(4fb) Data Driven Discovery of Novel Functional Materials and Process Understanding

Current global challenges in health and environment require the rapid design and deployment of novel and tunable materials for a variety of purposes. These include structural applications, energy storage, catalysis, and peptide and drug design. In addition to the domain-specific challenges, certain issues arise when conceptualizing new materials regardless of their function such as the inherent sparsity of data resulting from the large combinatorial space of composition and structure. Furthermore, not only is the discovery of new materials a significant challenge, but the development of a suitable manufacturing process is an essential step towards the translation of lab-scale technology to a usable form. My future research program will address materials discovery and process optimization by applying data driven techniques towards the discovery and development of novel material and process systems and optimization of reaction conditions for the development and manufacture of materials and biochemical molecules (both small molecules and biologics).

The rational design of specialized materials and processes requires a synergy between ab initio methods, experimental data, and learning algorithms. My research aims to generate and screen potential chemistries with a particular focus on biological, structural, energy storage and catalytic materials and molecules. By leveraging machine learning techniques and the vast amounts of experimental and computational data that exists today–due to improvements in measurement systems and computational capabilities–we will elucidate the underlying relationships that exist within the data. I thus hope to uncover correlations between chemical structure and function of multicomponent materials. This can improve our understanding of the mechanisms of functional materials such as the relationship between structure and stability in perovskite cells.

My research thrusts will include:

1. Development of computational pipelines for the rapid generation and exploration of novel material candidates

2. Mapping structure-property and structure-activity relationships using machine learning techniques

3. Data representation that encode multiple facets of multicomponent materials including the crystallographic and electronic structure for use as descriptors within our computational pipelines, and

4. Optimization of manufacturing processes using data driven approaches

Research Background

My graduate studies with Prof. Krishna Rajan at the University at Buffalo focused on the development of an informatics approach to suggest new potential chemistries for multicomponent materials using statistical analysis and high dimensional representations of multicomponent materials. I used the test case of bulk metallic glasses to highlight the key issues in the field of high throughput predictions and propose a new probabilistic analysis of rules for glass forming ability using rough set theory. I applied this approach to a broad range of binary alloy compositions in order to predict new metallic glass compositions. Our data driven approach considers not only a broad variety of thermodynamic, structural and kinetic based criteria, but also incorporates qualitative and descriptive attributes associated with eutectic points in phase diagrams ^1,2. We also demonstrated our methods on single atom alloy systems, wherein we used outlier detection techniques to identify candidates that possess the potential to overcome linear scaling laws, that signify a theoretical limit to catalytic activity. This is derived from the surface electronic effects that govern these reactions. Our techniques were thus able to generate a large space of compounds and screen for targeted properties rapidly ³. We have also shown the application of machine learning approaches, such as convolutional neural networks, to quantitatively find and extract characteristics in the material fingerprints generated using Hirshfeld surfaces to develop rapid classifications across multiple material classes, chemistries and properties.

My research focus during my postdoc with Prof. Connor Coley at the Massachusetts Institute of Technology is the optimization of biomanufacturing processes using data driven techniques. I have been developing a variety of machine learning models to predict cell culture performance and product quality attributes for biologic drugs such as monoclonal antibodies and enzymes. The goal of this work is to shorten process development times for biopharmaceutical manufacturing by facilitating the selection of optimal process parameters.

Teaching Interests

My background with a Bachelors and Masters in Chemical Engineering and a PhD in Materials Design and Innovation prepares me to teach introductory and advanced chemical engineering and materials engineering courses that are considered part of the core curriculum within these engineering programs. In addition, I am also interested in developing curricula for and teaching subjects dealing with Data Science, Statistical Analysis, Machine Learning and applications thereof including interdisciplinary applications within the domains of Materials Science and Chemical Engineering. I am particularly interested in the development of a chemical engineering curriculum that is based on the pedagogical principles of active learning and backward design. Currently, I am also a participant of the 2021 Kaufman Teaching Certificate Program (KTCP) cohort. The KTCP certificate program has helped me to develop and formalize my teaching skills based on evidence-based pedagogical techniques of learning.

Publications List:

Aparajita Dasgupta, Scott R. Broderick, Connor Mack, Bhargava U. Kota, Ramachandran Subramanian, Srirangaraj Setlur, Venu Govindaraju, and Krishna Rajan. Probabilistic Assessment of Glass Forming Ability Rules for Metallic Glasses Aided by Automated Analysis of Phase Diagrams. Scientific Reports, 9(1):357, January 2019. Number: 1 Publisher: Nature Publishing Group.

2. Bhargava Urala Kota, Rathin Radhakrishnan Nair, Srirangaraj Setlur, Aparajita Dasgupta, Scott Broderick, Venu Govindaraju, and Krishna Rajan. Automated Analysis of Phase Diagrams. In2017 14th IAPR International Conference on Document Analysis and Recognition (ICDAR), pages17–18, Kyoto, Japan, November 2017. IEEE.

3. Aparajita Dasgupta, Yingjie Gao, Scott R. Broderick, E. Bruce Pitman, and Krishna Rajan. Machine Learning-Aided Identification of Single Atom Alloy Catalysts. The Journal of Physical Chemistry C, 124(26):14158–14166, July 2020. Publisher: American Chemical Society.