Lam, Isaac Keat2023-03-172023-03-172022https://udspace.udel.edu/handle/19716/32464Three primary topics are investigated in this work: fabrication and characterization of high Ga Cu(In,Ga)Se2 (CIGS) made by the precursor reaction method, quantification of Si surface passivation quality by capacitance spectroscopy, and machine learning prediction of defect characteristics in perovskite solar cells. ☐ Research on the precursor reaction method thus far has been restricted to relatively low Ga compositions with (Ga/(In+Ga) ≈ 25%). By increasing the ratio of Ga, it is possible to increase bandgap and thus increase open circuit voltage. This work develops and characterizes the precursor reaction process for use with Ga/(In+Ga) ≈ 50%, with the goal of improving open circuit voltage and efficiency. It is shown that with increased Ga ratio, increased Voc is achieved, but the rate of conversion from precursor to absorber phases is decreased. Additionally, increased Ga improves film adhesion at increased selenization temperatures as well as improving film morphology. ☐ Passivation of c-Si surfaces using H2S gas has been demonstrated to yield high quality passivation and high minority carrier lifetime. Metal-insulator-semiconductor (MIS) devices are prepared on n-type Si wafers with three different passivation types, SiNx, H2S, and a-Si. Capacitance-voltage characteristics and lifetime data are collected for the various samples and used to extract fixed charge, density of interface traps, and capture cross sections. Samples that received H2S passivation show low interface trap density, implying high quality c-Si surface passivation. ☐ Capacitance measurement techniques are powerful methods for characterizing semiconductor devices. Voltage dependent admittance spectroscopy (C-V-f) has recently been used to characterize electronic loss mechanisms in CIGS solar cells. However, it can be difficult to analytically extract meaning from this data. Drift-diffusion modeling can be used to simulate a large dataset of C-V-f maps with known defect characteristics. By training a machine learning algorithm with simulated data, dominant loss mechanisms can be identified in solar cells. This work demonstrates the technique through application to perovskite solar cells.Solar cellAbsorber fabricationCapacitanceVoltageCIGSThesis1373237913https://doi.org/10.58088/zsjb-py232023-02-14en