(4ey) Deciphering Catalyst Structural Evolution in Heterogeneous Catalysis: Machine Learning Accelerated Nanoparticle Modeling Under Environment-Driven Reconstruction

Heterogeneous catalysis is known for its intricate structural complexity, making it challenging to understand the reaction mechanisms on catalyst surfaces. One significant difficulty arises from the intangible interaction between adsorbate and catalyst active sites during surface reactions. Notably, multicomponent (bimetallic, trimetallic...) catalysts are renowned to frequently exhibiting high, synergistic performance as compared to their metal constituents, and can undergo adsorbate-induced surface segregation and reconstruction upon exposure to reaction conditions. Meanwhile, the morphology change of metallic nanoparticles under such reconstruction continues to be a significant concern with great research interest in the field of heterogeneous catalysis.

Tremendous progress has been witnessed in computational catalysis, where numerous important catalytic systems have been explored on single-crystal facets. Still, an actual catalytic system is inevitably much more complicated than single-crystal models. To date, nanoparticle-based modeling studies that examine adsorbate-induced segregation fall broadly into multiple approaches (e.g. Wulff constructions, molecular dynamics (MD) simulations, Monte Carlo simulations), all of which have limitations (e.g. trade-off between computational resources, accuracy, and efficiency). Coupling machine learning (ML) techniques with atomic simulations have rapidly evolved over recent years, with successful applications of structure-property determination and reaction pathway identification. Additionally, ML potentials can be constructed and further employed in classical MD. The combination of ML potentials with MD emerges as a promising tool to simulate large-scale nanoparticle systems over extended time periods while maintaining quantum mechanical (QM) accuracy. Such a powerful tool enables a priori examination of dynamic, reconstructing nanoparticles under reactive gas exposures.

During my PhD in computational catalysis, I led projects on the development of multiscale modeling approaches to capture environment-driven bimetallic nanoparticle reconstruction. Using density functional theory (DFT), a concise mean-field nanoparticle model (kubic harmonics) was developed to identify the nanoscale interactions leading to bimetallic nanoparticle reconstruction induced by reaction conditions. This work was published and further selected as front cover in Catalysis Science & Technology. Although such multiscale method was successfully benchmarked by experimental data using Rh₅₀Pd₅₀ nanoparticle as a case study, two follow-up questions have arisen:

• What is the dynamic morphology of bimetallic nanoparticles during adsorbate induced reconstruction under reactive environments?
• Is it possible to simplify complex nanoparticle models (constructed as full atom systems) by concise mean-field results (easy to obtain), while still maintaining a consistent accuracy?

To address these questions, I am investigating the induced, dynamic morphological evolution of nanoparticles under cyclic oxidizing/reducing reaction conditions through several multiscale modeling methods. My approach combines density functional theory (DFT), mean-field models, Wulff construction, and MD with ML interatomic potentials (MD+ML-IAP). This method will allow me to benchmark the accuracy for mean-field models from MD+ML-IAP, which enables a thorough description of morphological evolution of nanoparticle reconstruction as well as examines the effectiveness of concise mean-field tools that requires less time and computational resources. During this project, I exploit my deep background knowledge of interdisciplinary subjects from computational catalysis to chemistry to statistical mechanics to data science with advanced analytical ML tools. Overall, my goal is to advance current knowledge of bimetallic nanoparticle catalyst structural evolution and accelerate the potential of tunable nanoparticle engineering as a novel catalyst design strategy.

Research Skills:

• DFT (VASP)
• Ab initio phase diagram construction
• Multiscale nanoparticle modeling
• Mean-field model construction
• Coverage, configuration dependent energy mapping
• ML-IAP construction (FitSNAP)
• Ab initio MD
• Classical MD (LAMMPS)
• Supervised ML model construction for multicomponent catalyst screening
• Potential energy surface (PES) mapping from transition state calculations
• High learning ability
• Creative problem reasoning and solving
• Critical thinking
• Leadership and mentorship
• Efficient communication and collaboration
• Scientific writing and visualization

Successful Proposals:

Center for Functional Nanomaterials (CFN) at Brookhaven National Lab, 2023

Determining Promoter Effects in the Nanoscale Surface Structure and Stability of NiO-based Catalysts for Ethane Oxidative Dehydrogenation (130,000 CPU hours)

Center for Nanoscale Materials (CNM) at Argonne National Lab, 2022

Determining Promoter Effects in the Nanoscale Surface Structure and Stability of NiO-based Catalysts for Ethane Oxidative Dehydrogenation (440,000 CPU hours)

Center for Nanophase Materials Sciences (CNMS) at Oak Ridge National Lab, 2022

Enabling the Rapid Prediction of Reaction Intermediate Stability via Linear Scaling Relations for Ethane Oxidative Dehydrogenation on Transition Metal Oxides (450,000 CPU hours)

Presentations:

1. Wang, Shuqiao; Hensley, A.J.R., Nanoscale Structure and Stability of Doped NiO-Based Catalysts for Ethane Oxidative Dehydrogenation, ACS Middle Atlantic Regional Meeting (MARM), Jun. 2024. (poster)
2. Wang, Shuqiao; Hensley, A.J.R., Elucidating the Nanoscale Driving Forces for Environment-Driven Rh-Pd Nanoparticle Reconstruction, 2023 AICHE Annual Meeting, Nov. 2023. (oral)
3. Wang, Shuqiao; Hensley, A.J.R., Determining Promoter Effects in the Nanoscale Surface Structure and Stability of NiO-Based Catalysts for Ethane Oxidative Dehydrogenation, 2023 AICHE Annual Meeting, Nov. 2023. (oral)
4. Wang, Shuqiao; Hensley, A.J.R., Estimating Stability of Ethane-Derived Species on NiO(100) and NbO(100) from Density Functional Theory, ACS Northeast Regional Meeting (NERM), Oct. 2022. (poster)
5. Wang, Shuqiao; Hensley, A.J.R., Estimating Stability of Ethane-Derived Species on NiO(100) and NbO(100) from First Principles, CNMS User Meeting, Aug. 2022. (poster)

Publications:

1. Wang, Shuqiao; Hensley, A.J.R., Rotational Symmetry Effects on Multibody Lateral Interactions between Co-Adsorbates at Heterogeneous Interfaces. ACS Physical Chemistry Au 2024, In Press. DOI: 10.1021/acsphyschemau.4c00019.
2. Wang, Shuqiao; Hensley, A.J.R., Probing the Nanoscale Driving Forces for Adsorbate-Induced Rh₅₀Pd₅₀ Nanoparticle Reconstruction via Mean-Field Models of Multi-Faceted Nanoparticles. Catalysis Science & Technology 2024, 14, 1122-1137.
3. Garzon, A.; Wang, Shuqiao; Omoniyi, A.; Tam, L.; Che, F.; Hensley, A.J.R., Temperature and Pressure Driven Functionalization of Graphene with Hydrogen and Oxygen via Ab Initio Phase Diagrams. Applied Surface Science, In Review.
4. Furrick, I.; Omoniyi, A.; Wang, Shuqiao; Robinson, T.; Hensley, A.J.R., Integration of Facet-Dependent, Adsorbate-Driven Surface Reconstruction into Multiscale Models for the Design of Ni-Based Bimetallic Catalysts for Hydrogen Oxidation. ChemCatChem, In Review.
5. Wang, Shuqiao; Hensley, A.J.R., Dopant Effects on the Geometric Structure and Electronic Properties of NiO(100). In Preparation.
6. Leitner, K.; Wang, Shuqiao; Hensley, A.J.R., Identification of Selectivity Determining Steps for Ethane Oxidative Dehydrogenation on NiO(110). In Preparation.