Researchers from Azure Quantum Microsoft, Physical and Computational Sciences Directorate Pacific Northwest National Laboratory, and Microsoft Surface have successfully integrated artificial intelligence (AI) models and cloud high-performance computing (HPC) to expedite the discovery of new materials. The team used this approach to sift through over 32 million candidates and predict approximately half a million potentially stable materials in under 80 hours. The research focused on solid-state electrolytes for battery applications, identifying 18 promising candidates with new compositions. This method could revolutionize materials discovery, reducing the time from concept to solution.
Accelerating Computational Materials Discovery with AI and Cloud Computing
A team of researchers from Azure Quantum Microsoft, Physical and Computational Sciences Directorate Pacific Northwest National Laboratory, and Microsoft Surface have demonstrated the potential of integrating artificial intelligence (AI) models and cloud high-performance computing (HPC) in accelerating the discovery of new materials. The team used these technologies to navigate through more than 32 million candidates and predict around half a million potentially stable materials in less than 80 hours.
High-Throughput Computational Materials Discovery
High-throughput computational materials discovery has been a promising field for many years. However, the constraints imposed by large-scale computational resources present a significant bottleneck. The team addressed this issue by combining state-of-the-art AI models and traditional physics-based models on cloud HPC resources. This approach allowed them to quickly navigate through a vast number of candidates and predict potentially stable materials.
Focus on Solid-State Electrolytes for Battery Applications
The team focused their research on solid-state electrolytes for battery applications. Their discovery pipeline identified 18 promising candidates with new compositions. They also rediscovered a decade’s worth of collective knowledge in the field as a byproduct. The team synthesized and experimentally characterized the structures and conductivities of their top candidates, demonstrating the potential of these compounds to serve as solid electrolytes.
Utilizing Cloud High-Performance Computing
The team employed around one thousand virtual machines (VMs) in the cloud for this process. Cloud HPC has recently become widely known due to its ability to train and host large-scale AI models. In the realm of computational materials discovery, these advancements imply that cloud computing can effectively manage smaller-scale complex calculation jobs, greatly increasing the number of material candidates that can be evaluated computationally.
The Role of Artificial Intelligence in Materials Discovery
AI models for materials science have the potential to vastly expedite the computational discovery process. State-of-the-art AI models can predict the results of physics-based quantum mechanical calculations but are several orders of magnitude faster, making them ideal for predicting general material properties. The significant speed advantage of AI-based techniques over direct simulation has made it possible to explore materials across a vast chemical space that greatly exceeds the number of known materials.
Experimental Validation of New Material Compositions
The team synthesized and experimentally characterized the structures and conductivities of their top candidates. The outcomes of this process are noteworthy for their rapid identification of promising material candidates, reflecting insights that have emerged over the last decade in this field. This efficient screening and discovery process significantly reduces the overall time from concept to solution in materials discovery.
The Future of Materials Discovery
The team’s approach of integrating AI models and cloud HPC not only accelerates materials discovery but also showcases the potency of AI-guided experimentation in unlocking transformative scientific breakthroughs with real-world applications. This unprecedented approach could offer more examples of the computational discovery of new phases of Li and Na-conducting solid electrolytes, marking a new age of materials discovery where computational approaches take the lead in the prediction-synthesis-characterization cycle.
The article titled “Accelerating computational materials discovery with artificial intelligence and cloud high-performance computing: from large-scale screening to experimental validation” was published on January 8, 2024, by authors Chi Chen, Dan Thien Nguyen, Shannon Lee, Nathan A. Baker, Ajay Karakoti, Linda Lauw, C.D. Owen, Karl T. Mueller, Brian A. Bilodeau, Vijayakumar Murugesan, and Matthias Troyer. The article, which was published on arXiv (Cornell University), discusses the use of artificial intelligence and high-performance computing in the field of computational materials discovery. The DOI reference for the article is https://doi.org/10.48550/arxiv.2401.04070.
