Production of graphene on an industrial scale: High-Performance Computing helps refine the process for improved efficiency

Researchers have used GCS HPC resources at the Technical University of Munich to create more efficient methods of producing graphene on an industrial scale.

Graphene is one of the most important scientific discoveries of the past century. Although graphene may be familiar, it is an allotrope carbon substance. This means that graphene has a similar structure to graphite. However, graphene opens up new possibilities for creating and building new technologies.

The graphene material is two-dimensional. Each “sheet” is one atom thick. However, its bonds give it the strength of some of the most complex metal alloys in the world while being lightweight and flexible. Scientists from many fields have been intrigued by graphene’s unique properties. They are researching using it for next-generation electronics and coatings on tools and industrial instruments.

Graphene’s enormous potential has led to one of its most significant challenges: graphene is hard to produce in large quantities, and the demand is constantly growing. Although recent research suggests that a liquid copper catalyst might be an efficient and fast way to make graphene, researchers need to learn more about the molecular interactions that occur during graphene formation. This means that they cannot use this method to produce graphene sheets consistently.

A team of researchers from the Technical University of Munich (TUM) has used the SuperMUC-NG and JUWELS high-performance computing systems (HPC) at the Julich Supercomputing Centres (JSC) and Leibniz Supercomputing Centres (LRZ) to simulate graphene formation on liquid Copper.

A window into the experiment

Graphene’s attractiveness stems primarily from its uniform crystal structure. Producing graphene with impurities would be a waste of effort. Researchers can use scotch tape to remove graphite layers from graphite crystals in laboratory situations or when a minimal amount is required. This technique removes pet hairs from clothes with tape or other adhesives. This method produces flawless graphene layers but is not practical for large-scale graphene production.

The industry requires high-quality graphene. One promising method is to use a liquid metal catalyst that allows the self-assembly of carbon atoms from molecular precursors into one graphene sheet. Although the liquid can scale up graphene production quickly, it introduces many complications, such as the high melting temperatures of Copper and other metals. Researchers use experiments to study how atoms interact in various conditions when designing new materials. Although technological advancements have made it possible to gain insight into the atomic-scale behavior of materials even in extreme situations like very high temperatures, researchers still use experiments to determine the exact changes that occur. Computer simulations can help, but simulating dynamic systems such as liquids has its complications.

Andersen stated that you must use molecular dynamics simulations (MD) to obtain the correct sampling to describe anything like this. “Then there’s the size of the system — you need to have enough system to simulate liquid behavior accurately.” Molecular dynamics simulations, unlike experiments, allow researchers to view events at the atomic level from many angles. Researchers can also pause the simulation to concentrate on specific aspects.

MD simulations can provide insights into individual atoms’ movements and chemical reactions that experiments cannot observe. However, there are some challenges. The compromise between accuracy, cost, and time is the most important. MD simulations require accurate ab initio models to drive them. It is costly to obtain simulations large enough to model these reactions accurately.

For the most recent simulations, Andersen and her coworkers used approximately 2,500 cores of JUWELS over more than a month. The team was able to simulate about 1,500 atoms in just a few picoseconds despite the enormous computational effort. While they may seem small, these simulations were the most extensive ab initio MD simulations ever done on graphene and liquid Copper. The team uses these simulations to develop more affordable methods to drive MD simulations. This allows for simulations of larger systems over longer timeframes and with greater accuracy.

Stabilizing the links in the chain

The record-breaking simulation work was published in the Journal of Chemical Physics. They then used the simulations to analyze experimental data from their latest paper, ACS Nano.

Andersen stated that the current-generation supercomputers, such as JUWELS or SuperMUC-NG, allowed the team to run their simulation. The next generation of machines would offer even greater possibilities as researchers can more quickly simulate larger systems or numbers over extended periods.

Andersen earned her Ph.D. in 2014. She indicated that graphene research had exploded over the same period. She said that it was fascinating that graphene is a new research focus. “People have looked at it closely in my scientific career,” Andersen stated that although there is still much to be done on liquid catalysts for graphene production, she believes the dual-pronged approach of HPC and experiment will be crucial to graphene’s further development and use in industrial and commercial applications. She said there was a lot of interplay between theory, investigation, and this research.