Multicomponent Gas Adsorption Behavior of Kerogen: A Molecular Investigation

The gas adsorption capacity of kerogen is typically determined experimentally by isothermal adsorption experiments. Mathematical models, known as adsorption isotherms, are utilized to tie experimental data to a theoretical framework to provide a continuous description of the underlying phenomena. In this work, we present a molecular investigation of the single-, binary-, and multicomponent adsorption behaviors of methane (CH4), ethane (C2H6), and carbon dioxide (CO2) based on a representative kerogen nanostructure to aid the assessments of the predictive power of theoretical adsorption models. In particular, we elucidate the applicability of the Langmuir, Toth, and Langmuir-Freundlich model for single-component methane, ethane, and carbon dioxide systems, as well as the nonrevised Langmuir, revised Langmuir, extended Freundlich, and Langmuir ratio correlation (LRC) for an equimolar binary of methane/ethane and Langmuir ratio correlation for a ternary mixture methane/ethane/carbon dioxide systems. Isotherm parameters are optimized in a mean square/root-mean-square sense and supported by statistical inference in the form of 95% confidence intervals complemented by 95% prediction bounds. In addition, we delineate how the maximum adsorption capacity can be calculated directly from the remaining model parameters, thereby reducing the complexity of the nonlinear regression problem and providing analytical solutions for the adsorption behavior of multicomponent models at infinite pressure. The key results show that Langmuir, Toth, and Langmuir-Freundlich models are capable of delineating the overall adsorption trend reasonably well for the single-species cases. For the binary (CH4/C2H6) and ternary (CO2/CH4/C2H6) equimolar mixture, the LRC model is the only isotherm capable of capturing the phenomena.

Automated summary

This study investigates the adsorption behavior of methane, ethane, and carbon dioxide in kerogen and evaluates the predictive power of different theoretical models. The results show that the performance of the models varies under different conditions.