Disentangling auroral, cloud and magnetic spot driven variability in three early L-dwarfs with HST/WFC3

Abstract

Variability monitoring provides an unparalleled insight into the atmospheric processes of brown dwarfs and directly imaged exo-planets. Inhomogeneous clouds, aurorae and magnetic spots have all been postulated as potential drivers of variability. While objects at the L/T transition have had their variability studied extensively, the variability of early L-dwarfs remains an understudied region of the parameter space. We use observations from the Hubble Space Telescope in the near-infrared, using WFC3/G141 to disentangle the drivers of variability in three known variable early L-dwarfs: 2MASS J1721039+334415, 2MASS J00361617+1821104 and 2MASS J19064801+4011089. We find that all three objects exhibit significant variability at all wavelengths, with white-light amplitudes of 0.53-1.41 %. We find that their colour variations are brighter and bluer compared to later spectral types, except for 2MASSJ19064801+4011089 which exhibits largely grey variations. We report a new period for 2MASS J1721039+334415, of 4.9^{+0.4}_{-0.2} hours. We find evidence of long term light curve stability in each object, which may indicate the presence of long lived features on their surfaces. We create a flexible modelling framework to model three potential drivers of variability: clouds, aurorae and magnetic spots. We fit our models to the spectral variability amplitude from 1.1-1.67 μm of each object. We find that changing cloud properties or magnetic spots are the most likely drivers of variability in each object. Auroral models do not reproduce the variability within the HST wavelengths, however future observations at longer wavelengths that probe higher in the atmosphere may be more sensitive to auroral effects. This work provides a foundation for future variability studies of early L-dwarfs and directly imaged exoplanets to disentangle auroral, cloud and magnetic spot driven variability.

Figures (18)

• Figure 1: Left: The broadband G141 filter light curves of each of our observations. A sample of 100 posterior fits of each light curve are shown in grey, while the best fit light curve is shown in colour for each observation. All four light curves are variable. Right: The periodogram for each light curve. The periodogram for the window function of HST is shown by the dash-dot black line in each plot. The period of each object from the literature is shown by the purple vertical line in the periodogram, while the period output from our celerité2 models are shown by the black vertical lines. The uncertainties for the period values from both the literature and this work are represented by the shaded regions in purple and black respectively. Our model periods agree with the literature for each observation except 2M1721+33, where we estimate a longer period of $4.93^{+0.36}_{-0.22}$ hours.
• Figure 2: Long term stability of 2M1721+33 across two different wavebands. On top is the metchev2015weather Spitzer 3.6 $\upmu$m light curve with both their fit in red and our Gaussian process celerité2 fit in blue applied to the Spitzer and HST light curves independently. Both models fit the data very similarly. On the bottom is the HST white light curve from our work with our Gaussian process celerité2 over-plotted (green). Despite these two light curves being observed over 9 years apart and at different wavelengths, there is a common shape to both light curves. They both exhibit a primary dip, followed by a secondary dip in each period. This may suggest long term stability in 2M1721+33's light curve.
• Figure 3: Top: Phase folded white light curve of 2M1721+33 using the the 2.6$\pm$0.1 hour period reported by metchev2015weather. The photometric points from each HST orbit are colour-coded and the best fit celerité2 model is plotted in black. Using this period for phase folding does not give a clean light curve. Bottom: Phase folded white light curve according to the period of $4.9^{+0.4}_{-0.2}$ hours from this work. While this observation does not cover more than 1.5 rotations, the phase folded regions (orbits 1 and 4, and orbits 2 and 5) overlap with each other in both the HST data and the celerité2 model. This cleaner phase folded light curve, along with Figure \ref{['fig:Spitzer LC']} motivate our reasons to adopt this new period for 2M1721+33.
• Figure 4: Top: The phase folded white light curve of epoch 1 of 2M0036+18, using the period from this work ($3.13^{+0.9}_{-0.8}$ hours). The best fit celerité2 model folds neatly as does the HST white light curve. Bottom: The phase folded white light curve for epoch 2 of 2M0036+18 using the same period. As with the first epoch, both the best fit celerité2 model and the HST white light curve overlap neatly. Despite both epochs covering slightly different phases of the light curve, the same period works well for both.
• Figure 5: Left: Colour modulation of 2M1721+33 (green), 2M0036+18 (blue) and 2M1906+40 (red) compared with the sample from Lew_colour (black circles and lines). The grey circles represent brown dwarfs with known parallaxes across spectral types, taken from the UltracoolSheet ultracool_sheet. The slopes of 2M1721+33 and 2M0036+18 in particular are shallower in comparison to the rest of the sample, showing that as these objects rotate, they are becoming brighter and bluer. Right: The 2MASS J vs 2MASS J-H' magnitudes for each observation (dark circles are binned down to a 441 second cadence) along with the individual fits for each brown dwarf using orthogonal distance regression (ODR), with their $\theta$ values and associated uncertainties. The colour modulation plot for 2M1721+33 appears as nearly two individual regions, however, this is due to a lack of phase coverage as a result of HST observations. The two observations of 2M0036+18 are plotted separately and they share a very similar slope, despite observations occurring 16 months apart.