Our research highlights serve as a collection of feature articles detailing recent scientific achievements on GCS HPC resources.
Since the dawn of the Information Age in the middle of the 20th century, humanity has seen rapid developments in the realm of electronics and materials science. In the 1950s, the UNIVAC-I became the first commercially available, general-purpose computer, capable of just under 2,000 calculations per second—a far cry from a modern iPhone, capable of more than 10 trillion calculations per second.
Whether it is new medical devices, materials for advanced manufacturing applications, information technology innovations, or simply the vast array of consumer devices, our rapid technological advancements were born out of a better understanding of how atomic particles behave and interact with one another at a fundamental level.
Understanding these interactions is the work of researchers dedicated to fundamental science. The centers that comprise the Gauss Centre for Supercomputing (GCS)—the High-Performance Computing Center Stuttgart (HLRS), the Jülich Supercomputing Centre (JSC), and the Leibniz Supercomputing Centre (LRZ)—are dedicated to supporting fundamental research in the interest of laying the groundwork for tomorrow’s great technological advancements.
To that end, researchers at the Julius-Maximillians-Universität Würzburg (JMU) have been long-time users of high-performance computing (HPC) resources at LRZ to illuminate the complex, mysterious world of solid-state physics—a scientific domain focused on understanding how particles interact with one another and their environments at the atomic and subatomic levels. Recently, the team investigated a previously poorly studied quantum system dubbed Kondo heterostructures, which reveal a host of fascinating emergent collective properties that hold promise for further theoretical, numerical and experimental investigations.
“The point of our research is to understand the quantum world and manipulate it,” said Prof. Dr. Fakher Assaad, JMU professor and lead researcher on the project. In view of applications down the road, we have to bear in mind that the quantum effects that we consider take place at very low temperatures. A huge challenge is to realize these effects at room temperature. Before we can do that, though, we must be able to more fully understand and play with these systems.”
Experiments and simulations work in concert toward new insights
In 2016, a multi-institutional team of Dutch experimentalists published an article in Nature Physicsstudying cobalt adatoms—small numbers of magnetic atoms that are adsorbed, or stuck, to a material surface rather than being absorbed into a material—on a copper surface. The team used an experimental technique called scanning tunneling microscopy (STM), which uses an ultra-sharp tip as a microscope to both observe and manipulate individual atoms into specific patterns, or structures, in order to better understand their magnetic properties and quantum behavior under certain conditions.
Understanding atomic systems’ behavior is not as simple as just pointing a microscope at them, though—experimentally, it is impossible to know both an electron’s speed and position at any given moment. This becomes even more daunting when looking at systems of many atoms and their many constituent electrons. In order to fully understand how nanosecond changes can impact these systems, researchers often turn to computational modelling to verify what they think they see experimentally.
In order to computationally model such a system, researchers rely on Monte Carlo simulations, which use statistical physics to sample all possible particle positions at a given moment. While the method is relatively straightforward, even a modest number of atoms has millions or billions of possible configurations, meaning researchers must have access to HPC resources to finish simulations in a reasonable amount of time. For quantum systems such as these, Assaad and his team do quantum Monte Carlo simulations. This translates quantum physics observed in the simulation into classical physics, but one dimension higher—a two-dimensional quantum system being translated into a three-dimensional classical system, for instance.
Using SuperMUC-NG at LRZ, Assaad and his collaborators applied their computational approach to the team’s experimental system and were able to model it with one hundred percent accuracy. However, the team wanted to take the work further and grow the system size from a handful of atoms to a much larger volume in order to see whether the behavior would change. In the process, they uncovered a new type of system where particles’ quantum spins in a metallic environment behave differently than previously observed. These so-called Kondo heterostructures offer physicists a promising lead in their pursuit of novel quantum phases.
“We have this simple model which reflects reality, but then you ask yourself, ‘What happens if, instead of having 10 cobalt adatoms on a metallic surface, we have an infinite chain?’” said Assaad. “This research started off motivated by a question that came from experiment. Since the model reproduced experimental data for a handful of cobalt adatoms, we know that it was correct.”
Then the work evolved into something where we could help guide experimentalists in their search for interesting new physics. This work is close to experiment, motivated by experiment, and shows strong feedback between numerics and experiment.”
Classical computing fuels the quantum revolution
As voracious HPC users, Assaad and his colleagues have been allocated time on both the CPU-centric SuperMUC-NG as well as the GPU-heavy JUWELS system at JSC, another GCS center. Assaad pointed out that in order to use different architectures, researchers must rework their applications to run efficiently on new machines. Luckily, they find good support for porting their applications at the centers. “It works well when you have people who really know these machines, with whom you can work closely, discuss things, and ultimately get an understanding of how to quickly make your program run better,” he said.
Having had long-term access to GCS resources, the researchers have developed a mature, stable computational workflow that remains flexible for studying a variety of quantum systems on diverse HPC architectures. This ultimately shines new light on the still-mysterious quantum world and brings research to life for a wider audience.
Moving forward, Assaad keeps dreaming: he indicated that he, like many other physicists, is always seeking out new and interesting problems that can be solved with today’s technologies, while also keeping an eye toward what could be possible tomorrow. With its computational approach, the team is interested in seeking new classes of model systems that connect to materials, exhibit novel phases and phase transitions, and inspire new applications. “The richness of physics is amazing. There is no limit to the variety of phenomena you can generate with materials, and we are pretty rudimentary in our understanding compared to what the quantum world offers. There is a huge potential for progress, but it takes time,” Assaad said.