Stellar Flares in the TESS Light Curves of Planet-hosting M dwarfs

TL;DR

This study leverages TESS 2-minute cadence light curves to quantify optical flaring on M dwarfs across a volume-limited 15 pc sample and JWST-targeted planet hosts. Using AltaiPony, it detects tens of thousands of flares and computes cumulative flare frequency distributions, fitting power-law relationships with intercept $\beta$ and slope $\alpha$ for stars at varying activity levels. The key finding is that the power-law exponent $\alpha$ remains near $\sim 2$ across low- and high-activity regimes, suggesting a largely uniform flare-production mechanism among M dwarfs. The results provide critical context for JWST transmission spectroscopy analyses by characterizing stellar contamination risks and informing target selection and observational strategies. Overall, the work connects stellar magnetic activity to exoplanet atmosphere studies, highlighting the need to account for flares in interpreting high-precision transit spectra.”

Abstract

M dwarfs are magnetically active stars that frequently produce flares, which have implications for both stellar evolution and exoplanet studies. Flare occurrence rates and activity levels of M dwarfs correlate with stellar characteristics such as age, mass, and rotation period. We search TESS observations of a known active population of M dwarfs as well as a volume-limited sample of M dwarfs within 15 parsecs. We detect flares in the light curves of these stars, including 276 of 538 M dwarfs within 15 pc, and calculate cumulative flare frequency distributions (FFDs) for each star. Based on flaring behavior, we categorize stars into relatively higher and lower activity groups and fit power laws to their FFDs to compare the power law exponent ($α$) across activity levels. We find $α=1.99 \pm 0.07$ for the combined FFD of the lower activity M dwarfs, compared to averages of $α= 1.94 \pm 0.58$ for highly active stars with 10-100 detected flares, and $α= 2.03 \pm 0.43$ for those with > 100 detected flares, suggesting little evolution in the power law distribution of flares as M dwarfs transition from high to low activity states. The uncertainties for the active star groups reflect the standard deviation of $α$ values across individual stars within each subset. Because stellar flares and associated stellar activity complicate exoplanet observations, we also examine the subset of M dwarfs with JWST transmission spectroscopy follow-up observations in Cycles 1-3. The flares we detect for these targets are consistent with the broader 15 pc sample, providing context for interpreting planetary atmosphere retrievals from JWST spectra.

Figures (8)

• Figure 1: The recovery probability (successful AltaiPony detection) of injected flares in the TESS light curves of GJ 1132 (top left), TOI-2445 (top right), TRAPPIST-1 (bottom left) and the average of recovered flare durations (in minutes) and flare SNR for all three stars (bottom right). From tests on approximately 6500 injected flares per star, these heatmaps show the probability that AltaiPony detects a flare of a given duration and amplitude relative to the light curve noise. The color of each bin represents the recovery probability from 0 to 1, with black contours indicating the approximate curves of of 25%, 50%, 75% and 90% recovery probability. Although the noise levels within each stars' light curve varies greatly, we see consistent recovery probability of flares.
• Figure 2: Clustering of stellar flare activity. Left panel: Each point represents one star, plotted in log‑log space of mean equivalent duration (ED) in seconds, versus cumulative flare frequency ($\nu$/d). Means are computed after excluding the lowest 20% of flares by ED to reduce the effects of low-energy flare detection incompleteness. A two‑component Gaussian Mixture Model (GMM) separates the stars into two clusters: lower activity (blue) and highly active (gold). Right panel: Full flare frequency distributions (FFDs) for each star, colored by their GMM-assigned label.
• Figure 3: Flare frequency distributions for all M dwarf targets from detected flares in their SPOC light curves. In both panels, the blue and gold contours represent the flares from the combined Gunther_2020 and 15 pc samples, split by relative activity, based on the GMM clustering shown in Figure \ref{['fig:PLsplit']}. The density plots on the outer edges represent the normalized distributions for flare equivalent duration in seconds (ED) on the x-axes, and cumulative flare frequency on the y-axes. Each scatter point represents a flare we detect for the different JWST targets which are differentiated shape and color. The left panel contains our computed FFDs of 11 M dwarf transmission spectroscopy targets from JWST GO cycles 1 and 2 and the right panel shows FFDs we computed for 4 M dwarf targets in Cycle 3. The Cycle 1 and 2 targets largely fall in the lower activity regions aside from TOI-2445 and TRAPPIST-1, whereas the Cycle 3 targets largely fall in the highly active region with the exception of GJ 1151. Note that TRAPPIST-1 has been observed in each JWST Cycle, which is why it is included in both panels. We expect increased stellar contamination in planet atmosphere transmission spectra obtained from more active stars.
• Figure 4: The power law fit to the flares of the lower activity stars in our M dwarf samples (170 stars), indicated by the blue points in Figure \ref{['fig:PLsplit']}. We treat these stars as one distribution (containing 858 flares) and use the multiple_star argument of the AltaiPony FFD module to average each portion by the number of stars that contribute to it. We employ edmcmc to find optimal power law parameters $\alpha$ and $\beta$, with the posteriors of the algorithm's search shown in the left panel. The right panel shows the flare frequency distribution in blue with the power law (Equation \ref{['eqn:pl']}) fit to it using the optimal parameters from the MCMC run in pink. Based on the posteriors, this distribution is best fit with a power law exponent, $\alpha$, of $1.99 \pm 0.031$, and a $\beta$ parameter of $0.033 \pm 0.001$. This value for $\alpha$ is in agreement with literature findings for power laws fit to flare frequency distributions in the optical regime. See our discussion in Section \ref{['sec:ffdPLs']}, or Fig. 13 of AltaiPony, Tables 1 & 2 of Aschwanden_2021, and more recent findings for power law exponents fit to M dwarf FFDs Yang_2023Lin_2024 for example.
• Figure 5: Histogram of the power law exponents ($\alpha$ values) fit to 362 M dwarfs from both the Gunther_2020 and Gaia DR3 volume-limited 15 pc samples in the highly active category defined by Figure \ref{['fig:PLsplit']}. The gold distribution represents exponents fit to highly active stars with more than 10 detected flares (362 stars), which has an average value $\alpha _{\mathrm{avg}}$ of $1.81$ and a standard deviation $\pm 0.32$, indicated by the pink vertical line. The vertical blue line represents the $\alpha$ value best fit to the combined lower activity star FFD shown in Figure \ref{['fig:mcmc']}) with the shaded region representing the uncertainty estimated with Equation \ref{['eqn:uncert']}. The averages are consistent with each other and with literature findings for $\alpha$ values of optical flares for M dwarfs.