The atomic nucleus is a small, dense region made of closely packed protons and neutrons situated at the center of the atom. Understanding how atomic nuclei emerge from the individual interactions among their constituents is a complicated quantum many-body problem, whose solution has fundamental implications to our understanding of the universe.
Researchers at Argonne in collaboration with the École Polytechnique Fédérale de Lausanne (EPFL) have developed a novel method to solve the nuclear quantum many-body problem. The quantum-mechanical wave function of the nucleus is compactly represented by artificial neural networks, which are efficiently trained by minimizing the energy of the system.
Detailed comparisons with existing numerical methods have proven that artificial neural networks can reproduce the binding energy of light nuclei and the spatial distributions of protons and neutrons. Work is underway to extend this approach to larger nuclei that are essential for constraining key features of the nuclear force. Applications range from helping analyze the next generation of neutrino experiments, such as DUNE, and the search for neutrinoless double beta decay, to predicting the structure of neutron stars and the gravitational wave signature of their mergers.