2025 Fritz Kutter Prize awarded to USI Master graduate for the thesis carried out at IRSOL
Istituto ricerche solari Aldo e Cele Daccò
1 December 2025
Samuel Corecco, Master in Artificial Intelligence graduate from the Faculty of Informatics at Università della Svizzera italiana (USI), is the winner of the 2025 Fritz Kutter Prize, which recognises high-quality and innovative theses in Computer Sciences from Master and PhD graduates of a Swiss university.
Corecco’s thesis “Using Artificial Intelligence for retrieving physical parameters of solar plasma from spectropolarimetric lines” was carried out at Istituto ricerche solari Aldo e Cele Daccò (IRSOL) under the supervision of Prof. Svetlana Berdyugina. The award was presented by the Fritz Kutter Fund in a ceremony held at ETHZ on 24 November.
Space weather is a field that is at the forefront of applied solar-terrestrial research and, above all, fundamental to the safety of human civilisation. In fact, geomagnetic storms triggered by the Sun can put electrical grids on Earth and space activities in orbit at risk.
The mechanism that triggers these storms lies in the Sun's magnetic field. To monitor it, scientists analyse solar light polarization, which reveals the temperature, velocity and magnetic properties of solar plasma. However, decoding them – a mathematical challenge known as the “inversion problem” – is currently a slow, computationally demanding process that is difficult to apply on a large scale. As next-generation telescopes prepare to generate petabytes of multi-dimensional data every day, traditional analysis methods are reaching a bottleneck. Prof. Svetlana Berdyugina emphasises: “It is not only computational speed that is important. Modeling polarimetric data requires the complex physics of light propagation through the magnetized solar atmosphere.” This is where SunTransformer, the subject of Samuel Corecco's thesis, comes in.
SunTransformer is a new sequential process based on artificial intelligence that revolutionises the way we process solar data and offers a perfect blend of speed, accuracy and physical information. Unlike traditional software, which can take hours to process data, the new approach is incredibly fast and employs the most advanced models of atomic and molecular polarization. It can predict spectral line profiles with almost zero error and process a complete map of a large magnetic region on the Sun (e.g., from the JAXA Hinode satellite, used in the thesis) in just 36 seconds. Amazingly, it achieves this efficiency using standard hardware and requires only 1.5 GB of memory. SunTransformer's speed, accuracy and physics-based modelsallow for a near-instantaneous analysis of high-resolution data, enabling scientists to observe changes in the Sun as they occur. Since rapid magnetic, temperature and velocity mapping is crucial for predicting solar flares and coronal mass ejections, processing data quasi-instantly makes it possible to better anticipate space weather events. What's more, the system is designed to handle the massive influx of data from future observatories, such as the 4.2m European Solar Telescope (EST, under development).
In short, SunTransformer represents a fundamental step from raw data to concrete physical information, ensuring that as our view of the Sun becomes clearer and sharper, our understanding keeps pace.