Abstract Stellar flares are important indicators of stellar magnetic activity and have been widely studied. Detailed studies of flares have raised new questions. For example, the quasiperiodic pulsation (QPP) phenomenon in flares challenges standard flare models, yet the underlying mechanism of this phenomenon remains poorly understood. We utilize the 2 minute cadence light-curve data from TESS Sectors 1–74 prior to 2024 January 30. By combining the previously published convolutional neural network–based machine learning flare identification method with traditional flare detection techniques, we have obtained a flare sample consisting of 208,280 flare events from 29,280 flare stars. In our flare events sample, flares with rise times within 10 minutes and decay times within 50 minutes account for more than 90% of the total number of flares. The proportion of flare stars is higher in binaries, the main sequence, and stars with KM spectral types compared to single stars, giants, and stars with spectral types other than KM. The fitted α index results for the flare frequency distributions across different evolutionary stages indicate a general trend of decreasing α index from main-sequence stars to giants. Except for M-type stars, the α index also gradually decreases from K-type to O-type stars. Using the flare QPP identification method, based on the fully convolutional networks published by previous researchers, we have also identified flare events with QPP, and through flare parameter constraints, we selected 10,465 flares with QPP features from M-type stars and qualitatively investigated their correlations with stellar parameters.
Tianhao Su (2025) studied this question.