The Materials Virtual Lab (MAVRL) is an interdisciplinary group of scientists who aims to bring forth a data-driven future for materials science. The figure below shows the five pillars of the MAVRL.
- Science: We integrate first principles calculations, i.e., calculations based on solving the Schrödinger equation, with phenomenological materials theory to predict materials properties.
- Data: We build sophisticated software frameworks to automate calculations at unprecedented scales and develop intuitive data user interfaces for exploring large materials data sets.
- Learning: We develop state-of-the-art artificial intelligence (AI) and machine learning techniques (ML) to extract chemical insights and accelerate materials predictions and simulations.
- Application: We focus on technologically-relevant materials problems that can have a major impact on societal well-being.
- Community: We believe that open data, open APIs, and open-source software are key to democratizing and enhancing reproducibility in materials research.
Creating Open Materials Data and Software
Electronic structure calculation codes have reached a level of maturity that it is now possible to reliably automate and scale first principles calculations across any number of compounds.
In MAVRL, we develop the infrastructure to facilitate the creation and analytics of rich materials datasets. This infrastructure involves the development of flexible materials data representations for state-of-the-art database technologies, coupled with robust platforms for high-throughput computing. We founded the Python Materials Genomics (pymatgen) materials analysis package.
We are a key partner of the Materials Project, an open science initiative to make the data for all known inorganic compounds publicly available to all materials researchers to accelerate materials innovation. A recent web application developed by MAVRL is Crystalium, a database of the computed surface energies and Wulff shapes of all elemental crystals.
AI and Machine Learning
Machine learning (ML) is the branch of artificial intelligence that deals with the development of algorithms and models that can automatically learn patterns from data and perform tasks without explicit instructions. In recent years, ML has come to the forefront as a means to mimic, or even surpass, human performance in a variety of tasks, from playing board games such as chess and Go to medical imaging to autonomous driving.
In the MAVRL, we develop state-of-the-art ML models to address three major bottlenecks in materials science:
- Discover. We develop highly accurate ML models that can vastly improve our ability to predict novel materials with superior properties. An example is the MatErials Graph Network (MEGNet), a graph-based deep learning model (see schematic above) that has been shown to achieve universally high prediction accuracies in a broad range of properties in both molecules and crystals.
- Scale. We develop ML interatomic potentials that enable us to simulate complex materials, e.g., multiple principal element, aka high-entropy, alloys, at large length and time scales.
- Accelerate. We develop ML approaches to radically accelerate the interpretation of experimental characterization data, such as X-ray absorption spectra.
Effective materials design often involves trade-offs between multiple properties. In the MAVRL, we collaborate with a vast network of experimental scientists to develop novel materials for a variety of applications. Our focus is on materials driven technologies that can have a major impact on societal well-being. Here are some areas we are currently working on.
Rechargeable alkali-ion batteries
Since its development in the 1970s, the rechargeable alkali-ion battery has proven to be a truly transformative technology, providing portable energy storage for devices ranging from small portable electronics to sizable electric vehicles. Despite its great success, there still remains many areas for improvement. To truly displace gasoline combustion engines, we need batteries that have energy densities far higher than those available today. We also need batteries that are safer and less prone to bursting into flames.
To increase energy density, one potential solution is to develop multi-electron cathodes that can reversibly cycle more than one electrode per transition metal. MAVRL is a proud member of the NorthEast Center for Chemical Energy Storage (NECCESS), an Energy Frontier Research Center funded by the US Department of Energy. Through NECCESS, we collaborate with the Whittingham and other groups to develop new multi-electron cathode materials.
To improve safety as well as to enhance energy density, one potential solution is to replace the flammable organic solvent electrolyte used in today’s lithium-ion batteries with a non-flammable solid electrolyte. In the MAVRL, we view all-solid-state batteries as ultimately a multi-component optimization problem. We use first principles methods to study and design superionic conductors that are not only fast conducting, but also electrochemically stable and mechanically compatible with the electrodes. A holy grail is to enable the use of lithium metal, which is the highest energy density anode possible for lithium-ion battery technology.
Lighting is one of the largest consumers of energy today. An estimated 20% of total US electricity output is consumed in lighting our homes, offices and industries. Highly efficient light emitting diodes (LEDs) are one of the most promising technologies for next generation lighting. However, current LEDs still suffer from a fundamental tradeoff between luminous efficacy and color rendering.
In MAVRL, we use first principles calculations and ML to understand and develop novel phosphor materials for LEDs. We seek to understand how the crystal and electronic structure affect the emission properties of phosphors, as well as their response to temperature and composition. In doing so, we aim to develop novel phosphor compositions that can significantly outperform today’s LED phosphors. A particularly illustrative example is our prediction of Sr2LiAlO4, the first known quaternary compound in the Sr-Li-Al-O chemical space, as a novel phosphor. Synthesized by the McKittrick group, Sr2LiAlO4 was confirmed to be indeed an effective phosphor for high color-rendering index LEDs.
Structural alloys for extreme environments
Multi-principal element alloys (MPEAs), colloquially also known as “high entropy” alloys, are alloys comprising four or more elements, usually in nearly equiatomic concentrations. They have drawn rapidly growing interest due to their exceptional mechanical properties under extreme conditions. In the MAVRL, we develop ML interatomic potentials to investigate the complex relationships between segregation, short range order and the mechanical properties of these alloys.
The Materials Virtual Lab gratefully acknowledges the generous funding provided by the US Department of Energy, the National Science Foundation, and the Office of Naval Research.
You must be logged in to post a comment.