Minimal set of crystallographic descriptors for sorption properties in hypothetical Metal Organic Frameworks: Role in sequential learning optimization

Abstract

Several studies have been recently reported in the literature on sorption properties of MOFs with a number of organic sorbates, such as ethanol and methanol. Surprisingly, still few studies have been reported on water sorbate despite its large availability, low cost and environmental sustainability, and the screening of a large number of hypothetical MOFs-water working pairs for engineering applications is still challenging. Based on a recently reported database of over 5000 hypothetical MOFs, a first contribution of this study is the identification of the minimal set of crystallographic descriptors underpinning the most important sorption properties of MOFs for \ch{CO2} and, importantly, for \ch{H2O}. Furthermore, a comprehensive comparison of several Sequential Learning (SL) algorithms for MOFs properties optimization is carried out and the role played by the above minimal set of crystallographic descriptors clarified. In sorption-based energy transformations, thermodynamic limits of important figures of merit (e.g. maximum specific energy) depend both on operating conditions and equilibrium sorption properties in a wide range of sorbate coverage values. The access to the latter properties is often incomplete, with essential quantities such as equilibrium adsorption isotherms spanning over the full sorbate coverage range and values of the isosteric heat being only partially available. As a result, this may prevent the computation of objective functions during the optimization procedure. We propose a fast procedure for optimizing specific energy in a closed sorption energy storage system with the only access to the water Henry coefficient at a fixed temperature value and to the specific surface area.

Figures (13)

• Figure 1: Overview of the methodological protocols to identify and test the minimal set of ruling crystallographic descriptors of sorption properties in several sequential learning algorithms. Over 5000 hypothetical MOFs from ref. boyd2019data are first featurized by CFID, with the corresponding full set of descriptors provided to AutoMatminer for a preliminary descriptor reduction and machine learning models training of sorption properties of interest. The Kernel SHAP interpretation algorithm is thus used to finalize the identification and ranking of a reduced sub-set of ruling descriptors ( genetic code of the chosen property). Several sequential learning schemes are tested using both the full set of descriptors and the reduced one for comprehensive comparison.
• Figure 2: Predictions and corresponding normalized cumulative curves for the coefficients of importance of the four regression AutoMatminer models trained for Henry coefficient for CO2, working capacity for CO2, Henry coefficient for H2O, surface area. Model performance is shown in terms of coefficient of determination $R^2$, Mean Absolute Error (MAE), Root Mean Squared Error (RMSE).
• Figure 3: 10 most important features according to SHAP ranking of Henry coefficient for CO2. For each feature (i.e., each line), 100 dots are shown, representing the 100 samples of the testing set used for computing 100 different SHAP values (impacts on the model output, horizontal axis). The color represents the corresponding feature value. The features are sorted according to the mean over the absolute SHAP values.
• Figure 4: 10 most important features according to SHAP ranking of working capacity for CO2. For each feature (i.e., each line), 100 dots are shown, representing the 100 samples of the testing set used for computing 100 different SHAP values (impacts on the model output, horizontal axis). The color represents the corresponding feature value. The features are sorted according to the mean over the absolute SHAP values.
• Figure 5: 10 most important features according to SHAP ranking of Henry coefficient for H2O. For each feature (i.e., each line), 100 dots are shown, representing the 100 samples of the testing set used for computing 100 different SHAP values (impacts on the model output, horizontal axis). The color represents the corresponding feature value. The features are sorted according to the mean over the absolute SHAP values.