Fast Surrogate for the Earth Matter Effect on Solar Neutrinos

This paper presents a deep-learning surrogate model for the Earth matter effect on solar neutrinos using a residual convolutional network conditioned on neutrino oscillation parameters. The model is trained on numerical solutions computed over a grid of parameter values and predicts Earth-induced transition probabilities across neutrino energy and zenith angle. It provides a computationally fast approximation to direct numerical calculations, demonstrating the application of neural network surrogates in neutrino physics.