NERSCPowering Scientific Discovery Since 1974

NERSC Initiative for Scientific Exploration (NISE) 2011 Awards

Accelerated Materials Design Towards the Materials Genome

Kristin Persson, Lawrence Berkeley National Laboratory

Associated NERSC Project: The Materials Genome (matgen)

NISE Award: 2,500,000 Hours
Award Date: March 2011

One of the foremost reasons for the long process time in materials discovery is the lack of comprehensive knowledge about materials, organized for easy analysis and rational design. The goal of the proposed work is to accelerate the way materials discovery is done by creating a high-throughput computing environment together with a searchable, interactive database of computed materials properties. The idea is driven by our urgent need fornew materials to realize an economy based on renewable energies. Today, it takes 18 years on average from materials conception to commercialization. Unless we implement new approaches to the materials discovery and development process, we cannot expect to have a significant impact on our energy economy in the 20-30 year time frame needed to address the climate issues.

Materials innovation today is largely done by intuition and based on the experience of single investigators. Google has demonstrated the value of organizing data and making the data available and searchable to large communities. We propose to leverage the information age for materials using the only tool that can efficiently scan multiple properties for all known materials in a reasonable amount of time: computations. First-principles calculations have reached the point of accuracy where many materials properties, relevant for photovoltaics, batteries, and thermoelectrics can be reliably predicted. Most importantly, computational approaches are by the nature of their scalability and objectivity the most efficient and consistent way of scanning Nature in search for optimal material solutions. Once computed, we will store and organize the data so that it may be readily accessed for data mining and statistical learning algorithms. This extensive knowledge of materials’ properties, linked to their structure and chemistry, will enable us to invert the design problem and specify the needed arrangement of atoms and how it can be made to target desired properties and constraints.