Author ORCID Identifier

https://orcid.org/0000-0003-1719-5281

Date of Award

2026

Document Type

Thesis (Ph.D.)

Department or Program

Engineering Sciences

First Advisor

Jifeng Liu

Abstract

The discovery of new photovoltaic semiconductors is challenging because performance depends on the simultaneous optimization of band gap, optical absorption, carrier lifetime, and defect tolerance. We address this challenge through an experiment-theory feedback loop, testing a promising group of pnictide semiconductors identified through high-throughput computational screening. Our experimental results provide more insights to further improve the theoretical screening process. We present an investigation of the optoelectronic properties of the Zintl pnictide BaCd2P2 (BCP), showing that it exhibits bright band-to-band photoluminescence emission, strong photoconductive-current generation, a room-temperature carrier lifetime of up to 300 ns, and an implied open-circuit voltage comparable to that of GaAs, despite having substantial impurities. We then use first-principles calculations to study the defect properties of BCP, identifying the origins of both deep- and shallow-defect-related transitions and showing that the nonradiative recombination rate associated with its dominant deep defect is lower than that of GaAs. These results suggest that BCP is unusually tolerant to impurity-assisted nonradiative recombination and provide a microscopic explanation for its long carrier lifetime. We then extend the search for defect-tolerant photovoltaics to Zn-based pnictides, eliminating the toxic Cd component of BCP. We investigate the optical properties of CaZn2P2 thin films as a potential wide-gap top-cell absorber, demonstrate p-type doping through Cu ion exchange, and observe rectification in a CaZn2P2/Si heterojunction. We also examine ZnP2 and analyze its carrier lifetime, including preliminary studies of trapping and recombination dynamics. Finally, we develop a graph neural network architecture for machine-learning interatomic potential (MLIP) models that accelerate density functional theory (DFT) calculations. We achieve over 3×acceleration in DFT computation time when relaxing structures using our MLIP model prior to DFT calculations. We also present multiple new datasets of hybrid functional DFT calculations, curated over multiple years of continuous computation. This thesis demonstrates how experiment and theory can be combined to accelerate the discovery of new photovoltaics.

Download

Share

COinS