The largest ground-based catalogue of M-dwarf flares

TL;DR

This work addresses the need for large, homogeneous samples of M-dwarf flares to study magnetic activity and its implications for exoplanet environments. It introduces a multi-stage, ML-powered pipeline trained on flare templates injected into real ZTF light curves, followed by rigorous post-filtering and visual inspection to produce a high-purity catalogue of 1,229 flares. For 655 Gaia-distance-bearing flares, bolometric energies range from $10^{31}$ to $10^{35}$ erg and a power-law relation $E_{ ext{flare}} = 10^{32.045} \ t_{ ext{FWHM}}^{1.306}$ links energy to light-curve width. The results reveal a sharp increase in flare frequency toward later M-dwarfs (peaking near M4–M5) and a decreasing flaring fraction with Galactic height, providing a comprehensive resource and methodology for upcoming wide-field surveys such as LSST to study stellar activity, its evolution, and implications for habitability.

Abstract

We present the largest ground-based catalogue of M-dwarf flares to date, comprising 1,229 time-resolved events identified in Zwicky Transient Facility Data Release 17. Using high-cadence ZTF observations collected between April 2018 and September 2020, we analyzed over 93 million variable light curves containing 4.1 billion photometric measurements. Flare candidates were identified through a machine-learning pipeline trained on simulated light curves generated by injecting TESS-based flare templates into ZTF data and then refined through an extensive post-filtering stage combining an additional classifier, metadata checks, and human inspection. For 655 flares with reliable Gaia-based distances and well-sampled light curves, we derived bolometric energies ranging from 10^31 to 10^35 erg. A clear correlation is observed between flare frequency and spectral subtype, with a sharp increase toward later M dwarfs, particularly near M4-M5, coinciding with the transition to full convection. Using 680 flaring stars with known vertical distances from the Galactic plane, we find that the fraction of flaring stars decreases with increasing Galactic height. The resulting catalogue provides the most comprehensive ground-based sample of M-dwarf flares and establishes a framework for flare detection and classification in upcoming wide-field surveys such as the Vera C. Rubin Observatory Legacy Survey of Space and Time.

Figures (12)

• Figure 1: Distribution of magnitudes in r-passband from ZTF DR17 high-cadence data.
• Figure 2: Flowchart of the flare light-curve simulation process. Original TESS flare templates are converted from flux to magnitude, interpolated, and injected into randomly selected ZTF high-cadence time grids with realistic noise and brightness distributions. Further details of the simulation procedure are provided in Section \ref{['sec:flare-simulation']}.
• Figure 3: Precision Recall curve for validation and test datasets.
• Figure 4: Light curve of a $\delta$-Scuti star misclassified by the classification pipeline. The grey line shows the period-folded light curve of all observations, smoothed with a median filter. The light grey shaded area indicates the average deviation. Red points highlight high-cadence observations only, which were incorrectly classified as a stellar flare by our classification algorithm.
• Figure 5: Representative examples of found stellar flare activity in M dwarf stars. The light curves demonstrate the diverse morphology of flare events, ranging from simple single-peak flares to complex multi-peak structures, with amplitudes varying from subtle brightness increases to dramatic outbursts exceeding several magnitudes.