(525j) Prediction of Metal-Organic Framework Adsorption Isotherms Using Ab Initio Derived Neural Network Potentials

Metal-organic frameworks (MOFs) form an extensive class of porous materials of interest for new applications, as they show high surface area, high thermal stability, and strong chemical tunability. As a consequence, predicting their properties computationally is crucial in advancing the exploration of such a wide chemical space. Particularly, MOFs are promising materials in the carbon capture space, as they present low gas transport limitations, and specific MOFs have been shown to selectively and reversibly bind to carbon dioxide [2]. In this space, a predictive and generalizable method for the calculation of adsorption isotherms in MOF systems could not only accelerate the discovery of new promising materials for carbon capture but also reveal the underlying mechanism of adsorption processes.

While empirical force fields can be combined with grand canonical Monte Carlo (GCMC) simulations to estimate adsorption isotherms in specific MOFs [1], this approach relies on fixed functional forms as well as an extensive system-specific parameterization, limiting its applicability to a broader set of MOFs or to automated force field generation. Ab initio neural network potentials (NNPs), derived from density functional theory calculations, are a promising and generalizable alternative for accurate binding energetics and isotherm prediction. Furthermore, its ability to capture interactions in a diversity of chemical environments (upon further training) provides an opportunity to accurately elucidate adsorbate pore dynamics, MOF framework flexibility as well as adsorbate-adsorbate interaction effects.

In this work, we combine ab initio NNPs and GCMC simulations to predict CO2 adsorption isotherms for well-studied MOF-274, Mg2(dobpdc) (dobpdc4- = 4,4’-dioxidobiphenyl-3,3’-dicarboxylate), comparing to prior simulations and experiments. In the attached figure, we present preliminary results for this system, which shows a promising ability to recover experimental results from the proposed method. Ultimately, this method will be used to predict adsorption isotherms for different gases, such as water, nitrogen, and methane, as in application conditions additional adsorbents may be present.

In the short run, we aim at applying this prediction methodology to MOF systems which require molecular dynamics steps throughout the GCMC simulation in order to fully capture the adsorption physics. For instance, some MOFs present cooperative mechanisms related to adsorbent-adsorbent interactions, while others undergo framework mobility upon adsorption. Simulating either of these systems with parametrized force-field methods would be quite challenging, as it would be necessary to accurately capture all physical interactions in the system to correctly evolve it over time. The NNPs we propose, on the other hand, can predict the dynamics of these systems at an ab-initio accuracy, and therefore, constitute a promising method in this application space. In the long run, we hope this contribution will accelerate the understanding of gas adsorption in MOF systems, and the discovery of more efficient carbon capture materials.

[1] Mercado, Rocio, et al. "Force field development from periodic density functional theory calculations for gas separation applications using metal–organic frameworks." The Journal of Physical Chemistry C 120.23 (2016): 12590-12604.

[2] Sumida, Kenji, et al. "Carbon dioxide capture in metal–organic frameworks." Chemical reviews 112.2 (2012): 724-781.