Neutron stars possess a superfluid core where neutrons flow without resistance. Scientists traditionally relied on complex models involving "Cooper pairs" to understand this phenomenon. A new study leverages artificial neural networks to accurately predict neutron star behavior without such assumptions. The study was published in Physical Review Research.
Researchers study neutron star crusts by simulating neutron matter then adding “hidden” neutrons to mediate interactions between “real” neutrons. Next, neural networks construct a quantum wave function of neutron matter’s normal and superfluid phases. Image Credit: Jane Kim, Ohio University
The study modified the conventional "single-particle" approach by introducing "hidden" neutrons that facilitate interactions among the "real" neutrons and encode quantum many-body correlations, allowing Cooper pairs to naturally emerge during the calculation.
The Impact
Understanding neutron superfluidity is crucial for gaining insights into neutron stars, including their cooling mechanisms, rotation, and phenomena like glitches—sudden changes in their rotation rate. While direct experimental access to neutron star matter is impossible, the fundamental interactions governing this matter are the same as those that govern atomic nuclei on Earth.
Researchers are focused on developing nuclear interactions that are both simple and predictive. Solving the quantum many-body problem accurately is an essential part of evaluating the quality of these interactions. This work employs simple interactions that align well with previous calculations that used much more complex interactions.
Summary
Low-density neutron matter exhibits intriguing emergent quantum phenomena, such as the formation of Cooper pairs and the onset of superfluidity. Researchers studied this density regime using artificial neural networks combined with advanced optimization techniques.
Using a simplified neutron interaction model, they calculated the energy per particle and compared these results to those derived from more realistic interactions. This approach proves to be competitive with other computational methods while significantly reducing the cost.
The research was funded by the Department of Energy (DOE) Office of Science, Office of Nuclear Physics, Office of Advanced Scientific Computing Research, Scientific Discovery through Advanced Computing (SciDAC) NUCLEI program, and the US National Science Foundation.
Numerical calculations were performed using the Laboratory Computing Resource Center at Argonne National Laboratory and the computers of the Argonne Leadership Computing Facility, a DOE Office of Science user facility.
Journal Reference
Fore, B., et al. (2023) Dilute neutron star matter from neural-network quantum states. Physical Review Research. doi.org/10.1103/physrevresearch.5.033062.