In some exotic magnetic materials, the geometry of the atomic lattice prevents spins from ordering magnetically in the way their interactions suggest that they should. This frustration of magnetic interactions can lead to magnetic phases known as spin liquids, spin glasses or spin ices, in which the arrangement of spins mirrors the atomic structure of a liquid or glass: globally disordered, but locally correlated. Understanding the nature of these local spin correlations is key to understanding many fundamentally-important materials, including high-Tc superconductors and multiferroics. The problem is that local spin correlations cannot be determined by conventional crystallographic methods. Instead, a model of magnetic interactions is proposed for a given system, and its predictions compared with single-crystal magnetic neutron scattering data. But this approach is time-consuming at best, and impossible if single-crystal samples can’t be made.
We have shown that reverse Monte Carlo analysis of the magnetic component of powder data can recover much of the information for which single-crystal samples have been preferred. Considering as case studies a number of frustrated systems — differing in lattice symmetry, dimensionality, and form of magnetic interaction — we find that in each case the RMC approach recovers the intricate features of the single-crystal diffuse scattering pattern. RMC also produces spin configurations which allow us to probe directly the real-space spin correlations, readily picking out spin anisotropy. In some systems where the degree of local magnetic order is strong (such as spin ice) we get better results by making use of a constraint that the magnetic environments of different spins should be as similar as possible — demanding, in effect, that magnetic structure should be the simplest allowed by the data. The method is very general, and is equally applicable to molecular systems.
We hope that these results will open up avenues for research on frustrated magnetic systems for which single-crystal samples don’t yet exist, and demonstrate an efficient way of extracting more of the information contained within diffuse powder scattering.