Pseudogap and Mott transition studied by cellular dynamical mean-field theory

Abstract

We study metal-insulator transitions between Mott insulators and metals. The transition mechanism completely different from the original dynamical mean field theory (DMFT) emerges from a cluster extension of it. A consistent picture suggests that the quasiparticle weight Z remains nonzero through metals and suddenly jumps to zero at the transition, while the gap opens continuously in the insulators. This is in contrast with the original DMFT, where Z continuously vanishes but the gap opens discontinuously. The present results arising from electron differentiation in momentum space agree with recent puzzling bulk-sensitive experiments on CaVO3 and SrVO3.PACS numbers: PACS: 71.30.+h; 71.10.Fd; 71.10.Pm Mechanisms and nature of correlation-induced metalinsulator transitions (MIT) are fundamental challenges in condensed matter physics 1 . Although the metals and insulators far away from the MIT are relatively well understood, electronic states dominated and controlled by the proximity of the MIT are far from complete understanding. We find many challenging phenomena such as the high-T c superconductivity in this region, which waits for better understanding of underlying proximity of MIT.The dynamical mean-field theory (DMFT) offers a clear picture of the MIT by taking the limit of high dimensions 2 . In the DMFT, the MIT is approached from metals by the reduction of the quasiparticle weight Z at the Fermi level and the transition is driven by vanishing Z 3 . At the transition, a nonzero insulating gap is already open in the density of states (DOS) as a result of the vanishing Z of the coherent peak isolated from the well-separated upper and lower Hubbard bands. Early photoemission spectroscopy (PES) and inverse-PES of Ca 1−x Sr x VO 3 4 showed a three peak structure, called a sharp coherent band at E F and the incoherent ones corresponding to the upper and lower Hubbard bands at a few electron volts above and below E F , which is consistent with the above DMFT scenario. However, recent more bulk-sensitive PES have revealed a qualitatively different and puzzling feature with a new broad peak at around 0.2eV connected to the lower Hubbard band around 1.6eV and a pseudogap formation around the Fermi level 5 . This is in marked contrast with the DMFT results.In this paper, we show that a scenario of the MIT completely different from the original DMFT emerges in two and three dimensional systems by extending the DMFT to allow the momentum dependence of the self-energy. The MIT is now not driven by a continuous reduction of Z to zero, but by the quasiparticle poles moving away from the Fermi surface, leading to a discontinuous jump of Z to zero at the Fermi level. Such a MIT is the consequence of an anisotropic extinction of the Fermi surface with the topological change, namely the electron differentiation in momentum space. We show the results by considering a cluster extension of the DMFT up to four sites for the Hubbard model on the square and cubic lattices. Our results indicate an opening of pseudogap in the DOS a...

Correlated Dirac fermions on the honeycomb lattice studied within cluster dynamical mean field theory

“…This finding is also consistent with the behavior seen on the square lattice. 16 We close this section by commenting briefly on the relevance of the results discussed earlier for graphene. There exists at present considerable debate on the role of screening and on the magnitude of the short-range Coulomb repulsion in this material.…”

“…A similar pseudogap induced by short-range correlations at half-filling was observed below the Mott transition in the Hubbard model for the square lattice. [15][16][17] Neglecting the small insulating term of (K,iω n ) sufficiently far below the transition, the effective-mass enhancement of the quasiparticle bands near the Dirac points is given by m * /m = 1 − a = 1.0, . .…”

“…The opening of the pseudogap below the transition, and the variation of the effective mass with Coulomb energy, are qualitatively similar to analogous results obtained within cluster DMFT calculations for the square lattice. [15][16][17] The outline of this paper is as follows: In the next section, we briefly outline the application of ED/CDMFT to the Hubbard model for the honeycomb lattice. Section III provides a discussion of the results.…”

Sum rules and bath parametrization for quantum cluster theories

“…In such a situation details of the fitting procedure are important and are accordingly discussed in the literature. [11][12][13][28][29][30] To fit the Anderson parameters V l and l we use the distance function…”

Quantum Phase Transition in the Two-Band Hubbard Model

“…Even in standard Mott insulators such as V 2 O 3 and LaTiO 3 , the presence of inequivalent orbitals leads to strong orbital polarization where in the vicinity of the transition nearly localized electrons coexist with weakly itinerant bands [4 -6]. In the case of cuprates, one of the most fascinating aspects is the coexistence of strongly and weakly correlated regions of the Brillouin zone, giving rise to so-called hot spots and cold spots, with breakdown of Fermi-liquid behavior in the former and Fermi-liquid properties in the latter [7][8][9]. Since the momentum variation of the self-energy is associated with spatial fluctuations, an intriguing analogy exists between multisite interactions within a single band and single-site interactions among inequivalent orbitals.…”