Abstract by Hayden Oliver
Physics and Astronomy
Use of Moment Tensor Potentials as a high-throughput method for evaluating materials
The discovery of a cobalt-based γ/γ′-forming superalloy in 2006 inspired our computational investigation of the Co-Al-W system. We use the Moment Tensor Potentials (MTP) active-learning framework to predict the total energy, forces, and stresses for nearly 300,000 derivative superstructures in the Co-Al-W system. We report several new structures on the convex hull with the ordered L12-Co3(Al, W) crystal structure, which may correspond to the experimentally discovered γ′ phase. Small additions of tantalum, titanium, and vanadium lower the formation enthalpy of the L12 structure, which could act as γ′ stabilizers at high temperature. While the total cost of the MTP analysis is about 1/1000th the cost of a similar result using DFT, the difference between calculated and predicted results is less than 3 meV/atom. MTP can be used for high-throughput phase stability analysis for many kinds of materials, such as high-temperature superalloys, multi-principle element alloys, high-Tc magnetic materials, superhard materials, and electrochemical materials.