Infrared Spectroscopy Data-driven Machine learning for Unveiling Thermolysis Pathways of Metal-Organic Frameworks

Abstract

This project aims at developing machine-learning approach to unravel the thermolysis pathways of metal-organic frameworks (MOFs) into atomically doped metal oxide catalysts. The research methodology encompasses two main components, namely density functional theory (DFT) and a machine learning method based on the least absolute shrinkage and selection operator (LASSO). Starting with small amounts of experimental infrared spectroscopic data, the proposed machine-learning approach can extrapolate a more comprehensive infrared spectroscopy dataset. This augmented dataset is then employed to predict the pyrolysis pathways of MOF materials, providing valuable insight into the synthesis of single-atom catalysts (SACs). The innovation is that using infrared spectroscopy data-driven machine-learning for unveiling thermolysis pathways of MOFs is a powerful tool for understanding the behaviour of these materials and developing new applications for them. Chapter 1 is the introduction and Chapter 2 presents a literature review. The advances and challenges of machine learning assisted reaction pathway finding in SACs. Then the application of machine learning tools, IR spectra, and MOFs thermolysis is studied and presented. Chapter 3 the Formation Mechanism of a Single-Atom Catalyst via Infrared Spectroscopic Analysis. The synthesis of single-atom catalysts (SAC) through the pyrolysis of zeolitic imidazolate frameworks (ZIFs) represents a crucial pathway, and the mechanism can be examined using infrared (IR) spectroscopy. The results showed that the Pearson correlation exceeding 0.7 when compared to experimental data, the algorithm furnishes correlation coefficients for the chosen structures. This substantiates essential structural changes over time and temperature. Extends the study to other SACs formation from MOFs and the conclusions are drawn in Chapter 4, following the discussions of challenges and perspectives of machine learning on experimental graph recognition in reaction pathway exploration. The novel MOFs producing single atom catalyst provides a new platform for electrocatalyst development. This approach possesses substantial potential for robustness and has the capability to be applied across a wide spectrum of applications for intelligent analysis of in situ experimental characterization data in the future.