Hawk Helps Researchers Identify a New Class of 2D Materials

Keyvisual image main
Semitransparent structure of germanium manganese oxide (GeMnO3), one of the newly discovered 2D materials, superimposed on its magnetic surface structure. The color pattern highlights regions in the vicinity of magnetic ions where the magnetization points out of the plane (red) or into the plane (blue). This strong spatial variation of the magnetic information could play a crucial role for data storage applications, for instance. Image: Rico Friedrich, HZDR.

Scientists at the Helmholtz Zentrum Dresden-Rossendorf are using HPC to find graphene-like materials for novel electronic and catalysis applications.

In 2004, researchers from the University of Manchester demonstrated that a single atomic layer of carbon atoms arranged in a particular hexagonal pattern could be peeled from graphite into ultra-thin layers to form graphene.

This two-dimensional (2D) nanomaterial has been of great scientific interest because it exhibits distinctive stability, flexibility, transparency, and conductivity, properties that make it a good candidate for creating new electronic devices and other advanced materials.

Since that time, materials scientists have not only focused on graphene, but have also been seeking other 2D materials that can complement graphene's desirable traits under different temperature, pressure, or environmental conditions. The focus has been on 2D systems, which are the structural units of the existing bulk, but layered, materials, where the layers are held together by weak forces, with the best example being graphene/graphite. A fundamental question has been raised: do 2D materials, which do not have layered bulk analogues, actually exist? The first experiments indicated that this is indeed possible.

To that end, researchers led by Dr. Rico Friedrich and Dr. Arkady Krasheninnikov at the Helmholtz Zentrum Dresden-Rossendorf (HZDR) have been using high-performance computing (HPC) resources at the High-Performance Computing Center Stuttgart (HLRS) to screen candidates for new 2D materials.

“2D materials are very interesting from a fundamental point of view for applications in things such as future information technology and nanoscale interface systems,” said Dr. Rico Friedrich, HZDR researcher and head of the project. “These so-called non-van der Waals 2D systems are a fascinating new research direction with very few representatives yet. There were no solid, data-driven, theoretical calculations done for seeking out new candidates, and that is why we wanted to focus on finding materials that exhibit these properties.”  

With the help of HLRS’s Hawk supercomputer, the team recently found promising new candidates for such 2D materials. It published its results in Advanced Electronic Materials.

Seek and you shall find

Graphene ushered in a new era of advanced material design, but it also comes with limitations. It is considered a “van der Waals” material, meaning that carbon atoms are aligned into 1-atom-thick layers by forming weak, non-chemical bonds due to the so-called van der Waals force between the layers. The bond is so weak, in fact, that a graphene layer can be removed from the graphite used in pencils simply by carefully peeling it away with scotch tape. While simple to extract at a small scale, it is brittle at a larger scale, making manufacturing challenging.  

It is the atomic-level uniformity and organization that both gives graphene its distinctive properties and presents challenges for bulk manufacturing processes. To develop complementary 2D materials that would be suitable for different environments, researchers must identify compounds whose atomic structures are not as atomically uniform but can nonetheless deliver similar characteristics in two dimensions.

That task can prove daunting. Friedrich noted that there are over 3.5 million entries in the AFLOW materials database, one of the largest materials catalogs in the world. To find viable alternatives to graphene, the researchers search through all these materials to find correlations and patterns in their properties that could make them candidates for creating “non-van der Waals” 2D materials. Once potential candidates are identified, the researchers must then computationally model those candidates’ behaviors under different settings to determine if they exhibit beneficial traits in situations where they would be needed. Some materials behave quite similarly to graphene when they are still attached to a substrate but break apart once removed. Other candidates exhibit desirable properties in limited temperature or pressure ranges. Using high-performance computers like HLRS’s Hawk system, researchers like Friedrich can efficiently search through millions of candidates and quickly screen individual materials’ properties.

“You need to investigate a huge space to figure out which materials exhibit these properties in 2D,” Friedrich said. “You can only do that with high-throughput calculations, because after you identify candidates, you still must study their energetic and dynamic properties, which influences how easily 2D sheets can be extracted from a bulk material, to make sure that they will remain stable under the ambient temperature. Hawk is really well-suited to this task because it is a very efficient, stable machine that helps get our results for individual candidates fast — it is almost a real-time, interactive process.”

In 2021 and 2022, the team searched the AFLOW materials database to identify promising oxide materials. Using Hawk, it found 28 possible candidates for further investigation.

In work published in early 2023, the team expanded its search beyond the realm of oxides into other material spaces like sulfides and chlorides in pursuit of candidates that exhibit an ultra-low “exfoliation energy”, meaning that they could be most easily separated from bulk materials into 2D sheets. The research also involved rigorously verifying materials’ stability in different temperature conditions. This approach is already paying dividends, as the researchers found two chemical compounds — SbTlO3 and MnNaCl—that have extremely low exfoliation energy. Both materials are revealing as next-generation electronic materials, as their properties show promise for information technology, data storage, advanced functional structures, and magnetic applications.

Friedrich noted that this approach is only beginning, and HPC will remain an essential part of the team’s search for new 2D materials. “HPC access is essential to screen the candidates and to be able to study their properties. We want to extensively generalize the search for such systems in the future, which will likely yield thousands of candidates that can only be studied with HPC,” he said.

Cutting-edge research requires high-end expertise

With its method in place, the team plans to broaden its search. Friedrich was recently named as a Dresden-concept Junior Research Group Leader, and he will use the next 5 years to completely catalog all possible new 2D material candidates, which he anticipates could be thousands of compounds. He will also extensively study their chemical functionality and molecular interfaces to explore how these materials could be used for advanced electronic applications.

Once the team has a sufficient catalog of candidates, he anticipates working closely with HPC experts to train dedicated machine learning models that could help the researchers quickly find similar traits among disparate classes of materials, accelerating their search for relevant 2D compounds. Friedrich noted that developing these complex computational workflows is only possible for scientists like himself when working in concert with HPC centers.

 “There was a point when it seemed that every group in computational chemistry and physics had to administer its own cluster, so today it is a big advantage to work with HPC centers that have a huge degree of expertise and provide access to workshops and tutorials,” he said. “One of my former supervisors once explained that HPC is like driving a Formula 1 car; setting a Formula 1 car in motion is not that hard, but becomes very difficult at its maximum efficiency. It’s the same with computational research. As scientists we have ideas about materials, but to investigate them at the cutting edge of science we need the expertise at HPC centers and integration of hardware infrastructure is crucial.”

—Eric Gedenk

This article was published in the Spring 2023 issue of InSiDE Magazine.

Related publication

Baronwsky T, Krasheninnikov A, Friedrich R. 2023. A new group of 2D non-van der Waals materials with ultra low exfoliation energies. Adv Electron Materials. 202201112.

Funding for Hawk was provided by Baden-Württemberg Ministry for Science, Research, and the Arts and the German Federal Ministry of Education and Research through the Gauss Centre for Supercomputing (GCS).