Assessed uncertainty in the Ti-W phase diagram combining experimental and ab-initio data

Abstract

The binary Ti-W system is reassessed with uncertainty quantification using the open source software ESPEI (Extensible, Self-optimizing Phase Equilibria Infrastructure). Mixing energy data at 0 K from ab-initio are used to supplement the phase diagram data measured at 1373 K and above. For the parameter optimization, Bayesian Inference with Markov Chain Monte Carlo is used as implemented in ESPEI, which allows for an inherent uncertainty quantification of the model parameters and a propagation of the uncertainty to the phase diagram, mixing enthalpy, and the position of invariant points. The new phase diagram shows a higher solubility of Ti in BCC_A2 W at low temperatures compared to older phase diagrams obtained from phase diagram data only.