Stellar flares are observed in large numbers with space observatories such as the Transiting Exoplanet Survey Satellite (TESS). The well-sampled profiles of these flare light curves may contain valuable information about the physical processes in the stellar atmospheres. To investigate this, we searched for flares in TESS light curves from the first five years of the mission, and created a pure sample of 120 000 flares with a combination of a recurrent neural network and manual vetting. By scaling these flares in both timescale and amplitude, we reveal continuous changes in the average flare profiles with spectral type. These trends are not apparent for just a few flares, they only emerge when averaging thousands of events. We use dimensionality reduction to explore this flare shape space, and present some applications.
Seli et al. (Sun,) studied this question.