Spin state transition and covalent bonding in LaCoO3

Abstract

We use the dynamical mean-field theory to study a p-d Hubbard Hamiltonian for LaCoO3 derived from ab initio calculations in local density approximation (LDA+DMFT scheme). We address the origin of local moments observed above 100 K and discuss their attribution to a particular atomic multiplet in the presence of covalent Co-O bonding. We show that in solids such attribution, based on the single ion picture, is in general not possible. We explain when and how the single ion picture can be generalized to provide a useful approximation in solids. Our results demonstrate that the apparent magnitude of the local moment is not necessarily indicative of the underlying atomic multiplet. We conclude that the local moment behavior in LaCoO3 arises from the high-spin state of Co and explain the precise meaning of this statement

Excitonic instability at the spin-state transition in the two-band Hubbard model

“…The role of Hund's coupling in correlated electron systems has been recently theoretically studied in the context of Hund's metals 1,2 and the spin-state transitions driven by pressure 3,4 as well as temperature 5,6 or doping 7 . Competition of different spin states was also linked to the peculiar magnetic properties of iron pnictides 8 .…”

Improved Green’s function measurement for hybridization expansion quantum Monte Carlo

“…Then, estimator for G C α (iω n ) is derived analogously to equation (13). The only difference here is that the Z : sum on the right hand side goes over complementary indices and τ ′ and τ ′′ must be properly ordered.…”

“…This property makes the algorithm stable at any regime. In principle, we can store estimator (13) in any basis during the simulation. Nevertheless, the imaginary time representation is inefficient because we must either introduce a large discretization error or use a fine imaginarytime grid which requires many numerically costly evaluations of expression (13).…”

“…In principle, we can store estimator (13) in any basis during the simulation. Nevertheless, the imaginary time representation is inefficient because we must either introduce a large discretization error or use a fine imaginarytime grid which requires many numerically costly evaluations of expression (13). Despite its clear advantages, the recently proposed orthogonal polynomial representation for the Green's function [19] is not well suited for our purpose either and we use the Matsubara basis.…”

“…Such a situation is common in materials containing orbitals with fluctuating occupation coupled by interaction to filled or empty orbitals, e.g. the t 2g orbitals in nickelates [10], e g orbitals in early transition metal oxides [11,12], t 2g orbitals in the low-spin state of LaCoO 3 [13], or Hubbard model with crystal-field splitting [14,15]. Although the filled orbitals experience little fluctuations themselves and their influence on the active (partially filled) orbitals is well approximated by a static potential, it is their Green's functions that may become prohibitively noisy.…”