AI Algorithm Provides a Better Way to Search for and Build Nanoporous Materials
One day, nanoporous materials could help solve some of society's most serious problems, from absorbing carbon dioxide or methane from the air to storing hydrogen gas as fuel and detecting toxic compounds in the air.
Thanks to tiny nanometer-sized pores, these materials are useful for many sustainability applications, but because chemists create them in labs, molecule by molecule, their development is cumbersome and expensive.
A team of researchers from Washington State University and Oregon State University developed a unique computer algorithm that plays a 20-questions-style game, quickly narrowing thousands of possible molecular structures to find the optimal one with minimal cost and effort. The study was published in the Molecular Systems Design and Engineering Journal.
“The key challenge is that nanoporous materials are a mix of different chemical elements, which must be assembled to find the best combination. Nanoporous materials have a huge variety of potential molecular building blocks and structures that can be mixed almost endlessly. If we tested new configurations of these elements and their structures in the lab each time, it would be very expensive, so the computational task is how to identify the right combination of elements with the properties you are interested in. This is where our AI-based algorithmic work comes in”
In a proof-of-concept study, the scientists selected the best nanoporous material candidate for absorbing methane, a powerful greenhouse gas that contributes to global warming. After evaluating just 120 possible candidates, they found the already known best candidate from a library of 70,000 materials, significantly better than traditional algorithms manage.
"Nanoporous materials (NPMs) can be used to store, capture and detect many different gases. When addressing the adsorption problem, we often want to find the one in the library with optimal adsorption properties. The high cost of synthesizing NPMs and measuring gas adsorption, whether these experiments are conducted in the lab or through simulation, often prevents exhaustive searching," the study states.
One of the study's authors and an associate professor at the School of Electrical Engineering and Computer Science, Janardhan Rao Doppa, said: "Aryan's algorithms find the best material with fewer evaluations." One reason the algorithm works well is that it takes into account the three-dimensional structures of the material.
“We are trying to make a somewhat smarter search, while existing methods that are used did not try to use models of the relationship between a material's structure and its properties. We explicitly build statistical models that allowed us to predict the properties of unknown materials and have well-calibrated uncertainty, which means you know what you don't know, so when we explored the space, we explored it in a much more reasoned way rather than at random”
As their algorithm went through each new iteration of the material, it virtually ran an experiment, updated its understanding of the structure and property relationships, and then based on that chose another nanoporous material.
Now the researchers aim to automate and generalize the methodology even further. They have already made fundamental progress toward this goal in a new paper to be presented at the 2021 Conference on Neural Information Processing Systems (NeurIPS). The scientists hope to use their unique algorithms to improve searches in other types of real-world applications, such as developing catalysts used in industrial processes.
A nanoporous material is a material containing pores whose sizes fall within the nanoscale range (~1-100 nm). The term is used to indicate that the material's specific properties (sensing, adsorption, catalytic, diffusion and others) are related to the presence of nanopores. Most known membranes, sorbents and catalysts can be classified as such materials.
Many natural nanoporous materials exist, but they can also be made artificially. To do this, polymers with different melting points are combined so that when heated, one polymer decomposes. A porous material with pores of uniform size has the property of letting through only certain substances while blocking others.
Nanoporous materials are used in gas masks, autonomous life support systems on submarines and crewed space stations. They can be used in filters for fine purification of air and water from contaminants (in particular chemical and bacteriological toxic agents). For the material to work effectively, it is important to exceed the percolation threshold, beyond which unconnected pores and channels begin to form through passages for fluid flow.
