Exploring DFT-based phase diagram calculations using a modified CALPHAD approach

Phase diagrams calculated entirely from first-principles provide important thermodynamic information to new material systems at the prediction stage, and also have the potential to reduce both time and expense in investigations for materials design. Combining DFT calculations with a modified CALPHAD approach we demonstrate how a limited number of first-principles calculations can be used predictively to produce a complete thermodynamic model.

The aluminium-nickel system was chosen as a test case as it contains many features commonly found in phase diagrams, including ordered and disordered phases of various structures including bcc and fcc. The progress in reproducing an experimentally-based phase diagram using DFT calculations combined with the CALPHAD approach is shown, and some of the shortcomings and suggestions for improvements discussed.

Commonly, thermodynamic descriptions made using the CALPHAD approach use the Bragg-Williams approximation for the configurational entropy of a solid, which is a point correlation model ignoring the pair and higher order interactions. By considering a toy model of a binary system with fcc-based phases, we introduce pair and tetrahedral contributions using Cluster Variation Method (CVM) to the configurational entropy and examine the effects of these changes on the order-disorder transitions in the resulting phase diagram. It is hoped that introducing these atomic correlations to the configurational entropy model in the thermodynamic description will provide some additional insight for the DFT-based phase diagram.