AI Design of Disordered Materials for Clean Energy and Sustainable Applications

Artificial intelligence (AI) is revolutionizing the discovery of materials, particularly for clean energy and sustainable chemical processes. Multicomponent crystals—solids made from several elements in repeating patterns—can power better batteries, catalysts, and other technologies. The challenging part is predicting how atoms arrange themselves (atomic ordering) and how that ordering affects their properties. Traditional simulations struggle in this huge design space, so modern machine learning is needed.

In this project, students will build and apply graph-based neural networks—graph convolutional models and crystal-graph transformers—that treat a crystal as a network of atoms and bonds. You will train models to recognize and compare alternative atomic arrangements across different crystal families, check what the models are learning, and test whether predictions hold on new, unseen compounds. The goal is to deliver fast, trustworthy tools that guide design choices: which elements to combine, how to arrange them, and what properties to expect. Outcomes include a clean, reproducible code repository, clear benchmarks, and (when results warrant) posters or short papers—contributions that help accelerate the search for next-generation materials. 

1. Get hands-on experience at the intersection of AI, data science, and materials informatics for clean energy and chemical processes. Build proficiency in modern ML (especially graph neural networks and crystal-graph transformers) and learn how atomic structure influences materials properties.
2. Grow core research skills: organize and visualize datasets; run and interpret models; read and critique papers; and explain results clearly through brief updates, well-designed figures, and lightning talks. You’ll work closely with the PI and a graduate mentor and collaborate with partners at UB and beyond.
3. Share your work: present at UB symposia, apply to national conferences, and, when results warrant, co-author a paper. You’ll leave with a polished GitHub repository, a poster or talk, and strong letters and experiences that bolster graduate-school and/or job applications. 

The student should learn the basics of Python coding and linear algebra. The PI will also guide the reading of a few fundamental research papers based on this project to help the student(s) understand the background and goal of this project. 

materials design and innovation, computer science and engineering, chemical and biological engineering, chemistry, physics, energy, sustainability, data science, artificial intelligence