Challenge
Accelerating materials innovation will enable rapid deployment of advanced energy and industrial technologies that are essential for American competitiveness, from structural materials to materials for energy storage to other functional materials for advanced technologies. The identification and commercialization of new materials with transformative properties that dramatically improve performance, energy efficiency, reliability, and resilience, however, is a time- and resource-intensive process due to the inherent complexity of materials science and the practical limitations of traditional simulations, synthesis, and characterization techniques that still require significant trial and error.
AI Solution
The convergence of current and emerging AI technology with the growing availability of large, curated datasets may be a tipping point for materials discovery, design, and qualification. The development of physics-aware AI frameworks that exploit the complementary strengths of foundation models, deep learning, computer vision, generative AI, and agentic AI will enable entirely new capabilities for materials design that iteratively couple prediction, synthesis, characterization, and analysis to yield closed-loop learning systems that are interpretable, trustworthy, and capable of bridging large scales in space and time. The ultimate goal of inverse design (designing materials for given property specifications) requires advanced experimental and simulation capabilities as well as AI reasoning and explainability.
Justification
DOE’s suite of world leading and unique experimental and computational capabilities for materials research, including X-ray light sources, neutron scattering facilities (and their associated characterization equipment), nanoscale science research centers, materials databases, and exascale computers, is collectively the most comprehensive and performant in the world. These capabilities, along with the availability of very large materials data sets coupled with sustained investments in the development of AI-enabled physics-informed models, has positioned DOE to take a leadership role in implementing the materials by design vision.
National Impact
Tight integration of AI into the materials discovery-to-product workflow could significantly reduce time to market in manufacturing—from many years to decades down to months to a few years. This acceleration will dramatically reduce development timelines for critical technologies including batteries, energy systems, structural and functional materials, strengthening American technological leadership and enabling faster deployment of innovations that create jobs and strengthen economic and national security.
Aligned Actions from the National Laboratories
Accelerating Multiscale Materials Modeling with Machine Learning
National Challenge: Designing Materials with Predictable Functionality
Leading Lab: Sandia National Laboratories
Density Functional Theory is a gold standard for calculating the electronic-structure of materials, but its computational demands limit the size of practical simulations to only a few hundred atoms. Sandia National Laboratories and the Center for Advanced Systems Understanding (CASUS) developed Materials Learning Algorithms, or MALA, a machine-learning software framework that predicts materials’ electronic structure with chemical accuracy while scaling computationally with system size. By combining machine learning with physics-based calculations, MALA enables accurate electronic-structure modeling at length and time scales that were previously impractical. Researchers have demonstrated calculations involving more than 100,000 atoms and achieved more than a 5,000-fold speedup over Density Functional Theory for 2,000-atom systems. These capabilities open the door to faster, larger-scale simulations that can help researchers design and understand materials with predictable functionality.
MALA was recognized with a 2023 R&D 100 Award. Its award-winning models, data sets, and workflows will be included in Genesis Mission’s American Science and Security Platform, making this established capability available to future platform users.