Mapping phase diagrams with deep generative models

Computing free energy differences is a computationally demanding task, requiring a rigorous sampling of the phase space. Here, we train a machine learning model based on conditional normalizing flows to map between probability distributions of condensed phase systems at different thermodynamic conditions. Using the trained flow model, a large number of uncorrelated configurations can easily be generated to efficiently estimate ensemble averages. This allows us to assess free energy differences over a wide range of temperatures and pressure, needed to evaluate the relative stability of different phases and reconstruct phase diagrams.