Diffusion of Carbamazepine in Hydrophobic Zeolites: A Comparative Study Using Classical and Machine-Learned...

Area of Science:

• Environmental Science
• Materials Science
• Computational Chemistry

Background:

• Hydrophobic zeolites show promise for adsorbing persistent pollutants like carbamazepine (CBZ).
• Understanding the diffusion dynamics of CBZ within zeolite pores is critical for optimizing their adsorption capabilities.
• Current computational methods often use static interaction energies, limiting insights into dynamic processes.

Purpose of the Study:

• To investigate the free energy surfaces (FES) governing carbamazepine diffusion in hydrophobic zeolites.
• To compare the accuracy of classical force fields versus a neural network potential (MACE) for simulating CBZ diffusion.
• To elucidate the complex energetic landscape influencing CBZ transport within zeolite pores.

Main Methods:

• Umbrella sampling simulations were performed to calculate FES for CBZ diffusion.
• All-silica zeolites were used as model systems for the simulations.
• Results from a classical force field were compared against a fine-tuned MACE neural network potential.

Main Results:

• Both classical force fields and MACE showed qualitative agreement, but MACE predicted higher activation barriers.
• MACE's accuracy is attributed to better representation of atomic contact energy penalties at transition states.
• CBZ was observed to undergo kinetic trapping in metastable orientations, indicating a complex diffusion pathway.

Conclusions:

• Carbamazepine diffusion in zeolites is governed by a complex landscape of translational and rotational barriers.
• Neural network potentials like MACE provide a more accurate description of diffusion dynamics than static interaction energies.
• This study supports the rational design of shape-selective zeolites for effective environmental remediation of pollutants like CBZ.