RESEARCH HIGHLIGHTS

Our research highlights serve as a collection of feature articles detailing recent scientific achievements on GCS HPC resources. 

Research Snapshot: Scientists Expand Study of Ultrafast Electron Dynamics with Help of SuperMUC-NG
Research Snapshot –

Researchers at the University of Regensburg are pushing the study of ultrafast electron dynamics to new levels of precision. Thanks to their multi-year effort to improve upon algorithms based on quantum mechanics, the team succeeded in running significantly more accurate simulations of electron orbits across 2D materials.

Battery-powered electronic devices are ubiquitous in today’s world. Everything from smart phones and watches to cars and even internal medical devices rely on compact, efficient power sources to function. But despite this rapid growth in demand, improving battery technology—especially battery lifespan—has remained a stubborn challenge for scientists and engineers thanks to the limitations of experimental observation and current battery design.

Thankfully, more recent technological innovations in the world of high-performance computing (HPC) have provided physicists with new tools to examine this problem at the atomic scale, which not only has implications for better batteries, but also could ultimately help improve solar cells and design other next-generation electronics. A team of researchers led by Dr. Jan Wilhelm from the Institute of Theoretical Physics at the University of Regensburg recently employed the Leibniz Supercomputing Centre’s (LRZ’s) supercomputer, SuperMUC-NG, to better understand how electrons move through a conductive medium and what factors affect that conductivity.

2D Materials Hold the Key

One big problem in studying what causes batteries to lose potency over time is the fact that today’s batteries are sealed cells: One cannot observe what is happening to the electrons inside the battery cell without disassembling it and thus stopping the electron motion. And although simulations can help to an extent, even the most powerful HPC architectures can only compute quantum mechanical simulations on the scale of nanoseconds, not the months or years that it would take to observe battery degradation.

“We cannot probe inside the battery, so often we don’t know how the atoms are really arranged,” Wilhelm said. “It is therefore very hard for a quantum-mechanical simulation to show why a battery ages and deteriorates.”

These hurdles have led researchers like Wilhelm to focus their attention increasingly toward 2D materials, which only consist of a single layer of atoms. What makes them especially useful for this research is that one can experimentally observe electron dynamics in 2D materials keeping the sample intact. Simulating ultrafast electron dynamics that govern energy movement in 2D materials may help inform researchers about the larger problem that prevents batteries from holding their charge over time: how do electrons transfer energy through the medium, and what inhibits the efficiency of that energy movement?

For example, researchers can run experiments to observe the motion of electrons excited by a laser pulse in the 2D material, similar to electron dynamics inside a solar cell. The collected data can be either directly fed into or compared with computer simulations to better understand the principles governing electron dynamics. “We work towards simulating this whole experiment,” Wilhelm said. “The simulations are absolutely key, because often the experimental data that is recorded is very difficult or even impossible to interpret without a simulation.”

Mathematical Approaches

Because there are many possibilities for how a negatively charged electron may interact with a positively charged unoccupied space within the medium—known together as an electron-hole pair—calculations become more computationally demanding as more particles are added to the system. Approximations are necessary even for very small samples, so methods for accurately charting electron motion within a medium are still evolving. Prior to Wilhelm’s work, many of these computational models were limited to just a few atoms and a couple dozen electrons despite having world-class HPC resources available.

To tackle this issue, Wilhelm spent more than five years writing and perfecting his computational codes and improving approximate algorithms to better simulate more complex electron-hole movements. His team focused particularly on improving the scope and accuracy of the GW method, a common method for calculating the energy gap between electrons and unoccupied electronic states known as holes. “This was the biggest achievement, in my opinion,” Wilhelm said. “We have reduced the computing time of the GW method by more than a factor of 10,000. Now, we can apply the GW method to larger surface areas and get the correct electronic levels for larger systems of atoms, which will be crucial for simulating electron dynamics.”

The results of this fundamental work are not only illuminating new knowledge about electron-hole pairs and their motion, but also establishing more precise algorithms and codes for other researchers tackling similar problems with HPC. Wilhelm’s mathematical work and code is now available to other scientists in the open-source software CP2K, and he is also working with Intel to optimize CP2K for a variety of computing nodes, including porting it to GPU architectures like the forthcoming SuperMUC-NG Phase 2 at LRZ.

Looking Ahead

With these improved methods in hand, researchers like Wilhelm are now focused on how to further scale the system size and improve the accuracy of calculations. “We built a kind of computational microscope that allows us to see the motion of the electron-hole pairs, but we are still limited only 1,000 atoms in the model,” Wilhelm pointed out. He acknowledges that quantum computing could potentially overcome this problem in time and would not have to rely on approximations, but it is still an infant technology that will take years or even decades to develop.

For now, researchers are mining for more experimental data to find new clues. New facilities like the Regensburg Center for Ultrafast Nanoscopy (RUN) at the University of Regensburg, for example, are giving a fresh boost to research like Wilhelm’s by providing state-of-the-art experiments to test and observe electron motion in 2D materials. Next-generation HPC systems at LRZ and the other Gauss Centre for Supercomputing centers will serve as valuable, complementary resources in the search for improved electronic materials.

-Sarah Waldrip

Related Publication: Graml, M., K. Zollner, D. Hernangomez-Perez, P. Faria Junior, and J. Wilhelm. “Low-Scaling GW Algorithm applied to Twisted Transition-Metal Dichalcogenide Heterobilayers,” Journal of Chemical Theory and Computation 20, 2202 (2024). https://doi.org/10.1021/acs.jctc.3c01230

 

Funding for SuperMUC-NG was provided by the Bavarian State Ministry of Science and the Arts and the German Federal Ministry of Education and Research through the Gauss Center for Supercomputing (GCS).

Tags: Physics Materials Science Universität Regensburg LRZ