(617c) Performance Analysis of Metal-Organic Frameworks (MOFs) for CO2 Capture By Smart Connectionist Models Using Combined Structural and Thermal Descriptors

The combustion of fossil fuels is continuously increasing CO₂ concentration in the atmosphere, which mainly leads to global warming and climate change ¹ (e.g., the main source of anthropogenic emission of CO₂ is coal-fired power plants ²). Its concentration has risen from 375.82 ppm in February of 2003 to 420.82 ppm in February of 2023 ³. CO₂ emission contributes to 55% of the total greenhouse gases ⁴. Based on the recent statement of Intergovernmental Panel on Climate Change (IPCC), the reduction of CO₂ emission by 45% from the levels of 2020 by 2030 is inevitable to limit global warming to 1.5 ^oC ². Effective strategies for carbon management (e.g., capture, storage, conversion, and utilization) are needed to curb the global temperature increase ⁵. CO₂ can be captured from the atmosphere through employing biological, physical, or chemical processes. The available carbon capture approaches have undeniable limitations. For instance, biological processes are ineffective and slow, and the chemical ones may result in further environmental pollution, intricate post-treatment units, and equipment corrosion due to the volatilization of organic solvents and carbon resources waste ⁴. The current methods that involve utilizing amine-based solvents/sorbents suffer from the tremendous cost of energy at the amount of 60-80% of the total cost of operation ⁵. Overcoming the drawbacks, more convenient, efficient, and energy saving methods are required to substitute the conventional methods of CO₂ capture ⁴. Adsorption separation as a potential method is relatively inexpensive, simple in terms of equipment and operation, and has comparatively low energy consumption when an adsorbent regeneration unit exists in the process ⁴. Adsorption techniques using solid porous materials have attracted significant attention because of their low energy consumption and cost of operations. Particularly, nano-porous materials have been extensively studied as promising adsorbents since they have different topologies/structures and pore sizes that can be efficiently optimized to capture CO₂ ⁶. Metal Organic Frameworks (MOFs) are in the category of nano-porous solids synthesized by the self-assembling of metal ions or clusters and polydentate organic linkers that act as structural building units to form open crystalline frameworks ⁷. An infinite number of MOFs can be synthesized by combining organic linkers and metal nodes ⁸. MOFs are made with various ranges of surface area, geometry, chemistry, structural robustness, and functionality suitable for diverse applications including gas separation and storage, non-linear optics, heterogeneous catalysis ⁹, controlled drug release, sensing, and light-harvesting ⁷. MOFs have high specific surface area (>8000 m²/g ¹⁰) and large porosity; they are extensively used in gas adsorption and separation ⁴. With advancement of artificial intelligence tools, machine learning (ML) is a developing research paradigm that will inspire material and chemistry discovery ¹¹. In fact, first principle model-based simulations of CO₂ capture are not practical owing to high computational costs and some unrealistic assumptions. Recently, ML algorithms have been employed to effectively develop new MOFs for CO₂ capture ¹². In this research, different ML techniques such as Decision Tree (DT), Random Forest (RF), and Support Vector Machine (SVM) along with hybrid techniques such as Artificial Neural Network-Particle Swarm Optimization (ANN-PSO), coupled simulated annealing-least squares support vector machine (CSA-LSSVM), and adaptive neuro-fuzzy inference system (ANFIS) are employed to quantitatively assess the CO₂ uptake of proper MOFs including BIO-MOF-1, BIO-MOF-11, B-PEI-300, CALF-20, CAU-1, CoBDP, CPM-33a, and Cu3(BTC)2. Moreover, a Gene Expression Programming (GEP) is used to develop a mathematical model to relate the operating parameters involved in the process of CO₂ adsorption. Several MOF structural descriptors such as surface area, pore volume, and porosity and thermal descriptors including heat capacity, heat conductivity, and thermal expansion along with the operating parameters such as pressure and temperature are considered as the input parameters for the developed ML models, and the CO₂ uptake is considered as the output/target parameter. After data processing, 80% of the data is used as the training set, and the remaining data is utilized as the validation and testing set. The outputs of the designed models are compared with the experimental data extracted from the literature, and the accuracy of the models is determined by the statistical measures such as the coefficient of determination (R²), mean square error (MSE), and average absolute relative error percentage (AARE%). The most accurate model is selected to perform the parametric sensitivity analysis to determine the degrees of impact of the input parameters on the MOFs’ ability for CO₂ uptake. The developed models can be used to analyze the performance of common and effective MOF for capturing CO₂, and they can be employed to develop more accurate modeling and optimization models to evaluate CO₂ capture processes in terms of practical, environmental, and economic prospects.

(1) Amar, M. N.; Ouaer, H.; Ghriga, M. A. Robust smart schemes for modeling carbon dioxide uptake in metal-organic frameworks. Fuel 2022, 311, 122545.

(2) Dureckova, H.; Krykunov, M.; Aghaji, M. Z.; Woo, T. K. Robust machine learning models for predicting high CO₂ working capacity and CO₂/H₂ selectivity of gas adsorption in metal organic frameworks for precombustion carbon capture. J. Phys. Chem. C. 2019, 123 (7), 4133-4139.

(3) Global Monitoring Laboratory. Trend of Carbon Dioxide Concentration. National Oceanic and Atmospheric Administration, 2023. https://gml.noaa.gov/webdata/ccgg/trends/co2_trend_mlo.png (accessed 24 March 2023).

(4) Deng, X.; Yang, W.; Li, S.; Liang, H.; Shi, Z.; Qiao, Z. Large-scale screening and machine learning to predict the computation-ready, experimental metal-organic frameworks for CO₂ capture from air. Appl. Sci. 2020, 10 (2), 569.

(5) Zheng, B.; Oliveira, F. L.; Neumann Barros Ferreira, R.; Steiner, M.; Hamann, H.; Gu, G. X.; Luan, B. Quantum Informed Machine-Learning Potentials for Molecular Dynamics Simulations of CO₂’s Chemisorption and Diffusion in Mg-MOF-74. ACS NANO 2023.

(6) Zhang, X.; Zhang, K.; Yoo, H.; Lee, Y. Machine learning-driven discovery of metal–organic frameworks for efficient CO₂ capture in humid condition. ACS Sustain. Chem. Eng. 2021, 9 (7), 2872-2879.

(7) Fernandez, M.; Boyd, P. G.; Daff, T. D.; Aghaji, M. Z.; Woo, T. K. Rapid and accurate machine learning recognition of high performing metal organic frameworks for CO₂ capture. J. Phys. Chem. Lett. 2014, 5 (17), 3056-3060.

(8) Majumdar, S.; Moosavi, S. M.; Jablonka, K. M.; Ongari, D.; Smit, B. Diversifying databases of metal organic frameworks for high-throughput computational screening. ACS Appl. Mater. Interfaces 2021, 13 (51), 61004-61014.

(9) Chong, S.; Lee, S.; Kim, B.; Kim, J. Applications of machine learning in metal-organic frameworks. Coord. Chem. Rev. 2020, 423, 213487.

(10) Altintas, C.; Altundal, O. F.; Keskin, S.; Yildirim, R. Machine learning meets with metal organic frameworks for gas storage and separation. J. Chem. Inf. Model. 2021, 61 (5), 2131-2146.

(11) Li, S.; Zhang, Y.; Hu, Y.; Wang, B.; Sun, S.; Yang, X.; He, H. Predicting metal-organic frameworks as catalysts to fix carbon dioxide to cyclic carbonate by machine learning. J. Materiomics. 2021, 7 (5), 1029-1038.

(12) Burner, J.; Luo, J.; White, A.; Mirmiran, A.; Kwon, O.; Boyd, P. G.; Maley, S.; Gibaldi, M.; Simrod, S.; Ogden, V. ARC–MOF: A Diverse Database of Metal-Organic Frameworks with DFT-Derived Partial Atomic Charges and Descriptors for Machine Learning. Chem. Mater. 2023, 35 (3), 900-916.