Carolin Wahl defends Thesis
Carolin Wahl defends her thesis, "High-Throughput Materials Discovery with Nanomaterial Megalibraries" on January 10th, 2024.
PhD Abstract:
Although the discovery of new materials drives most areas of technological development, it is traditionally a slow, serial trial-and-error process. Indeed, since the beginning of time, researchers have only structurally characterized ~500,000 inorganic materials, which represents a small subset of the vast number of possibilities. For example, if one considers the number of possible combinations of just four out of the 61 metals with stable isotopes and then includes a consideration of particle size over the 1-100 nm length scale, there are over 1010 possibilities. Therefore, developing high-throughput materials discovery platforms is essential for fully evaluating nanomaterials and identifying the structures that have the properties most impactful for a given technological need.
This dissertation describes developments in the high-throughput synthesis and screening of nanomaterials using a platform termed “nanomaterial megalibraries”, a spatially encoded centimeter-scale chip that contains millions to billions of individually addressable nanomaterials. Megalibrary preparation relies on scanning probe lithography (SPL) techniques that can be easily parallelized to prepare millions to billions of materials in a single experiment. This dissertation begins with an introduction to materials discovery, synthetic strategies for preparing materials libraries, and corresponding high-throughput screening techniques (Part I). Then, a machine learning approach is applied to megalibraries for the first time by building a model that predicts structural features of complex nanoparticles from their compositions, enabling the targeted design of nanomaterials with unprecedented chemical complexity (Part II). One of the major achievements to come out of this work is the expansion of the elements that are synthetically accessible via scanning probe lithographies, thus enabling stable metal in the periodic table to be prepared in a megalibrary (Part III). Through the design of a phase-separating nanoreactor system, megalibraries of metal or metal oxide nanoparticles with arbitrary complexity are synthesized. Characterization and screening are key areas of this work and thus a high-throughput screening platform for megalibraries using automated electron microscopy was developed (Part IV). The development of an electron-transparent substrate for megalibrary synthesis that is compatible with atomic-resolution characterization is described, and machine learning for the classification of nanoparticles is developed in pursuit of automating the data acquisition from megalibrary samples. Finally, an automated analytical framework for the identification of a crystal system of arbitrary unknown structures from nanoparticle electron diffraction data by accumulating information across datasets is introduced. Advanced electron microscopy techniques are used to elucidate the nature of paracrystalline motifs on the surfaces of iridium oxide electrocatalysts, which resulted in the development of a derived strontium iridate material that vastly outperforms commercial IrO2 in terms of stability (Part V). The thesis concludes with a summary and discussion of future directions based on this research. Overall, this work outlines an end-to-end framework for high-throughput materials discovery, comprised of synthesis, screening, and AI-integrated analytics. It may serve as a starting point for further developments in these three areas towards a fully autonomous materials discovery platform.