Vanadium Oxides

From Icamipedia

Vanadium oxides are considered as prototypes of electronic strongly correlated systems. In particular, vanadium dioxide (VO2) and vanadium sesquioxide (V2O3) have received a lot of interest for many decades for their rich phase diagrams. In particular they show metal-insulator transitions (MIT) by varying the temperature or with doping.

Vanadium dioxide

At a temperature just above room temperature, VO2 undergoes a double transition, from a metal to an insulator and from a rutile to a monoclinic structure. It has been long debated on which of the two components is the main responsible in driving the transition. The issue is whether the electronic correlation is so strong to localize the electrons into a Mott-Hubbard insulator, or the structural distortions alone are enough to induce the insulating phase (Peierls model); in other words, whether the effects on the electronic properties of VO2 derive from the structural transition or the electronic transition itself induces also a structural distortion; whether the gap in the insulating phase is due to the structural deformation or to electronic correlation. Experimental results themselves are often contradictory and no unique conclusion about the character of the transition can be firmly drawn. Also most of the state-of-the-art theoretical approaches were unable to describe this complex transition. Modelistic approaches are able to catch some of the features of the electronic properties of VO2, but a consistent intepretation of the transition calls for a parameter free calculation.

In the insulating phase of VO2 one finds that the Kohn-Sham LDA density of states is metallic. This result is often interpreted as a sign of strong correlation. But, by definition, Kohn-Sham eigenvalues (and density of states) are not meant to describe photoemission. On the other side, DFT-LDA reproduces VO2 ground-state properties well. DFT is an exact theory for the ground state and LDA is, even for VO2, a good approximation.

In a broad range of materials a perturbative GW approach corrects LDA underestimation of bandgap. But in VO2 pertubative GW corrections fail to open the gap, in seeming support to the hypothesis of strong correlation. By performing a parameter-free self-consistent quasiparticle calculation, one finds that KS-LDA wave functions are not a sufficiently good approximation for the quasiparticle wave functions at the Fermi level. This explains the failure of the perturbative GW approach. In fact self-consistent quasiparticle calculations (at the COHSEX level) do suceed in opening a gap. The final result is in quantitative agreement with experiment. The change of the wavefunctions with respect to the LDA ones is of utmost relevance The quasiparticle band paradigm is still valid for VO2. Importantly, in order to obtain quantitative agreement with experiments it is crucial to take into account self-consistency including the degrees of freedom beyond the t2g orbital subset (which are instead neglected in DMFT calculations). Electronic correlation has indeed to be adequately treated. A Hartree-Fock calculation, where no correlation at all is taken into account, yields an insulator with a huge gap. In the insulating phase the system becomes more electronically one dimensional with a stronger polarization along the c axis and this leads to the gap opening. If this orbital redistribution is underestimated, as happens in LDA, the system remains metallic.

Concerning the quasiparticle description of the metal, the task for KS-LDA is easier. The metal is electronically more isotropic than the insulator, and LDA and quasiparticle wave functions are more similar. The metallic phase shows a satellite in the photoemission spectrum at a binding energy of 1.3 eV. Such a satellite cannot be described by a quasiparticle DOS. In the DMFT model it is interpreted as a lower Hubbard band. A parameter-free calculation of the electron energy loss in the metallic phase demonstrates the presence of a plasmon resonance in the same range of energies, due to a group of V d-d transitions. Therefore, these findings in the EELS allows one to assign the structure at a binding energy of 1.3 eV to a plasmon satellite.

References

• V. Eyert, Ann. Phys. (Berlin) 11, 650 (2002), .
• S. Biermann et al., Phys. Rev. Lett. 94, 026404 (2005).
• Matteo Gatti, Fabien Bruneval, Valerio Olevano, and Lucia Reining, Phys. Rev. Lett. 99, 266402 (2007).