Researchers discovered information from under the Sun’s surface about the emergence of solar storms using the LUMI supercomputer

This is partly due to the sheer complexity of the physical problem to be solved: simple pen and paper are not enough; numerical modelling is needed. This kind of modelling is also challenging because of many different time and other scales, and it requires supercomputer resources, describes professor Maarit Korpi-Lagg from the Department of Computer Science at Aalto University, Finland.

Korpi-Lagg leads a nine-person multidisciplinary research group involving computer science researchers from Aalto University and astrophysics researchers from the German Max Planck Institute. The group was one of the LUMI supercomputer’s GPU pilot phase users.

LUMI pilot project proved the theory

Solar data is continuously collected through terrestrial and satellite observatories. Researchers also develop computational models that can be compared with observations. However, so far it has proved impossible to build a numerical model that could reproduce all relevant real-world observations and help predicting solar storms better.

In the LUMI pilot project, a physical mechanism the existence of which could explain the differences between previous models and observations was studied.

– In our LUMI pilot project called VISSI (VerIfying Small-Scale dynamo actIon in the Sun), we continued to search for an answer to the question of whether a fluctuating dynamo can occur in the Sun’s convection zone. Dynamo mechanisms are a bit like a bicycle dynamo, but kinetic energy does not turn into electrical energy, i.e., a light in the lamp, but rather into solar magnetic field energy, Korpi-Lagg illustrates.

Before the LUMI pilot project, there was no certainty whether a fluctuating dynamo actually is significant in the deep solar convection zone.

– A theoretical prediction was that it should exist, but numerically no one had yet found it being relevant in sun-like conditions. With the LUMI pilot project, we were able to numerically verify the existence of this mechanism using simulation models that were the most accurate of their own type so far to be created. The goal of the project was to solve this significant sub-question in order to make clear the direction for further research, Korpi-Lagg continues.

The research will continue on the LUMI supercomputer with computing resources awarded through Finland’s own LUMI Extreme Scale Call. In this project, the research group begins to investigate how the solar subsurface small-scale dynamo mechanism verified in the LUMI pilot project affects the larger scale solar phenomena and, ultimately, the emergence of solar storms.

– In our follow-up project called SISSI (Studying Small-Scale dynamo action in the Sun), we will study the effects of the dynamo process verified during the pilot project on larger scale solar phenomena, such as preservation of kinetic energy, thermal convection, and a large-scale dynamo. All of these interact to generate the Sun’s magnetic field, which forms active regions, some of which produce solar flares. These then, in their turn, cause geomagnetic storms that can damage our technologically advanced society, states Korpi-Lagg.

More time to prepare

Solar flares and storms occur most when the Sun’s activity is high. Solar activity has an 11-year cycle with the next peak in activity predicted to be in 2025.

There are space weather service centres around the world predicting space weather. The predictions are at the moment based on monitoring the changes in the solar surface, which means that predictions can only be made when the active region has already appeared on the Sun’s surface. It can often be too late to react by then.

– The roots of all these disturbances are located in the convection zone under the Sun’s surface, which our group studies. It is difficult to predict the most energy-containing eruptions because disturbances travel close to the speed of light. We are currently developing methods to monitor sub-surface changes. These could possibly give us as long as a couple of days more time to react, Korpi-Lagg says.

A project of technical and scientific interest

The month-long LUMI pilot project was challenging but rewarding for the research group.

– We have recently completed the implementation of a new paradigm. That means switching from CPU-based computing to graphics accelerator-based computing. This enables the transition from petascale to exascale computing. The LUMI pilot proved that this is possible and scientifically useful also in practice, not just in theory. The project also paved the way for new challenges. Before we were mainly concerned about whether we can utilize thousands of graphics accelerators efficiently. This fear proved groundless, but during the project, the fast storage and transfer of colossal amounts of date caused us trouble. We must get back to these challenges and find even better solutions for the follow-up project, Korpi-Lagg notes.

The group uses Astaroth code developed at Aalto University, which was optimised for AMD’s GPUs.

– The work was worthwhile: we can now use LUMI’s GPU capacity to do in a day what we used to need a month for when using CPUs, Korpi-Lagg gives an indication of the computing power of GPUs and how the work has been accelerated.

Korpi-Lagg is pleased with the fact that the transition to GPU architectures will also save energy.

– Faster calculations also mean considerable energy savings. For years I have felt bad thinking about how much environmental resources supercomputers require, but now working on one of the most environmentally friendly supercomputers in the world, means this is a great weight off my mind, Korpi-Lagg concludes.

Watch the video interview of Maarit Korpi-Lagg:

Writer: Anni Jakobsson