2024 AIChE Annual Meeting
(54e) Investigation of Pd-Based Alloy Membranes for Hydrogen Separation: An Application of a Universal Neural Network Potential
Authors
In this work, we report our investigation of the surface composition and structures of the Pd-based alloys and their impact on the adsorption of gas species on the metallic membranes. Our unique proprietary universal neural network potential, called Preferred Potential (PFP), was applied to perform necessary atomistic simulations. One of the advantages of using PFP is the capability of simulating chemical reactions at near-DFT accuracy on much larger system or much higher speed.
We examined the surface segregation behavior of the alloys as a function of temperature and composition using Monte Carlo simulations. Pronounced segregation of the solute was found in fcc Pd3Ag(111) and Pd3Au(111), but solvent segregation was found in Pd3Pt(111). The segregation behavior is found to be significantly altered by the presence of adsorbate species and absorbed hydrogen in the bulk phase. These findings are consistent with the previously known facts in the literature.
MD simulations are also utilized to explore the dynamics of the gaseous species as well as the surface metal atoms. The MD simulations were found to allow structural relaxation in which the simple MC simulations cannot explore. The surface perturbation observed do vary depending on the composition of the alloy and the adsorbed species.
The outcome of this study will provide some important insights to the design of Pd-based alloy membranes and the method of evaluating their performance with atomistic detailed information.