(556a) Physics-Based Universal Representation of Single and Multi-Component Equilibria to Simulate Complex Adsorption Processes

Adsorption equilibrium data of a single component or that of a multi-component mixture on an adsorbent, can be determined either experimentally or via grand canonical Monte Carlo (GCMC) simulations at a set of discrete fluid phase concentrations of the component(s). With the increase in machine-learning tools, we will likely rely on computational methods for determining single and binary adsorption equilibria. Concurrently, we are also witnessing the use of performance-based screening where sorbents are selected not based on simple metrics but using process-level models. The use of theoretical functional forms for isotherm models that closely describe the equilibrium data is the link between GCMC and process models, as demonstrated in software packages like RUPTURA that are limited to simulating isothermal adsorption column breakthrough profiles^[1].

We propose using data-driven spline models to seamlessly represent single and multi-component equilibria for novel sorbate-sorbent interactions that challenge traditional isotherm models. The use of uninformed data-driven models like splines to describe sparsely measured experimental equilibrium measurements, is susceptible to spurious artefacts from overfitting, when used to either interpolate or extrapolate equilibrium loadings at other compositions (partial pressures of the component) and temperatures. This is proposed to be regularized by using a thermodynamic basis to supplement sparse experimental data in making space-filling equilibrium loading predictions, before using a data-driven spline model. The Clausius-Clapeyron relation is used to enforce isosteric predictions that factor in pressure dependence of loadings at arbitrary temperatures where data is not available. The following sorbate-sorbent combinations have been investigated using the abovementioned framework for both unary and multi-component adsorption equilibria that exhibit unique inflections and steps, particularly in case of the metal organic framework (MOF) adsorbents: N₂ /CO₂ on zeolite-13X^[2], CO₂/H₂O on CALF-20^[3], and CO₂/N₂ on tetramine-appended MOF^[4]. This is followed by experimentally validating both the simulated thermal and composition fronts, using spline models for thermodynamically enriched experimental adsorption equilibrium data. Based on the success of predicting non-isothermal experimental single and binary breakthrough profiles, we extend the technique to predict pressure swing adsorption processes.

References

[1] Sharma, S. et al. RUPTURA: simulation code for breakthrough, ideal adsorption solution theory computations, and fitting of isotherm models. Molecular Simulation, 49(9), 93–953 (2023).

[2] Wilkins, N. S. & Rajendran, A. Measurement of competitive CO 2 and N 2 adsorption on Zeolite 13X for post-combustion CO 2 capture. Adsorption 25, 115–133 (2019).

[3] Nguyen, T. T. T. et al. Competitive CO2/H2O Adsorption on CALF-20. Ind Eng Chem Res 63, 3265–3281 (2024).

[4] Evaluation of a tetramine-appended MOF for post-combustion CO2 capture from natural gas combined cycle flue gas by steam-assisted temperature swing adsorption. doi:10.26434/chemrxiv-2023-gs8tg.