A D V A N C E D M A T E R I A L S & P R O C E S S E S | N O V E M B E R / D E C E M B E R 2 0 2 2 1 0 MACHINE LEARNING | AI MODELING METAL-ORGANIC FRAMEWORKS The nanosized pores of metal-organic frameworks (MOFs) generate expansive internal surface areas, making them extremely versatile for applications such as separating petrochemicals and gases, mimicking DNA, and producing hydrogen. Professor Berend Smit at EPFL’s School of Basic Sciences, Switzerland, and his group are using machine learning (ML) to advance the discovery, design, and categorization of the numerous MOFs filling up chemical databases. Smit and his colleagues developed a ML model that predicts the heat capacity of MOFs: It works by forecasting how the local chemical environment changes the vibrations of each atom in a MOF molecule. “These vibrations can be related to the heat capacity,” says Smit. “Before, a very expensive quantum calculation would give us a single heat capacity for a single material, but now we get up to 200 data points on these vibrations. So, by doing 200 expensive calculations, we had 40,000 data points to train themodel on how these vibrations depend on their chemical environment.” To demonstrate the practical impact of the research, engineers at Heriot-Watt University, U.K., simulated the MOF performance in a carbon capture plant. “We used quantum molecular simulations, machine learning, and chemical engineering in process simulations,” explains Smit. “The results showed that with correct heat capacity values of MOFs, the overall energy cost of the carbon capture process can be much lower than we originally assumed. Our work is a true multiscale effort, with a huge impact on the techno-economic viability of different solutions to tackle climate change.” www.epfl.ch. MACHINE LEARNING LOOKS AT LIGNIN A new research project is demonstrating how artificial intelligence (AI) can improve production of renewable biomaterials. The study focuses on the extraction of lignin, a papermaking byproduct produced in large quantities but mainly used as an inexpensive fuel. Developing valuable materials and chemicals from lignin is the ultimate goal of the project. Scientists from Aalto University and the University of Turku, both in Finland, are working together to find the best extraction conditions for various lignin-based products with the help of Bayesian optimization. Their machine learning approach employs a computer model that, for a given combination of experimental conditions, can predict both the amount of extracted lignin and its properties. As with other AI methods, Bayesian optimization requires data to learn from, but in contrast to approaches such as neural networks, data collection is guided by the algorithm itself. The result is that the computer tells the scientist working in the lab which conditions to use for the next experiment. By choosing the variables in an intelligent way, the AI guarantees that only a small number of experiments are necessary to create an accurate model. The successful application of Bayesian optimization to the challenge of lignin extraction suggests that AI may soon become a standard tool alongside traditional statistical tools for planning and predicting experimental outcomes. The team is now collaborating with several other research groups at Aalto to expand their methodology to a wider set of problems in materials science. www.aalto.fi. Metal-organic frameworks capturing CO2 from flue gasses. Courtesy of EPFL. Photo and graphic with birch tree. Courtesy of J. Löfgren.
RkJQdWJsaXNoZXIy MTYyMzk3NQ==