Bi 5d and Te 4d of t^*eff*_RABBITT = 4 +/- 14 as is in perfect agreement with the streaking delay. Also in a one-dimensional toy model photoemission delays obtained from both methods agree with each other. However, close to Cooper minima different delays are obtained by RABBITT and streaking because different XUV spectral distributions are involved.