From Bronze Age alloys to advanced ceramics, technological progress heavily relies on the development and use of polycrystalline materials whose performance is largely dictated by crystallite interfaces (i.e. grain boundaries). Interfaces between different materials are also key to designing heterostructures for use in a range of modern applications (e.g. solar cells, photocatalysts, quantum dot displays). Here, the formation of well-ordered interfaces is achieved by controlled deposition of one semiconductor on the surface of another which is largely constrained by epitaxial matching.
Recently, top-down manipulation of two-dimensional (2D) materials has created a new class of layered heterostructures in which epitaxial constraints are less pronounced due to the relatively weak van der Waals interfacial interactions. The resulting freedom to carefully tune the interfaces in such systems (e.g. relative in-plane twist angles of layers) is highly promising for developing the next generation of 2D nanodevices and has already yielded spectacular new phenomena.
Although clearly playing a huge role in established and emergent technologies, interfaces are highly complex systems whose properties are typically difficult to predict and/or rationalise. Computational modelling is playing an increasingly important role in helping to analyse and understand interfaces. Recent methodological advances have tended to focus on approaches for searching for detailed low- energy atomic/electronic structures of selected interfaces. However, given the huge number of possible ways in which two surfaces can interact, efficient and accurate screening of the energetic/structural landscape of viable interfaces is a pre-requisite for more in-depth investigations. Machine learning has been used to screen the structures and energies of metallic tilt grain boundaries, but required prior training with 10,000’s of calculated examples. Here we address the screening challenge with a simple, powerful and direct modelling approach which, in principle, allows for rapid, unconstrained and systematic exploration of energies and structures of interfaces between arbitrary solid surfaces.
