Accelerating computational materials discovery with artificial intelligence and cloud high-performance computing: from large-scale screening to experimental validation (2401.04070v1)

Abstract: High-throughput computational materials discovery has promised significant acceleration of the design and discovery of new materials for many years. Despite a surge in interest and activity, the constraints imposed by large-scale computational resources present a significant bottleneck. Furthermore, examples of large-scale computational discovery carried through experimental validation remain scarce, especially for materials with product applicability. Here we demonstrate how this vision became reality by first combining state-of-the-art AI models and traditional physics-based models on cloud high-performance computing (HPC) resources to quickly navigate through more than 32 million candidates and predict around half a million potentially stable materials. By focusing on solid-state electrolytes for battery applications, our discovery pipeline further identified 18 promising candidates with new compositions and rediscovered a decade's worth of collective knowledge in the field as a byproduct. By employing around one thousand virtual machines (VMs) in the cloud, this process took less than 80 hours. We then synthesized and experimentally characterized the structures and conductivities of our top candidates, the Na$x$Li${3-x}$YCl$_6$ ($0 < x < 3$) series, demonstrating the potential of these compounds to serve as solid electrolytes. Additional candidate materials that are currently under experimental investigation could offer more examples of the computational discovery of new phases of Li- and Na-conducting solid electrolytes. We believe that this unprecedented approach of synergistically 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.

• The paper presents an AI and cloud HPC framework that screens over 32 million candidate materials, narrowing them down to 18 promising compositions.
• It integrates machine learning potentials with physics-based models, completing high-throughput screening in under 80 hours using 1,000 cloud virtual machines.
• Experimental validation confirmed Li and Na-based compounds, such as Na2LiYCl6, exhibiting viable ionic conductivity for advanced battery applications.

Review of "Accelerating Materials Discovery with AI and Cloud HPC"

This paper presents a pioneering approach to computational materials discovery, leveraging AI models alongside cloud high-performance computing (HPC) to enhance the efficiency of identifying novel materials. The paper addresses the pressing need for accelerated discovery of new materials, specifically targeting applications such as solid-state electrolytes in battery technology, where experimental progress often lags behind computational predictions.

Summary and Key Findings

The research encapsulates a multifaceted strategy combining AI with traditional physics-based models to expedite the materials discovery pipeline. Utilizing state-of-the-art ML potentials integrated with cloud HPC resources, the authors effectively sift through over 32 million candidate materials. This profound computational undertaking results in the prediction of approximately half a million candidates with potential stability, markedly restricting the focus to a smaller subset of 18 promising materials compositions. These exhibit properties conducive to use as solid-state electrolytes, thereby enhancing the landscape of battery materials research.

The robust computational framework is noteworthy, deploying an unprecedented volume of cloud-based virtual machines—around 1,000—completing the high-throughput screening in under 80 hours. The subsequent experimental synthesis and characterization lend empirical support to the computational predictions. In particular, the paper identifies a series of Li and Na-based compounds demonstrating viable ionic conductivity, such as Na$_2$LiYCl$_6$, which exhibits dual ion conduction, underscoring its potential utility in both lithium and sodium battery applications.

Implications and Future Directions

Practically, this research underscores the viability of cloud HPC in democratizing access to large-scale computational resources, reducing barriers to entry for researchers in the field of materials science. The successfully validated integration of AI models with cloud computing sets a precedent for future research in computational materials science, enabling rapid assessment and experimental validation of candidates with complex property requirements.

Theoretically, the paper opens new avenues for exploration in materials design and discovery, emphasizing the importance of carefully curated datasets and ML model training. The introduction of flexible, efficient ML models capable of evaluating diverse chemical spaces without the constraints of traditional methodologies is a significant advancement.

Future developments may see the refinement of ML models to incorporate more nuanced features, such as site disorder and defects, common in high-performance materials but underrepresented in the data used for this paper. Additionally, expanding the framework to directly generate materials with specific desired properties, building on recent generative model approaches like MatterGen, could further enhance the discovery process.

Conclusion

The paper demonstrates a compelling intersection of AI-powered methodologies with cloud-based computational capabilities, pushing the boundaries of materials discovery. By effectively reducing the cycle time from concept to experimental validation, this paper paves the way for more efficient exploration of the materials space, offering significant promise for advancements in energy storage technologies and beyond. Through ongoing development and expansion of computational strategies, the potential for uncovering novel, functionally transformative materials remains vast.