(2cz) Physics-Informed Material Discovery Tools for Energy and Space Applications

A major bottleneck to enabling next-generation energy, health, space, and sustainability technologies is the lack of efficient discovery tools for advanced functional materials. This inefficiency is often a result of the overwhelmingly large combinatorial space of candidate materials, which is sparsely observed. Furthermore, existing approaches to explore this space are often biased by expert “knowledge” and tend to favor material configurations similar to the ones that are already known to perform well. This may well lead to the discovery of suboptimal materials. Additionally, the high cost associated with experimental characterization and first principles quantum mechanical calculations restricts the size of the available datasets. In light of these challenges, my research vision is to accelerate the discovery of new chemicals, materials, and devices that are useful to society through machine learning (ML) and automation. To realize this vision, my group will integrate information from disparate sources(quantum mechanical calculations, coarse-grained and multi-scale models, and data science methods) into a Bayesian optimization pipeline that is suited to expensive sparse datasets. Our primary focus will be to develop new ML tools that blend Bayesian optimization and neural network methods to model and design energy-harvesting materials and thermal protection systems for energy and space applications. The initial research directions of my lab will be:

1. Bayesian optimization-accelerated discovery of perovskite alloys and tandem perovskite devices for next-generation space solar cells
2. Development of deep learning models for the design of multi-scale materials and devices
3. Multi-scale modeling of the chemistry and transport within porous materials employed in space applications, with a particular emphasis on the new woven thermal protection systems developed by NASA.

Research Contributions and Relevant Prior Work

My expertise in computational chemistry, multi-scale modeling, hypersonics, and machine learning provides a unique skill set enabling me to tackle the proposed research topics effectively. My postdoctoral research introduces a class of material discovery methods that combine physics-informed belief models with Bayesian optimization. The first method I have co-developed is called a Physical Analytics pipeline, PAL 2.0 [Sharma Priyadarshini et al., in preparation, Clancy, Le, Sharma Priyardarshini, HEMI Seed Grant, 2023]. The key contributing factor of my proposed framework is the creation of a physics-based hypothesis by combining Neural Networks and Gaussian Processes Regression. PAL 2.0 has been successfully applied and validated on various materials, including metal halide perovskites and thermoelectric semiconductors. We are now using PAL 2.0 in a closed-loop collaboration with experimentalists to discover high-temperature shape memory alloys [Clancy (PI), Gienger (co-PI), Sharma Priyadarshini (Senior Personnel), Space@Hopkins Seed Grant]. In my doctoral research, I worked on an innovative and novel approach involving the doping of Thermal Protection Systems (TPS) with heat-absorbing molecules. This technique enhances the payload capacity of the spacecraft, resulting in more efficient and cost-effective space missions. As part of this effort, extensive quantum chemistry calculations were carried out for carbon clusters and hydrocarbons.

Funded Grants

1. Clancy, P. (PI), Nam, L. Q. (co-PI), Sharma Priyadarshini, M. (Senior Personnel), Towards Proton Radiation-Resistant Perovskite Solar Cell Materials for Space Applications, internal JHU funding: HEMI/APL Seed Grant ($50,000), 2023.
2. Clancy, P. (PI), Gienger, E. (co-PI), Sharma Priyadarshini, M. (Senior Personnel), Machine Learning-accelerated Discovery of New High-Entropy Shape Memory Alloys for Space Actuation, internal JHU funding: Space@Hopkins Seed Grant ($25,000), 2023.

Selected Publications

1. Sharma Priyadarshini, M.*, Venturi S.*, Zanardi I. and Panesi M., Efficient Quasi-Classical Trajectory Calculations by means of Neural Operator Architectures, Physical Chemistry Chemical Physics, 25(20), 13902-13912, 2023.
2. Sharma Priyadarshini, M., Jo, S. M., Venturi S., Schwenke, D. W., Jaffe, R. L. and Panesi M., Comprehensive Study of HCN: Potential Energy Surfaces, State-to-State Kinetics, and Master Equation Analysis, The Journal of Physical Chemistry A, 126(44), 8249-8265, 2022.
3. Sharma Priyadarshini, M., Jaffe, R. L. and Panesi M., Carbon Clusters: Thermochemistry and Electronic Structure at High Temperatures, The Journal of Physical Chemistry A, 125(32), 7038-7051, 2021.
4. Venturi S., Sharma Priyadarshini, M., Lopez, B., and Panesi M., Data-Inspired and Physics-Driven Model Reduction for Dissociation: Application to the O₂ + O System, The Journal of Physical Chemistry A, 124(41), 8359-8372, 2020 (Featured as Journal Cover).
5. Sharma Priyadarshini, M., Liu, Y. and Panesi M., Coarse-Grained Modeling of Thermochemical Non-Equilibrium using the Multigroup Maximum Entropy Quadratic Formulation, Physical Review E, 101(1), 013307, 2020.

Teaching Interests

I am interested in teaching core chemical engineering subjects such as transport phenomena, thermodynamics and statistical mechanics, heat transfer, general chemistry, renewable energy technologies, and kinetic processes.

I also look forward to creating a new elective course on Data Science and Applied Machine Learning for Chemical and Biomolecular Engineering Applications. This elective course will have a theory component in which the basics of machine learning will be taught and a lab component that will focus on applying the techniques to real ChemBE problems. I will make the course accessible to students with different academic backgrounds. My aim is to highlight the interdisciplinary nature of these tools and enable students to use them in their future research and career endeavors.