Search and Characterization of New Materials For Renewable Energy Applications Through First Principles Calculations and Machine Learning

Baldassarri, Bianca

doi:https://doi.org/10.21985/n2-mw1a-d586

Work

Search and Characterization of New Materials For Renewable Energy Applications Through First Principles Calculations and Machine Learning

Public

Download PDF

Download All Files (.zip)

Large scale adoption of sustainable technologies for energy productionand storage can be greatly facilitated by scientific advances impacting efficiency, cost and availability. The study of materials is instrumental in both upgrading the performance of existing technologies and enabling the development of new ones, and ab-initio methods and machine learning represent powerful tools in accelerating the process of materials discovery and characterization. This thesis presents multiple works leveraging computational methods to enable increased understanding and predictions of different properties across various classes of materials for renewable energy applications. A large portion of the discussion is dedicated to the thermodynamics of oxygen loss, with particular focus on solar-thermochemical water splitting applications for the production of green hydrogen. After successfully confirming the predictive accuracy of DFT computations of oxygen vacancy formation energy through comparison with experimental data, multiple high-throughput DFT studies are conducted surveying different classes of materials and identifying hundreds of new candidates compounds for solar-thermochemical hydrogen (STCH) applications. The data generated and insight gained through the high-throughput studies are then leveraged to construct machine learning models predicting the oxygen vacancy formation energy, and uncover additional hundreds of new STCH candidate materials. The knowledge acquired from such works is subsequently applied in a different context by exploring the stability of oxygen in cathode materials for Li-ion batteries. Finally, the focus is shifted from bulk to surface phenomena by studying segregation and adsorption behaviours of interest for catalytic applications. A dataset of hundreds of DFT computed segregation energies is constructed and used to train a predictive model, and segregation and oxygen adsorption behaviours in mixed transition metal carbides are investigated to guide the search for new corrosion resistant supports for fuel cell applications.

Creator