Enhancing POMSimulator Applications in Heteropolyoxometalates: A Statistical and Data-Driven Approach

Identifier:  TDX:4558

Handle: https://hdl.handle.net/20.500.11797/TDX4558

Authors:  Buils Casasnovas, Jordi

Abstract:
The study of polyoxometalates (POMs) has evolved since the 19th century, showcasing their versatility in fields like medicine, electrochemistry, and catalysis. However, the processes that lead to the self-assembly of these inorganic oxo-clusters remain poorly understood, prompting the need for improved tools to explore their speciation. A novel methodology, POMSimulator, has been developed to model the complex multi-species multi-equilibria processes in aqueous solutions of POMs. Utilizing Density Functional Theory (DFT) calculations, Graph Theory, and Chemical Equilibrium, POMSimulator generates chemical reaction networks (CRNs) to establish multiple speciation models and derive formation constants (Kf) for various POMs. This thesis enhances POMSimulator through two major approaches. Firstly, a data-driven method predicts the linear scaling parameters necessary for adjusting DFT-derived formation constants without experimental references. This involves a Multi-Linear Regression (MLR) model based on an extensive database of speciation models. Secondly, a clustering process is introduced to manage the vast data generated, embedded within a statistical pipeline. Both methods have been validated using previously studied systems and align with the FAIR data principles by making the POMSimulator code publicly available.