What Defines a Perovskite?

After reading the recent publication of Jodlowski et al.1 with great interest, we would like to address a misconception, that is unfortunately common in the field. Jodlowski et al. give a great and comprehensive overview over the different possibilities of overcoming the intrinsic problems of lead based halide perovskite semiconductor materials, including cation and anion substitutions.1 The fundamental problem, however, lies in the use of the term “perovskite” for all of these structures. The perovskite structure type is a clearly defined term that should be used in its correct way, and we would like to urge the research community to obey these rules. In order to define the term perovskite properly, one should note the following fact: The mineral named perovskite, CaTiO3, was first described by Rose, a German mineralogist who described the habit and some reactions of what he named perovskite after Russian mineralogist Lev Perovski.2 We would argue that materials that can be counted to this structural family should exhibit a close structural similarity to the crystal structure of the defining mineral Perovskite (CaTiO3) and some criteria for this are given below. It might be, that some of the confusion could stem from the fact that at ambient conditions, CaTiO3 itself does not crystallize in the cubic aristotype structure, but rather in an orthorhombic structure which can be derived from the cubic aristotype in a group-subgroup relationship.3 Generations of crystallographers and solid state chemists have dedicated their work to the elucidation of the structural relations between the cubic aristotype of the perovskite type structure and its many lower symmetry variations (hettotypes).4 In fact, the perovskite type is one of the classical examples for the use of group–subgroup relationships and we would like to highlight the work of Bock and Müller5 and Bärnighausen6 on the thorough understanding and rationalization of these structural relationships and ultimately, a perovskite needs to adopt a crystal structure of the perovskite aristotype or a hettotype thereof. The reason for these rules lies in the close relationship between structure and properties. This is in the case of semiconductors predominantly the relationship between atomic and electronic structure: Once the octahedra are no longer organized in a 3D network, the electronic properties will become significantly anisotropic and are hence no longer comparable to those of the original structure type. It is unfortunate that we need to insist on the fact that the B-cations should be octahedrally coordinated, as we found examples in the literature, where “perovskite” materials were made of isolated metal-halide tetrahedra.7 We do agree with the authors that compounds in the elpasolite-type structure with a general formula of A2BB'X6 could be understood as double perovskites, since the corner sharing network of octahedra is still in place. As an example, we would like to highlight that structures in the K2PtCl6-type, such as Cs2SnI6 should by no means be understood as perovskites. While Jodlowski et al. clearly state that the K2PtCl6 structure type is composed of isolated [BX6] octahedra, they base their reasoning for its categorization as perovskite on two main points1: 1) the proximity of the [BX6]2− octahedra and 2) “the vacancy-ordered double perovskites experience the same loss of symmetry in phase transitions upon cooling as that occurring in ABX3.”1 Unfortunately, we are unable to agree to both of these points: 1) the closest inter-octahedral I–I distance in Cs2SnI6, for instance, is 4.18 Å,8 which is significantly longer than twice the corresponding van-der-Waals radius of iodine (2∙1.98 Å = 3.96 Å)9 and hence no bonding would be expected between the isolated octahedra. 2) The fact the K2PtCl6-type structures show symmetry reduction upon cooling is nothing that is mutually unique with the perovskite-type structures but it is rather a general effect that low temperature phases are of lower symmetry than high-temperature phases. Analogous to the above-mentioned case, neither of the structure types further discussed in the paper obeys a 3D network of corner sharing octahedra (or distorted octahedra) and therefore cannot be regarded as of perovskite-type. In particular for the Ruddlesden–Popper phases, we would like to draw the attention of the reader to the original publications of Ruddlesden and Popper,10, 11 where the authors point out that these compounds contain perovskite layers. Perovskite layers are, however, not layered perovskites and no matter how small the change in semantics may appear, the latter would not be correct. Finally, we do understand the incentive of the research community to rationalize this category of materials under a common naming and do believe that this facilitates the discussion amongst the research community. With the reasoning laid out above, however, we do not believe that the term “perovskite” is appropriate for all of the materials but should only be used for materials with a clear relation to the perovskite-type structure. We would like to instigate a more general debate in the community as to which term should be used and is scientifically sensible. It appears to us that “halide semiconductor” materials may be useful, since the halide anions appear to be a common feature in this family of compounds. We would like to insist on the fact that our comment and concerns are solely concerned with the terminology used in this article, and unfortunately many others, but not with the scientific quality of this review. The authors declare no conflict of interest.