2025 AIChE Annual Meeting

(465l) Hydrogen Adsorption and Membrane Separation Performance in Amorphous Pd-Based Alloys Using a Universal Machine Learning Interatomic Potential

Authors

Taku Watanabe - Presenter, Georgia Institute of Technology
Akihiro Nagoya, Preferred Computational Chemistry
Membrane materials for highly efficient separation and purification of hydrogen are attracting attention for the realization of a sustainable energy society. In particular, Pd-based materials exhibit high permeability and selectivity as hydrogen separation membranes, but conventional crystalline Pd membranes have issues such as reduced lifespan due to hydrogen embrittlement and high costs. On the other hand, amorphous alloys, with their diverse local structures, are considered to be materials that can overcome some of the challenges of crystalline materials, and experimental and theoretical reports have shown the possibility of high hydrogen permeation performance and improved durability [1-4]. However, their performance is highly dependent on composition, and there are many unresolved issues regarding the microscopic hydrogen permeation mechanisms and performance expression mechanisms.

In this study, we aim to understand the hydrogen adsorption mechanism and membrane separation performance in amorphous alloys at the molecular level using our universal machine learning interatomic potential, Preferred Potential (PFP), on a cloud service called Matlantis, and to obtain design guidelines leading to high performance.

Multicomponent amorphous alloy structures, mainly Pd-Si based, were constructed by molecular dynamics (MD) simulations simulating the rapid quenching method. The quenching rate was systematically varied to examine the effects on local structure and hydrogen adsorption/permeation performance. The generated structures were locally characterized by Voronoi analysis, coordination number distribution, and SOAP descriptors. Hydrogen adsorption energies were extensively evaluated at thousands of sites using the MLIP, and representative sites were compared and verified with first-principles calculations (DFT). Membrane permeation performance was quantitatively analyzed by non-equilibrium molecular dynamics (NEMD) to analyze hydrogen permeation behavior under chemical potential differences.

As a result of simulations based on MLIP, hydrogen adsorption in Pd-based amorphous alloys preferentially occurs at locally Pd-rich irregular structural sites, and particularly high adsorption energy was observed around Pd clusters with low coordination numbers. This result is consistent with existing research. It was also found that the size distribution and connectivity of free volume in the structure decisively contribute to hydrogen permeation performance and that local electronic states are important for permeation path selection. Furthermore, the addition of subsidiary elements such as Ni and Cu increases the free volume. It improves hydrogen permeability while also expanding the diversity of adsorption sites, suggesting that high performance can be expected in specific composition ranges. As a result, composition candidates for the Pd-Ni-Cu-P system, which is expected to have high performance, were theoretically proposed.

The theoretical knowledge obtained in this study regarding the relationship between atomic-level structure and hydrogen adsorption/permeation performance is useful as a molecular design guideline for high-performance Pd-based amorphous hydrogen separation membranes. In the future, we plan to conduct comprehensive performance evaluations, including not only permeation performance but also thermal stability, which is an issue for amorphous alloys, as well as theoretical exploration of further high-performance materials and advancement of simulation methods. We hope this work will also showcase the performance of our MLIP and it's applicability to complex problems.

References

[1] Zhang, K., & Way, J. D. (2017). Sep. Purif. Technol., 186, 39–44.

[2] Bosko, M. L., et al. (2021). Int. J. Hydrogen Energy, 46, 7271–7286.

[3] Itoh, N., et al. (1998). J. Membr. Sci., 139, 29–35.

[4] Palumbo, O., et al. (2016). Membranes, 6(3), 43.

[5] Sarker, D., et al. (2018). Sci. Rep., 8, 6084.

[6] Takamoto, S., et al. (2022). Nat. Commun., 13, 2991.

[7] Takamoto, S., et al. (2023). J. Materiomics, 9(1), 90–100.