A Machine-Learning Approach for Identifying CME-Associated Stellar Flares in TESS Observations

TL;DR

This work addresses whether solar flare–CME relationships extend to other stars by training a Sun-as-a-star flare classifier on GOES 1–8 Å data to distinguish eruptive from confined events and applying it to 41,405 TESS white-light flares from FGKM stars. The authors build a robust feature set combining manual flare-morphology metrics with deep image features from a ResNet50, reduced by PCA, and evaluate multiple classifiers, with Random Forest performing best. They demonstrate cross-band transferability between solar soft X-ray and stellar white-light flares, and find that ~47% of stellar flares exhibit CME-like morphologies, with the intrinsic CME association fraction estimated between 0.35 and 0.60 when accounting for model uncertainty. A key result is that the predicted CME fraction decreases with flare energy, suggesting stronger magnetic confinement on active stars, with important implications for exoplanet space weather and a scalable approach for future multi-band CME studies.

Abstract

Coronal mass ejections (CMEs) are major drivers of stellar space weather and can strongly influence the habitability of exoplanets. However, compared to the frequent occurrence of white-light flares, confirmed stellar CMEs remain extremely rare. Whether such flares are commonly accompanied by CMEs is a key question for solar-stellar comparative studies. Using Sun-as-a-star soft X-ray flare light curves observed by the GOES XRS 1--8~Å channel, we compiled a sample of 1,766 M-class and larger solar flares and extracted features with both deep convolutional neural networks and manual methods. Five machine-learning classifiers were trained to distinguish eruptive from confined flares, with the random forest model achieving the best performance (true skill statistic; TSS = 0.31). This TSS value indicates that the model possesses a moderate ability to discriminate between eruptive and confined flares. Normalized white-light and GOES XRS flare light curves show broadly consistent temporal evolution, reflecting their shared energy-release history and supporting a probabilistic transfer of the model to white-light flare data. We applied the best-performing RF model to 41,405 TESS-detected flares on FGKM-type main-sequence stars, predicting that approximately 47% of events show CME-like morphological characteristics, with the model-implied intrinsic association fraction lying in the range 35%--60%. Intriguingly, the CME occurrence rate decreases with increasing flare energy, indicating that the most energetic flares may be more strongly confined by overlying magnetic fields. These results provide new insight into flare-CME connections in diverse stellar environments and have important implications for assessing the impact of stellar eruptive activity on exoplanetary atmospheres.

Figures (11)

• Figure 1: Normalization and feature extraction of a solar flare light curve. Panel (A) displays the normalized light curve (red solid line). Panel (B) shows a subset of extracted features, including the normalized curve (black line), a linear fit to the rise phase (green solid line), an exponential decay fit to the decay phase (red dashed line), and the full width at half maximum (FWHM; blue solid line). Shaded areas indicate the integration intervals for the rise phase (light yellow), decay phase (light purple), and FWHM (light blue hatched).
• Figure 2: Comparison of normalized light curves for white-light flares (solid blue) and the GOES XRS 1–8 Å flares (dashed red). Panels (A) and (B) correspond to events 20110907_222745 (GOES class X1.8) and 20131110_043815 (GOES class X1.1), named after the start times of the white-light flares. The x- and y-axes represent normalized time and normalized flux intensity, respectively. Pearson correlation coefficients of 0.995 and 0.993 indicate a strong similarity in light curve profiles.
• Figure 3: Distribution of the model-predicted eruptive-flare fraction as a function of flare energy for FGKM-type main-sequence stars at a flare blackbody temperature of 9000 K. Panels (A)–(H) correspond to F-, G-, K-, and M-type stars, with two panels for each spectral type. The left panels show flare-count histograms, separating confined flares (blue) and eruptive flares (orange), with the corresponding eruptive fractions indicated in the legends. The right panels display the eruptive-flare fraction as a function of flare energy, together with a linear-fit trend (orange dashed line). Panels (I) and (J) show the combined results for all FGKM-type stars. Overall, the eruptive-flare fraction decreases with increasing flare energy.
• Figure 4: Distribution of the model-predicted eruptive-flare fraction as a function of flare energy for FGKM-type main-sequence stars at a flare blackbody temperature of 12000 K. Panels (A)–(H) correspond to F-, G-, K-, and M-type stars, with two panels for each spectral type. The left panels show flare-count histograms, separating confined flares (blue) and eruptive flares (orange), with the corresponding eruptive fractions indicated in the legends. The right panels display the eruptive-flare fraction as a function of flare energy, together with a linear-fit trend (orange dashed line). Panels (I) and (J) show the combined results for all FGKM-type stars. Overall, the eruptive-flare fraction decreases with increasing flare energy.
• Figure 5: Distribution of the model-predicted eruptive-flare fraction as a function of flare equivalent duration (ED) for FGKM-type main-sequence stars. Panels (A)–(H) correspond to F-, G-, K-, and M-type stars, with two panels for each spectral type. The left panels show flare-count histograms, separating confined flares (blue) and eruptive flares (orange), with the corresponding eruptive fractions indicated in the legends. The right panels display the eruptive-flare fraction as a function of flare ED, together with a linear-fit trend (orange dashed line). Panels (I) and (J) show the combined results for all FGKM-type stars. The eruptive-flare fraction decreases with increasing flare ED.