(385bc) Computational Investigation of Physical and Chemical Phenomena in Pristine and Defective Metal-Organic Framework Uio-66 Using Quantum Chemistry and Machine Learning

1. Materials’ atomic modeling: My primary objective is to focus on realistic material modeling across diverse material classes. I am deeply interested in understanding the chemistry of materials and their real-world applications. To this end, I have developed skills in modeling software, namely VESTA, Chemcraft, and ASE.
2. Transport properties: My Ph.D. work has dealt with mass transport in porous materials to search for efficient materials. I am interested in applying the concept of molecular transport to other types of transport at different scales, such as heat transfer (Thermal conductivity), ion diffusion, and surface diffusion.
3. Machine learning methods for chemistry/Cheminformatics: I gained experience coupling state-of-the-art machine learning techniques with chemistry, which I would like to keep working on later in my career. I am willing to extend my knowledge in machine learning and artificial intelligence with their application in chemistry. I am also interested in cheminformatics for material science as structure-activity relations (SAR)/ structure-properties relations (SPR) studies, which have proved to be a powerful tool in biological sciences.
4. Data Engineering: Data quality is paramount when transitioning from conventional computational methods to machine learning frameworks to train an accurate model. Along with machine learning techniques, my focus would be to improve data quality regarding accuracy and correct modeling.

Atomistic modeling of physical processes like diffusion and adsorption provides relevant trends in the mechanism without performing physical experiments in the laboratory. With the appropriate modeling of the material, the simulated results are comparable to the experimental results. The advancement in computing efficiency over the decades has made atomistic modeling a quick, reliable, and powerful technique in material science, and the incorporation of machine learning concepts has increased its effectiveness.

Metal-organic frameworks (MOF) showcase a range of promising applications, from gas adsorption to catalysis, some of them in commercial use. They can potentially decompose the organophosphate chemical weapons into chemically inactive intermediates attributed to the Lewis acid sites in the MOF. Current alternatives are commercial adsorbents like metal-supported activated carbon, which do not decompose the agents but only adsorb them. The efficacy of adsorbents depends on the adsorption capacity and the adsorption kinetics. Current adsorbents inadequately fulfill these criteria. MOFs with superior adsorption properties are favored candidates for the next-generation adsorbents. This work focuses on the Zr-based metal-organic framework UiO-66 and its topological variants. The representative polar molecule, isopropyl alcohol, was used in pristine hydroxylated UiO-66, and the defect-engineered UiO-66 instead of chemical agents to understand the concentration effects on self and transport diffusivity in the defect-engineered MOFs. Along with diffusion, the simulated adsorption capacity of pristine and defect-engineered MOFs was calculated at ambient conditions, which showed improved performance compared to current alternatives.

To address the reactivity of MOFs, monocarboxylic acid capping groups were used during the MOF synthesis to correlate with the reactivity of UiO-66. Using ab initio methods is essential to understand chemical reactivity, but these methods are limited to the system size and computing resources required to explore longer timescales effectively. Thus, a machine learning interatomic potential (MLIP) was proposed along with additional training to enable MLIP training on reaction intermediates and products, which was implemented to explore the chemical reactivity of chemical agents like nerve and blistering agents. The work aimed to identify a material that can both adsorb and react with the agents to mitigate the threat.

Overall, the work showcases a complete overview of adsorption, diffusion, and chemical reactions in MOFs to engineer them for chemical warfare agent destruction. The scope is not limited to this system but applies to multiple complex chemical systems that are difficult or impossible to study experimentally.