(356c) Molecular Modeling and Machine Learning-Based Design and Discovery of Nanoporous Materials for Energy and Environmental Applications

I am interested in design and discovery of next-generation of materials using molecular simulations, ab-initio calculations and machine learning for gas storage and separation, sensing, and environmental applications. I also want to work on development of novel data science-assisted modeling protocols for material discovery.

Abstract

Though conventional experiment-based material design had been successful in finding novel materials in the past, it would become more challenging to discover next-generation of materials through this route alone. This is evident as experiments requires a plethora of physical resources to examine a single potential material and often, they can only test a limited number of samples. In this prospect, computational research can offer right insights and can lead the way in exploring and discovering novel materials. And with the advent of parallel computing, and smaller transistors, we can test thousands of molecules within days using the proper algorithm and identify the best candidates for a particular application; thus, saving huge overhead costs and time. Towards this goal. I have been working on a variety of projects using molecular modeling and machine learning-based protocols.

• Navigating gas adsorption landscapes in MOFs using Active learning — We implemented an Active learning protocol to model gas adsorption in a Cu-BTC metal-organic framework (MOF). AL is a methodology which iteratively adds samples to the model until the uncertainty in the prediction is low enough to be accepted as a surrogate model. The sampling process for adsorption simulation has been done using grand canonical monte carlo (GCMC) simulations and the surrogate was initiated using a prior dataset. The AL protocol has been tested for both isotherms (constant temperature) as well as for a pressure-temperature phase space. Introducing AL to predict adsorptions especially for multiple features is extremely beneficial as we can get a comparable accuracy to GCMC with an order of magnitude smaller number of samplings. This would help to explore a vast phase of operating conditions for gas adsorption and separation in MOFs and can be transferred to other adsorption systems.
• Nano-porous material design and discovery for gas sensing using Henry’s coefficients — Metal-organic frameworks (MOFs) are known for their high internal surface area and pore volume and have shown potential in multiple applications including their use in gas sensors. In this work we study adsorption of six different gases (H₂O, CO₂, N₂, N₂O, CH₄, and O₂) at dilute conditions on a MOF database (CoRE MOF database > 9000 structures) using the Widom particle insertion technique. We conducted this study in a high-throughput screening fashion for each MOF adsorbent–adsorbate pair and evaluate the Henry’s coefficients. The MOF structures are selected from the CoRE MOF database which consists of structures that have been synthesized in laboratory. Using these Henry’s coefficients, we calculate the Selectivity of all the possible binary gas mixtures that can form out of these six gases. Then, we analyze the geometric and chemical properties of the MOFs which provides the best performance in terms of sensing these gases. Thereafter, we use a ranking criterion to find MOFs that are best candidates for sensing multiple gases.
• Understanding the role of charge distribution and pore size for water vapor adsorption in idealized nano-porous materials — Water vapor is present in major gaseous streams such as industrial flue gas, air, etc., separating it efficiently can augment water supply, increase efficiency for proton conduction, better CO₂ capture performance in presence of water and many others. MOFs, having a high pore volume and surface area, can act as a media for separating water vapor from these mixtures. In this work, we sought to provide further understanding on the role of electrostatic interactions in water vapor adsorption in nanoporous materials. To do so, we designed idealized carbon-based cylinders (ICC) and performed Monte Carlo simulations to understand the role of electric potential and pore size in the water vapor adsorption. Our study revealed that charge arrangement on pore surface plays a significant role in influencing the hydrophilicity of ICCs. We also observed that a surface was found to be more hydrophilic when charges are alternated (from positive and negative) along the circumference than when they were alternated along the length of the cylinder. Through the electric potential, we found a strong relationship between the electric potential inside a pore and the water vapor adsorption sites. We hypothesize that MOFs having similar electrostatic characteristics to the ICCs would also exhibit identical water vapor adsorption behavior.