Metal-insulator transitions and dilute electron and hole doping in the extended Hubbard (d-p) model.

Abstract

Metal-insulator transitions in the three-band extended Hubbard model are studied within the mean-field theory of the Kotliar-Ruckenstein slave-boson formalism. It is shown that the charge-transfer (CT) insulator transition (\ensuremath{\Delta}\ensuremath{\ll}U) takes place when twice the CT-energy loss 2\ensuremath{\Delta} exceeds eight times the average kinetic-energy gain per site E, contrary to the Mott-Hubbard (MH) transition (U\ensuremath{\ll}\ensuremath{\Delta}) which occurs when the Coulomb-energy loss U exceeds E. The positions of the renormalized d levels of quasiparticles formed on dilute doping are analyzed. A dilute hole doping into the insulator phase gives rise to several qualitatively different doping regimes, e.g., those with d-hole or d-electron doping determined by some simple rules, in which dilute hole doping just inside the boundary always induces d electron doping. We will show that the dilutely hole-doped phase diagram is remarkably symmetric with respect to U and \ensuremath{\Delta}, i.e., the MH and CT insulators. The optical conductivity gap ${\mathit{E}}_{\mathrm{gap}}$ of the MH and CT insulators is interpreted as a jump in the chemical potential \ensuremath{\Delta}\ensuremath{\mu} between dilute electron and hole doping, and is strongly renormalized from the bare Hubbard gap U or CT gap \ensuremath{\Delta} near the metal-insulator boundary, while it is roughly given by the bare gap well inside the insulator phase.