The Largest M Dwarf Flares from ASAS-SN

We obtained follow-up VRI photometry for the majority of our sources, both to aid in source classification (combined with survey photometry) and to provide quiescent V-band measurements to compare to the V-band flare detections. Typically, the photometry was obtained when acquiring sources for spectroscopic observations. We obtained optical imaging using the Inamori-Magellan Areal Camera and Spectrograph (IMACS; Dressler et al. 2011) on the Baade-Magellan 6.5 m telescope, the Ohio State Multi-Object Spectrograph (OSMOS; Martini et al. 2011) on the MDM Observatory Hiltner 2.4 m telescope, and the Wide Field Reimaging CCD Camera (WFCCD) on the Irénée du Pont 100-inch Telescope. All data were reduced with standard routines in the IRAF ccdred package. We performed aperture photometry on the reduced images using the IRAF apphot package. The data were calibrated using SDSS (York et al. 2000) Data Release 7 (Abazajian et al. 2009) data when available and using the AAVSO Photometric All-Sky Survey (APASS; Henden & Munari 2014) when SDSS was unavailable. When using SDSS, the Sloan filters were transformed onto the Johnson–Cousins magnitude system using transformations presented by Robert Lupton.¹⁶ The resulting photometry is presented in Table 1.

We initially identified flares based on quick photometry from ASAS-SN images. To examine the light curves in more detail, we performed aperture photometry at the location of each flare as described in Kochanek et al. (2017). Briefly, the photometry is done using the IRAF apphot package and calibrated to APASS (Henden & Munari 2014). However, for ASASSN-13be there is a nearby blended star, and so we performed aperture photometry on subtracted images, calibrated to APASS, and then added back in the quiescent flux as measured from our higher-resolution OSMOS images. We also examined the ASAS-SN light curve for each source to look for additional flares and to characterize the flares detected by ASAS-SN (see Section 4). In addition to the original flares that triggered an ASAS-SN classification, we detected six additional significant (${\rm{\Delta }}V\lt -1.0$) flares on five stars, which we include in our analysis. The photometry of each flare, along with 3σ upper limits for images within ∼24 hr of each flare event, is presented in Table 3.

To confirm the stellar nature of the M dwarf candidates and to characterize them in terms of spectral type and quiescent chromospheric activity (through the Hα emission line), we obtained low-resolution (R ∼ 2000–4000) optical spectra from the du Pont, MDM, Baade-Magellan, and LBT telescopes. The instruments and telescopes used for each target are listed in Table 2. We adopted the spectra for ASASSN-13cb and ASASSN-16ae described by Schmidt et al. (2014) and Schmidt et al. (2016), respectively.

We obtained optical spectroscopy for bright southern sources on the Irénée du Pont telescope at the Las Campanas Observatory (LCO). We used the Wide Field CCD in long-slit grism mode yielding a wavelength range of 370–910 nm and a resolution of 8 Å. We obtained optical spectroscopy for faint southern sources using IMACS (Dressler et al. 2011) on the Baade-Magellan 6.5 m telescope. We observed candidates with the F/2 camera, 07–09 slits, 300 line mm⁻¹ grism at a blaze angle of 175, yielding a wavelength range of 390–800 nm and a dispersion of ∼1.3 Å pixel⁻¹. Optical spectroscopy for bright northern sources was obtained using OSMOS (Martini et al. 2011) on the MDM Observatory Hiltner 2.4 m telescope. We observed candidates with 09–12 slits and VPH grism, yielding a wavelength range of 390–680 nm and R ∼ 1600.

We reduced the spectra from Magellan, du Pont, and MDM using standard IRAF routines, supplemented by cosmic-ray removal using the LA cosmic detection algorithm (van Dokkum 2001). Wavelength calibration for Magellan and DuPont used arc lamps taken at the position of the observations and resulted in solutions good to 0.1–0.3 pixels. For MDM, arc lamps were typically taken once per night, resulting in wavelength solutions good to 0.5 pixels with 1–4 pixel offsets that we corrected using the Hα emission line. No radial velocity standards were taken with any telescope, so measuring velocities from the spectra was not attempted. For all three telescopes, we performed flux calibration using observations of a single flux standard per night.

We obtained spectra for the faintest northern sources with the Multi-Object Double Spectrograph 1 (MODS1; Pogge et al. 2010) mounted on the dual 8.4 m Large Binocular Telescope (LBT) on Mt. Graham. The MODS data were reduced with a combination of the modsccdred¹⁷ python package and the modsidl pipeline.¹⁸ The reductions included bias subtraction, flat-fielding, 1D spectral extraction, wavelength calibration using an arc lamp, and flux calibration using a spectroscopic standard taken the same night.

We assigned optical spectral types to each star based on comparison to the Bochanski et al. (2007) SDSS spectroscopic templates, deconvolved to the resolution of each observed spectrum. The majority of spectra covered the wavelength range 5500–9000 Å, and spectral types were selected based on matching spectral slope, molecular bands, and atomic features over that entire spectral range. The MDM spectra covered 4600–6800 Å, and due to inconsistent normalization and low flux on the blue end, spectral types were assigned based on 5500–6800 Å. Of the 53 stars, 47 were classified as M dwarfs with types M2–M9, as listed in Table 1, along with one K5 dwarf (ASASSN-14he) and one L0 dwarf (ASASSN-16ae; Schmidt et al. 2016). Spectra for all M and L dwarfs are shown compared to their best-fit templates in Figure 1. Spectral classifications for the stars are given in Table 1, and photometry for the other objects is given in Table 4. We include the L0 dwarf in our statistics and results because its chromospheric activity is not expected to be distinct from that of the 10 M7–M9 dwarfs included in the sample. We exclude the K5 dwarf because it is significantly more massive than the rest of the sample (types M2 and later).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

View figure set (48 panels)

Tarball of figure set images

In addition to these ASAS-SN flare stars, we also observed one previously identified T Tauri star (see Section 6.3) and one newly identified young star (see Section 6.5). For one faint object, ASASSN-14aa, the spectrum of a V = 23.4 mag source close to the ASAS-SN position is not stellar, and upon further examination the real source of the outburst is likely to be a V = 25.2 mag point source nearby. Follow-up Magellan photometric observations of another faint source, ASASSN-14be, was unable to identify any source consistent with the ASAS-SN position. We place 5σ limits of V > 25.6 mag, R > 24.9 mag, and I > 23.6 mag at the position of ASASSN-14be, making its counterpart significantly too faint to obtain a spectrum of sufficient quality to provide a spectral type. Because the stellar photometry meets color selection criteria for M dwarfs, ASASSN-14aa and ASASSN-14be are likely to be ΔV < −10 mag flares on distant M dwarfs. We do not include them in this analysis owing to the lack of spectroscopic observations.

The survey strategy of ASAS-SN—acquiring two or three 90 s dithers on a field before moving on to another field—does not allow for a detailed analysis of flare light curves, but the data are sufficient to place accurate lower limits on the V-band flare energy. Given a single detection that indicates a flare-type brightening, the lowest-energy flare that could correspond to that observation is a simple, classical flare with the peak of the flare occurring during that single observed point. We calculate the flare energy based on the fit of a simple, classical flare to the data and refer to it as a lower limit because other light curves that fit the points have a higher energy (see Section 4.4.1).

For the simple, classical flare, we draw on the Davenport et al. (2014b) empirical flare template, which parameterizes the flare as a fast rise phase

$$\begin{eqnarray}&&\displaystyle \frac{{F}_{\mathrm{rise}}}{{F}_{\mathrm{amp}}}=1+1.941{t}_{1/2}-0.175{t}_{1/2}^{2}-2.246{t}_{1/2}^{3}-1.125{t}_{1/2}^{4}\end{eqnarray} \tag{ 1 }$$

and a two-component exponential decay phase

$$\begin{eqnarray}&&\displaystyle \frac{{F}_{\mathrm{decay}}}{{F}_{\mathrm{amp}}}=0.689{e}^{-1.600{t}_{1/2}}+0.303{e}^{-0.278{t}_{1/2}}.\end{eqnarray} \tag{ 2 }$$

The flare is entirely characterized by the flare amplitude (F_amp) and the half-light decay timescale (t_1/2). Davenport et al. (2014b) created this flare template from the co-added light curves of the 885 simple, classical flares found in Kepler data for M4 GJ 1243, the most active M dwarf in the original Kepler field. In addition to F_amp and t_1/2, the agreement between the flare template and the flare observations depends sensitively on the position of the template peak relative to the integration times of the observed data point. Because we are calculating the lower limit on the V-band flare energy, we assume that the actual flare peak occurred during the ASAS-SN observations, and let the exact position of the template peak vary between the times of the first and last observed data points.

To estimate the best fit of the flare template to the observations, we first generated 400 flare light curves with t_1/2 varying from 10 to 2000 s in increments of 5 s. We integrated the flare light curve over the exposure time for each data point (90 s), varying the position of the peak from the beginning of the first exposure to the end of the last exposure in increments of 1 s. For each combination of peak position and t_1/2 we compared the integrated fluxes to the data points, accepting the parameters where the normalized integrated light curve fell within the 1σ uncertainties of the observed fluxes.

The parameter space where the flare template and observations are in agreement is not distributed normally with respect to t_1/2 or peak position, so we adopt the median values as the best-fit result and the interquartile range as the asymmetric uncertainties. We obtain the V-band energy released from the best-fit flare template by integrating over total flare time and multiplying by the distance factor (d²) and V-band central wavelength (5500 Å). The resulting V-band energy is then a lower limit of the observed ASAS-SN flare. These lower limits on the V-band energy range from $\mathrm{log}({E}_{V}/\mathrm{erg})=32$ to 35 and are shown as a function of spectral type in Figure 2 and listed in Table 6.

Zoom In Zoom Out Reset image size

We examined the observations and the range of best-fit template curves for each flare to classify each set of data points as occurring during the rise or decay phase of each flare. We found that of the 54 flares with more than one photometric observation, we observed the rise phase 16 times, observed the decay phase 37 times, and caught the peak in the middle of three observations for one flare. Over the course of an entire flare, the decay phase is roughly 12 times longer than the rise phase (Davenport et al. 2014b). It is unlikely that we are observing the end of the decay phase, however, since flares are much fainter then. The ratio of decay phase to rise phase observed for the ASAS-SN sample (2.31) corresponds to the ratio of decay phase to rise phase for the flare template if it is only visible at 52% or more of its peak value. Hence, the balance of observations occurring during rise versus decay is a good indicator that we are likely observing most flares close to their peak values.

The timescales for each flare are determined from one or two points. To test whether the resulting flare timescales are reasonable, we compare to the correlations between flare energy and flare length found in high-cadence M dwarf observations from Kepler data (Hawley et al. 2014; Silverberg et al. 2016). They derive two linear relationships, one for classical flares and one for multipeaked complex flares. Each relationship has an expected error of a few percent, but the data span an order of magnitude above and below the best-fit line. We convert the relationship between flare duration and Kepler-band energy to t_1/2 and V-band energy assuming t_1/2 = 0.14 × t_duration (Davenport 2016) and a flare spectral energy distribution of a 10,000 K blackbody (see, e.g., Kowalski et al. 2013) to convert energies between the two bands (${E}_{{Kepler}}=1.9{E}_{V}$). Figure 3 shows the relationship between timescale and flare energy compared to the timescales and energies derived from the fitting procedure described in the previous section. We also calculate energies for each flare based on that relationship; because flare energy and t_1/2 are dependent on each other, we calculate them iteratively until the change in the log of energy is less than 0.01 between steps. The final values (given in Table 6) fall within an order of magnitude of those based on the fitting procedure and are on average a few percent stronger.

Zoom In Zoom Out Reset image size

Compared to the relationship from Kepler data, approximately half of the flares have timescales and energies that are in the expected range based on the Kepler flare data, but the remaining half have shorter timescales than expected for flares with their peak flux. The majority of the ASAS-SN flares are higher energy than those observed in Kepler data of GJ 1243, however, so the relationship between duration and energy has been extrapolated to higher energies than it may be applicable. In a sample of ∼20 well-characterized flares, Kowalski et al. (2013) found that larger flares were more impulsive (shorter t_1/2) than smaller flares, indicating that the relationship between duration and energy may be shallower at higher flare energies. The relationship between half-light timescale (t_1/2) and total duration is also based on a single flare peak and so does not hold for complex flares that have multiple peaks. Complex flares would have shorter measured t_1/2 than the expected values for simple flares. Each of these factors leads to an underestimate of the total flare energy, so the V-band energies derived from the light-curve fits are conservative as lower limits.

Three of the ASAS-SN flare M dwarfs had additional observations from the same flare separated by more than a few minutes, and we examine these to better understand the limitations of our fit. For ASASAN-13bg and ASASSN-14jy, this occurred because the stars are positioned on the overlap strip between two different ASAS-SN fields, resulting in two sets of observations separated by ∼15 minutes. For ASASSN-13cb, the additional observations were an effort to obtain follow-up data for this dramatic transient event (see Schmidt et al. 2014). During our fitting procedure, we excluded points farther than 1000 s from the peak to focus on the data from a single telescope pointing, so the fits for those three objects are based on a subset of the available data. The light curves and fits for these objects are shown in Figure 4.

Zoom In Zoom Out Reset image size

The best fit to ASASSN-13bg is based on two observations at the end of the rise phase of the flare, and there are two additional observations taken before the peak of the flare that fall above the model. To fit the later observations would require a longer t_1/2 than the model shown, but the current best-fit parameters already have a long half-light timescale (t_1/2 = 1373 s) compared to the V-band energy ($\mathrm{log}({E}_{V}/\mathrm{erg})=32.7$). Because they fall so far from the best fit to the peak observations, it is more likely that the points excluded from the fit are part of a different peak in the same flare. While the precursor peak could be very strong, a minimum fit to it would also be nearly an order of magnitude less energetic than the main flare and only contribute an additional few percent to the total energy.

The best fit to ASASSN-13cb is based only on the two observations near the peak, not including two additional observations taken over 2 hr later. This resulted in a relatively short half-light timescale (t_1/2 = 270 s) and a V-band energy ($\mathrm{log}({E}_{V}/\mathrm{erg})=32.7$) much lower than the original estimate from Schmidt et al. (2014), which included all points but did not have an empirical model for the flare. The timescale calculated from the Kepler relationship between energy and timescale (t_1/2 = 2563 s) is an excellent fit to the later data points, however, and corresponds to a much higher energy ($\mathrm{log}({E}_{V}/\mathrm{erg})=34.4$). That timescale is a poor fit for the steep drop between the first two data points.

The match between the best-fit flare template and the observations is much better for ASASSN-14jy, with the points not included in the initial fit falling along the curve representing the lower quartile in uncertainty. This could indicate that the simple flare model is a good fit to at least some of the flares, but additional complexity could have occurred between the detected points.

Given that none of the data points for any of these three flares fall significantly below the best-fit lines, it is reasonable to continue to assume that the fits represent solid lower limits on the V-band flare energies.

While sparse observations can place a strong lower limit on flare energies, the upper limits of the observed flares are less clear. We examine possible upper limits from two different perspectives. First, we examine the variations in flare shape that would also fit the data in the context of previous work. Second, we use FFDs to estimate the probability of more luminous flares.

4.4.1. Constraints on the Flare Shape

In our energy estimates, we used the highest photometric observation as the peak of a classical flare template. We argue that these assumptions are both more likely to lead to a lower estimate of the flare energy and so are valid lower limits. In Figure 5, we use ASASSN-14gj as a reference to demonstrate some possible higher-energy templates. One of the simplest higher-energy flare templates is a classical shape with an earlier peak at a higher flux. We show a flare with a peak at 10 times the observed peak flux, with a decay timescale of 107 s and a total energy of E_V = 5.8 × 10³³ erg. The timescale is very short for this template, due to the strong difference between the first two points, and would fall well below the relationship shown in Figure 3. The combination of timescale and energy is unlikely to be physically possible because the volume of hot plasma needed to produce the peak luminosity cannot cool that quickly. Thus, assuming a classical flare template and the observed decay between dithered observations, it is likely that we discovered ASASSN-14gj near its peak. Because the timescales of the majority of ASAS-SN flares are already low compared to their energies, a similar argument can be made for the majority of the flares observed.

Zoom In Zoom Out Reset image size

While it seems likely that most of these flares were observed near their peak, the assumption that all these flares are simple, classical flares is less likely to be true. Davenport et al. (2014b) find that approximately 15% of the flares detected on GJ 1243 are complex, multipeaked events. Hawley et al. (2014) and Silverberg et al. (2016) found that the fraction of complex flares increases with flare energy, with around half of flares above 10³³ erg typically showing complex behavior. The higher incidence of complex events at high energies is, in part, because very energetic flares cover a relatively large area of the stellar surface and are more likely to trigger sympathetic flaring (e.g., Anfinogentov et al. 2013). Because the ASAS-SN flares are mostly higher energy than 10³³ erg, nearly half of the 55 flares are probably complex and so would be poorly fit by the empirical classical flare template. Figure 5 shows two of many possibilities for complex, multipeaked events, drawing inspiration from previously observed M dwarf flares (Hawley & Pettersen 1991; Kowalski et al. 2013). It is not currently possible to estimate the presence or strength of multiple peaks based on sparse observations near the peak, but since the addition of more peaks increases the energy, the classical flare assumption continues to be a conservative lower limit.

4.4.2. The Frequency of Larger Flares

The energies of flares and distribution of spectral types are essentially a scatter plot, with the shape of the distribution being set primarily by the underlying stellar population and the observational constraints rather than a fundamental property of flares.

One limit on our detection of flares in early M dwarfs is the contrast effect; on a dim, late M dwarf, an energetic flare will have a much higher contrast with the underlying photosphere than it would on a bright, early M dwarf. To show this effect, we calculated a characteristic energy for a 1 mag flare (ΔV = −1) as a function of spectral type. To convert peak flux to energy, we used the relationship between flare peak and flare energy discussed in Section 4.4.1. That limit (shown in Figure 2) falls below the majority of the flares for dwarfs earlier than M5. For dwarfs later than M5, flares smaller than $\mathrm{log}({E}_{V}/\mathrm{erg})=32$ would still have a contrast of greater than 1 mag. Due to the current classification methods of ASAS-SN, a flare of less than 1 mag is unlikely to have made it on our list.

We also examined the effect of the faint detection limit of ASAS-SN on the range of energies detected. The detection limit varies from night to night but is typically above V = 16 (and can be as faint as V = 17.5). At a distance of 100 pc, that limit corresponds to a flare energy of $\mathrm{log}({E}_{V}/\mathrm{erg})=33$, and there are fewer flares than expected that fall below that energy. Those flares are likely to have occurred during good conditions or on more nearby stars.

To understand the distribution of spectral types, we calculated the distribution expected if the 49 M dwarfs were drawn evenly from the solar neighborhood (shown in Figure 2; based on the mass functions of Cruz et al. 2007; Bochanski et al. 2010). Overall, there are fewer M3 and earlier dwarfs than expected and more later types. This is reasonable, given that flares of the same energy are more difficult to detect on these bright M dwarfs.

Kinematics are often used as age indicators for populations of stars; young stars typically have cold orbits characterized by low velocities with respect to the Sun and positions near the Galactic plane, while older stars have warmer orbits and are more likely to have higher velocities and be located further from the Galactic plane. The majority (45 of 49 objects) of ASAS-SN flare stars are located within 200 pc of the Galactic plane, with only two M dwarfs (ASASSN-13dj and ASASSN-15ll) found at Galactic heights greater than 400 pc (in this case, both below the plane).

We adopted Gaia proper motions for the 42 stars with matches in the Gaia database. For five of the seven remaining stars (excluding ASASSN-13dj and ASASSN-15ll owing to the lack of 2MASS detections) we calculated proper motions based on the difference between 2MASS and WISE coordinates. The Gaia and 2MASS/WISE proper motions are given in Table 5. We combined these proper motions with distances to calculate tangential velocities (V_tan) for 47 dwarfs, which are also given in Table 5. The overall mean tangential velocity is 29.8 km s⁻¹, with a standard deviation of 22.6 km s⁻¹. There is a single M dwarf with a velocity larger than 100 km s⁻¹ (ASASSN-13be with V_tan = 117.5 km s⁻¹). If this outlier is removed, the mean tangential velocity is 27.9 km s⁻¹, with a standard deviation of 18.7 km s⁻¹.

Comparing these to the data from Faherty et al. (2009) indicates that the M dwarfs in the ASAS-SN M dwarf sample are slightly younger than M dwarfs in the solar neighborhood, implying ages younger than ∼5 Gyr and likely membership in the thin-disk population. But with a few dwarfs at high Galactic heights and fast V_tan, it is not likely that they are a universally young population, nor are they significantly younger than the Galactic disk.

As a manifestation of strong magnetic fields, the flare rate of M dwarfs is frequently correlated with other indicators of magnetic activity (Kowalski et al. 2009; Hawley et al. 2014). Our spectroscopic observations include the region surrounding the Hα emission line, frequently used as an indicator of activity in these low-mass stars (Gizis et al. 2000; West et al. 2008). We measured Hα equivalent widths (EW) using a custom IDL code that adopts the wavelength ranges and method described by West et al. (2011). Of our 49 stars, only three (K dwarf ASASSN-14hc and two M dwarfs) had no detectable emission in the Hα line. Both of the M dwarfs had spectra with sufficiently high signal-to-noise ratio that an emission line with an EW of ∼0.75 Å would be detectable; they either have Hα in relatively weak emission, weak absorption, or an absence of the activity indicator.

The activity fraction as a function of spectral type is shown in Figure 6 compared to the activity fraction of a reference sample of field M and L dwarfs that was selected based on color and not activity level (the SDSS DR7 M dwarf sample and the BOSS Ultracool Dwarfs sample; West et al. 2011; Schmidt et al. 2015). The two M dwarfs without detected activity (13dj and 15ll) have spectral types of M2 and M4; at those spectral types, the mean activity fractions (∼5% and ∼15%, respectively) are relatively low, so the overall activity fraction of the ASAS-SN flare sample still falls well above the active fraction for those spectral types. These two dwarfs also have two of the strongest flares and are the two most distant M dwarfs. Given their strong flares, it is surprising that they are not Hα emitters, but it is possible that they are slightly older M dwarfs that have a lower base level of activity and flare much more rarely.

Zoom In Zoom Out Reset image size

The Hα line strength can also be used to quantify the activity level of M dwarfs. EWs measure the strength of the line compared to the surrounding (pseudo-)continuum, but because the continuum varies with spectral type, we instead use the ratio of line luminosity to bolometric luminosity. We use the χ factor as a function of spectral type (West & Hawley 2008; Schmidt et al. 2014) to convert the EW to the ratio of Hα emission line flux to the stellar bolometric luminosity. Figure 7 shows the activity strength of the flaring ASAS-SN M dwarfs compared to that of the non-activity-selected reference sample (West et al. 2011; Schmidt et al. 2015). We also show a parameterized activity strength, calculated by subtracting the median per spectral type from every value and then dividing by the difference between the 30% and 70% values. This parameterized activity strength is meant to aid comparisons of the ASAS-SN flare sample to the non-activity-selected sample.

Zoom In Zoom Out Reset image size

The majority (42 our of 48) of the M and L dwarfs fall at activity levels above the median value for their spectral types. Those that have less than the median activity level include the two inactive dwarfs, as well as a few more M4 dwarfs and one M6 dwarf. There are 27 M dwarfs with activity strength that falls above the 90th percentile values for their spectral type. This indicates that the ASAS-SN M dwarf flare sample is a significantly more active population than that of the field stars, consistent with previous results showing that stars without Hα emission flare less often than those with (Hawley et al. 2014).

GJ 3039 (LP 525-39) was first identified as an Hα-emitting, high-proper-motion source (Stephenson 1986a, 1986b) before it was included in the third catalog of nearby stars (Gliese & Jahreiß 1991) and assigned an M4 spectral type (Reid et al. 1995). McCarthy et al. (2001) found a companion to GJ 3039 with a separation of 073 (∼30 au) and similar colors but a magnitude difference of 0.7 in J band. Schlieder et al. (2012) identified GL 3039 as a candidate member of the β Pictorus young moving group owing to a combination of kinematics-detected emission at X-ray in UV wavelengths. Previous photometric monitoring of GJ 3039 (Messina et al. 2016) identified two small R-band flares. The ASAS-SN light curve of GJ 3039 also includes a few small (ΔV < −0.2) flares, but none as large as the ΔV = −1.0 flare examined here.

SDSS 1022+19 was observed to flare in Fall 2013. Because the flare both was very large and occurred on a very red object, follow-up observations were obtained rapidly, and the object and its flare were analyzed by Schmidt et al. (2014), who classified it as a field age M8 dwarf with a lower-limit flare energy of log(E_V) = 34.4 [log(erg)]. We reanalyze the flare here using an updated fitting procedure and the Davenport et al. (2014b) flare template (unavailable during the original analysis) and find an energy of log(E_V) = 33.7 [log(erg)]. As discussed in Section 4, this fit does not show strong agreement with the late-stage (∼2 hr later) decay observed for ASASSN-13cb but does show agreement with the flare light curve calculated from the Kepler relationship between timescale and energy.

ASASSN-15af was identified as T Tauri star 2MASS J05181685−0537300 (2M 0518−05) as part of a photometric Hα survey in Orion (Wiramihardja et al. 1991). A spectrum from Lee & Chen (2007) revealed many emission lines, including Hα (EW = −57.3 Å), O i λ6300, Fe λ4924, Ca H and K, and He I (λ5876), sufficient to classify 2M 0518−05 as a T Tauri star, but there was no discussion of the continuum or spectral type. It is marked as a likely T Tauri star in the ASAS-SN transient page but was placed on our observing list as a result of its red colors. We obtained three spectra of 2M 0518−05 on 2016 August 2 and 2017 February 23, shown in Figure 8.

Zoom In Zoom Out Reset image size

The spectrum of 2M 0518−05 is shown compared to an M6 template spectrum; while the blue continuum is not consistent with an M6 dwarf, there are molecular bands throughout the redder portion of the spectrum (7000–9000 Å) that are well matched by the template spectrum. The subtracted spectrum shows less evidence of these bands and appears to primarily be composed of blue continuum and emission lines. While the combination of emission lines and blue continuum would be expected in a magnetically triggered flare, the line profiles are wider than those typically found during a flare event, most notably the Hα emission, which is ∼50 Å wide and double peaked. It is more likely that the observed spectra are elevated owing to accretion and other star formation processes.

The ASAS-SN light curve of 2M 0518−05 shows at least 14 distinct outburst events, including the initial 2015 event that triggered the observations (all shown in Figure 9). We obtained two additional photometric points for the object, obtaining V = 17.2 mag both times. This is likely to be the quiescent value, as ASAS-SN detects 2M 0518−05 in quiescence with a similar magnitude when the observing conditions are particularly good. While some of the outburst events were only detected on a single day and could be consistent with either magnetically induced flares or accretion, others last for multiple days at a time and so are far more likely to be accretion events. In addition to this activity, 2M 0518−05 also shows strong variation in the CRTS database and has excess UV emission (Sanchez et al. 2014). Because any magnetic activity on this object is difficult to separate from accretion, we exclude 2M 0518−05 from the sample of flaring M dwarfs.

Zoom In Zoom Out Reset image size

SDSS 0533+00 was observed to flare in 2016 January. Because the flare was both very large and occurred on a very red object, follow-up observations were obtained rapidly. The object and its flare were analyzed by Schmidt et al. (2016), who classified it as a field age L0 dwarf with a minimum V-band flare energy of $\mathrm{log}({E}_{V}/\mathrm{erg})=33.7$. The updated fitting method used in this work finds a longer t_1/2, resulting in the slightly larger energy of $\mathrm{log}({E}_{V}/\mathrm{erg})=33.9$.

ASASSN-16hl (2MASS J16045515–7223199) was marked in the ASAS-SN transients database as a CV or an outburst from a red star and was selected for follow-up owing to its particularly red colors in 2MASS and WISE. The spectrum of ASASSN-16hl (shown in Figure 10) has molecular absorption similar to an M1 dwarf but has a blue continuum and emission lines similar to a flare or other hot outburst. No other spectra were obtained, so it is unclear whether the single spectrum represents a continuous or transient state.

Zoom In Zoom Out Reset image size

The light curve of ASASSN-16hl shows multiple strong outbursts, including a strong event that resembles a flare (in part due to the lack of observations within a few days), two multiday outbursts, and a few smaller, longer outbursts (shown in Figure 11). Our follow-up photometry yielded a magnitude of V = 14, which is likely to have been during another outburst. The median ASAS-SN value is V = 16.1 mag, which is possibly the quiescent value. In further support that this is a peculiar, young object, the WISE database flagged the photometry as variable, and a SIMBAD search reveals a possibly associated X-ray source in ROSAT within a few arcseconds. ASASSN-16hl is most likely a previously unclassified young star. We used the Banyan Σ tool¹⁹ (Gagné et al. 2018) to check whether it is a member of any young moving group or association and found that it is most likely a field star.

Zoom In Zoom Out Reset image size