Development of Stable, High Optoelectronic Quality Perovskites Using Photoluminescence, Photoconductivity, and Machine Learning

Abstract

Improving the economics of solar cells is a key component to enable rapid adoption of clean electricity. Reducing the cost of photovoltaics can be attained by improving power conversion efficiency, increasing module lifetime, and reducing processing and CAPEX costs. However, mature technologies such as silicon are presently approaching fundamental efficiency limits, and cost reductions have plateaued. Hybrid Perovskites (HPs) are an emerging material class that show promise since they are inexpensive to fabricate via solution processing, and they have broadly tunable material properties. The HP bandgap can be precisely tuned to pair with silicon in a tandem solar cell, increasing the theoretical power conversion efficiency limit with minimal additional processing costs. The work presented in this dissertation is broadly focused on improving the performance and reliability of perovskite materials for photovoltaic applications. First, I demonstrate a photoluminescence-photoconductivity (PL-PC) technique that directly measures HP quasi-Fermi level splitting (∆EF) and carrier diffusion length (LD), which are predictors of device open-circuit voltage and short-circuit current respectively. I use this PC-PL technique to highlight several cases where ∆EF and LD are anti-correlated and show the importance of quantifying LD in determining overall absorber optoelectronic quality. Next, I focus on development of high-bandgap HPs for tandem applications. Although halide mixing can increase the bandgap to make an ideal bandgap pairing with silicon in tandem applications, the mixed-halide HPs suffer from lower relative voltages and phase instabilities under illumination. Using high-throughput combinatorial exploration, I identified a new compositional motif using larger guanidinium and smaller cesium to form a stable HP structure with enhanced lattice strain. This class of mixed-halide HPs with enhanced strain show modified band structure, higher film ∆EF, and device open-circuit voltage. I expanded the combinatorial exploration of HP compositions and collect a dataset of 13,000 photoluminescence (PL) spectra indexed by composition. Additionally, I show a proof of concept of using a machine learning approach to predict material bandgap and optoelectronic quality fraction. I also study the origin and impact of phase segregation in mixed-halide HPs and discover phase segregation occurs due to excess charge carriers, which can be generated either by photoexcitation or by current injection. Finally, I study degradation of HP films and devices in various combinations of light, atmosphere, humidity, and thermal stresses and use a machine learning model to link early time behavior to the time it takes the material or device to degrade. This effort reveals simple optical measurements such as transmittance and dark field microscopy have considerable utility in identifying early signs of HP degradation. Collectively, the work included in this dissertation demonstrates considerable progress toward understanding optoelectronic performance and reliability of a wide class of HP materials, with focus on high-bandgap HPs useful for tandem applications.