Combining High-Throughput Calculations with Machine Learning to Predict Phase Preferences of Transition Metal Dichalcogenides

Layered transition metal dichalcogenides (TMDs) are a chemically distinct and technologically important family of materials, which can adopt one of the several known crystal structures. A question that naturally arises is: <i>What dictates the observed phase preference of TMDs?</i> An answer to this important question can help to understand and hence, manipulate composition-structure-property relationships, paving the way for engineering properties of TMDs. In fact, this is an old problem, which was actively debated in the 1960s to1980s. It has once again become an important and relevant question with the emergence of 2D quantum materials. By combining high-throughput density functional theory-based calculations with machine learning techniques, we address this six-decade old question about TMDs for a much larger chemical phase-space than what was considered in the earlier works. Our analysis not only rediscovered known physicochemical attributes considered by earlier researchers, but also discovered other factors that were not previously known to influence the structural preferences displayed by TMDs. This work [Phys. Rev. Materials <b>6</b>, 094007 (2022)] demonstrates how machine learning can be used to tackle old problems in Condensed Matter Physics.<br/><br/><b>Funding:</b> This work is supported by the National Science Foundation (DMR-1752840). The computational support is provided by XSEDE under Project PHY180014.