Single crystal growth and electron spectroscopy of d1-systems

One of the most intriguing aspects of transition metal oxides is the wide variety and richness of their physical properties. Microscopic understanding of the unconventional behavior of this class of materials is at the heart of modern solid state physics research. The many body nature of the problem forms a true intellectual challenge requiring modern experimental and theoretical methods. There is a need for well-defined materials which can serve as model systems. The class of the RETiO3 (RE = rare earth) materials forms in this context a very interesting 'playground' for the quantitative study of the properties and excitation spectra of correlated oxides. It has the relatively simple perovskite crystal structure, and the Ti ions have (formally) only one electron in their 3d shell so that complications related to atomic multiplet effects can be avoided. Yet, the orbital degeneracy together with the presence of a small band gap lead to a number of interesting physics which are then subject of a flurry of detailed experimental and theoretical studies. Efforts are being made for a quantitative analysis as to test the accuracy of various theoretical approaches. Here we report on a detailed experimental study of the electronic structure of the RETiO3 system. The study aims to address one of the long standing topics in theoretical solid state physics, namely the single-particle spectral weight distribution in Mott-Hubbard systems in the vicinity of the metal-insulator transition. In view of the existing discussions in the literature, it turned out to be crucial to use well defined samples: we have to grow titanate single crystals with well defined stoichiometry and doping. This activity in fact forms the basis of this thesis work and much effort is also put into the characterization of the crystals using various methods including x-ray and neutron diffraction, as well as magnetic, transport and thermodynamic measurements. We have used photoelectron spectroscopy to investigate the electronic structure of the RETiO3 system. Although many photoemission studies have been reported in the literature, it is also realized more and more that those published results may not be representative for the material due to the extreme surface sensitivity of the particular photoemission technique used in those studies. This is important since the electronic structure of the surface is very different from the bulk, especially for strongly correlated systems. Essential aspect of the work is therefore to carefully optimize the conditions for this type of experiments as to make sure that the spectra obtained are truely representative for the bulk material. Only in this manner we can do a critical and quantitative evaluation of the various advanced many-body models currently available trying to describe the excitation spectra of strongly correlated systems. The main result of our spectroscopic study is as follows: we have utilized bulk-sensitive x-ray photoelectron spectroscopy to study the valence band spectral weight distribution of d1 Mott insulators LaTiO3 and YTiO3. We observed appreciable differences in the spectra, reflecting the difference in the one-electron band width W. We also found that the Ti 3d spectra of both materials are much broader than the occupied 3d bands calculated by band theories. The mean-field inclusion of the Hubbard U explains the band gap but produces even narrower bands, indicating the complete breakdown of standard mean-field theories in describing excitation spectra. We associate the observed spectra with the propagation of a hole in a system with surprisingly well suppressed charge fluctuations thereby showing characteristics of a t-J model. Upon doping we observe the creation of the new state at the Fermi level which is accompanied by a rapid decrease of the spectral intensity at about 1 eV binding energy. This rapid transfer of spectral weight characterizes the correlated nature of this material, and distinguishes this material clearly from ordinary band semiconductors. Surprisingly, we find also a large difference in the speed of transfer of spectral weight when comparing doping using excess oxygen or using Sr substitution. LDA+DMFT calculations suggest that this could be related to the fact that the Sr material has a smaller crystal field splitting in the t2g levels than the oxygen excess sample, yielding an effectively larger degeneracy and thus a smaller effective U versus W ratio.