Materials Informatics: Infrastructure and Methods

The recent decade has been a "golden age" for development of novel methods and tools in materials informatics. Among many other advances, the web-based and distributed databases with various properties of molecules and compounds of interest for chemical and biological areas of research have been expanding at a rapid pace. Alongside these advances, design of novel and optimization of the existing computational methods, based on computational high-throughput screening and machine learning, have been on the rise. These advances have opened multitude of novel venues in the research areas encompassing computational materials design and optimization within the paradigms of computational large-scale high-throughput screening and Big Data-based machine-learning.

Our group embarks upon the following themes of research, with the particular focus put on the computational design and optimization of liquid electrolytes for energy storage devices.

Electrolyte Genome Project

The Electrolyte Genome Project (EGP) was conceived to provide an essential infrastructure for computational design of novel materials and materials optimization for battery applications. It is envisioned that perspective EGP will employ a combination of density functional theory and classical molecular dynamics simulations to the end of assessing suitability of candidate electrolyte molecules and/or molecular compounds as potential building blocks for battery components (liquid electrolyte, anode, cathode). As a part of the project, our aim is to contribute to development of computational modules for integration of classical molecular dynamics into automated screening process. The short- and long-time objectives are the following:

  • Developing computational modules for integrating classical molecular dynamics into automated high-throughput screening framework.
  • Development of databases for storing information on properties of molecules and molecular compounds relevant for battery applications.
  • Design and development of computational modules for large-scale data processing.

Machine Learning-driven Electrolyte Discovery and Optimization

To facilitate discovery and optimization of novel constituent materials and modules for energy storage, we aim to develop machine learning-based models and concomitant databases. This enterprise encompasses a diverse array of projects involving applications of machine learning methods and big data. Among the priority venues are the following:

  • Big data analytics for establishing structure-property relationships in materials for battery applications.
  • Applications of machine learning methods for large-scale quantum and classical molecular dynamics simulations of molecular compounds.
  • Deep learning methods for design of molecules with desired properties.

Hybrid Energy Storage Devices
The development of powerful mobile devices, electrically driven cars, and renewable intermittent energy production fuels the ever-increasing demand for higher power and energy density in electrochemical energy storage devices. One proposed way to achieve this goal is the creation of a device which uses materials from batteries and supercapacitors, called a hybrid energy storage device. This project focuses on the computational screening of ionic liquid electrolyte for multivalent hybrid ion capacitors with the goal of being able to discover promising candidates for use in such a device. In addition to running bulk calculations to elucidate the solvent structure and calculate properties such as electrochemical window, viscosity, and diffusivity, we are considering electrolyte confined in different nanoporous carbon electrodes.

Design of Optimal Electrolytes and Electrode-Electrolyte Interfaces for Next Generation Lithium-Sulfur Batteries
Lithium-Sulfur (Li-S) batteries are one of the most promising candidates for next-generation batteries due to their higher theoretical energy density and lower cost compared to state-of-the-art Li-ion batteries. However, commercialization of Li-S batteries is hindered by their poor cycling performance and capacity retention caused by the formation and dissolution of lithium polysulfides (LiPSs) during discharging/charging in the electrolyte, the lifeblood of the battery. More specifically, continuous diffusion of LiPSs towards Li metal anode and their nucleation and sluggish kinetics on the host carbon materials at the cathode result in loss of the active material. Hence, tailoring the atomistic interactions between LiPSs, salt anion, and solvent, in addition to the functional properties of porous carbon materials is critical in controlling deleterious side reactions, designing optimal electrolytes, and improving the performance and longevity of Li-S batteries.

Predicting and understanding the physicochemical behavior of bulk electrolytes and electrode-electrolyte interfaces is achieved via a multi-scale-data-driven approach that combines density functional theory (DFT) and molecular dynamics (MD) simulations. The project is also done in close contact with experimentalists from PNNL, resulting in a better interpretation of experimental results and an enhanced predictive power of the next iteration of simulations. The result is a database consisting of a large number of materials with optimized parameters to be used in electrolytes that achieve all the performance metrics for a given battery application.