Investigation of atomic environments by computational thermodynamics: Applications in intermetallic catalysts and molten salts

• A Dissertation in Materials Science and Engineering by Rushi Gong
• Submitted in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy in December 2024
• Final Version PDF

Abstract

Understanding atomic environments is essential for optimizing the performance of materials by providing insights into underlying structure-property relationships. For complex multi-component solution phases, advanced thermodynamic models are required to capture the inherent complexities of atomic environments in the phase, including both long-range and short-range ordering. CALPHAD modeling, which describes the Gibbs energies of individual phases, is a key tool for investigating multi-component thermodynamics. One of the challenges in CALPHAD modeling is comparing various thermodynamic models and selecting the most appropriate one for each phase. Open-source software tools, PyCalphad and ESPEI, have enabled high-throughput CALPHAD modeling with uncertainty quantification. The integration of Bayesian parameter estimation and the Markov Chain Monte Carlo approach into ESPEI facilitates Bayesian model selection in CALPHAD modeling, enabling systematic identification of the most appropriate thermodynamic models for phase description.

This dissertation discusses the selection and statistical comparison of thermodynamic models for describing atomic environments in individual phases, focusing on intermetallic catalysts and molten salts. Employing a four-sublattice model for the Pd-Zn-based $\gamma$-brass phase successfully predicts the site occupancy of active metals within intermetallic structures. This atomic control of active-site ensembles in Pd-Zn-based $\gamma$-brass intermetallic catalysts improves activity and selectivity for hydrogenation reactions. For complex molten salts, candidate models such as the associate model, two-sublattice ionic model, and modified quasichemical model with quadruplets approximation are considered to describe the short-range ordering. Bayesian statistics are applied to compare these models and identify the most favorable one for molten salt liquids, further enhancing the accuracy of thermodynamic modeling and predictions of salt characteristics. Additionally, a template generator has been developed in the present dissertation to allow users to add customized thermodynamic models within PyCalphad. These advancements provide the community with new advanced opportunities to comprehensively evaluate thermodynamic modeling with uncertainty quantification and model selection, accelerating the optimization process in data-driven materials design and discovery.

List of Chapters

• Chapter 1 - Introduction
• Chapter 2 - Computational methodology
• Chapter 3 - Thermodynamic modeling of the Pd-Zn-M system for intermetallic catalysts design
• Chapter 4 - Thermodynamic modeling of fluoride and chloride molten salts with model selection, uncertainty quantification, and uncertainty propagation
• Chapter 5 - Exploring and implementing thermodynamic models for complex liquid in open-source software PyCalphad
• Chapter 6 - Conclusion

Vita

Rushi Gong was born on October 15, 1997, in Yiyang, Hunan Province, China. She grew up in Changsha, Hunan, and later moved to Shenzhen, Guangdong. In 2019, she graduated from Beihang University in Beijing, China, with a B.S. in Materials Science and Engineering and a minor in Mathematics. In 2020, Rushi began her Ph.D. studies at the Pennsylvania State University, advised by Dr. Zi-Kui Liu.

As of 2024, Rushi has published several scientific publications listed under Google Scholar (id:naO8fOgAAAAJ) including 3 first-author publications and 5 co-author publications.