(689a) Thermodynamics of Al Substitution in CHA with and without Organic Structure-Directing Agents | AIChE

(689a) Thermodynamics of Al Substitution in CHA with and without Organic Structure-Directing Agents


Hoffman, A. - Presenter, University of Florida
Xie, M., Imperial College London
Paris, C., Instituto de Tecnologia Quimica
Moliner, M., ITQ (CSIC-UPV)
Gomez-Bombarelli, R., Massachusetts Institute of Technology
Zeolites are porous aluminosilicate materials that are commonly used as catalysts. The active sites in zeolites form when substituting Al3+ for Si4+, which introduces an anionic charge that is frequently balanced by H+ to produce a Brønsted acid site. Changes in the proximity or location of Al within zeolites can change turnover rates (per H+) for reactions like methanol dehydration [1, 2]. Despite its importance, the factors that govern Al distribution remain undetermined. Recent work suggested that a combination of the relative stability of the crystallographic position of the Al in a framework and a sum of coulombic interactions between cationic organic structure-directing agents (OSDAs) and Al influenced Al placement [3]. Here, we enumerate all unique Al arrangements in CHA zeolites for Si/Al=5–35 (1–6 Al substitutions in a 36 T-atom unit cell). There is only one symmetrically unique Al position in CHA, but the number of unique structures increases with each Al substitution (Fig. 1a). This example shows that, even in high-symmetry frameworks, Al positioning drastically alters the number of symmetrically equivalent structures and illustrates how brute-force introduction of Al without removing duplicate structures can make computational studies combinatorically infeasible. We compute the energies of these arrangements with and without cationic OSDAs balancing framework anions using density functional theory (DFT). Energies of structures without OSDAs depend linearly on the sum of inverse Al-Al distances in the unit cell (Fig. 1b), indicating that predominantly coulombic interactions govern the relative stability of Al arrangements. Finally, we use these data and similar calculations with cationic OSDAs present to train a graph neural network model to evaluate the stabilities of Al arrangements with cations present without costly DFT calculations.


[1] Chem. Mater. 2016, 28, 2236–2247.

[2] ACS Catal. 2017, 7, 6663–6674.

[3] Chem. Mater. 2020, 32, 9277–9298.