Obtaining highly transferable parameters for empirical tight-binding simulations: from pure materials to polytypic heterostructures and dilute alloys

The tedious parametrization process remains a key bottleneck hindering broader adoption of the empirical tight-binding method. Here, we discuss the challenges and requirements of finding parameters for Slater-Koster-based empirical tight-binding schemes, and explain the shortcomings of the parametrization procedures commonly used in the literature. We then propose our way to address these challenges, through the presentation of an open-sourced parametrization toolbox that breaks up the task into several, incremental stages. The prototype of our toolbox is tailored especially to deal with a recently proposed, highly transferable scheme, but can be easily adapted to any Slater-Koster-based scheme. A case study of fitting GaN parameters is considered. Remarkably, the obtained GaN parameter set shows high transferability when applied to various contexts: from simple zincblende or wurtzite bulk cases, to polytypic heterostructures and even highly mismatched alloys in their dilute limits. This confirms the effectiveness and robustness of our parametrization approach.