Artificial intelligence (AI) has the potential to radically change how we conduct X-ray and neutron science. A new theme at LINXS, GAIN, will bring together researchers, facility scientists, and AI experts to identify where AI can make a genuine difference, from improving facility reliability and experimental efficiency to enabling more accessible experiments and faster data analysis.

“Performing experiments at large-scale facilities like MAX IV and ESS can be challenging. Our goal is to lower some of the barriers researchers face when using such facilities. This means looking at the entire experimental pipeline, from accelerator and beamline operations to data analysis,” says theme leader Pablo Villanueva Perez, Associate Professor in Physics at Lund University.

Theme leader Pablo Villanueva Perez is excited to explore how AI can be used to identify solutions that can have real impact for X-ray and neutron science.

Identify bottlenecks in X-ray and neutron science

GAIN (Generating Advances with AI for neutrons/X-rays) will bring together three communities: scientists working at large-scale facilities, X-ray and neutron users, and experts in AI and scientific computing. The challenge is not simply to develop new AI methods, but to connect them with real scientific and operational needs, Pablo Villanueva Perez notes.

GAIN will create a platform where facility scientists who understand the instruments, users who encounter challenges during experiments, and AI and scientific-computing experts can identify problems and solutions. These challenges occur across the whole experimental workflow, from loss of beam during an ongoing experiment and the need for extensive training and support to operate certain beamline instruments, to challenges in data processing and analysis.

Pablo Villanueva Perez highlights several examples of how AI could be used to tackle these issues.

“Losing the beam can be extremely disruptive during an experiment because every interruption means losing valuable beamtime. AI could help in several ways. It could identify unusual patterns in operational data that indicate a potential problem before an interruption occurs. When a problem does happen, an AI agent could also search facility documentation and previous cases to help identify possible solutions.”

Freeing up time for science and opening facilities to new users

“AI can also be used to optimise experiments at large-scale research facilities”, says Pablo Villanueva Perez.

One example is to develop AI-assisted tools that can help researchers interact with complex beamline control systems and perform operations that today require extensive training or the assistance of an instrument scientist. Already, scientists have begun to develop autonomous beamline experiments in which AI can control parts of the experimental workflow. Such approaches could also allow experiments to adapt in real time, using data being collected to help decide what to measure next.

By handling routine operations and helping users navigate complex controls, such tools could allow instrument scientists to devote more time to scientific collaboration, challenging experiments, and developing new capabilities. In addition, by reducing some of the technical barriers associated with operating complex instruments, AI could also make X-ray and neutron techniques more accessible to researchers from different scientific fields.

Use AI where it can make a meaningful difference

“There are many exciting possibilities for AI, but the important question is where it can make an actual difference. To answer that, we need to bring together the people who understand the facilities, the researchers who experience the challenges, and the people developing AI methods. GAIN is a place where we can match real problems with new possibilities and turn them into meaningful solutions,” says Pablo Villanueva Perez.

Builds on Villanueva Perez’s own research interests

The aim of the theme is also close to Pablo Villanueva Perez’s own research interests. Originally from Spain, he came to Lund seven years ago to explore the new capabilities offered by MAX IV and ESS. He develops novel X-ray imaging methods and instruments, with AI as a key ingredient in enabling new imaging capabilities.

“AI became a tool that enabled us to come up with solutions for how to develop new imaging methods. It allows us to tackle problems that are very difficult to address with conventional computational methods and helps us think in new directions.”

“Our work in X-ray imaging made us interested in exploring how we can integrate AI into the whole pipeline for X-ray and neutron science and motivated the theme application. With GAIN, we are looking forward to working with colleagues across these different communities to identify where AI could be used smartly and to accelerate its adoption.”

About GAIN

GAIN will start its work in 2027. The theme will work across three interconnected working groups spanning the full scientific workflow: Working Group 1: AI for sustainable accelerators; Working Group 2: Beamline controls, autonomous experimentation, and online data analysis; and Working Group 3: AI for knowledge. Through these working groups, GAIN aims to create a meeting point for facility scientists, X-ray and neutron users, and experts in AI and scientific computing.