A mini-survey for variability in early L dwarfs

We report differential I-band photometry of four early L-dwarfs obtained to study variability. We detect variability on the timescale of hours in two objects, 2M0746425+200032 (at a level of 0.007 mag -- 6.5 sigma) and 2M1108307+683017 (0.012 mag -- 5 sigma). We also place upper limits of 0.02 mag (1 sigma) on the variability of two others.


INTRODUCTION
The recently catagorised L-dwarfs  have rapidly become an important part of stellar astrophysics. L-dwarfs lie below M-dwarfs in terms of their temperatures, and their spectra are dominated by molecules such as TiO, FeH and CrH. Dust formation also plays a very significant rolé in the atmospheric physics of early Ldwarfs (by L5-7, the major dust species have begun to settle below the photosphere). The physics of dust formation is a complicated subject, but one which is critical to our understanding of L dwarfs. Variability studies give us a tool to probe the sub-global differences in the composition of the atmosphere. Other processes, such as the flaring seen in M dwarfs, may also contribute to variability in L dwarfs.
In this paper we report time-series photometry of four field L dwarfs. Two objects, 2MASS1108307+683017 and 2MASS0746425+200032 are found to be variable at the level of 0.012 mag (5σ) and 0.007 mag (6.5σ) respectively on the timescale of hours. We did not detect any variability above the noise level of ∼0.02 mag (1σ) from the other two targets (2MASS1146345+223053 and DENIS0909571-065806).

OBSERVATIONS AND REDUCTION
Data were obtained with PFCam on the 3-m Shane telescope at Lick Observatory on the nights of 2002 January 21 and 23 UT. Conditions were photometric on the first night, but patchy cloud interupted observations on the 23rd. PFCam has a 2048×2048 CCD array with a pixel scale of 0.296 ′′ /pixel. The full field of view, 10.1 ′ ×10.1 ′ , is more than sufficent for our purpose (high cadence, high precision differential photometry). We therefore used only a small section of the chip to improve read-out times. Chip windows were choosen to select the target, and 5 or 6 comparison stars of similar magnitude. The chip was also binned where possible to further reduce overheads, such that we collected images at 2 minute intervals on average. All observations were made through the I-band filter where the photon count from the extremely red L dwarfs is maximised. Calibration frames, consisting of bias frames, dome and sky flats were obtained for all chip configurations. A log of observations, including window sizes and chip binning, is given in table 1.
Data were reduced with IRAF 1 . Ten bias frames were median combined, and this was then subtracted from all data and flat field frames. An average normalised flat field was then constructed from sky flats. We divided the science data by the appropriate flat field. This provided images flattened to the 0.25% level. Fringing is visible at the ∼1% level on the scale of ∼10 ′′ . We could not remove this, as we did not obtain enough pointings to build a fringe map.
During the first night's observing, we discovered PF-Cam's shutter was working incorrectly, with only one blade of the traveling shutter system functioning. As the travel time of the blade is on the order of the milliseconds, a part in 10 6 for most science frames, this contributes neglibly to the photometric errors. The effect is however apparent in short (0.5-4s) dome flats. For these frames, we fitted the gradient with a linear function, and subtracted the fit.
In some science frames, the peak counts in a few pixels approached the full-well depth of the CCD (65536 ADU). We obtained data to test the linearity of the chip, and found it to be linear to ∼60000 ADU. Any frames with peak counts above this level were removed from the analysis below.
Photometry was performed with the APPHOT package within IRAF. We determined the FWHM for each frame within an observing sequence, and picked an aperture ∼1.5× larger than the average FWHM for that sequence. This aperture was used on all frames within that sequence. Magnitudes and errors were calculated from the fluxes given by APPHOT in the manner described by Everett & Howell (2001), including an estimate of informal errors from fringing and imperfect flatfield correction. An ensemble of bright stars were combined to give a zero-point magnitude for each frame (mens), which was then subtracted from the measured magnitude of the target (mtar) to build a differential lightcurve. Hence, an increase in δm (mtar-mens) represents a dimming of the object. A selection of comparison stars (not included in the ensemble stars) were also analysed to ensure any variability detected is intrinsic to the target. In no case did any of the ensemble stars display any variability.

2MASSJ1108307+683017
2M1108+6830 is an L1 dwarf. Gizis et al. (2000) report an H-α equivalent width of 7.8Å, which they suggest indicates it is possibly an older member of the stellar population, rather than a youthful brown dwarf. We observed 2M1108 for 4.5 hours with a sampling rate of 2 minutes (100s integrations). To obtain sufficent ensemble stars (at least 3), we used a field of 5.1 ′ ×5.1 ′ with 2x2 pixel binning (Figure 1). Figure 2 shows the resulting I band lightcurves for 2M1108 and a comparison star (which was not part of ensemble stars). It is clear from this figure that 2M1108 becomes ∼0.012 mag fainter during the last 1.5 hours of observation, whilst the comparison star stays constant to within ±0.003 mag throughout the observation. The standard deviation of the binned comparison lightcurve is 0.0025 mag. Our detection of variability in the binned lightcurve is therefore significant at the 5σ level. There may be a slight brightening in 2M1108 just before it fades (AJD 2 =6.03), but this is detected only at the 2σ level due to an increase in noise in this region.
Although the comparison star shows no variability, we must consider the possibility that 2M1108's variability is caused by second order colour effects (see Young et al. (1991) for a full discussion). The L dwarfs have much redder spectral energy distributions than the ensemble stars, and hence have a longer effective wavelength in the I filter. This means the L dwarfs have a smaller effective extinction coefficent, which would manifest itself as an apparent brightening in the differential lightcurve with increased airmass. However, the data exhibit the opposite behaviour, since the observation started at an airmass of 1.15 and ended at 1.3. The detected dimming of 2M1108 we see is therefore unlikely to be caused by airmass effects.
The H-α emission from 2M1108 implies chromospheric activity. However, it is not clear if the variability we detect is related to this. The variability may alternatively be due to photospheric features (magnetic spots or dust clouds) which modulate the brightness as the object rotates. There is no periodic signal in our data, but our temporal sampling 2 AJD = HJD-2452290.0 clearly does not cover a sufficent range to exclude this possibility. Based on v sin i measurements (Basri et al. 2000), we expect the typical rotation period of an L dwarf to be on the order of 6-12 hours.

2MASSJ0746425+200032
2M0746+2000 is infact a close (0.22 ′′ ) binary of roughly equal mass ). We do not resolve the binary in our observations, and therefore refer to the whole system as 2M0746 rather an either individual component. We discuss how this may effect our results later.
2M0746 has a spectral type of L0.5. Schweitzer et al. (2001) detect no Li in its spectra, and therefore estimate a minimum mass of ∼0.06M⊙for each component. Due partly to its binary nature, 2M0746 is one of the brightest known L dwarf systems, having an I band magnitude of 15.2. It also lies in a rich stellar field, making it an ideal target for accurate differential photometry. We observed this target for nearly 6.5 hours with a sampling rate of 1-2 minutes. Changes in seeing required us to adapt our observing strategy (to avoid saturating the target). The majority of the usable images measure 2.2 ′ ×2.2 ′ , and contain the target and 4 bright comparison stars (Figure 1). These stars are also visible in all previous configurations used on this object.
The derived I band lightcurves are shown in Figure 3. 2M0746 shows variability in two of the three configurations. In the second configuration (AJD=5.64-5.70), 2M0746 brightens (δm decreases) by ∼0.01 mag while the comparison stays constant to within 0.005 mag. However, due to the increased noise in this configuration, this is only signficant at the 2σ level. During the third configuration (AJD=5.70-5.89), 2M0746 fades by ∼0.007 mag around AJD=5.78 before returning to its original brightness. The comparison object is constant to ∼0.002 mag, and its binned lightcurve has a standard deviation of 0.0011 mag. Our detection of variability in the third configuration is therefore significant at the 6.5σ level.
It is possible that the brightening we see from 5.66 to 5.7 is analogous to the brightening from 5.78 to 5.82. This  ). D0909-0658 was discovered by the DENIS survey (Delfosse et al. 1997).   (Table 1). Each configuration has a different zeropoint, and all three have been individually mean subtracted. Differences between the configurations are therefore not significant.
may suggest that we are seeing rotational modulation of the lightcurve on a period of ∼3 hours. If this were the case, we would expect to see the next minima at AJD≃5.9 -just after the end of our observations. Three hours would be a fast, but not unprecedented rotational period for a brown dwarf. For example, Kelu-1 has a rotational period of 1.8 hours (Clarke, Tinney & Covey 2002). Unfortunately, the binary nature of 2M0746 complicates matters significantly. As we are observing an unresolved binary, we are essentially sampling the combined lightcurve of two objects. It is not therefore possible to attribute specific events in the lightcurve to either object. In addition, the binarity will tend to "wash-out" variability from either component of the binary; for example, 1% variability from one component will result in only 0.5% measured variability from the system.
2M0746 is clearly an interesting target for future variability observations, especially if the binary can be resolved. Martín et al. (1999) classify D0909-0658 as an L0 dwarf. Figure 1 identifies it along with the comparison star whose lightcurve is plotted in Figure 4. Seeing on this night was ∼2 arcsec poorer than the first night, requiring longer (300s) exposures to obtain the necessary signal to noise ratio. D0909-0658 was observed in two batches, seperated by ∼90 minutes. Figure 4 shows no significant indication of variability either between or within the two sequences. We therefore place an upper limit on the variability of D0909-0658 of σ i <0.02 mag (1 sigma) on the timescale of 0.5-4 hours.

2MASS1146345+223053
2M1146+2230 is actually a close equal brightness binary with a separation of 0.3 ′′ (Koerner et al. 1999). Schweitzer et al. (2001) have detected Li in the combined spectrum of  2M1146, confirming its status as a brown dwarf binary with each component having M<0.06M⊙. In addition, an earlier type background star is located 1 ′′ away from the binary . The seeing during our observations (3 arcsec) did not allow us to resolve the binary or to detect the background star.
Bailer-Jones & Mundt (2001) have previously made a marginal detection of variability in this object at the level of ∼0.015 mag. Poor seeing did not allow us to reach this level of accuracy. The lightcurve ( Figure 5) shows 2M1146 is constistent with being non-variable at the noise level of ∼0.02 mag (1 sigma) during our observations.

CONCLUSIONS
We have detected variability in time series I band observations of two field L dwarfs (2M0746425+200032 and 2M1108307+683017), and placed upper limits on the variability of two more (2M1146345+223053 and D0909571-065806). The cause of variability is unclear, but it may be due to imhomogenous structures within the photosphere. In the case of 2M0746+2000, weak evidence exists for a periodicity of ∼3 hrs. This could be explained by a photospheric feature and a 3 hour rotation period.
In now seems clear that a significant fraction of, if not all, L-dwarfs are variable at the 1-2% level, and that very few have much larger variability (Clarke, Tinney & Covey 2002, Bailer-Jones & Mundt 2001, Martín, Zapatero Osorio & Lehto 2001. In this study, both stars for which we obtained photometry better than 1% exhibited variability. Future surveys for variability in L-dwarfs require a photometric precision better than 0.5%.