Integration of Coarse-Grained Simulation and Topological Data Analysis to Advance Mesostructure Analysis of CO₂ Separation Membrane Materials

In recent years, the use of data science in materials research and development has been actively promoted. However, its application to polymeric materials has been limited to predicting properties influenced by low hierarchical chemical structures, such as monomer structure. The physical properties of many polymeric materials depend on their mesostructures, which presents a significant challenge in accurate prediction. Predicting mesostructural properties requires machine-readable quantification of mesostructural features. The experimental generation of mesostructure data is time-consuming and costly. Therefore, simulation plays a crucial role in generating mesostructured data. Advanced simulation methods capable of generating plausible mesostructures are necessary. Coarse-graining (CG) methods are effective for large space-time scale objects, such as polymeric material mesostructures. Among CG methods, dissipative particle dynamics (DPD) is advantageous in terms of computational cost. In DPD, interactions between coarse-grained particles are represented by the Flory-Huggins χ parameter. The method for calculating the χ parameter by quantum mechanical (QM) calculations has only recently been developed, with few examples considering QM effects in DPD simulations. It is noteworthy that there are no examples of analysis comparing different ionization states of polymers with ionic functional groups.<br/> <br/> In this study, to validate the effect of ionization of the CG structure in DPD, DPD of CO₂ separation membrane materials was performed. We chose the CO₂ separation membrane because pH during the polymerization are known to affect the mesostructure of the membrane due to the effect of ionization of the ionic groups, thereby the performance of CO₂ separation. Here, ionic functional groups with different ionization states were defined as distinct CG particles. In DPD, pH variations during polymerization were expressed as differences in the ionization state of ionic functional groups. Persistent homology (PH), a topological data analysis method, was applied to analyze the mesostructures obtained by DPD to elucidate ionic state-dependent differences in the distribution of ionic functional groups. The distribution of ionic functional groups is difficult to verify experimentally and serves as a first step in understanding and predicting CO₂ separation performance.<br/> <br/> First, CG particles of CO₂ separation polymers composed of N-[3-(Dimethylamino)propyl]methacrylamide (DMAPM), N-tert-butylacrylamide (TBAm), and N-Isopropylacrylamide (NIPAm) were prepared, and χ parameters were calculated using QM calculations. The ionic functional group of these polymers is the amino group. DPD with these χ parameters yielded a mesostructure of CO₂-separating polymers aggregated into particles, consistent with the experimental results. The coordinates of amino groups were then extracted from the obtained mesostructures, and detailed analyses were performed using PH. Based on the PH results, a logistic regression model was constructed to classify whether a given mesostructure was a neutral or ionized amine. The coordinates of amino groups corresponding to the criteria of the classification model were visualized using PH inverse analysis, showing that neutral amino groups were mostly present in the interior of the particle, while ionized amino groups were mostly present on the particle surface. This may be due to the stronger interaction between the amine groups and the surrounding water due to ionization. This result indicates that DPD with χ parameters including the QM effect can generate mesostructures of polymers with ionic functional groups. Furthermore, it suggests that by combining with PH, we have succeeded in finding that amino groups show different distributions depending on their ionization state. This is a major advance in simulation that generates data and knowledge that would be difficult to obtain through experiment.