UB engineer awarded $110,000 to use AI, simulation for new nanomaterials

BUFFALO, N.Y. – University at Buffalo engineer Kaihang Shi specializes in using artificial intelligence, molecular simulation and other approaches to create nanomaterials for energy and sustainability applications.

He recently received $110,000 from the American Chemical Society to produce microporous polymers with the goal of making natural gas processing more energy efficient.

Microporous polymers are substances with small holes or nanoscale pores. They mimic some of the porosity found naturally, like in shale rocks, wood and human bones. The surface area of microporous polymers can be large, despite the nanoscale level of the small pores, according to Shi.

“The surface area for one gram of metal-organic framework — a type of nanoporous material — can be as high as 8,000 m², which is commensurate to half the area of UB’s football stadium,” says Shi, PhD, assistant professor in the Department of Chemical and Biological Engineering, who joined UB in 2023.

The grant, from the society’s petroleum research fund, will support experiments using complex computer simulations that quickly analyze the behaviors of many molecules all at once, as well as machine learning, a form of AI. The goal is to create new polymers that can make methane diffusion, which is a key part of natural gas processing, more eco-friendly and less expensive.

Using nanoporous materials as membranes

Nanoporous materials, including microporous polymers, have helped address critical societal challenges as a key component in energy storage, carbon capture, nanomanufacturing, drug delivery, tissue regeneration, and chemical separation. Shi and his research group aim to develop advanced computational methods to discover and design materials for similar applications.

“Microporous polymers are promising membrane materials to meet the industrial demands for large-scale, energy-efficient natural gas processing,” Shi says. “The intricate pore networks in the polymers are crucial for chemical separation, but the impact of these networks on methane diffusion is not fully understood.”

Shi and his research group aim to close this gap. Using microporous polymers for methane diffusion has potential as an energy efficient way to process natural gas, but the materials’ intricate channels of pores make the movement of methane difficult to predict. Shi and others are developing a novel graph-based computational method to characterize the polymers and predict their properties regarding industrial applications like methane diffusion. Their findings have potential to advance the understanding of molecular diffusions in natural geological materials, which could enhance oil and gas recovery and geologic carbon dioxide capture.

“This project will lead to a new graph-based package tailor-made for porous materials,” Shi says. “We expect this package will be adopted by the community to enhance the visualization, characterization, and understanding of nanoporous materials.”

Machine leaning for this project and beyond

Many machine learning models are black boxes — they can produce results, but often do not explain how they developed the results. Shi and team are developing their machine learning architecture to provide not just results, but a comprehensive understanding of how it achieved those results.

“Predicting molecular diffusion in nanoporous materials is a challenging task, usually involving lengthy molecular dynamics simulations that are time consuming,” Shi says. “We will develop our machine learning model to predict molecular transport behavior in nanoporous materials rapidly and accurately. It will also help identify porous structures that favor molecular diffusion.”

He continued, “We expect our machine learning work will help pave the way for using machine learning for scientific discovery.”