�ڰ�������

Optional
Meeting ID: 962 9148 7208
Passcode: 762670

Abstract: Materials selection has been a challenge since the dawn of time. With the advent of computational methods and high throughput databases, materials can be discovered and assessed before they are even synthesized. In this thesis I will discuss my work towards designing new materials for clean energy uses, especially alloys which have been underexplored from computational methods. To begin, I present a case study explaining the synthesis of the Yb14MnSb11 thermoelectric material. Next, I explore the class of AM2Pn2 Zintl materials. From the previously identified BaCd2P2, it is suspected that other materials of this stoichiometry may also exhibit its same “defect-tolerance.” I find that there is a large number of other AM2Pn2 materials that are stable and isostructural to BaCd2P2, while possessing a large range of bandgaps which could make them useful for a variety of devices, such as infrared detectors, thermoelectrics, single junction photovoltaics, and tandem solar cells. Extending this class of materials, I turn to the idea of creating alloys among the AM2Pn2 to take advantage of their isostructural nature. Focusing on solar absorber materials for tandem top cells, I search among all possible quaternary alloys in this family and screen for the most promising candidates. I identify Mg-alloying as a key strategy to increase the bandgap high enough for a tandem top cell solar absorbers. I show Ca(Cd0.8Mg0.2)2P2, one of the most promising candidates, has the ideal bandgap of 1.8 eV for optimal efficiency and find that its bandgap is direct in nature. Finally, I present a comparison of the cluster expansion and machine learned interatomic potential (MLIP) methods for predictions of Fe-Mn-Ni-Al-Cr high entropy alloys. From a set of about 5,000 DFT calculations for training, I show that each method is able to achieve a similar level of accuracy of about 30 meV/atom, but that the performance of MLIPs from the Multi Atomic Cluster Expansion (MACE) converge faster. This aids in addressing the outstanding challenge of including disordered phases in computationally predicted phase diagrams.

Thesis Committee: Geoffroy Hautier (Chair), Jifeng Liu, Emme Burgin, Wenhao Sun (University of Michigan)