User-friendly and robust Calphad optimizations using Calphad Optimizer in FactSage

Abstract

The Calphad Optimizer available since FactSage 8.2 is presented. It contains both a user-friendly graphical interface, as well as input and output files that are easily readable by humans and machines. For a number of typical experiments (liquidus and solidus temperatures, two-phase equilibria, heat of mixing, heat of reaction, enthalpy increment) tailored recipes exist to enable beginners and non-specialists the input of the relevant data. For many other experimental data, user-defined experiments can also be created. Two different optimizers are currently implemented. While running the optimization, the user is presented with graphs showing the convergence progress, both globally (error sum vs. iteration), as well as for each experiment (comparison between experimental and calculated data) and for each optimization parameter (error sum vs. parameter value). As a first post-processing option, phase formation maps can be calculated.

In the latest version of the Calphad Optimizer, available since FactSage 8.3, the error sum is calculated primarily based on Gibbs energy differences, not the raw experimental data. The details of the implementation are described. Using the Pb-Sn and Cr-Ni systems as examples, we show how the change in the error sum calculation increases robustness and speed of the optimization, making one-shot Calphad database optimizations possible.