Manufacturing and research often need the use of chemical separation techniques, including gas separation. It generates millions of tons of carbon emissions and makes up a staggering 15% of the energy consumed in the United States.
If the proper materials could be found to produce them, the separation of gases by passing them through membranes might be an effective, ecologically acceptable substitute for the methods used now.
A group of computer scientists, chemical and mechanical engineers, and researchers at the University of Notre Dame have discovered, synthesized, and tested polymer membranes that can separate gases up to 6.7 times more effectively than membranes that have previously been synthesized. This was accomplished by using a graph-based machine learning approach.
Their results have been published in Cell Reports Physical Science.
The microscopic porosity of the material is what determines the membrane’s performance, stated Agboola Suleiman, a PhD student in Ruilan Guo’s lab, the Frank M. Freimann Collegiate Professor of Engineering.
Co-author Suleiman stated that the optimal membrane material is one that is both selective and permeability-balanced, allowing some gases to pass through while keeping some out.
The group employed graph neural networks (GNN), a kind of machine learning that is especially well-suited to showing a material’s molecular structure as well as its interactions with other molecules, to identify this Goldilocks material. Two polymers with the appropriate characteristics to surpass previously produced membranes were found by GNN after it was trained on datasets.
According to Tengfei Luo, co-author of the article and the Dorini Family Professor for Energy Studies and associate chair of the Department of Aerospace and Mechanical Engineering, “our machine learning algorithms led us to materials that had previously only been used for electronics applications.” Subsequently, we created these materials and conducted laboratory experiments to confirm their exceptional ability to separate gases. It was like finding hidden gems.
Since it can be expensive and time-consuming to synthesize polymers, there is a dearth of information about their molecular structure and chemical properties.
Nevertheless, this issue was resolved by algorithmic advances developed by co-authors and computer scientists Meng Jiang and his doctoral student Gang Liu.
As stated by Jiaxin Xu, a PhD student in Luo’s lab and co-author of the paper, “they were able to augment and improve our data by using machine learning techniques.” They were able to forecast the best membrane materials and provide an explanation for their superiority thanks to the graph-based model that was enhanced with details about the molecular characteristics of each material.
The best-performing polymers produced by the group could be utilized to make membranes that can separate multiple gas pairs—a crucial feature for industrial applications.
