Domain walls are natural borders in ferromagnetic, ferroelectric or ferroelastic materials. It seems that they can also be reactive areas that produce crystallographic phases never before observed in bulk materials. See Letter p.379
Interfaces between different oxide materials have been attracting much attention. They are seen as a source of new properties and functionalities. Engineering these borders has already revealed phenomena such as conductivity and superconductivity at the frontier between insulating compounds, and magnetism between non-magnetic materials1. In fact, it is not even necessary to combine different materials to create interfaces. Ferroic materials such as ferromagnets, ferroelectrics and ferroelastics naturally break into domains characterized by different orientations of the material's spontaneous ferroic order — magnetization for ferromagnets, electric polarization for ferroelectrics and macroscopic deformation for ferroelastics. These domains are separated by interfaces called domain walls2. On page 379 of this issue, Farokhipoor et al.3 highlight that, instead of being a passive region that accommodates the different ferroic-order orientations of the domains that border it, a domain wall can also be a reactive area that generates and stabilizes new two-dimensional crystallographic phases not achievable by conventional means.
Owing to spatial symmetry breaking and local mechanical constraints, a domain wall may exhibit properties distinct from the two domains that surround it. The physics of domain walls is expected to become remarkably complex in multiferroics (materials that have two or more forms of ferroic order), in which different types of domain wall coexist and can be coupled2. Clear understanding of what happens at domain walls remained elusive until the last few years, when imaging techniques such as atomic force microscopy and high-resolution transmission electron microscopy, combined with first-principles calculations, provided access to atomic-scale characterization of the walls. This characterization brought to light unexpected phenomena such as the conducting behaviour of ferroelectric domain walls in the insulating multiferroic bismuth ferrite4 (BiFeO3) and other non-conducting oxides5,6, and opened perspectives for domain-wall nanoelectronics2.
In their study, Farokhipoor et al. focused on ferroelastic domain walls in terbium manganite (TbMnO3). Beyond revealing a new functionality of domain walls related to the local stabilization of an exotic two-dimensional crystallographic phase, the authors also explained the appearance of a net magnetization in the otherwise antiferromagnetic low-temperature phase of thin films of TbMnO3 and related compounds; in a purely antiferromagnetic phase, spins of neighbouring electrons point in opposite directions, producing no net magnetization.
TbMnO3 is a distorted perovskite, a compound of general formula ABO3, where A and B are two cations of different size and O is oxygen. At low temperatures, bulk TbMnO3 develops a spiral spin structure that breaks spatial inversion symmetry and induces an electric polarization, making it a magneto-electric multiferroic7,8. At the structural level, it adopts at low temperature a common 'Pbnm orthorhombic' lattice configuration, which can be viewed as a distortion of the ideal, high-temperature cubic structure. The distortion primarily involves in-phase rotations of oxygen octahedra about the vertical (z) axis with amplitude Rz+ (Fig. 1a) and anti-phase rotations of oxygen octahedra about the horizontal (x) and (y) pseudo-cubic directions with equal amplitude (Rx− = Ry−; Fig 1b).
When TbMnO3 is grown in epitaxial thin-film form on a strontium titanate (SrTiO3) substrate, as in the present study, it preserves such an oxygen rotation pattern, with the Rz+ rotation axis aligned along the growth direction; in epitaxial growth, the film's atoms are 'aligned' with atoms in the underlying substrate. As discussed by Farokhipoor et al., to accommodate the mechanical constraint induced in the film by the epitaxial growth process, the film naturally develops '90°' ferroelastic domain walls, which are associated with a reversal of one of the Rx− or Ry− rotation patterns (Rx− in Fig. 1d).
By combining experimental and first-principles techniques, Farokhipoor and colleagues demonstrate that, to release the specific mechanical stress inherent in such a domain wall, a systematic chemical substitution of Tb atoms by smaller Mn atoms occurs in every other row along the film's growth direction, producing a new phase with unexpected square-planar MnO4 groups. The extra Mn atoms at the domain wall are responsible for unusual magnetic properties: being located between consecutive MnO2 planes that are antiferromagnetically coupled, these atoms are magnetically frustrated — that is, they cannot simultaneously align or anti-align their spins with those of both neighbouring planes. Such magnetic frustration leads to canting of neighbouring spins and produces a net magnetization.
It is important to understand that the driving force for the chemical substitution at the domain wall is not the motions of the oxygen octahedra themselves, but the presence of additional Tb displacements in the horizontal direction. In ABO3 perovskites, such anti-polar A-cation motions (here, Tb displacements; Fig. 1c) of amplitude DxA = DyA are intrinsic to the Pbnm phase: they are naturally induced by the oxygen rotations through linear coupling of Rx− (Ry−), Rz+ and DxA (DyA) distortions9. As previously discussed in another context10, such odd coupling of three distortions mandates that the reversal of Rx− at the domain wall (while keeping Rz+ unchanged) produces a reversal of DxA. This latter reversal translates into opposite motions of the Tb cations on the left and right sides of the domain wall (Fig. 1d), and creates a non-uniform steric (geometric) effect that is responsible for the selective chemical substitution. This effect is therefore not restricted to TbMnO3, but should be generic for this type of domain wall in orthorhombic perovskites.
It is usually understood that domain walls adjust to release the stress they are subjected to. This is true, but what happens here is more subtle than a simple elastic relaxation. The stress produced at the domain walls studied by Farokhipoor et al. is far from homogeneous. The anti-polar Tb motions produce a peculiar saw-like steric effect modulated at the atomic scale. The domain wall therefore seems to be a unique confined environment that is able to generate and stabilize new crystallographic phases not necessarily achievable by other means.