(Nanowerk news) A large number of 2D materials that graphene can have nanopores – small holes formed by missing atoms through which foreign substances can pass. The properties of these nanopores dictate many of the materials’ properties, allowing them to detect gases, filter seawater, and even help sequence DNA.

“The problem is that these 2D materials have a wide distribution of nanopores, both in terms of shape and size,” says Ananth Govind Rajan, assistant professor in the Department of Chemical Engineering, Indian Institute of Science (IISc). “You don’t know what will form in the material, so it’s very difficult to understand what the resulting membrane property will be.”

Machine learning models can be a powerful tool to analyze the structure of nanopores to discover new tantalizing properties. But these models struggle to describe what a nanopore looks like.

Govind Rajan’s lab has now devised a new language that encodes the shape and structure of nanopores as a sequence of characters, in a study published in J.Journal of the American Chemical Society (“Machine-learned language for the chemical space of nanopores enables structure-property relationships in 2D nanoporous materials”). This language can be used to train any machine learning model to predict the properties of nanopores in a wide variety of materials. A selection of nanopores that might be present in graphene along with their powerful names. The white arrow indicates the starting atom, and each STRONG is written by traversing the nanopore edge atoms counterclockwise. (Image: Piyush Sharma)

Called STRONG – String Representation Of Nanopore Geometry – the language assigns different letters to different atom configurations and creates a sequence of all the atoms on the edge of a nanopore to specify its shape. For example, a fully bonded atom (having three bonds) is represented as “F” and a corner atom (bonded to two atoms) is represented as “C” and so on. Different nanopores have different types of atoms at their edge, which dictate their properties. The STRONGs allowed the team to develop rapid ways to identify functionally equivalent nanopores with identical edge atoms, such as those linked by rotation or reflection. This drastically reduces the amount of data that needs to be analyzed to predict nanopore properties.

Just as ChatGPT predicts textual data, neural networks (machine learning models) can “read” the letters in STRONG to understand what a nanopore will look like and predict what its properties will be. The team turned to a variant of a neural network used in natural language processing that works well with long sequences and can selectively remember or forget information over time. Unlike traditional programming where the computer is given explicit instructions, neural networks can be trained to figure out how to solve a problem they have never encountered before.

The team took a series of nanopore structures with known properties (such as energy of formation or barrier to gas transport) and used them to train the neural network. The neural network uses this training data to determine an approximate mathematical function, which can then be used to estimate the properties of a nanopore when given the structure in bold letters.

This also opens up exciting possibilities for reverse engineering—creating a nanopore structure with the specific properties you’re looking for, something that’s particularly useful in gas separation. “Using STRONGs and neural networks, we looked for nanoporous materials to separate CO2 from flue gas, a mixture of gases released when the fuel burns,” says Piyush Sharma, former MTech student and first author of the study. This process is essential for reducing carbon emissions. The researchers were able to identify several candidate structures that could effectively capture CO2 from a mixture that includes oxygen and nitrogen.

The team is also looking into the idea of ​​creating digital twins of 2D materials. “Suppose you collect a lot of experimental data on a material. Then you can try to see what would have been the collection of nanopores that would have led to this performance,” says Govind Rajan. “With this digital twin material, you can do a lot of things—predict performance for separating a different set of gases, or you can come up with completely new use cases for the same material.”