PARALLAXES OF SOUTHERN EXTREMELY COOL OBJECTS. I. TARGETS, PROPER MOTIONS, AND FIRST RESULTS

The first brown dwarf, GD 165B, was discovered in 1988 (Becklin & Zuckerman 1988) but was not recognized as such until 1995 when Gl229B (Nakajima et al. 1995) and other objects with the same characteristics were found. Rapidly, many examples were discovered primarily in the large infrared surveys, i.e., the Two Micron All Sky Survey (2MASS; Skrutskie et al. 2006) and Deep Near-Infrared Survey (DENIS; Epchtein et al. 1999), and the deep optical Sloan Digital Sky Survey (SDSS; York et al. 2000). It was soon realized that new spectral types, L and T, were needed (Kirkpatrick et al. 1999). Since then over 700 L/T dwarfs have been discovered¹⁰ by various authors and just recently a sample of 210 new L dwarfs from the SDSS was announced (Schmidt et al. 2010a). These objects have heralded a whole new sub-field of astronomy.

Interest in brown dwarfs has been particularly prominent in the interpretation of their spectral and photometric properties. Theory has been led in unexpected directions by unpredicted behaviors: the very strong evolution of spectral type with age (Burrows et al. 1997), the "hump" in the J-band magnitude as a function of spectral type at the L/T transition (Tinney et al. 2003), notable differences between infrared spectra of optically classified objects and vice versa (e.g., Figure 3 in Kirkpatrick 2008), and a turnaround in color at the T8/T9 spectral type (Warren et al. 2007). We are also slowly uncovering significant numbers of L and T sub-dwarfs (Sivarani et al. 2009; Burgasser 2004; Bowler et al. 2009, 2010) and other peculiar objects that challenge the theoretical models.

Parallax is a crucial parameter for understanding these objects as it is the only direct way to calculate an absolute magnitude and hence energetic output. In brown dwarf structure models, particularly for T dwarfs, the determination of metallicity and surface gravity from spectra is degenerate (Leggett et al. 2009) and hence luminosity, which requires a parallax, is used to constrain either the radius or the temperature and helps break this degeneracy. Precise absolute velocities that in turn provide age and origin indications require precise parallaxes.

In light of the role of distance, it is important that for these new objects we have a significant sample with measured parallaxes. As shown in Figure 1, only a small fraction of known L/T dwarfs have measured parallaxes—the black histogram—which limits any calibrations and generalizations we can make. To increase the current sample, in 2007 the Osservatorio Astronomico of Turin and Observatório Nacional of Brazil began the PARallaxes of Southern Extremely Cool objects (hereafter PARSEC) program to determine parallaxes for 140 bright L and T dwarfs. In Figure 1, we include the PARSEC objects that illustrate our goal to attain at least 10 objects per spectral bin for L dwarfs and to increase the current sample for the fainter T dwarfs. A number of other programs are also underway to address this shortfall (for example, C. G. Tinney 2009, private communication, and J. K. Faherty as part of the Brown Dwarf Kinematics Project, Faherty et al. 2009), but even including the expected additional objects the results of the PARSEC program will at least double number of L and bright T dwarfs with parallaxes.

Zoom In Zoom Out Reset image size

In this paper, we present the PARSEC program: Section 2 describes the instrument and target selection; Section 3 details the observational and reduction procedures; and Section 4 describes about the procedures used to produce a catalog of standard proper motions and preliminary parallax solutions for 10 objects. Finally, in Section 5 we discuss some uses we have made of this catalog and future plans.

2.1. The Instrument

2.2. Observations

The observational procedure is as follows.

• 1.  
  For each field we make one quick 50 s exposure and locate the target.
• 2.  
  Using the WFI move-to-pixel procedure we offset the telescope to move the target to pixel (3400,3500), which is in a flat part of CCD 7 (Priscilla) at less than 1/4 of the diagonal leading to the optical center of the mosaic.
• 3.  
  We make the first science exposure of 150 s for objects with z<18.0 and 300 s for z⩾18.0.
• 4.  
  The camera is then slightly offset, 24 pixels in both directions, and the second science image of the same exposure time is automatically begun.
• 5.  
  We check the counts of the target in the first image. If the signal-to-noise ratio of the target in the first image is less than 100, in real time we increased the exposure time accordingly. This is usually only the case in particularly poor sky conditions.

This procedure is very efficient and the dead time for the telescope is minimal. The total time for a target is 10–25 minutes depending on magnitude and other overheads, enabling us to observe three to four objects per hour. Our time allocation usually results in always having grouped nights and, as multiple observations in the same run are of limited value, to both increase the sample and allow some redundancy, the target list has sex to eight objects per hour. We attempted to observe the majority of targets close to the meridian except during the twilight hours when we wish to include rising or setting targets at their maximum parallactic factor.

Observations began in 2007 April and, as of 2009 September, targets have between 1.5 and 2.5 years of observations. The frequency of observations has been reasonably constant with a two to three night run every two months.

Figure 2 displays the sky distribution of the targets. Table 1 summarizes the dates and nights of the observations taken up to 2009.

Zoom In Zoom Out Reset image size

Table 1. Observations of the PARSEC Program at the ESO2p2/WFI up to 2009

Date	Nights
2007 Apr	09, 10, 11, 12
2007 Aug	31
2007 Sep	01, 02
2007 Oct	05, 06, 07
2008 Jan	04, 05
2008 Feb	26, 27
2008 Apr	02, 03
2008 May	27, 28
2008 Aug	21, 22
2008 Oct	24, 26
2008 Dec	18, 20
2009 Mar	01, 02, 03
2009 Apr	30
2009 May	06, 08
2009 Jul	21, 22, 23
2009 Dec	15, 18, 21

Download table as:  ASCIITypeset image

2.3. Target Selection

2.4. Image Reduction Procedures

2.5. Centroiding and Feasibility Tests

The WFI, having a large field of view, has significant astrometric distortions and the CCDs have significant relative tilts (see the WFI section at www.eso.org). However, the fundamental requirements for relative astrometry, which underlie all small field parallax determinations, are stability and repeatability. For this reason, we use the WFI move-to-pixel routine to put the target on the same pixel for each science exposure and only consider astrometric distortion changes over the observational campaign. The move-to-pixel position, (3400,3500), is sufficiently inside CCD 7 that reference stars from only the top third of the chip are needed to make a low-noise astrometric transformation between different epochs. The move-to-pixel procedure introduces a significant overhead, but as shown by Platais et al. (2002), on a similar mosaic, the chips move relative to one another and this introduces a change in the astrometric deformation that was impossible to model at the mas level as required by our parallax goal.

We have tested three centroiding routines: a two-dimensional Gaussian fit to the psf as used in the Torino Observatory Parallax Program (Smart et al. 1999, TOPP), a one-dimensional Gaussian fit to the marginal distributions, and the Gaussian psf fit provided by daophot in IRAF. In a comparison of object positions observed on 14 nights over an 18 month period of the field around the object 0719–50, we found the TOPP routine worked best. In Figure 3, we plot the rms of the position differences as a function of instrumental z magnitude for the frames in question. The median rms for all objects is 23 mas and for the brighter objects from 12 to 16 is 10 mas. If we apply the same test on images that have not been fringe-corrected as described above, we find that the median precision has deteriorated to 28 mas. Applying the fringe map supplied by the WFI calibration team provided a precision that was intermediate between a nightly fringe correction and no correction.

Zoom In Zoom Out Reset image size

To determine parallaxes for our targets we consider only the top third of CCD 7. This is sufficiently large that, at these magnitudes, we have enough reference objects for a transformation to a common system and sufficiently small that we can assume that the variation in astrometric distortion over the observational campaign is smaller than the errors of a linear transformation. Once the (x, y) coordinates have been determined, the parallax and proper motions are derived using the methods adopted in TOPP (Smart et al. 2003, 2007). Schematically, we transfer the base frame to a standard coordinate system using the UCAC2 stars, adjust all subsequent frames to this base frame using all common stars and a simple linear transformation, and find the relative parallax and proper motion of the target star by a fit to the resulting observations in the system of the base frame. In the z band, the atmospheric refraction is small and we assume the differential color refraction to be negligible (Stone 2002). To calculate the correction from relative to absolute parallax, we use the galaxy model of Mendez & van Altena (1996) in the z band to estimate the mean distance of the common stars. For the reference stars in these fields, this parallactic mean distance is of the order of 1 mas with an error of <30%. For more details on the parallax determination procedure and this correction, the reader is referred to Smart et al. (2003, 2007).

In Figure 4, we reproduce the solutions for two targets, 0719–50 and 1004–33. These two objects were also found to have companions as discussed below. In Table 3, we report the parallaxes for 10 objects observed in the early runs of the PARSEC program. Listed are: object ID, position, number of stars, number of frames, absolute parallax, absolute proper motions, epoch difference, and correction applied from relative to absolute parallax. In the following, we discuss some of these objects in more detail.

Zoom In Zoom Out Reset image size

0539-00. This object was found to have a photometric variability on a timescale of 13.3 hr (Bailer-Jones & Mundt 2001); we note that in our sequence the instrumental magnitude decreases by 0.05 ± 0.03 mag over the two year sequence. The radial velocity was found not to vary for two observations spaced four years, excluding a companion of mass greater than 10 M_J (Zapatero Osorio et al. 2007); the astrometric residuals also show no evidence for binarity.

0719-50. For this object we find μ_α, μ_δ = 198.1 ± 3.2, −61.4 ± 3.9 mas yr⁻¹, and it is identified by the PARSEC observations as a common proper motion companion of 2MASS 07193535−5050523 with μ_α, μ_δ = (200.3 ± 8.9, −67.6 ± 5.7). Both the parallaxes agree within the errors confirming the binary nature of this system. Both these objects have previous proper motion estimates, 0719–05 of 199.11 ± 20.49, −46.440 ±13.78 mas yr⁻¹ (Casewell et al. 2008) and 2MASS 07193535−5050523 of 206.2, −64.2 (Finch et al. 2007), but they were not noted as a common proper motion system. The analysis of the proper motion distributions in the range of magnitude and in the sky loci surveyed by the entire PARSEC program gives a probability smaller than 0.002 for a chance occurrence of such common pair of large proper motions. This chance becomes even smaller if the common distance is also considered. The brighter star has GSC 2.3 magnitudes of B_J= 16.01, R_f= 13.50 and I_n = 11.60 and 2MASS magnitudes of J = 10.33, H = 9.74, and K = 9.482. Combining the magnitudes and distance with the calibrations in Hawley et al. (2002), we find that the most consistent spectral type for this object is an M3–M4 dwarf.

0835-08. Cruz et al. (2003) find a spectroscopic distance of 8.3 pc in agreement with our distance of 8.5 pc, the nearest target in this sample and one of the nearest known L dwarfs to date. The astrometric residuals present no evidence of binarity, which, combined with the consistency of the photometric and parallactic distances, allows us to confidently say that this is a single system.

0909-06. This is considered the prototypical L0 object, marking the beginning of the L dwarf sequence with a temperature of 2200 K (Basri et al. 2000). We expect by the end of the program to have the distance to this object to better than 5%.

1004-33. We confirm, as suggested in Casewell et al. (2008) and Seifahrt et al. (2005), that this object is a binary companion of the nearby bright object LHS 5166. The proper motions and parallaxes of the two objects are both within 1σ of each other. The brighter star has GSC 2.3 magnitudes of B_J= 15.51, R_f= 13.33, and I_n = 11.29, and 2MASS magnitudes of J = 9.85, H = 9.30, and K = 9.03. Combining the magnitudes and distance with the calibrations in Hawley et al. (2002), we find the spectral type for this object is an M3 dwarf consistent with the dM4.5e found in Seifahrt et al. (2005).

1705-05. Reid et al. (2006) consider the possibility that this object is part of a binary system with a companion at a position angle of 5° and distance 136. This proposed companion was too faint to be frequently observed as part of our program; however, the color indicates a spectral type of T1–T2 that is inconsistent with the spectral type indicated by the apparent magnitude and distance of T7–T9. Hence, we conclude that this object is more likely to be a background late M dwarf at ∼200 pc rather than a companion to 1705–05.

1750-00. Based on spectroscopic observations, Kendall et al. (2007) found a distance of 8^+0.9_−0.8 pc and due to discrepancies in the spectral type indicators suggest that it may be a binary system of L5–L6 and L8–L9 dwarfs. We find that the trigonometric distance is consistent with the photometric one and do not find any evidence of binarity in the residuals which implies that it is a single system and the discrepancies must have some other explanation.

0641-43, 1018-29 and 1539-05. The first two objects are of type L1, and the third is of type L2. They are all in the 20–30 pc distance range. 0641-43 and 1539-05 are fast moving objects, with one of the proper motion components larger than 600 mas yr⁻¹. None of them was the subject of any particular discussion in the literature.

Figure 5 compares the trigonometric parallaxes from Table 3 versus the corresponding spectroscopic parallaxes based on the calibration of Knapp et al. (2004). The small number of sources precludes any direct conclusion; yet the targets asymmetry relatively to the equal values diagonal brings support to the importance of trigonometric parallaxes as the fundamental calibrators for spectroscopic distances.

Zoom In Zoom Out Reset image size

The parallax determination of the targets uses only the upper third of CCD7; however, the reduction pipeline is applied to the entire mosaic of eight CCDs. From these data, we have constructed a proper motion catalog, sampling the whole of the southern hemisphere with the exception of the lowest galactic latitudes where the number of known L/T dwarfs is significantly reduced. This proper motion survey can be used to search for companions to the targets, and other fast moving objects which will usually be nearby and/or sub-dwarfs. Combined with the magnitudes, the proper motion survey can also be used to build a reduced proper motion diagram to search for brown dwarf candidates. This catalog contains proper motion determinations for 195,700 objects.

Independently for each CCD and each observation, we have determined an astrometric reduction relative to the Second US Naval Observatory CCD Astrograph Catalog (UCAC2, Zacharias et al. 2004). The average number of reference stars was 20, with which polynomial functions were adjusted on right ascension (R.A.) and declination (decl.). Depending on the number of reference stars the polynomial degree was 2 or 3 and cross terms have been included. The rms errors of the solutions did not show any dependence on the type of the polynomial employed.

The proper motion determination was made from a match to the 2MASS point source catalog. In principle, the program frames should be complete with respect to 2MASS and the epoch difference is small, so a nearest neighbor match should be sufficient to not mismatch high proper motion objects. As a safety measure, the proper motions were determined for each observation combination and later averaged. Deviant values were removed from these averages, as they either came from unrecognized frame defects or faulty measurements. A more robust algorithm is being developed using the GSC 2.3 positions at different epochs. At the targets galactic latitudes, blending is rare and its effects would be negligible to the relatively bright 2MASS stars.

The histograms of the R.A. and decl. proper motion distributions are presented in Figure 6. The mean value for α is −2.8 mas yr⁻¹ (standard deviation σ = 12.1 mas), and −4.0 mas yr⁻¹ (standard deviation σ = 12.3 mas) for δ. This compares well with the corresponding values for the UCAC2 catalog in the same regions, which are: −2.7 mas yr⁻¹ (standard deviation σ = 14.6 mas) for α, and −3.6 mas yr⁻¹ (standard deviation σ = 30.1 mas) for δ. Zonal averages (3^h × 30°) also produced similar means for the PARSEC program and the UCAC2 catalog stars. Figure 7 compares the proper motions of stars in common with the PARSEC program and the UCAC2 catalog. Pearson's linear correlation coefficient is 0.95 both on R.A. and on decl. The largest difference appears for the smallest proper motions ( 25 mas yr⁻¹), the central parts in Figure 7, where the PARSEC values typically exceed those from the UCAC2 by 6%.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 8 is a similar comparison of the program target proper motions against values found in the literature. The linear correlation is 0.82. There are five outliers: 0523–14, 0559–14, 0624–45, 1828–48, and 1956–17. For each of the outliers a local comparison against the UCAC2 proper motions within the target's CCD has shown the same level of agreement found for the program as a whole. In the case of 0523–14 and 0624–45, the original proper motion paper (Schmidt et al. 2007) does not have any special discussion of these sources. Our results are obtained, respectively, from two and three PARSEC observations, and more data must be added on to reach a clearer understanding. For 0559–14, the original proper motion paper (Dahn et al. 2002) indicates that the proper motion was taken within just 2.1 yr, and the results should be taken as preliminary. For 1828–48 the original proper motion paper (Burgasser et al. 2004) indicates troublesome observations at high air mass and under patchy skies. For 1956–17 the original proper motion paper is from the SuperCOSMOS Sky Survey (Hambly et al. 2001) which for individual objects may have large errors, while our results include seven consistent PARSEC observations.

Zoom In Zoom Out Reset image size

The histograms of the proper motion errors are shown in Figure 9. The mean values are 5 mas yr⁻¹ both for R.A. and decl. The similarity of the behavior in both coordinates, already seen in the comparison to UCAC2 values and for the errors distribution, is shown in Figure 10, where the pair-wise R.A. and decl. proper motions are plotted for all objects. The actual proper motion distribution reflects several factors, prominently the galactic rotation, and a uniform distribution modulated by the inverse square of the distance. The combination of factors may result either in a Poisson-like or in a Gaussian-like distribution. We want to investigate whether the peculiar geometry of each CCD, and/or some artifact left by the astrometric reduction made independently for each of them would reflect on the proper motions distribution. In order to not assume an a priori model, the cumulative density distribution of the proper motions was fitted by an exponential decay, either along the R.A. or the decl. directions, and indeed when those are quadratically combined to produce the apparent sky motion. The exponential decays are characterized by the one free parameter which we call scale length. It is an uncomplicated, robust estimator borrowed from the description of stochastic processes in which events can occur continuously and are independent of each other. We calculated the scale length in density steps of 10 mas yr⁻¹ in order to investigate the presence of clumps, voids, or systematics within the parent proper motions population. This was done for each CCD, as shown in Table 4. From one CCD to another, the variation of the least-squares-adjusted scale length was always smaller than the rms of the adjustments, implying consistent astrometric precision from the different CCDs.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The proper motion catalog will be provided upon request.¹² We have chosen not to distribute it to the data centers until the observations are finished, and we have a final product. The above evaluation and our first release is based on the first year of the program, comprising six observation periods, from 2007 April up to 2008 April.

In Figure 11, we plot the reduced proper motion, H(K) = K + 5 × log(μ_tot) + 5, as a function of the z − K color. The z magnitudes come from a zero-point correction to the instrumental magnitudes of the first observations, the K are 2MASS magnitudes. The targets appear as white diamonds and delineate the brown dwarf zone. The field stars are represented by contour levels, from a 100 × 100 matrix of stars count. From the field stars furthermost in the targets zone possible brown dwarf candidates can be identified for spectroscopic follow-up.

Zoom In Zoom Out Reset image size

We have presented the first parallaxes from the PARSEC program. The results bode well for the whole program which is expected to finish in early 2011. We have confirmed a candidate binary (1004–33) and discovered another (0719–50), both of which will make good benchmark systems. The WFI has a large field of view, so we are able to put very good constraints on the wide binary systems and the parallaxes allow us to immediately isolate unresolved binaries because of their over luminosity with respect to their color. Given these two properties, and the large sample, we expect to be able to put sensible constraints on the binary fraction of brown dwarfs, a quantity that is critical for estimating the substellar mass function. We have produced a catalog of proper motions sampling the whole of the southern hemisphere. This catalog provides an independent validation of the UCAC2 proper motion system. The proper motion distributions are shown to be statistically well behaved; it follows that the proper motions for the fainter objects will have the same precision. We will continue to update this catalog online until the end of the program and plan to improve it including also GSC 23 database observations.

The authors acknowledge the support of the Royal Society International Joint Project 2007/R3, the PARSEC International Incoming Fellowship, and IPERCOOL International Research Staff Exchange Scheme within the Marie Curie 7th European Community Framework Programme. A.H.A. thanks CNPq grant PQ-307126/2006-0. J.I.B.C. acknowledges CNPq financial support 477943/2007-1. D.N.S.N. thanks FAPERJ grant E-26/110.177/2009.

This research has made use of the SIMBAD database operated at CDS France, the Second Guide Star Catalog developed as a collaboration between the Space Telescope Science Institute and the Osservatorio Astronomico di Torino, the 2MASS which is a joint project of the University of Massachusetts and the Infrared Processing and Analysis Center/California Institute of Technology, and the M, L, and T dwarf compendium housed at DwarfArchives.org and maintained by Chris Gelino, Davy Kirkpatrick, and Adam Burgasser.