Researchers from the High Energy Nuclear Physics Laboratory at the RIKEN Pioneering Research Institute (PRI) in Japan and their international collaborators have achieved a groundbreaking discovery that bridges artificial intelligence and nuclear physics. By applying deep learning techniques to a vast amount of unexamined nuclear emulsion data from the J-PARC E07 experiment, the team identified, for the first time in 25 years, a new double-Lambda hypernucleus. This marks the world’s first AI-assisted observation of such an exotic nucleus—an atomic nucleus containing two strange quarks. The finding, published in Nature Communications, represents a major advance in experimental nuclear physics and provides new insight into the composition of neutron star cores, one of the most extreme environments in the universe.

All ordinary matter is composed of atoms, whose nuclei consist of protons and neutrons. These particles, known as hadrons, are made of more fundamental building blocks called quarks. A strong nuclear force binds hadrons together, maintaining a delicate balance that both pulls them together and keeps them far enough apart to prevent collapse. Understanding the origin of this force is crucial to explaining how matter and the universe were initially formed.

To probe this force more deeply, scientists study hypernuclei, atomic nuclei that contain hyperons—particles that include a strange quark. In extremely rare cases, two hyperons can be bound within the same nucleus, forming a double-Lambda hypernucleus. These systems allow researchers to directly measure the interaction between the two Lambda particles and to explore how the nuclear force behaves when strange quarks are involved. This knowledge is also essential for understanding the properties of matter under the extreme densities found in neutron stars, where hyperons are expected to exist.

Detecting double-Lambda hypernuclei has long been a major challenge because they are produced very rarely and exhibit complex decay structures. In the J-PARC E07 experiment, nuclear emulsion plates recorded the tracks of particles resulting from hypernuclear formation and decay. However, only a very small fraction of the emulsion data has been analyzed because conventional methods are time and labor consuming. As a result, a vast amount of valuable information remains unexplored.

The RIKEN-led team developed a deep learning–based analysis framework to process this enormous dataset. By training neural networks to recognize the subtle signatures of double-strangeness events, the researchers were able to automatically extract candidate images indicating the potential formation and decay of double-Lambda hypernuclei. These candidates were then examined under a microscope, and one event was confirmed through detailed kinematic analysis as the production of a double-Lambda hypernucleus of boron-13 (13ΛΛB), in which two Lambda particles are bound to a boron-11 nucleus. This identification is only the second unambiguous observation of a double-Lambda hypernucleus in history and the first such discovery in nearly 25 years. Moreover, it is the first time that the interaction between two Lambda particles has been measured in a nucleus other than helium.

Remarkably, this breakthrough was achieved by analyzing just 0.2 percent of the total emulsion data from the J-PARC E07 experiment. Based on this detection rate, the researchers estimate that the full dataset could contain more than 2,000 double-strangeness events awaiting discovery. The team plans to continue refining their deep learning–based analysis methods and to extend their search to reveal the behavior of Lambda particles in other nuclei, as well as the interactions between xi hyperons and atomic nuclei.

According to Takehiko Saito, chief scientist of the High Energy Nuclear Physics Laboratory, “This achievement demonstrates how artificial intelligence can uncover extremely rare phenomena hidden within massive experimental datasets, revealing events that would be nearly impossible to find by human inspection alone. We believe that their approach will open the door to large-scale discoveries of double-strangeness hypernuclei and deepen our understanding of the nuclear force and the structure of matter in the universe.”