In this project of the research group we perform calculations as based on density functional theory (DFT) and the local density approximation (LDA). Our goal is to study in detail the interplay of structural properties and the electronic structure. This serves the purpose of deriving information about the origin of the metal-insulator transition (MIT) especially of VO2. This material undergoes a first order metal-insulator transition at 340 K with a conductivity change of more than five decades. This transition is accompanied by a structural transition from the high-temperature rutile phase to the low-temperature monoclinic (the socalled M1) phase. The crystal structure of the M1 phase is characterized by dimerization within the vanadium atom chains running along the rutile c axis plus a lateral displacement of the vanadium atoms in a antiferroelectric, zigzag-like mode. From our calculations we get indeed a significant response of the electronic structure to changes of the crystal structure, which consists of band splittings and shifts and reflects strong electron-lattice interaction. Nevertheless, due to the limitations of the LDA, we are not yet in a position to reproduce the optical bandgap of the insulating phase, but still get a small semimetal-like overlap of the bands. It should, however, be noticed that band below and above this overlap region do have a different orbital character. For this reason we performed a final check on the effect of the lattice distortion coming with the transition by performing an additional set of calculations, which used a hypothetical crystal structure, where the structural changes occuring at the metal-insulator transition were artificially amplified without further symmetry breaking. As a result, we obtained a finite bandgap.
The strong coupling of the crystal structure to the electronic properties motivated investigation of other insulating low-temperature phases of VO2. We looked in particular at the monoclinic M2 phase, which is regarded as a metastable variant of low-temperature VO2 and occurs on doping with very small amounts of Cr or Al or on the application of slight uniaxial pressure. In this phase parts of the structural characteristics of stoichiometric VO2 are suppressed. To be specific, while the latter modification shows both the dimerization and the zigzag-like mode on every vanadium chain, these two modes are separated in the M2 phase: Only half of the chains dimerize, whereas the other half displays the antiferroelectric mode. In addition, these latter chains behave like linear Heisenberg-chains. Starting with spin-degenerate calculations we get indeed different results for the two types of chains. While the partial densities of states DOS) for atoms located on the dimerizing chains resemble those obtained for the M1 phase, we get partial DOS similar to the rutile phase for the vanadium atoms located on the zigzag chains. In particular, no bandgap is obtained. Things change drastically when spin-degeneracy is lifted. In this case states originating from the zigzag-type chains likewise display a band splitting this leading to a full separation of bands with different orbital character. As for the M1 phase we are still faced with a small but finite overlap of the bands, which, however, does not principally hinder the insulating state, as additional calculations using hypothetically amplified structural modifications reveal.
To conclude, our calculations clearly reveal the decisive contribution of strong electron-lattice interaction to the metal-insulator transition of VO2, which may be interpreted in terms of Peierls-like instability. Our findings are thus in good agreement with the recent ultrasonic investigations, which likewise revealed strong structural changes at the phase transition. Due to the present limitations of density functional theory coming with the local density approximation we can, however, not yet decide if the structural changes suffice to drive the metal-insulator transition. For the same reason, the role of electronic correlations also remains an open question. Still, as our recent LDA+U calculations indicate, the metal-insulator transition may be a consequence of a cooperative effect of electron-lattice interaction and electronic correlations.
We have complemented our investigations for VO2 by calculations for MoO2, V2O5 and Ti2O3. The former compound is a metal at all temperatures and crystallizes in the same structure as VO2 in the M2 phase. Investigation of this material thus allowed to study the tendency of rutile related transition metal dioxides towards the monoclinic M1 structure independent from the issue of the metal-insulator transition. Our calculations showed, that as for VO2 the metal atom dimerization might be interpreted as a Peierls-like lattice instability, which results from optimization of the overall chemical bonding properties. Our results are in very good agreement with photoemission, X-ray absorption and Shubnikov-de Haas measurements, which were performed by the group of Prof. Horn.
For the semiconducting V2O5 we showed, that strong electron-lattice interaction transforms the deviations of the crystal structure from ideal geometry of the characteristic VO6 octahedra into significant band shifts. This is the reason for the relatively large optical bandgap of 2.2 eV.
Detailed accounts of the electronic structure calculations performed within the DFG Research Group are given in the Diploma thesis of Robert Horny (Experimentalphysik II), the Ph.D. thesis of Wei-Dong Yang and the Habilitation thesis of Volker Eyert. Please do not hesitate to contact us for further discussions (e-mail: eyert@Physik.Uni-Augsburg.DE).