Evidence for Sympathetic Flaring in TESS Data

Abstract

Most flares on the Sun occur at random, but there is a small percentage of "sympathetic flaring" -- the triggering of one flare by another. Previously there had been no widespread confirmation of sympathetic flares on other stars. In this work, we developed a new flare detection algorithm that is sensitive to closely-separated and overlapping stellar flares. We applied it to TESS data and discovered ~ 220,000 flares on ~ 16,000 stars, the majority of which are M-dwarfs. The wait time distribution between flares demonstrates an excess of closely-separated flares, relative to expectations from a Poisson process. We attribute this to sympathetic flares, occurring at a rate of between 4% and 9%, which matches the rate seen on the Sun. Our result is the first statistically robust detection of sympathetic flares on other stars, demonstrating a commonality between the Sun and low-mass stars.

Figures (15)

• Figure 1: Normalized distribution of the temperatures of the stars across our three samples. The temperatures were taken from the TESS Input Catalog v8.2 Paegert2022. The ranges of each stellar subtype are shown via the dashed black lines.
• Figure 2: Detrending and flare search process from raw lightcurve (top) to final flattened curve (bottom). Top shows the raw time resolved brightness of the star in black points with the quadratic trend coming from the orbit of the TESS telescope in blue. Middle shows the quadratic subtracted lightcurve overlaid with the trend found by wotan in red. The lightcurve shown in bottom is the final flattened lightcurve in black and the detected flares. Primary flares are colored in red with the peaks labeled as red stars and the secondaries labeled in blue with a blue star representing its peak. We also add an inset zoom-in around the secondary flare to show its morphology. Residual signal of the spot modulation is seen in the final lightcurve shown in bottom which leads us to consider the global spread of the points to calculate the flux threshold as opposed to the photometric error. We also note that cutting 100 cadences on either side of a break prevents classifying detrending atrifacts as flares as seen in the orbit break of this lightcurve.
• Figure 3: Demonstration of flare detection and modeling of toffee used to find secondary flaring events. Left demonstrates an example of a secondary flare event being found at an earlier time compared to the primary coming from TIC 295777692 Sector 10, and right demonstrates an example of a secondary flare being found at a later time compared to the secondary coming from TIC 100481123 Sector 5. Non-flare flux points are colored in black with a gray dashed horizontal line representing the $3\sigma$ threshold of the global spread used to find flares. The large primary flares with points identified as being associated with the flare are colored in red and points associated with the secondary are colored in blue with the peaks labeled with stars. The black line traces the best fits of the Gaussian rise and the double exponential decay models used to model the rise and the decay of the flare, respectively. The inset figures on each figure show residuals which are used by toffee to identify a secondary flare by finding the points above the $2\sigma$ threshold shown in a black dashed line. Left shows the residuals to the primary flare model in the rise portion of its respective flare. We note the Gaussian rise fitting procedure does not require the function to pass through the first point in the flare in order for secondary flare detections to be robust. Right shows a portion of the residuals in the decay portion of a flare around the detection of a secondary flare.
• Figure 4: Example of refitting decay portion of primary flare to find secondary. The solid line shows the first attempted fit using all flux points in the decay. The three blue points represent the flux points labeled as being marginally missed as a flare with residuals $\geq 1.5\sigma$. The dashed line shows the refitted double exponential decay after removing the blue points lessening their effect on the fitting. After refitting the new residuals of the three points lie above the $2\sigma$ threshold for recognition as a secondary and is thus added to the flare catalog. This shows the general trend that secondary flares recovered via a second fitting are generally low amplitude flares very close to $2\sigma$ signals.
• Figure 5: Cumulative amplitude counts of flares in the Feinstein (green), Yudovich (pink), and Seli (blue) samples. Flares amplitudes are expressed in terms of the spread of the lightcurve, $\sigma$, $A_{\sigma}$, or higher expressed in logspace. Deviations from a straight line in log-log space (a power-law) are emblematic of incompleteness at low amplitude (left of the vertical dashed line) or small number statistics at high amplitude.