Predicting vibrational entropy of fcc solids uniquely from bond chemistry using machine learning

Abstract

Despite a well-recognized contribution of vibrational entropy (Svib) in the phase stability of alloys, this remains a peripheral quantity due to its high computational cost. In this work, using a combination of density functional theory (DFT) calculations and machine learning (ML), we show that the expensive Svib computations can be completely circumvented, and the Svib can be predicted with high accuracy. This is possible because using DFT calculations in fcc solids, we show that there exists a unique force constant (FC) – bond length relationship for every bond (i.e., A-A or A-B) and the influence of the alloy composition on FCs can be captured with the change in bond lengths only. By building the FC-bond length DFT database coupled with ML models, we show that the FCs between any two elements can be predicted in any composition. This capability in turn enables predicting the Svib of any complex alloy without having to perform a single new FC calculation thereby significantly reducing the computational costs. This work opens a new avenue to predict Svib of complex HEAs thereby making Svib as readily available as the mixing enthalpy. Using this framework, vibrational entropies may now be included in the thermodynamic frameworks to improve the phase-stability predictions.