Date of Award

Spring 6-21-2019

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

First Advisor

Charles B. Musgrave

Second Advisor

Alan W. Weimer

Third Advisor

James W. Medlin

Fourth Advisor

Adam Holewinski

Fifth Advisor

Ronggui Yang

Abstract

The advancement of society has been historically predicated on the discovery or invention of new materials and in particular, inorganic solid-state materials have had transcendent influence on society – stone for tools, steel for structures, silicon circuits and solar cells, and so forth. Synthesizing a new material is time-consuming, costly, and frustrating for those tasked with the job. The success of solid-state synthesis can be greatly improved if one knows the thermodynamic stability of the material they are trying to make and those they are trying to avoid. This dissertation addresses the prediction of thermodynamic stability for solid-state materials primarily using a branch of quantum chemistry called density functional theory (DFT) and statistical approaches that fall under the umbrella of data analytics and machine learning.

We partitioned the pathways of solid-state decomposition into three types to quantify the success of DFT-based approaches for predicting thermodynamic stability in a high-throughput manner. By comparing with experiment, we find that in general, DFT performs quite well. Importantly, when the decomposition pathway type is elucidated for all known inorganic crystals, we find that the type that DFT performs the best on is the most prevalent, supporting the efficacy of DFT-based stability predictions.

Still, DFT is computationally expensive and not always practical for a given problem. This motivates the use of data analytics to accelerate the prediction of thermodynamic stability using so-called “descriptors”. We applied the SISSO (sure independence screening and sparsifying operator) algorithm to identify a new tolerance factor (descriptor) for predicting the experimentally realized stability of perovskites, which are a class of inorganic solids having significant utility as solar absorbers, catalysts, and capacitors.

This new tolerance factor was applied to identify new double perovskite solar absorbers in the cesium-chloride chemical space. In doing so, we gain insights into the stability of these materials, point out some pitfalls of common high-throughput approaches, and reveal a number of potential all-inorganic solar absorbers which may become active components in high-efficiency solar cells.

Much of the computational materials field is resigned to studying temperature-independent thermodynamics because of the expense of including the effects of vibrational entropy in the solid-state. To address this problem, we again used SISSO, this time to identify a simple descriptor for the Gibbs energy of an arbitrary inorganic crystalline solid. We show how this descriptor can be used for rapid predictions of temperature-dependent stability and thermochemical equilibrium.

As a demonstration of the utility of the Gibbs energy descriptor, we used it to screen for active materials that might be able to mediate the conversion of air, water, and sunlight into ammonia using chemical looping. These results provide a detailed thermodynamic analysis of the involved reactions for this process, highlighting the challenging tradeoff between metal oxide and metal nitride stability that must be met for the process to succeed.

This work helps reveal the lack of exploration of metal nitride compounds relative to their oxide counterparts. We show that the space of ternary metal nitrides that have been synthesized has the potential to double based on DFT-based stability predictions. We also developed quantitative descriptors for the bonding in metal nitrides to help rationalize their stability and highlight opportunities for synthesizing new nitrides with interesting technological properties.

Available for download on Thursday, January 27, 2022

Share

COinS