Deep-Learning-Based End-to-End Predictions of CO₂ Capture in Metal–Organic Frameworks

Abstract

Metal–organic frameworks (MOFs) have become an active topic because of their excellent carbon capture and storage (CCS) properties. However, it is quite challenging to identify MOFs with superior performance within a massive combinatorial search space. To this end, we propose a deep-learning-based end-to-end prediction model to rapidly and accurately predict the CO2 working capacity and CO2/N2 selectivity of a given MOF under low-pressure conditions. Different from previous methods, our prediction model relies only on the data from the Crystallographic Information File (CIF) rather than handcrafted geometric descriptors and chemical descriptors. The model was developed, trained, and tested on a dataset of 342489 topologically diverse MOFs. Experimental results on the dataset show that the proposed model achieves high prediction performance, i.e., R 2 = 0.916 for predicting the CO2 working capacity and R 2 = 0.911 for predicting the CO2/N2 selectivity. With regard to the identification of potential high-performing MOFs, 1020 of 1027 (top 3%) high-performance MOFs were recovered while screening only 12% of the entire dataset using our provided pretrained model, reducing the computation time by nearly an order of magnitude when the model was used to prescreen material prior to computationally intensive grand canonical Monte Carlo (GCMC) simulations while still capturing 99% of the high-performance MOFs. In the ab initio training task, the method can achieve R 2 = 0.85 with only 20% of the labeled data used for training and recover 995 of 1027 (top 3%) high-performance MOFs with only 12% of the entire dataset screened.

An XGBoost Algorithm Based on Molecular Structure and Molecular Specificity Parameters for Predicting Gas Adsorption

“…52 The trend of adsorption isotherms is also essential for screening MOFs. Therefore, we propose an index K (mean trend error) to measure the trend accuracy of adsorption isotherms: GCMC GCMC (9) where n is the number of adsorption points, v prediction is the predicted value of adsorption points, and v GCMC is the simulation value of adsorption points. The smaller the value of k, the more accurate is the prediction of the isotherm trend.…”

“…Besides, constructing descriptors, also known as feature engineering, is a complex task in machine learning. The relationship between the original data and the adsorption needs to be deeply understood to keep trying to find the best descriptors . The ability to obtain a model with high reliability and validity depends largely on whether the descriptors which reflect the structural properties of the MOFs were chosen adequately.…”

“…The relationship between the original data and the adsorption needs to be deeply understood to keep trying to find the best descriptors. 9 The ability to obtain a model with high reliability and validity depends largely on whether the descriptors which reflect the structural properties of the MOFs were chosen adequately. It places demands on the researcher's design skills and computational power.…”

AI for Nanomaterials Development in Clean Energy and Carbon Capture, Utilization and Storage (CCUS)

“…The application of AI in nanomaterials discovery and design toward developing cleaner energy technologies and achieving a carbon-neutral future has been explored and discussed in some reviews. − However, the majority of the previous reviews tend to focus solely on one specific carbon-neutral energy technology (e.g., solar cell , and battery), while others provide inadequate details on the application of AI in nanomaterials discovery and design for these technologies. , Given the extensive use of AI in materials research, a review that summarizes and discusses the application of AI in the design of nanomaterials for various major carbon-neutral energy technologies (e.g., solar cell, hydrogen energy, battery, and CCUS devices) is currently lacking.…”

Evaluating CH4/N2 Separation Performances of Hundreds of Thousands of Real and Hypothetical MOFs by Harnessing Molecular Modeling and Machine Learning

“…Machine learning (ML) has been very recently used with HTCS to predict the gas adsorption and separation performances of very large numbers of MOFs in a time-efficient manner and to extract the hidden correlations between the structural properties of materials and their gas separation performances . These ML studies focused on CO 2 /CH 4 , − CO 2 /N 2 , , CO 2 /H 2 , , O 2 /N 2 , H 2 S/CH 4 , C 3 H 8 /C 3 H 6 , and Xe/Kr separations. For instance, Anderson et al constructed a database of over 400 hypothetical MOFs and showed that ML algorithms can accurately predict the CO 2 /H 2 and CO 2 /N 2 selectivities of these materials using the structural and chemical features of the MOFs as the input.…”