Software tools for high-throughput CALPHAD from first-principles data

We present a set of software tools that largely automate the process of converting ab initio data into thermodynamic databases that can readily be imported into standard thermodynamic modeling softwares. These tools are based on the Special Quasirandom Structures (SQS) formalism, extended to transparently handle, not only traditional fcc, bcc and hcp solid solutions, but also multiple-sublattice structures with possible sublattice disorder. A large database of pre-generated SQS is provided that covers over 30 of the most common multi-sublattice structures and spans the composition ranges of each of their sublattices. In addition, we exploit a theoretically justified and robust method to address the issue of assigning free energies to mechanically unstable “virtual” phases, thus providing a compelling solution to a long-standing problem in CALPHAD modeling, especially in the context of ab initio data. We also propose a simple low-order approximation scheme to include short-range order effects that requires no additional ab initio input. The resulting thermodyamic database seamlessly combines ab initio data (formation energies and, optionally, vibrational free energies) with elemental Scientific Group Thermodata Europe (SGTE) data. The proposed tools provide a clear path to expand the coverage of high-throughput efforts towards non-stoichiometric phases and non-zero temperatures. The generated free energy models can also provide very good starting points to perform complex thermodynamic assessments, especially in cases where the available experimental data poorly constrain some thermodynamic parameters. The Cu-Pt-W phase diagram is calculated as an example.