Symposium SF03: Thermal Transport Under Extreme Conditions
Symposium SB07: Frontiers of Unconventional Polymer Networks

Symposium X—MRS/The Kavli Foundation Frontiers of Materials

Symposium X_Wednesday_800 wideKristin Persson, Lawrence Berkeley National Laboratory & University of California-Berkeley

The Era of Data-Driven Materials Innovation and Design

Written by Alison Hatt

At Wednesday’s Symposium X, Kristin Persson talked about the exciting era of data-driven materials innovation and design powered, in part, by the Materials Project, a research program that is revolutionizing materials research by leveraging the power of supercomputers and big data.

Creating a new material with specific properties has historically been slow, time consuming, resource intensive, and limited by human ingenuity. However, the past several decades have seen an explosion in computational materials capabilities, producing vast quantities of data and leading Persson to speculate, early in her career, about new ways of approaching materials design.

“Can we do accelerated learning on this simulated data?” asked Persson. “Can we correlate elastic tensors, dielectric tensors, and behaviors of materials, back to their crystal structure and their chemistry, and become smarter about where we go to look for new materials with specific properties?”

Symposium X_Wednesday_800 wide_3
Those speculations drove Persson to create the Materials Project, a program that combines high-throughput calculations and databases of calculated and measured materials properties to enable rapid screening and identification of promising candidates for various applications. This data-driven approach has the potential to accelerate the discovery of new materials with tailored properties, saving time and resources compared to traditional synthetic methods.

“Because we don’t have a hundred years to figure out some of our problems,” Persson said. “We have to be faster. We cannot do materials design as we have in the past.”

To illustrate the power of the Materials Project approach, Persson described a project to create a better photoanode for artificial photosynthesis, to meet precise criteria for the electronic structure and photochemical stability. Using the Materials Project tools, they identified a few thousand compounds that met some broader criteria, further filtered the list down to about 400 compounds, and then performed focused calculations to narrow those down to 34 compounds that met their more specific criteria. Working with John Gregoire at Caltech, the researchers rapidly synthesized the candidate compounds using inkjet printing, and identified 16 that met the stated criteria.

In a second example, Persson described efforts to find a lead-free piezoelectric material. Using a similar process, the team identified a potentially unstable phase of strontium hafnium oxide compound that met their criteria but presented substantial synthetic challenges. Collaborators at National Renewable Energy Laboratory eventually made the compound, but the entire process took two years, far slower than Persson envisions for the future of materials discovery.

Symposium X_Wednesday_800 wide_2

The two examples highlighted the impact of experimental bottlenecks on the accelerated materials design process. Persson shared a vision for improving the process through automated synthesis capabilities, wherein the details of large numbers of experiments, both successful ones and failures, are recorded and made available to the broader community. She went on to reveal that last month, Gerd Cedar’s group at Lawrence Berkeley National Laboratory took their new tool for automated synthesis online and synthesized 39 completely new compounds from Materials Project predictions in just three weeks.

However, Persson also acknowledged the challenges and limitations of data-driven materials research. She emphasized the need for continuous efforts to improve the accuracy and reliability of computational methods, as well as the availability and quality of materials data. She also addressed the glittering promise of machine learning, which seems poised to revolutionize the field, and cautioned that we must use ML judiciously and not attempt to extrapolate beyond the available, real data.

Persson closed by lauding the power of the Materials Project to democratize materials data, as its powerful tools and vast databases can be used, for free, by researchers in any part of the world.  Faster, more powerful, and more equitable: it’s a compelling vision for the future of materials research.   

Symposium X—MRS/The Kavli Foundation Frontiers of Materials features lectures aimed at a broad audience to provide meeting attendees with an overview of leading-edge topics.


The comments to this entry are closed.