<p>A wide range of optoelectronic applications demand semiconductors optimized for purpose.</p> <p>My research focused on data-driven identification of ABX3 Halide perovskite compositions for optimum photovoltaic absorption in solar cells.</p> <p>I trained machine learning models on previously reported datasets of halide perovskite band gaps based on first principles computations performed at different fidelities.</p> <p>Using these, I identified mixtures of candidate constituents at the A, B or X sites of the perovskite supercell which leveraged how mixed perovskite band gaps deviate from the linear interpolations predicted by Vegard's law of mixing to obtain a selection of stable perovskites with band gaps in the ideal range of 1 to 2 eV for visible light spectrum absorption.</p> <p>These models predict the perovskite band gap using the composition and inherent elemental properties as descriptors.</p> <p>This enables accurate, high fidelity prediction and screening of the much larger chemical space from which the data samples were drawn.</p> <p><br></p> <p>I utilized a recently published density functional theory (DFT) dataset of more than 1300 perovskite band gaps from four different levels of theory, added to an experimental perovskite band gap dataset of \textasciitilde{}100 points, to train random forest regression (RFR), Gaussian process regression (GPR), and Sure Independence Screening and Sparsifying Operator (SISSO) regression models, with data fidelity added as one-hot encoded features.</p> <p>I found that RFR yields the best model with a band gap root mean square error of 0.12 eV on the total dataset and 0.15 eV on the experimental points.</p> <p>SISSO provided compound features and functions for direct prediction of band gap, but errors were larger than from RFR and GPR.</p> <p>Additional insights gained from Pearson correlation and Shapley additive explanation (SHAP) analysis of learned descriptors suggest the RFR models performed best because of (a) their focus on identifying and capturing relevant feature interactions and (b) their flexibility to represent nonlinear relationships between such interactions and the band gap.</p> <p>The best model was deployed for predicting experimental band gap of 37785 hypothetical compounds.</p> <p>Based on this, we identified 1251 stable compounds with band gap predicted to be between 1 and 2 eV at experimental accuracy, successfully narrowing the candidates to about 3% of the screened compositions.</p>

In dieser Arbeit wurde die atomare Struktur von Si(100)- und Ge(100)-Oberflächen untersucht, die mit metallorganischer chemischer Gasphasenabscheidung (MOCVD) für anschließende Heteroepitaxie von III-V-Halbleitern präpariert wurden. An der III-V/IV Grenzfläche werden atomare Doppelstufen auf der Substratoberfläche benötigt, um Antiphasenunordnung in den III-V-Schichten zu vermeiden. Die MOCVD-Prozessgasumgebung beeinflusst die Domänen- und Stufenbildung der Si- und Ge(100)-Oberfläche sehr stark. Deswegen wurden in situ Reflexions-Anisotropie-Spektroskopie (RAS) und Ultrahochvakuum-(UHV)-basierte oberflächensensitive Messmethoden verwendet, um die verschiedenen Oberflächen zu charakterisieren. In situ RAS ermöglicht die Identifizierung der Oberflächenstruktur und somit Kontrolle über die Oberflächenpräparation, insbesondere der Domänenbildung auf Si- und Ge(100). Beide Oberflächen wechselwirken stark mit dem H2-Prozessgas, was zu Monohydrid-Bedeckung während der Präparation führt und sogar zu Si-Abtrag während Präparation unter hohem H2-Druck. Die Erzeugung von Leerstellen auf den Terrassen bewirkt eine kinetisch bedingte Oberflächenstruktur, basierend auf Diffusion von Leerstellen und Atomen. Dadurch kommt es zu ungewöhnlichen DA-Doppelstufen auf verkippten Si(100)-Substraten während auf exakten Substraten ein schichtweiser Abtrag stattfindet. Unter niedrigem H2-Druck bildet sich eine energetisch bedingte Domänen- und Stufenstruktur. Während das H2-Prozessgas keinen direkten Einfluss auf die Stufen- und Domänenbildung von verkippten Ge(100)-Oberflächen zeigt, ist der Einfluss von Gruppe-V-Elemente entscheidend. Die As-terminierten Ge(100)-Oberflächen bilden eindomänige Oberflächen unterschiedlicher Dimerorientierung und Stufenstruktur abhängig von Temperatur und As-Quelle. Angebot von P an Ge(100)-Oberflächen durch Heizen in Tertiärbutylphosphin führt zu einer ungeordneten, P-terminierten Ge(100)-Oberfläche, die instabiler als die Ge(100):As-Oberfläche ist. / In this work, the atomic surface structure of Si(100) and Ge(100) surfaces prepared in metalorganic chemical vapor phase deposition (MOCVD) ambient was studied with regard to subsequent heteroepitaxy of III-V semiconductors. At the III-V/IV interface, double-layer steps on the substrate surface are required to avoid anti-phase disorder in the epitaxial film. The MOCVD process gas ambient strongly influences the domain and step formation of Si and Ge(100) surfaces. Therefore, in situ reflection anisotropy spectroscopy (RAS) and ultra-high vacuum-based (UHV) surface sensitive methods were applied to investigate the different surfaces. In situ RAS enabled identification of the surface structure and the crucial process steps, leading to complete control of Si and Ge(100) surface preparation. Both surfaces strongly interact with H2 process gas which leads to monohydride termination of the surfaces during preparation and Si removal during processing in high H2 pressure ambient. The generation of vacancies on the terraces induces a kinetically driven surface structure based on diffusion of vacancies and Si atoms leading to an energetically unexpected step structure on vicinal Si(100) substrates with DA-type double-layer steps, whereas Si layer-by-layer removal occurs on substrates with large terraces. Processing in low H2 pressure ambient leads to an energetically driven step and domain structure. In contrast, H2-annealed vicinal Ge(100) surfaces show no direct influence of the H2 ambient on the step structure. At the Ge(100) surface, group-V elements strongly influence step and domain formation. Ge(100):As surfaces form single domain surfaces with different majority domain and significantly different step structures depending on temperature and As source, respectively. In contrast, exposure to P by annealing in tertiarybutylphosphine leads to a very disordered P-terminated vicinal Ge(100) surface which is less stable compared to the Ge(100):As surfaces.