(2mf) Machine-Learning Driven Exploration of Catalytic Reaction Networks

Various kinds of key industrial/environmental process (e.g. combustion, heterogeneous catalysis) involve multitude of chemical species and reactions which are represented with very complex reaction network. Correct understanding of reaction mechanism enables potential advances in engineering such as by finding better catalysts or reactor design. Nevertheless, our level of understanding the reaction network and the way we approach is still based on chemist’s intuition rather than being systematic leaving large room for future development. Mechanistic study can be carried out either in bottom-up(theoretical modeling) or top-down(experimental interpretation) directions, but I’d like to stress on bottom-up direction by combining microscopic first-principles data and microkinetic simulation. [1] However, preparation of the input data for microkinetic simulation is often compromised to pristine or idealized systems due to computational cost without properly addressing the complexity of real conditions like in finite temperatures. This requires consideration of statistical averaging which can be readily achieved by machine-learning driven enhanced sampling. [2,3] I’d like to focus on following points as future research direction.

(1) Development of automatic reaction network exploration protocol.

(2) Application of machine-learning to reduce computational cost for enhanced sampling and data analysis

(3) Microkinetic modeling based on first-principles input data and reaction mechanism elucidation

Postdoctoral project: “Data-driven systematic exploration of complex reaction networks in heterogeneous catalysis” (Advisor : Karsten Reuter, Fritz-Haber Institute of the Max Planck Society, Germany)

- Sponsored by Alexander von Humboldt research fellowship for postdoc

- Supervised by Johannes T. Margraf (Incoming professor in Bayreuth university, Germany)

PhD dissertation: “First-principles computational and machine Learning approach to understanding reaction mechanisms of hazardous chemicals” (Advisor : Byungchan Han, Yonsei university, Korea)

Research stay: (Host: Karsten Reuter, Technical university of Munich, Germany)

- Jointly sponsored by DAAD/NRF Summer institute program.

Teaching Interests:

I’m a chemical engineer by training but since graduate school, I’m more involved in atomistic or molecular scale modeling. Over the course of PhD and postdoc periods, I have supervised several master/PhD students for their researches and published a few papers with them which gave me experiences on scientific discussion and jointly progressing over a particular scientific problem. [7,8] As a graduate student, I took part in teaching assistant for “quantum chemistry for chemical engineers” taking in charge of several hands-on tutorial sessions and tutored python programming courses for undergraduates. I was also involved in several teaching sessions as postdoc helping PhD students applying machine learning techniques for their research purposes.

I’d like to help students to have integrated points of view on different scales and disciplines as chemical engineer. This will be achieved through programming their own simulation from scratch along with implementing the physical/chemical principles to their own suite of program. By doing so, students can have deeper understanding on both micro/macroscopic scales and related their knowledge to industrial point of view as a chemical engineer.

Selected Publications (Total: 12, 7 first author, 5 co-authored)

[1] Margraf, J. T., Jung, H., Scheurer, C., & Reuter, K. (2023). Exploring catalytic reaction networks with machine learning. Nature Catalysis, 6(2), 112-121.

[2] Jung, H., Sauerland, L., Stocker, S., Reuter, K., & Margraf, J. T. (2023). Machine-learning driven global optimization of surface adsorbate geometries. npj Computational Materials, 9(1), 114.

[3] Stocker, S., Jung, H., Csányi, G., Goldsmith, C. F., Reuter, K., & Margraf, J. T. (2023). Estimating Free Energy Barriers for Heterogeneous Catalytic Reactions with Machine Learning Potentials and Umbrella Integration.

[4] Jung, H., Stocker, S., Kunkel, C., Oberhofer, H., Han, B., Reuter, K., & Margraf, J. T. (2020). Size‐Extensive Molecular Machine Learning with Global Representations. ChemSystemsChem, 2(4), e1900052.

[5] Jung, H., Kang, J., Chun, H., & Han, B. (2018). First principles computational study on hydrolysis of hazardous chemicals phosphorus trichloride and oxychloride (PCl3 and POCl3) catalyzed by molecular water clusters. Journal of hazardous materials, 341, 457-463.

[6] Jung, H., Shin, T., Cho, N., Kim, T. K., Kim, J., Ryu, T. I., ... & Han, B. (2020). Thermochemical study for remediation of highly concentrated acid spill: Computational modeling and experimental validation. Chemosphere, 247, 126098.

[7] Park, S.*, Jung, H.*, Min, K. A., Kim, J., & Han, B. (2021). Unraveling the selective etching mechanism of silicon nitride over silicon dioxide by phosphoric acid: First-principles study. Applied Surface Science, 551, 149376. (*shared first author)

[8] Tao, W.*, Jung, H.*, Ryu, T. I., Hwang, S. R., & Han, B. (2021). Dramatic catalytic activation of kinetically inert disilane hydrolysis in metallic iron particulate via barrierless chemical dissociation: First-principles study. Applied Surface Science, 560, 149988. (*shared first author)

Keywords : First-principles calculation, Reaction network, machine-learning, global optimization, free energy sampling