Design of Multi-Principal Element Materials Using a Combination of First-Principles Calculations and Machine Learning
Date of Award
Doctor of Philosophy (PhD)
Multiple-principal element materials, including alloys (MPEAs), oxides and other compounds combine several elements on a sub-lattice, which increases the configurational entropy and contributes to their thermodynamic stability. The vast combinatorial design space of multi-principal element materials gives rise to diverse microstructures and unprecedent combination of properties. It also poses a challenge—how to rapidly screen MPEAs with stability and properties for targeted applications. Furthermore, fundamental composition-structure-property relationships in MPEAs are currently missing. In this dissertation, I have used MPEAs as a representative system and investigated the role of chemical composition on their phase-stability and mechanical properties. With a combination of first-principles calculations and materials informatics, I have investigated factors involving entropy, enthalpy, electronic structure, and distortion, and developed models that can be generalized to design new multi-principal element materials with desired phase(s). The first part of the dissertation focuses on the rapid identification of stable intermetallics. A machine learning model is developed to predict the formation enthalpy of intermetallics given their chemical composition. The model uses easily accessible elemental properties as descriptors and have a mean absolute error of 44 meV/atom on testing binary alloy. With the ML model, we identified new intermetallics that are subsequently confirmed with DFT calculations. We further extend this model to predict compositional complex intermetallics using adaptive transfer learning. We further probe the phase evolution in NbVZrTix (x = 0 – 1) and Al-Cu-Mn alloys with DFT calculations, which are confirmed experimentally by our collaborators.. In the second part, we develop a thermodynamic model that can predict the mixing enthalpy of multicomponent solid solutions using pairwise interactions. We establish a database that contains the mixing enthalpy of 17 refractory metals in BCC lattice from DFT calculations and fit them to accurately capture the mixing enthalpy of over 20,000 equimolar multicomponent BCC solid solutions. By comparing their energy with DFT-calculated enthalpy of intermetallics and using convex hull analyses, we identify the stable phase(s) of any refractory MPEA as a function of temperature. The predicted stability of NbTiZr, NbTiZrV, and NbTiZrVM ¬(M = Mo, Ta, Cr) systems agree well with prior experimental observations. We apply this model to assist experimental design in NbVZr-based MPEAs. The predicted phase evolution in NbVZr-Tix (0 < x < 1) matches well with laser-processed alloy libraries. We further derive several simple criteria for screening single-phase MPEAs. In the last part, we extend the above method to the design of multicomponent high-entropy ABO3 perovskite oxides (HEPO). We employ Sr2+ as A-site cations and consider 14 transition metals as candidate B-site cations. We use charge balance, tolerance factor, entropy and enthalpy obtained from two-component double perovskite oxides as descriptors to identify the stability of single-phase perovskites from over 3,000 HEPO compositions. Model predicted single-phase HEPOs, such as Sr3(NbZrFe)O9, Sr3(NbZrV)O9 and Sr3(NbZrTi)O9 have been confirmed with DFT calculations.
Katharine Flores, Kenneth Kelton, Zohar Nussinov, Li Yang,
Available for download on Monday, August 26, 2024
Computational Chemistry Commons, Materials Science and Engineering Commons, Mechanics of Materials Commons