The chemical industry is at a turning point, with rising demand for olefins, critical building blocks for plastics, polymers, and countless downstream materials. As traditional feedstocks become more volatile, producing these compounds efficiently is both a technological challenge and a major economic opportunity. Nonoxidative alkane dehydrogenation offers a promising route by converting abundant alkanes into high-value olefins, but it requires catalysts that are highly active, selective, and stable under harsh conditions. My research leverages first-principles calculations and multiscale modeling to understand reaction events at the surface level, guiding the rational design of next-generation catalysts that enable more efficient and cost-effective chemical processes.
A central theme of my work is the development of structure-activity relationships (i.e., linear relationships) to accelerate catalyst discovery, reducing reliance on trial-and-error experimentation. We developed a comprehensive alkane dehydrogenation model that captures the catalyst Lewis acid–base properties (through the binding energy of dissociated H2) and alkane stability (through carbenium ion stability). Both descriptors are readily accessible from experiments or can be computed using density functional theory (DFT) calculations. We identified Ga-doped γ-Al₂O₃ as a highly active catalyst for ethane, propane, and isobutane dehydrogenation, where Ga incorporation significantly lowered the activation barriers of the kinetically favored mechanism by tuning surface Lewis acidity. Extending this methodology to metal nitrides, we discovered aluminum nitride as an efficient dehydrogenation catalyst. Further refinement through Zn doping enhanced catalytic performance, lowering activation barriers and increasing the overall turnover frequency for propylene production. Proof-of-concept experiments confirmed Zn-doped AlN’s superior performance, achieving a significantly higher intrinsic reaction rate (255 vs. 223 mol h−1 g−1), lower apparent activation energies (154 vs. 214 kJ mol-1), and a 4.5% increase in propane conversion compared to undoped AlN.
Expanding into methane activation, we explored its thermo- and photocatalytic nonoxidative coupling to C₂ hydrocarbons on rutile TiO₂. By combining DFT calculations with kinetic Monte Carlo (kMC) simulations, we revealed that photogenerated holes facilitate methane activation via homolytic C–H bond cleavage, substantially lowering energy barriers compared to traditional thermocatalysis. While methyl radicals recombine to form ethane at ambient temperatures, surface poisoning by adsorbed hydrogen poses a challenge for sustained activity. Elevating the temperature aids in hydrogen desorption and catalyst regeneration, enabling a synergistic photo-thermocatalytic process. This work introduces a dynamic catalysis approach that integrates heat and light for more energy-efficient and selective hydrocarbon upgrading.
Currently, I leverage my experience in high-throughput computational chemistry, combined with my background in data science, to tackle challenges in electrochemical energy storage. I focus on understanding electrolyte decomposition and solid-electrolyte interphase (SEI) formation in lithium-ion batteries. In this work, we constructed the largest chemical reaction network to date, containing over 10,000 species and 209 million electrochemical reactions. Using stochastic network analysis and without prior knowledge on reaction mechanisms or end products, we recovered experimentally observed products and uncovered new molecules that were previously unknown, with the potential to significantly enhance battery performance. These findings enable the development of more reliable batteries, reducing the need for frequent replacements and ultimately lowering costs for industries reliant on large-scale energy storage systems. Whether in catalysis or battery research, my work couples first-principles-based multiscale modeling and data-driven approaches to drive innovation in sustainable energy solutions.
Research Skills
Computational Modeling: DFT (CP2K, VASP, Gaussian, QChem), kMC simulations, Microkinetic Modeling, Fluid Dynamics (COMSOL)
Programming: Python, BASH, MATLAB, Git/GitHub, Jupyter
Machine Learning & Data Science: PyTorch, NumPy, Pandas, SciPy, Scikit-learn, Pymatgen, Atomate, RDKit
Visualization: VESTA, Blender, Materials Studio, Avogadro, VMD, Crystal Maker, PovRay
High-Performance Computing: HPC workflows (FireWorks)