Interpretable Machine‐Learning and Big Data Mining to Predict Gas Diffusivity in Metal‐Organic Frameworks

Abstract

For gas separation and catalysis by metal-organic frameworks (MOFs), gas diffusion has a substantial impact on the process' overall rate, so it is necessary to determine the molecular diffusion behavior within the MOFs. In this study, an interpretable machine learing (ML) model, light gradient boosting machine (LGBM), is trained to predict the molecular diffusivity and selectivity of 9 gases (Kr, Xe, CH 4 , N 2 , H 2 S, O 2 , CO 2 , H 2 , and He). For these 9 gases, LGBM displays high accuracy (average R 2 = 0.962) and superior extrapolation for the diffusivity of C 2 H 6 . And this model calculation is five orders of magnitude faster than molecular dynamics (MD) simulations. Subsequently, using the trained LGBM model, an interactive desktop application is developed that can help researchers quickly and accurately calculate the diffusion of molecules in porous crystal materials. Finally, the authors find the difference in the molecular polarizability (𝚫Pol) is the key factor governing the diffusion selectivity by combining the trained LGBM model with the Shapley additive explanation (SHAP). By the calculation of interpretable ML, the optimal MOFs are selected for separating binary gas mixtures and CO 2 methanation. This work provides a new direction for exploring the structure-property relationships of MOFs and realizing the rapid calculation of molecular diffusivity.

Accelerating Metal–Organic Framework Selection for Type III Porous Liquids by Synergizing Machine Learning and Molecular Simulation

“…Ensemble learning addresses inherent flaws in a single model or a model with a certain set of parameters. Its fundamental idea is to combine weak learners to establish a strong model . The data set was randomly divided into a training set (80%) and a test set (20%).…”

“…Its fundamental idea is to combine weak learners to establish a strong model. 56 The data set was randomly divided into a training set (80%) and a test set (20%). During training, the kfold cross-validation technique 57 was also applied for a k-fold value of 5 to verify the accuracy and stability of the model.…”

Accelerated Discovery of Metal–Organic Frameworks for CO2 Capture by Artificial Intelligence

“…CO 2 adsorption and diffusion data of MOFs have been generally obtained from HTCS of thousands of materials and then used to train predictive ML models, while experimental CO 2 uptake data of MOFs are utilized to optimize the material synthesis procedures. 51−53 Figure 2(b) shows the four important aspects of AI in the discovery of novel MOFs: (i) ML algorithms are used to estimate the materials' performance metrics related to CO 2 capture and separation, such as CO 2 uptake, 54 diffusivity, 46 and permeability, 55 by reducing the need for complex simulations and time-consuming experiments through classifying high-or low-performing materials. 45,56,57 For example, the quantitative structure−property relationship (QSPR) classifiers generated based on support vector machines accurately identified the promising materials offering high CO 2 adsorption capacity (>1 mmol/g at 0.15 bar and >4 mmol/g at 1 bar, 298 K) among 292,050 different types of hMOFs.…”

“…Both types of algorithms are crucial in materials genomics for predicting properties and identifying patterns, but their limitations must also be considered. Figure (a) shows that ML algorithms, such as decision trees, random forests, gradient boosting, and ANN, have been used to predict CO 2 adsorption capacities − and CO 2 diffusion rates in porous materials based on materials’ structural properties, such as pore size, surface area, porosity, and topology. More complex algorithms, including convolutional neural networks, genetic algorithms, deep learning, and deep neural networks, have also recently been implemented in MOF research. − Achieving accurate ML predictions requires training the models based on a high-quality data set obtained for a large number and diversity of materials.…”

Computational Simulation of CO2/CH4 Separation on a Three-Dimensional Cd-Based Metal–Organic Framework

“…MOFs are porous inorganic materials that have received much attention recently owing to their outstanding characteristics, such as significant surface area and porosity, thermal/chemical stability, and tunability. − MOFs are considered promising materials for a variety of potential applications across various fields, including gas storage and separation, chemical sensing, biomedical applications, adsorption, and heterogeneous catalysis. − They are considered as one of the most promising physical adsorbent materials in the process of separating CO 2 /CH 4 . The engineering design of MOFs for gas separation applications is currently a rapidly growing area of research. − Numerous experimental and simulation-based studies have been documented involving the separation of CO 2 from CH 4 through the use of MOFs. − Considering the quadrupole moment and polarizability of CO 2 , current research efforts to enhance CO 2 uptake and selectivity primarily involves strategies aimed at improving the interaction between CO 2 and the frameworks. These include, but are not limited to, the utilization of various open metal sites, the insertion of functional groups, , the development of smart adsorbents, and ligand shortening in MOFs. , An investigation showed that MOF-801­(Ce) displayed improved separation performance for CO 2 /N 2 and CO 2 /CH 4 compared to MOF-801­(Zr/Hf) .…”