Monday 31 March 2025
A team of scientists has developed a new approach to predicting the properties of atomic nuclei, which could have significant implications for our understanding of the universe.
The researchers used a type of artificial intelligence called a Bayesian neural network to create an emulator that can predict the binding energy and charge radius of oxygen isotopes across a range of fidelities. Fidelity refers to the level of detail in the calculations, with higher fidelities requiring more computational power and lower fidelities providing coarser estimates.
The team’s approach is significant because it allows them to make predictions at much higher fidelities than would be possible using traditional methods, which are limited by their computational requirements. This means that they can explore a much wider range of possibilities and gain a better understanding of the underlying physics.
One of the key challenges in predicting the properties of atomic nuclei is the need to account for the interactions between protons and neutrons within the nucleus. These interactions are complex and difficult to model, but the emulator developed by the team is able to capture them accurately.
The researchers tested their emulator using a dataset of oxygen isotopes, which they divided into two parts: one for training the emulator and another for testing its performance. They found that the emulator was able to make accurate predictions across a range of fidelities, including those that were not included in the training data.
This approach could have significant implications for our understanding of the universe. By allowing us to predict the properties of atomic nuclei with greater accuracy, it could help us better understand the processes that occur within stars and other celestial objects. It could also enable us to design new experiments and make more precise measurements.
In addition to its potential applications in astrophysics, this approach could also have implications for our understanding of the fundamental laws of physics. By allowing us to test theoretical models against experimental data with greater precision, it could help us refine our understanding of the underlying laws that govern the behavior of particles and forces.
The team’s results are a significant step forward in the development of emulators for predicting the properties of atomic nuclei. They demonstrate the power of combining artificial intelligence with traditional computational methods to tackle complex problems.
Cite this article: “Predicting Atomic Nucleus Properties with Artificial Intelligence”, The Science Archive, 2025.
Artificial Intelligence, Bayesian Neural Network, Atomic Nuclei, Binding Energy, Charge Radius, Fidelity, Computational Power, Astrophysics, Fundamental Laws Of Physics, Emulators
