Problems with reconciling density functional theory calculations with experiment in ferropnictides

First-principles calculations of magnetic and, to a lesser extent, electronic properties of the LaFeAsO-based superconductors show substantial apparent controversy, as opposed to most weakly or strongly correlated materials. Not only do different reports disagree about quantitative values but there is also a schism in terms of interpreting the basic physics of the magnetic interactions in this system. In this paper, we present a systematic analysis using four different first-principles methods and show that while there is an unusual sensitivity to computational details, well-converged full-potential all-electron results are fully consistent among themselves. What makes results so sensitive and the system so different from simple local magnetic moments interacting via basic superexchange mechanisms is the itinerant character of the calculated magnetic ground state, where very soft magnetic moments and long-range interactions are characterized by a particular structure in the reciprocal (as opposed to real) space. Therefore, unravelling the magnetic interactions in their full richness remains a challenging but utterly important task.

Figure 1

(Color online) Fixed spin moment calculations for ${\text{LaFeAsO}}_{1-x}{\text{F}}_x$.

Figure 2

(Color online) A comparison of the band structure derived using experimental coordinates to that using optimized coordinates. The three-dimensional band that crosses the Fermi energy in the optimized structure is extraordinarily sensitive to As height, dropping below $E_F$ when it is raised to the experimental position. It is replaced by a two-dimensional band.

Figure 3

(Color online) GGA Fermi surfaces of ${\text{LaFeAsO}}_{1-x}{\text{F}}_x$ for (a) $x=0$ with optimized coordinates, (b) $x=0.1$ with optimized coordinates, (c) $x=0$ with experimental coordinates, and (d) $x=0.1$ with experimental coordinates. Note that the band structure is three dimensional at low doping in the optimized structure but becomes two dimensional with doping, while in the experimental structure it is two dimensional and insensitive to doping.