Abstract:Neutrinos can undergo fast flavor conversions (FFCs) within extremely dense astrophysical environments such as core-collapse supernovae (CCSNe) and neutron star mergers (NSMs). In this study, we explore FFCs in a \emph{multi-energy} neutrino gas, revealing that when the FFC growth rate significantly exceeds that of the vacuum Hamiltonian, all neutrinos (regardless of energy) share a common survival probability dictated by the energy-integrated neutrino spectrum. We then employ physics-informed neural networks (PINNs) to predict the asymptotic outcomes of FFCs within such a multi-energy neutrino gas. These predictions are based on the first two moments of neutrino angular distributions for each energy bin, typically available in state-of-the-art CCSN and NSM simulations. Our PINNs achieve errors as low as $\lesssim6\%$ and $\lesssim 18\%$ for predicting the number of neutrinos in the electron channel and the relative absolute error in the neutrino moments, respectively.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is the influence of fast flavor conversions (FFCs) on multi - energy neutrino gases in extreme astrophysical environments such as core - collapse supernovae (CCSNe) and neutron - star mergers (NSMs), and their reconstruction in these environments. Specifically, the paper focuses on how neutrinos of different energies share a common survival probability when the growth rate of FFCs significantly exceeds the vacuum Hamiltonian, and this probability is determined only by the neutrino spectrum after energy integration. In addition, the paper also explores how to use physics - informed neural networks (PINNs) to predict the asymptotic results of multi - energy neutrino gases after FFCs.
### Main research contents:
1. **Performance of FFCs in multi - energy neutrino gases**: The study found that when the growth rate of FFCs is much greater than the vacuum frequency, the influence of neutrino energy on its flavor conversion process becomes unimportant, and all neutrinos, regardless of their energy levels, have the same survival probability, which is determined by the neutrino spectrum after energy integration.
2. **Application of PINNs**: The paper proposes using PINNs to predict the asymptotic state of multi - energy neutrino gases after FFCs. PINNs use the zero - order and first - order moments of initial (anti - ) neutrinos and information in specific energy intervals to predict neutrino moments after FFCs.
3. **Performance evaluation**: Through simulated experimental data, the performance of PINNs in predicting neutrino number density and moments was evaluated. The results show that PINNs can achieve a relative error of about 6%, while in an ordinary neural network without additional features, this error is about 12%. For the prediction of neutrino moments, the mean absolute relative error of PINNs is about 18%, while the error of the ordinary neural network is about 22%.
### Key methods and techniques:
- **Physics - informed neural networks (PINNs)**: Combine the advantages of physical knowledge and neural networks, and improve the prediction accuracy by introducing additional features (such as the shape parameter of the survival probability distribution).
- **Feature engineering**: Define new features, such as α, Fνe, F¯νe, nνe,i, n¯νe,i, nνx,i, Fνe,i, F¯νe,i, Fνx,i, to enhance the performance of the neural network.
- **Loss function optimization**: Introduce an additional loss term to penalize the deviation of the neutrino number density in the electron channel, thereby ensuring the satisfaction of the physical conservation law.
### Conclusion:
The paper successfully demonstrates the effectiveness of PINNs in predicting the asymptotic state of multi - energy neutrino gases after FFCs, especially when dealing with neutrino flavor conversion problems in complex astrophysical environments, PINNs show higher accuracy and reliability. This provides a powerful tool for more accurately incorporating FFCs in future CCSNe and NSMs simulations.