Researcher Uses Artificial Intelligence to Discover New Materials for Advanced Computing

"Researcher Uses Artificial Intelligence to Discover New Materials for Advanced Computing"

—Undefined Trio

In a significant breakthrough for materials science, a team of researchers led by Trevor David Rhone, an assistant professor at Rensselaer Polytechnic Institute, has leveraged the power of artificial intelligence (AI) to identify novel van der Waals (vdW) magnets. Through the application of cutting-edge AI tools and techniques, the team has discovered transition metal halide vdW materials with large magnetic moments that exhibit promising stability, opening up exciting possibilities for data storage, spintronics, and quantum computing. The team's research, featured on the cover of Advanced Theory and Simulations, showcases the remarkable potential of materials informatics in advancing scientific discovery.

The emergence of two-dimensional (2D) materials has captivated researchers since their discovery in 2004. These materials, which can be as thin as a single atom, possess intriguing and unexpected properties. Of particular interest are 2D magnets, which retain their long-range magnetic ordering even when reduced to one or a few layers. This unique characteristic is attributed to magnetic anisotropy and presents opportunities for exploring exotic spin degrees of freedom, such as spin textures, that could revolutionize the development of quantum computing architectures. Furthermore, 2D magnets span a wide range of electronic properties and can be harnessed in high-performance, energy-efficient devices.

Rhone and his team employed a combination of high-throughput density functional theory (DFT) calculations and AI techniques, specifically semi-supervised learning, to uncover the properties of vdW materials. By leveraging labeled and unlabeled data, semi-supervised  learning enabled the identification of patterns and predictions. This approach addresses a common challenge in machine learning—the scarcity of labeled data—and paves the way for accelerated materials discovery processes.

"The use of AI not only saves time but also reduces costs," explains Rhone. "Traditionally,  materials discovery involved extensive simulations on supercomputers, which could take months. Laboratory experiments, on the other hand, could take even longer and incur higher expenses. Adopting an AI approach has the potential to significantly expedite the materials discovery process."

Using an initial dataset of 700 DFT calculations conducted on a supercomputer, the team trained an AI model capable of rapidly predicting the properties of thousands of materials candidates in mere milliseconds on a standard laptop. Through this approach, the researchers successfully identified promising vdW materials with large magnetic moments and low formation energy—a key indicator of chemical stability necessary for material synthesis and subsequent industrial applications.

Rhone emphasizes the versatility of their framework, stating, "Our approach can be easily extended to explore materials with different crystal structures. Mixed crystal structure prototypes, such as a dataset encompassing both transition metal halides and transition metal trichalcogenides, can also be examined using this framework."

Curt Breneman, the dean of Rensselaer's School of Science, applauds Rhone's application of AI in materials science, which continues to yield exciting results. "Not only has he advanced our understanding of novel properties in 2D materials, but his findings and methods are poised to contribute to the development of new quantum computing technologies," says Breneman.

This groundbreaking research showcases the immense potential of AI and materials informatics in driving scientific progress. By accelerating the discovery of new materials with tailored properties, researchers like Rhone are paving the way for technological advancements in fields ranging from computing and data storage to quantum technologies.

Reference:

Malatino, K. (2023). Researcher uses artif