(74b) Carbon Capture Using a Deep Learning- Based Group Contribution Framework for Targeted Design of Ionic Liquid

Ionic liquids (ILs), as promising green solvents, have received immense attention due to their tunable molecular structure as they can be tailored to meet needs for various industrial applications including catalysis, synthesis, electrochemistry, and gas separation [1], [2]. Conversely, their tunability makes it difficult to identify the optimum ILs from a large number of potential candidates. Therefore, a robust and less time-consuming computational method such as predictive models, especially deep learning (DL) algorithms is clearly needed for the optimal design of ILs and to accurately understand the dynamics of their physiochemical properties [3]. DL which is a subset of Machine learning (ML) are empirical model and it has been widely employed in many engineering applications [4]. DL employs many structures of a neural network to significantly boost the computing efficiency leading to an increase in neuronal complexity [5]. The main objective of this study is to construct a DL-based group contribution (GC) algorithm to predict the physicochemical properties of ILs, specifically density, viscosity, melting point, heat capacity, etc. The suggested methodology in this study comprises three major steps: Firstly, data collection where most of the dataset of the aforementioned physiochemical properties will be collected from NIST ThermoLit [6] over a different pressure and temperature as well as their critical properties (T_C, P_C, and ) as these data will be the basis for assessing the accuracy of the DL-GC based model. Secondly, model construction will be carried out in the programming language ‘Python’ by using the two widely used modules, ‘Scikit-Learn’ and ‘TensorFlow’ to develop an algorithm for IL property prediction. Thirdly, to evaluate the applicability of the DL-GC-based model, the proposed framework will be employed to optimize new IL molecule structure based on targeted physicochemical properties, cation-anion skeleton, and cation-substituent group to achieve maximum CO₂ solubility in a post-combustion carbon capture for Blue Hydrogen Production case study.

Reference

[1] W. Jiang, X. Li, G. Gao, F. Wu, C. Luo, and L. Zhang, “Advances in applications of ionic liquids for phase change CO 2 capture,” Chem. Eng. J., vol. 445, no. 3, p. 136767, 2022, doi: 10.1016/j.cej.2022.136767.

[2] Y. Ma, J. Gao, Y. Wang, J. Hu, and P. Cui, “Ionic liquid-based CO2 capture in power plants for low carbon emissions,” Int. J. Greenh. Gas Control, vol. 75, no. March, pp. 134–139, 2018, doi: 10.1016/j.ijggc.2018.05.025.

[3] J. Wang, Z. Song, H. Cheng, L. Chen, L. Deng, and Z. Qi, “Computer-Aided Design of Ionic Liquids as Absorbent for Gas Separation Exempli fi ed by CO 2 Capture Cases,” 2018, doi: 10.1021/acssuschemeng.8b02321.

[4] T. Deng, F. hai Liu, and G. zhu Jia, “Prediction carbon dioxide solubility in ionic liquids based on deep learning,” Mol. Phys., vol. 118, no. 6, pp. 1–8, 2020, doi: 10.1080/00268976.2019.1652367.

[5] S. Pezhman, S. Atashrouz, R. Nakhaei-kohani, and F. Hadavimoghaddam, “Modeling of H 2 S solubility in ionic liquids using deep learning : A chemical structure-based approach,” J. Mol. Liq., vol. 351, p. 118418, 2022, doi: 10.1016/j.molliq.2021.118418.

[6] “Ionic Liquids Database - ILThermo.” https://ilthermo.boulder.nist.gov/ (accessed Sep. 11, 2022).