The Catalyst Review July 2020 - 18

MOVERS & SHAKERS
Paulami Majumdar, PhD

Senior Research Specialist, Core R&D, Dow (Michigan)
Paulami received her bachelor's degree in Chemical Engineering at Jadavpur University in 2010 and her master's
degree at the Indian Institute of Technology (Bombay) in 2013, before going on to earn her PhD at Purdue University
in 2018-where she was supervised by Prof. Jeffrey Greeley. During her graduate studies, Majumdar applied quantum mechanical molecular modeling to transition metal and metal oxides as heterogeneous catalysts. Her research
focused on understanding coverage effects on scaling relationships and elucidating reactivity trends at metal-oxide
interfaces using the water-gas-shift as a prototype reaction. Her work provided fundamental insights into the role of
co-adsorbed species on the adsorption properties of a catalytic surface. In addition, her work also highlighted that
electrostatic interactions at the interface between a metal nanoparticle and oxide support might be crucial for catalytic properties
of the system. She joined Dow in 2018, where her research continues towards developing sustainable catalysts using fundamental
insights. Dr. Majumdar can be reached at PMajumdar@dow.com.

The Catalyst Review asked Dr. Majumdar to share her thoughts on the current opportunities and
challenges associated with the use of computational methods for catalyst development.
Heterogeneous catalysis is of paramount importance to chemical and energy industries for efficient and sustainable chemical transformations. However, the development of novel heterogeneous catalysts poses significant challenges because their
performance depends on reaction conditions (feed composition, temperature, pressure) as well as their atomic-level properties (surface termination, active site structure, coordination environment, particle size). These parameters, in turn, can change
under reaction conditions through processes such as surface reconstruction, oxidation, and particle sintering. The rational
design of catalysts requires a thorough understanding of reaction mechanisms and structure-activity relationships. In this regard, ab-initio modeling using quantum mechanical approaches such as Density Functional Theory (DFT) has rapidly emerged
as a promising tool to accelerate catalyst development. In recent years, a synergistic combination of experiments and quantum
modeling has become increasingly popular in catalysis research.
Despite significant developments in supercomputing capabilities, the computational cost of ab-initio methods is still prohibitively high for investigating a structurally complex heterogeneous catalyst under reaction conditions. Instead, these techniques
are primarily used to study simplified versions of active site structures at 0 K and under vacuum. Over the last decades, advancements in experimental in-situ and operando techniques for studying heterogeneous catalytic reactions have highlighted
the dynamic nature of a catalyst surface, which has spurred research towards developing multiscale modeling approaches.
Microkinetic modeling of reaction kinetics, fast ab-initio molecular dynamics, and machine learning are some of the promising
strategies that can potentially enable modeling of heterogeneous catalysts under reaction conditions.
Integrating machine learning methods with quantum mechanical methods promises to facilitate the rapid identification of catalysts in a wide composition and reaction space. The general idea is to use quantum methods on simple systems to generate
large data sets for training machine learning algorithms, which then can be used for more complex systems. Such approaches
can greatly accelerate obtaining structural information about catalysts as well as identifying transition states and mapping the
potential energy surface for a reaction. Over the years, the scientific community has generated a vast amount of experimental
and computational data on catalyst synthesis, characterization, and performance. Storing and organizing such data in a searchable format is key to the training and development of fast predictive machine learning algorithms. Machine learning approaches for fast catalyst design, therefore, is expected to work in tandem with the advancement in data or text mining, using natural
language processors. Machine learning combined with quantum modeling for heterogeneous catalyst discovery is a relatively
nascent field with much room for improvement, but the rapid progress currently being made in this area promises to accelerate the rational design of heterogeneous catalysts in future years.

The Catalyst Review - Subscribe Today!

http://www.catalystgrp.com/tcg-resources/studies-and-publications/the-catalyst-review/

The Catalyst Review

July 2020

http://www.catalystgrp.com/tcg-resources/studies-and-publications/the-catalyst-review/

The Catalyst Review July 2020

Table of Contents for the Digital Edition of The Catalyst Review July 2020