Impact of Chemical Features on Methane Adsorption by Porous Materials at Varying Pressures

Abstract

Superior performance in methane uptake capacity prediction by hypothetical metal organic frameworks has previously been accomplished using a novel combination of structural and chemical features with machine learning (ML) algorithms. This concept is extended for additional microcrystalline materials, focusing on 69 839 covalent organic frameworks (COFs) and 17 846 porous polymer networks (PPNs). For each material category, data was divided into train (80%) and test (20%) sets. Using the random forest (RF) algorithm, 10-fold cross-validation was carried out to evaluate the robustness of prediction for structural and chemical descriptors. Structural features included surface area, density, and void fraction. Chemical descriptors included the number and type of each atom, electronegativity, and degree of unsaturation among others. When chemical descriptors for adsorption at low pressures were included, significant improvements for predictions were observed compared to solely using structural descriptors. Specifically, adding chemical features increased the R 2 value from 0.66 to 0.87 for COFs and from 0.83 to 0.93 for PPNs. These results indicate that inclusion of chemical descriptors improves prediction across materials and pressures. While physisorption is the main driver for adsorption at these pressures, these results also imply contribution of surface chemical motifs on adsorption phenomena.

High-Performing Deep Learning Regression Models for Predicting Low-Pressure CO2 Adsorption Properties of Metal–Organic Frameworks

“…The chemical motifs descriptor was calculated as described by Anderson et al; however, this descriptor has been used frequently in the literature by other researchers. ,, The motifs used in the calculation of this descriptor are shown in Figure along with the percentage of MOFs that contain the structural motif. The bag-of-atoms descriptor was calculated according to the procedure outlined by Anderson et al Specifically, the unit cell of each MOF was split into 6 × 6 × 6 cuboids, and the sum of the framework atom ε and σ parameters as defined in the UFF force field was computed for each cuboid.…”

“…ML models have the advantage that once trained the models can screen materials in a matter of seconds per MOF versus hours in the case of molecular simulation. Thus, accurate ML models could be used as part of a prescreening protocol to filter out poorly performing materials and thereby reduce the number of computer-intensive GCMC simulations that are performed during high-throughput screening. − …”

“…We will refer to these features as descriptors for the ML models. Geometric descriptors, such as surface area, pore size, and density have been used extensively in the literature to predict gas uptake capacity in nanoporous materials using machine learning. ,− While these geometric descriptors correlate well with adsorption properties at higher pressures, they perform considerably worse at lower pressures. , One explanation for this observation is that at high pressure the adsorption is essentially a packing problem, so the geometry of the MOF can be correlated well to the adsorption properties. However, at low pressures, the chemical interactions of the gas with the MOF framework are likely to be more important in determining the adsorption properties.…”

Machine Learning Meets with Metal Organic Frameworks for Gas Storage and Separation

“…The use of the RF algorithm was extended to 69 839 hypothetical COFs to predict their deliverable CH 4 capacities utilizing structural and chemical descriptors. 75 The use of chemical and structural descriptors increased the accuracy of ML predictions for COFs, especially at low pressure. Those works emphasized the need for the consideration of chemical descriptors for accurate prediction of gas uptakes of nanoporous materials at low pressures.…”

Machine Learning-Guided Equations for Super-Fast Prediction of Methane Storage Capacities of COFs

“…The current focus, however, is to increase predictive accuracy of ML models by choosing different/sophisticated algorithms and/or generating novel input features. [44][45][46] Less attention has been given in generating features considering their accessibility, cost, measurability, and/or physical and chemical significance (i.e., interpretability). 47 Also, sometimes, feature generation could be daunting even for expert users.…”