Research

Computational Discovery and Design of Functional Materials

Our research focuses on the discovery and design of novel materials for energy conversion and storage, next-generation micoelectronics and environmental sustainability. We identify new composition–structure–property relationships to enable materials by design. Historically, materials have been discovered by Edisonian trial-and-error approaches and even, serendipitous “happy accidents.” Designing materials with tailored properties is challenging because of the astronomical number of possible compounds, and structures. We use computational tools, including first-principles methods and machine learning to discover and design functional materials.

Materials Discovery via High-Throughput Searches

The concept of high-throughput (HT) computational search is simple yet powerful — compute the properties of a large number of materials using first-principles, and then interrogate this database to search for materials with desired properties. Clearly, the predictions should be validated by reality. The HT experimental approach was pioneered over a hundred years ago by Edison; with the advent of accurate computational methods and availability of inexpensive computing resources, its computational counterpart is now a viable path for materials discovery. We are using HT search to find novel materials for applications that require contraindicated combination of materials properties.  For example, most known materials with high electrical conductivity also exhibit high thermal conductivity; thermoelectrics require materials with high electrical but low thermal conductivity.

Materials Design Through Doping and Alloying

Alloying and doping are common ways to design materials with optimized functional and structural properties. From alloying copper with tin to improve mechanical strength of tools and weaponry in the Bronze Age to alloying iron to produce steel during the Industrial Revolution, alloys have become the cornerstone of human advacement. Similarly, doping enabled silicon microelectronics during the Digital Revolution. Like materials discovery (Theme 1), alloying and doping are often driven by chemical intuition and laborious experimental trial-and-error. First-principles computations can accelerate alloying and doping decisions by identifying alloy combinations, optimal compositions, and suitable dopants and their effect on the functional properties. We are using computational approaches to design materials for energy storage and conversion, and next-generation microelectronics. 

Materials Discovery and Design in Large Chemical Spaces

Computational design of materials has become viable with the expansion of methods and databases. However, it remains challenging in practice to explore large chemical spaces with ab initio methods alone. It is estimated that there are more than a trillion plausible valence-balanced compositions when considering the elements in the periodic table and up to quarternary stoichiometry. Even more challenging is “inverse design” i.e., building a stable material that possess a specific set of properties. We are using a combination of ab initio methods and machine learning to search for new functional materials in vast chemical spaces. 

Open Science – Databases, Codes, Public Datasets

Thermoelectrics Design Lab: An open-access database of DFT-calculated thermoelectric properties of ~2700 inorganic materials

TEDesignLab: A Virtual Laboratory for Thermoelectric Material DesignComputational Materials Science 112A, 368 (2016)

www.prashungorai.org/tedesignlab

 

Pylada-defects: An open-source Python package for automation of first-principles point defect calculations

A Computational Framework for Automation of Point Defect CalculationsComputational Materials Science 130, 1 (2017)

github.com/pylada/pylada-defects 

 

Anisotropy Atlas: A dataset of calculated lattice thermal conductivity tensors of ~2500 layered (quasi-2D) structures from the Inorganic Crystal Structure Database. The calculation methodology is documented in this paper

github.com/prashungorai/anisotropy-atlas