Mode angular degree identification in subgiant stars with convolutional neural networks based on power spectrum

Abstract

ABSTRACT The identification of the angular degrees l of oscillation modes is essential for asteroseismology and it depends on visual tagging before fitting power spectra in a so-called peakbagging analysis. In oscillating subgiants, radial (l = 0) mode frequencies are distributed linearly in frequency, while non-radial (l ≥ 1) modes are p–g mixed modes that have a complex distribution in frequency that increases the difficulty of identifying l. In this study, we trained a one-dimensional convolutional neural network to perform this task using smoothed oscillation spectra. By training simulation data and fine-tuning the pre-trained network, we achieved 95 per cent accuracy for Kepler data.