A PULSATION SEARCH AMONG YOUNG BROWN DWARFS AND VERY-LOW-MASS STARS

The σ Orionis cluster was our primary target in the search for pulsation, with observations of candidate very-low-mass members in three fields. At approximately 3 Myr of age (Sherry et al. 2008), it should contain numerous stars and BDs that are still burning deuterium. Two 20' ×20' fields were initially selected for monitoring with the Cerro Tololo Inter-American Observatory 1.0 m telescope ("CTIO 1.0 m") from 2007 December 27 to 2008 January 7 and from 2008 December 14 to 24. We identified VLMSs and BDs in the cluster by mining the works of Béjar et al. (1999, 2001), Barrado y Navascués et al. (2001, 2003), Béjar et al. (2004), Caballero et al. (2004), Sherry et al. (2004), Scholz & Eislöffel (2004), Burningham et al. (2005), Kenyon et al. (2005), Franciosini et al. (2006), Caballero et al. (2007), Hernández et al. (2007), Caballero (2008), Luhman et al. (2008), and Lodieu et al. (2009); a total of 153 confirmed and candidate σ Orionis members fell in the two CTIO fields, including 15 spectroscopically confirmed young BDs. Further details on the selection, as well as a full list of objects and an image of the field of view, are provided in Cody & Hillenbrand (2010, hereafter CH10).

We re-observed a subset of the CTIO targets in the infrared with the Spitzer Space Telescope Warm Mission in 2009, focusing on a set of five confirmed and two candidate BD members of σ Orionis. Also included in the Spitzer 3.6 and 4.5 μm fields were seven additional higher mass (>0.1M_☉) cluster members. All targets were monitored continuously for a single 24 hr period, at a cadence of ∼30 s. An extensive description of these observations is available in Cody & Hillenbrand (2011).

Initial target compilation for the Chamaeleon I cluster was based on Luhman's (2007) work, plus additional members from Luhman & Muench (2008), Luhman et al. (2008), and Mužić et al. (2011). These studies indicate a median age for the cluster of ∼2 myr, making it slightly younger than σ Orionis, and a prime candidate for the deuterium burning pulsation search. The campaign on Cha I involved observations of a single 20' × 20' field with the CTIO 1.0 m from 2008 May 13–25. The field of view (FOV) was selected to maximize the number of BDs monitored and also avoid some of the dense nebulosity in this region; it is displayed in Figure 1.

Zoom In Zoom Out Reset image size

From the full source compilation we selected a total of 32 Cha I members for observation, of which 6 have spectral types consistent with substellar status (spectral type M6 and later) and 22 more are likely VLMSs with (M4 or later). The remaining four are higher mass stars and not expected to pulsate. We have compiled the existing photometric and spectroscopic data on the set of 32 objects, including optical through near-infrared photometry and spectral types, in Table 2.

IC 348 is an appealing target in the search for pulsation, since it is relatively compact (<1° square), and its membership is very well characterized (Luhman et al. 2003; Muench et al. 2007). At a 2–3 myr (Luhman et al. 2003), it is comparable in age to σ Orionis and Cha I. Several previous studies have identified numerous periodic variables, which are presumably the result of rotational modulation of spots (Cohen et al. 2004; Littlefair et al. 2005; Cieza & Baliber 2006, and references therein). The typical periods are near two to three days, but several objects have reported periods as short as five hours.

We observed IC 348 from the ground with the Palomar 60 inch telescope ("P60"; Cenko et al. 2006) and from space using the Hubble Space Telescope (HST) Wide Field Camera 3 (WFC3). The ∼125 × 125 ground-based field of view encompassed a significant spatial extent within IC 348, including the nebulous region in the cluster center. The WFC3 field is much smaller, with a full field of view of 162'' × 162''. To maximize the data cadence for HST we opted to observe in the subarray mode, for which only one of two 81'' × 162'' chips was used. Since the ground-based photometry preceded the space observations by more than three years, we were able to select several faint BD pulsation candidates that required photometry at the higher sensitivity levels afforded by HST. Therefore the WFC3 field did not cover an additional region, but rather fell within the previous ground-based FOV, encompassing four BDs and and two VLMSs. Two of these (L761 and L1434) do not have ground-based light curves since they suffered from low signal-to-noise. Both the ground- and space-based fields are illustrated in Figure 2.

Zoom In Zoom Out Reset image size

Selection of low-mass IC 348 cluster objects was carried out by considering the spectral types presented by (Luhman et al. 2003, 2005). A total of 194 members fell within the chosen ground-based FOV. However, we did not extract photometry for stars that fell on bad pixel columns, or for some BDs that were too faint for adequate signal to noise. There are also a number of brighter stars within the field for which we did not obtain data since their point-spread functions (PSFs) were saturated and distorted by nebulosity and scattered light within the central region of the cluster where several bright B stars lie. For the pulsation campaign, we ultimately monitored 147 low-mass IC 348 members, including 26 BDs (spectral types M6 or later) and 65 VLMSs (M4–M6). We present a compilation of basic target properties in Tables 3 and 4.

The Upper Scorpius (USco) region is one of the most spatially extended young associations, with stars and BDs spread over many tens of degrees on the sky. As a result, few variability studies have been performed here, apart from the work of Slesnick (2008). It is significantly older than the other three regions studied, with age estimates from 5 to 11 myr (Preibisch et al. 2002; Pecaut et al. 2012). Nevertheless, an examination of the H-R diagram of catalogued low-mass members shows that many objects have temperatures and luminosities overlapping the D-burning instability strip.

Because of the sparseness of this association, it is difficult to obtain data on more than one target at a time. We therefore selected fields carefully to maximize the number of pulsation candidates. We ultimately observed five different fields in Upper Scorpius, including 5 BDs and 11 VLMSs, with identifications and references listed in Table 5. None of these regions contained any nebulosity. Observations of three of the fields were abbreviated to three nights or less because of weather (the CTIO 1.0 m run), and tracking problems (USco members in the first field chosen for observation with the P60 fell too close to the edge of the detector and tended to wander out of the FOV). The list of dates is provided in Table 1. Since our observations in 2008 and 2009, further low-mass Upper Scorpius members have been discovered by Lodieu et al. (2011) and Dawson et al. (2011). We identified three of these objects from Lodieu et al. (2011) in our first FOV from CTIO 1.0 m Y4KCam monitoring in 2008 May, with no additional targets in any of the other observations.

We observed low-mass targets in σ Orionis, Chamaeleon I, and Upper Scorpius with the CTIO 1.0 m. The observational setup and data reduction procedures were the nearly same for all CTIO 1.0 m runs, and they are described in detail in CH10. The selected exposure time was 600 s for all runs, apart from the first set of observations on σ Orionis members (360 s).

The main distinction for the 2008 May observing run (see Table 1) was that sky conditions were not photometric, and just over two nights were lost to clouds. In addition, telescope building maintenance caused a new accumulation of dust specks on the detector each night, which resulted in inconsistent sky flatfield acquisition. We acquired dome flatfields at the beginning and end of each night to calibrate out dust "donuts," but these dome flats are known to misrepresent the true pixel sensitivity distribution by up to 10%. Therefore, we carried out flatfielding with sky flats on nights where at least seven were available, and when clouds precluded the acquisition of sky flats, we instead relied on the dome flats but performed an illumination correction using the high signal-to-noise composite provided by P. Massey.³

We used the P60 robotic telescope to observe BDs and VLMSs in the IC 348 cluster, as well as additional BDs in Upper Scorpius. Observations of IC 348 took place on a total of nine nights between 2008 November 17 and November 29. The chosen field center was R.A. = 03^h44^m197, decl. = +32°04'29''s (J2000), but since the P60 system was at that time subject to tracking inaccuracies, this position shifted up to 45'' throughout the run. We observed USco with the P60 during two different runs, from 2008 June 1 to 14, as well as 2009 May 14–30.

We chose exposure times of 240 or 300 s in the i' band to provide good sensitivity to the faint BDs without elongating the PSFs too much due to the lack of guiding. The established data reduction pipeline for the P60 performs basic calibrations, including bias subtraction, flatfielding, and fitting of the world coordinate system (Cenko et al. 2006). Although we obtained a series of sky flatfields, we determined that the domeflat images used by the pipeline were sufficient to correct interpixel sensitivity variations. Image alignment was also carried out with ease, since an accurate coordinate system was already superimposed on the calibrated images; we used the IRAF task wregister to complete this task.

We used the HST WFC3 ultraviolet/visible (UVIS) CCD to observe pulsation candidates in IC 348. The UVIS channel is comprised of two chips, each with 4096 × 2051 pixels; since we observed in subarray mode, we used only one of these (UVIS1). Each pixel is ∼004 across, for a total subarray field of view of ∼81'' × 162''.

Observations took place from 2011 January 29 to February 4, for just over seven hours of each day. Although HST is a space observatory, the Sun position and other observing constraints resulted in each block of observations ("visit") beginning at roughly the same time every day. Unfortunately much of visit 5 was compromised since the gyroscopic system failed and the field was lost for a number of hours. The viewing limits of HST are such that IC 348 objects may be observed for only 46 minutes of each 97 minute orbit. Therefore we designed exposure times to acquire as many images as possible per orbit, without exceeding the telescope's maximum data downlink rate. These varied among 128, 171, and 192 s. All observations were carried out through the F814W filter, which is centered near 8030 Å and similar to I band.

HST/WFC3 data are processed by pipeline, which includes standard bias and flatfield calibration, as well as cosmic ray rejection. The MultiDrizzle program corrects for geometric distortion and optimally combines sets of three of four consecutive images, even for undithered data such as ours.

With two nearly two-week observing campaigns on the CTIO 1.0 m telescope, we monitored over 150 VLMSs and BDs in the σ Orionis cluster. As detailed in CH10, we identified 38 periodically variable confirmed cluster members, along with 27 periodically variable candidate members. However, all but one of these variables had periods greater than 8 hr. The remaining single object (2MASS J053825570248370) had a period of 7.2 hr. This lack of detections in the one to four hour range was despite sensitivity to periods as short as 15 minutes, and amplitudes as low as several millimagnitudes down to a magnitude of I = 17 (and as low as 1 mmag at I = 14). We therefore concluded that there are no signs of deuterium-burning pulsation in our σ Ori data set. The full compilation of periodic variables, their phased light curves, and discussion of detection limits appears in CH10.

Using data from the CTIO 1.0 m telescope, we produced and analyzed light curves for all 32 observed BD and VLMS members of the Cha I cluster. In the interest of fully mining the data set and potentially identifying new cluster members, we additionally performed photometry on all 1548 objects in the Cha I field that were bright enough for detection in individual Y4KCam images (i ≲ 22).

These additional light curves enabled an assessment of the photometric noise as a function of magnitude, as shown in Figure 4. We have fit a trend to the cluster nonmembers, based on the expectations for Poisson and sky noise. Using the fit values as an indication of the magnitude-dependent photometric uncertainty, we have then selected variable objects via the χ² test. The full collection of these light curves is presented and discussed in the Appendix. The floor of the rms distribution reaches ∼3 mmag at an I-band magnitude of 14, as shown in Figure 4. Objects with I < 17 display a photometric precision of 1% or better.

Zoom In Zoom Out Reset image size

For each object in the sample, we carefully analyzed the periodograms for signs of periodicities on the few-hour timescales predicted for D-burning pulsation. Among the known cluster members, only four stars have estimated masses above ∼0.4 M_☉, and so most are candidates for the instability.

Based on the observing cadence of 600 s, we are sensitive to periodic signals with frequencies as high as 72 cd⁻¹ (i.e., P = 20 minutes). The majority of light curves contain low-frequency variations, most of which are due to intrinsic erratic variability; these shape the exponential rise in periodogram amplitude at <1 cd⁻¹. Despite sensitivity to few-millimagnitude levels, we do not find any evidence for variability with periods less than 16 hr, apart from one field object that appears to be a main sequence pulsator.⁴ A few periodograms display low-level signals in the range where pulsation is expected, but none of these meet the 99% significance level criteria, and we find most to be aperiodic variables. Furthermore, the light curves do not show clean trends when phased to these periods. We conclude that none of the 32 very-low-mass objects in the Cha I sample are periodic on one to five hour timescales, at least at or above the amplitude levels probed by the data.

On longer timescales, the final sample of variables contains 12 periodic objects in Cha I with clear variability. Observed and phased light curves for the cluster members among these are presented in Figure 5. We also identified periodic behavior among eight additional objects of unknown membership status. In most of these cases, the light curve shapes are characteristic of field pulsators or eclipsing binaries, and blue colors suggest locations in the background field. However, two relatively red objects with sinusoidal light curves may be bona fide Cha I members (2MASS J11073302−7728277 and 2MASS J11122675−7735183). The former (also known as CHXR 25) was classified by Luhman (2007) as a field dwarf, while the latter has no previous literature and would benefit from spectroscopic follow-up. The light curves for these new Cha I candidates are presented in Figure 6.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The measured properties for all variable sources in Cha I are listed in Table 6. This includes the aperiodic variables uncovered with the χ² test, as discussed in the Appendix. Of note, few Cha I members have been photometrically monitored previously, apart from a sample of 10 BDs and VLMSs presented by Joergens et al. (2003). Of the five with periods reported in that work, none are redetected as periodic variables here.

In IC 348 the photometry includes only known cluster members, since high extinction in the region blocks our view of most background field stars. We were therefore unable to perform a full assessment of photometric precision based on nonvariable field stars. However, a plot of rms light curves values versus magnitude (shown in Figure 7) provides a rough estimate of the performance for our P60 data set. Over the course of a single night, we reach a precision of ∼2 mmag at I = 13, and better than 0.01 mag down to I ∼ 17. The rms values measured over the entire week-long duration of the light curves show considerably higher spreads in photometry, ∼1% or higher in most light curves. We attribute much of this to intrinsic variation of the stars on >1 day timescales.

Zoom In Zoom Out Reset image size

We performed a period search analysis on all IC 348 cluster members observed with the P60 and HST. For the former, we are sensitive to periodicities on timescales from approximately 9 minutes to 12 days (resulting from a cadence of 270 s). Since we did not obtain data every night, our sensitivity to periods of more than a few days is highly nonuniform. We therefore focused on the short periods predicted for D-burning pulsation.

We generated periodograms from the light curves of all 147 unsaturated objects in the ground-based P60 field, including 91 BDs and VLMSs. A large fraction of these display variability by eye, much of which is erratic and dominated by low-frequency behavior. It is thus not surprising that many of the periodograms display excess power around 5–10 cd⁻¹, which we attribute to aliasing, based on the window function, which shows a series of peaks centered around 8 cd⁻¹. Further support for the spurious nature of these signals is that they do not reach the 99% significance level, and no coherent periodicity emerges when the light curve is phased.

In some cases, an obvious and statistically significant periodicity did appear on timescales of one day or more, suggesting rotation. When this occurred, we fit the overall trend and removed it from the light curve before searching for shorter timescale pulsation signals. In none of the IC 348 light curves did we uncover significant periodicities on the 1–5 hr timescales indicative of pulsation.

Similar to the ground-based monitoring data, the HST time series are subject to aliasing. This effect is associated with the 97-minute orbital timescales and is seen prominently in the periodogram of object L350, shown in Figure 3. Omitting the frequency regions subject to aliasing, we have again searched for periodicities on one to five hour timescales, in hopes of detecting pulsation. In general, we find that the periodogram values are consistent with noise at the 1–5 mmag level in the frequency range of interest (5–25 cd⁻¹).

Two exceptions in the HST data set are the IC 348 BDs L761 and L1434. The former is periodically variable, as shown in Figure 8. The period of 0.6 days is consistent with rotational modulation by one or more spots on its surface. L1434 is also variable (see Figure 9), but we are unable to determine whether it is periodic without a longer duration data set.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

As with the other young star regions, we produced discrete Fourier transforms to search for periodicities on a variety of timescales in Upper Scorpius. Targets here were observed during a number of runs (see Table 1), and we are sensitive to periodicities as short as 10 minutes in each of these data sets. The longest period detectable varies with the length of each observing run and ranges from less than 2 days to 16 days. Since some of the runs only had six or fewer total nights, the associated periodograms have lower frequency resolution than the data sets on other clusters, and the search for periodicities is more susceptible to systematic effects. While some of the objects display variability on night-to-night timescales, we cannot accurately quantify all possible periodicities. We therefore focused exclusively on the search for pulsation at one to five hour periods. Detection of signals in this period range is feasible given the number of datapoints (30–550, depending on the run) and the fairly low photometric uncertainties of the BDs (1%–3%).

We find that the data generally have sensitivity to signals in the few millimagnitude to 0.01 mag amplitude range. Despite this, none of the objects in USco showed significant periodic variability. However, the light curves of a few displayed night-to-night variations that may be indicative of accretion or variable circumstellar extinction.

We uncovered many cases of periodic variability in the collected time series, over a wide range of timescales. Our detection of both rotation on ∼1–3 day timescales in young cluster members and variability on hour timescales in background field pulsators and eclipsing binaries shows that our period detection algorithms are robust. Yet in the search for deuterium-burning pulsation, the data unanimously point to one conclusion: this instability is not present in young BDs and VLMSs above an amplitude of several millimagnitudes in the I band.

One might argue that objects in our data set simply do not exhibit pulsation because they are not situated on the H-R diagram instability strip. However, the large sample size makes this possibility highly unlikely. To show how improbable the chances are that none of our sample have H-R diagram positions overlapping the instability strip, we consider temperature–luminosity probability distributions for each object. We take these to be two-dimensional asymmetric Gaussians, normalized and centered at the adopted luminosities and temperatures. The Gaussian widths are given by the associated 1σ uncertainties, which are shown in the H-R diagrams in Figure 10. The position of each target then corresponds to a probability that it is susceptible to pulsation, which we determine by integrating its distribution over the entire region of the instability strip. For objects on or very close to the strip, this value is at least ∼20%–25%, whereas for the higher mass stars far from the strip it is close to zero. The probability that the position of a given object does not overlap with the instability strip is then 1.0 minus this quantity. The product of these values over all targets provides an estimate of the chance that no pulsators would be present in our sample.

We have performed this exercise for each of the clusters observed, and for alternate distances in cases where there is more than one possible value (IC 348 and σ Ori). Since the instability strip runs nearly along isochrones and varies slowly in luminosity, changes in the adopted distance can have a significant impact on the number of objects expected to pulsate. We take all possibilities into consideration.

In Cha I, we determined an expectation value of 3–4 objects on the strip and find a probability of 0.015 that no object positions actually overlap it. Turning this number around, there is a nearly 99% chance that at least one object should exhibit pulsation based on its position within the instability strip, assuming that the theoretical calculations underpinning it (PB05) do not suffer from gross systematic errors. Likewise in Cody & Hillenbrand (2011), we found a 70%–80% chance that at least one of our σ Ori targets observed in the infrared with Spitzer should exhibit pulsation. This is in contrast to the lack of short period variability in that data set.

USco does not have many targets overlapping the instability strip, and therefore the expectation is for only 1 or 2 objects to lie directly on it. In this region, we find a non-negligible probability of 0.22 that our sample did not include any pulsation candidates. For IC 348, on the other hand, we expect ∼11 objects on the strip and find a probability of 4×10⁻⁶ that none are actually on it. This is assuming a distance of 316 pc Herbig (1998). If we instead assume a lower distance of 260 pc based on Hipparcos parallaxes (Scholz et al. 1999), then the expectation is similar: nine objects on the strip and a probability of 5×10⁻⁵ that none are on it.

Finally, we have computed probabilities for σ Ori. The precise number of pulsation candidates depends on the adopted distance, which is debated to be either 350$^{+120}_{-90}$ pc, based on the Hipparcos parallax of σ Ori AB, or 440$^{+30}_{-30}$ from main sequence fitting (Sherry et al. 2008). Assuming a cluster distance of 440 pc (Sherry et al. 2008) we find that at least 4 targets are expected to be on the strip, with at most a 0.02 chance that none are. Substituting the alternate distance of 350 pc, we find nearly the same values (3, 0.06). The probabilities are upper limits since we do not have spectral types for part of the σ Ori sample and hence cannot reliably place these objects on the H-R diagram.

In conclusion, we expect with high confidence to have observed deuterium-burning oscillations if it is present at observable amplitudes. We now quantify the overall detection limits by considering the power-law fit to the periodograms of each observed young cluster member. These curves, of form A/(f + B) + C for frequency f and constants A, B, and C, trace out the noise level as a function of frequency (see Figure 3). For each object analyzed, we take the fit values at 5 cd⁻¹ (∼5 hr) and 25 cd⁻¹ (∼1 hr) as representative of the 1σ level above which no pulsation is observed. We display these values as a function of object magnitude in Figure 11 to illustrate the collective limit imposed by our entire data set.

Zoom In Zoom Out Reset image size

The median amplitude limit is several millimagnitudes. Objects with high-amplitude aperiodic variability (see the Appendix) are exceptions, as they have excess periodogram noise which is intrinsic. The rest of our targets, however, have maximum amplitudes in the periodogram of at most 0.002–0.004 mag. This represents the threshold above which we detect no periodicities. We conclude that if deuterium-burning pulsation is present in any of our sources, then its amplitude must be below this level.