Project Description
Using machine learning techniques, participants will explore the role of iron complexes in advanced energy storage, specifically focusing on understanding and predicting the redox potentials that govern the performance of flow batteries. Drawing on a documented scientific workflow, participants will move through the end-to-end data science process: curating chemical datasets, analyzing molecular structures, and training Graph Neural Networks (GNNs) to make accurate physical predictions. Utilizing Python and Jupyter notebooks, students will gain practical experience in machine learning while learning how to use computational tools to address urgent challenges in sustainable energy. By the end of the program, participants will have developed a foundational understanding of how AI-enabled research methods can be applied to solve complex problems in chemistry.
Learning Objectives
- Explain why redox potential matters for electrochemistry, especially for iron-complex design in flow-battery applications.
- Use Python and Jupyter notebooks to inspect, clean, organize, and visualize chemical or materials datasets.
- Distinguish between supervised learning and unsupervised learning in a scientific context.
- Understand the logic of graph-based molecular representations, where atoms are nodes and bonds are edges, as used in graph neural networks.
- Recognize how graph neural networks can be used to connect molecular structure with chemical properties.
