1.4 million euros for the digitalisation of fundamental research

From language models and autonomous experiments to machine learning: four research projects at the Heinz Maier-Leibnitz Centre (MLZ) are receiving funding of around 1.4 million euros from the Federal Ministry of Research, Technology and Space (BMFTR).

KI In the future, large datasets should be optimised using AI. © Grispb / Adobe Stock

In the future, large datasets should be optimised using AI. © Grispb / Adobe Stock

Funding began in 2025 as part of the ErUM-Data Programme, which ran for three years in total. ErUM stands for ‘Erforschung von Universum und Materie’ (German for ‘Research into the Universe and Matter’) and supports fundamental research in the natural sciences.

Big Data Becomes Smart Data

Research generates enormous datasets. However, before researchers can analyse them efficiently, the data must first be stored and filtered. Due to the sheer volume involved, such datasets are also referred to as ‘Big Data’. In future, however, AI should make it possible to easily extract the most promising data: thereby transforming ‘Big Data’ into ‘Smart Data’. Clustered according to patterns, this Smart Data is then made available to research teams for further analysis. In short: handling and working with large volumes of data paves the way for efficient research at major research facilities such as the Heinz Maier-Leibnitz Research Neutron Source.
Four of the 23 projects funded by the BMFTR focus on neutrons – all of them at the Heinz Maier-Leibnitz Centre. Dr Sebastian Busch from the Helmholtz Centre Hereon is leading two of these projects.

FRM II air Major research facilities such as the FRM II generate vast amounts of data – the funded projects are intended to improve data management.

Major research facilities such as the FRM II generate vast amounts of data – the funded projects are intended to improve data management.

On rapid feedback mechanisms and autonomous experiments

Thanks to modern neutron optics systems and detector technologies, neutron scattering experiments are generating more data than ever before – and in a very short space of time. The challenge here is not to lose track of the big picture. The team behind the FASTER project aims to ensure precisely that – the aim is to make autonomous experiments routine. “For researchers at large-scale facilities such as the FRM II, this technology also has the advantage of allowing them to make the most of the limited time available at the research instrument (or neutron beam),” says Sebastian Busch.

Information field theory for experiments at large-scale research facilities

Another project led by Sebastian Busch, ErUM-IFT-2, also deals with data from neutron scattering: “Data from such experiments are, by their very nature, difficult to express using simple, exact formulae – yet these are essential if we are to calculate from the results exactly what lies within a sample,” explains the physicist. In data analysis, therefore, approximations and other numerical methods are used, although these result in the loss of valuable information. A postdoctoral researcher will be working to further improve precisely these numerical methods.

Language models are already widely used in everyday life. They are now also being applied in research data management. © Antony Weerut / Adobe Stock

Language models are already widely used in everyday life. They are now also being applied in research data management. © Antony Weerut / Adobe Stock

Large Language Models (LLM) in Research Data Management

Data should be FAIR – Findable, Accessible, Interoperable, Reusable. The team behind DAPHNE4NFDI (Data from Photon and Neutron Experiments for National Research Data Infrastructure) and the MLZ are working to ensure that datasets meet these criteria: “Specifically, we achieve this by providing well-structured and machine-readable metadata from publications spanning the last few decades – through the integration of language models,” says Dr Michael Schulz, Deputy Scientific Director of FRM II. Together with Sebastian Busch and other colleagues, he leads projects known as Physics-LLM.