37 NEW T-TYPE BROWN DWARFS IN THE CANADA–FRANCE BROWN DWARFS SURVEY

Spectroscopic campaigns with GNIRS (Elias et al. 2006) on Gemini-South were conducted during semesters 2006A and 2007A in queue service mode (Table 1), using the cross-dispersed mode. The six candidates observed with GNIRS were confirmed as T dwarfs. CFBDS J095914+023655 (T3.5 ± 0.5), CFBDS J100113+022622 (T5.0 ± 0.5), and CFBDS J193430−214221 (T3.5 ± 0.5) were previously published with less accurate spectral types in Delorme et al. (2008b) while CFBDS J151114+060742 (drawn from an RCS-2 field) is an independent rediscovery of SDSS J151114.66+060742.9 (T0 ± 2; Chiu et al. 2006) for which we provide an improved spectrum and a more accurate spectral type of T2.0 ± 0.5. CFBDS J152655+034536 and CFBDS J150000−182407 are new discoveries, and were, respectively, uncovered from CFHTLS (Very Wide) and RCS-2 fields.

The 068 slit width, the short camera, and the 31.7 l mm⁻¹ grating, together result in a resolving power of 900, with full wavelength coverage between 0.9 and 2.4 μm. A–B (not ABBA) sequences were used, with individual 100 s exposures for SDSS J151114.66+060742.9 and 10 minute exposures for CFBDS J152655+034536 and CFBDS J150000−182407. The total on-source integration times are, respectively, 13.33, 80, and 40 minutes. An early A-type (ideally A0) star was observed immediately before each sequence for relative flux calibration and telluric absorption correction. We used our own IDL procedures to extract and calibrate the spectra, as described in Delorme et al. (2008a). The sequence of spectral images is flat-fielded using observation of an internal flat-field lamp obtained immediately after the science frames. The five useful cross-dispersed orders are then extracted into five separate images, which are each corrected for order curvature. The trace of most targets disappears at some wavelengths (e.g., in the methane and water absorption troughs), and we therefore determine the curvature of the orders from the matching telluric reference star spectrum. The straightened order frames are then pair subtracted, removing most of the sky, dark current and hot pixels contributions. Each frame is collapsed along the spectral dimension to determine the positions of the positive and negative traces, and the spectrum is then extracted using positive and negative extraction boxes that we force to have opposite integrals; this minimizes the contamination by sky line residuals that survived the pair subtraction. The spectra derived from the individual image pairs are then median combined into one spectrum for each object. The spectra of the A0 telluric calibration stars are identically processed, interpolated over the hydrogen recombination lines, corrected for interstellar reddening, and divided by a 10,000 K blackbody spectrum to produce a spectrum of the telluric absorption. Each target spectrum is then divided by the derived telluric transmission spectrum. A first-order wavelength calibration is obtained from an argon-lamp spectrum, and fine tuned by registering the bright OH⁻ lines detected in the sum (rather than the difference) of the pair of images of interest.

Following the failure of GNIRS, our spectroscopic follow-up program was moved to Gemini-North and NIRI (Hodapp et al. 2003), where we observed 23 candidates during semesters 2007A, 2007B, 2008A, and 2009B. All were found to be T dwarfs, and all are new discoveries except CFBDS J005910−011401, a T–Y transition object which we previously discussed in Delorme et al. (2008a), and ULAS J145935+085751 codiscovered by Burningham et al. (2010b). For nine targets which turned out to have early/mid-T spectral types, we limited our spectroscopic follow-up to the H band. For those where the H-band CH₄ spectral index confirmed a late-T type, we additionally obtained J-band, and sometimes K-band, spectra.

All NIRI observations use the low-resolution f/6 camera, with the centered 6 pixel (075) slit with the H and K grisms and the 6 pixel (0696) blue slit with the J grism. The total integration times varied between 40 and 80 minutes, with individual exposure times of either 285 or 300 s. The processing recipes which we used for the NIRI data closely mirror those described for the GNIRS data, with small modifications to match a different observing strategy: we used multiple dithers (usually six or more) along the slit, separated by typically 2–3 arcsec, which yielded cosmetically cleaner spectra and a better signal-to-noise ratio than simple repeats of an AB pattern. The spectral resolutions (λ/Δλ) of the final spectra are approximately 480, 520, and 520 for the J, H, and K bands.

Southern candidates were preferably observed at the VLT using ISAAC (Moorwood et al. 1998) in queue mode during semesters 2007B, 2008A, and 2008B. In total, 15 candidates were observed in the H band with ISAAC of which one (CFBDS J151803+071645) was a reobservation of a NIRI target, so 14 are new discoveries. Two of these targets (CFBDS J025805−145534, a T1.5 dwarf and CFBDS J022638−072831, T6.5) have full JHK coverage with ISAAC.

We used the low-resolution grating for each of the J, H, and K bands with a slit width of 08 and a pixel scale of 0147 which yielded spectral resolutions of 650, 635, and 560, respectively. The observing strategy and data processing recipes followed very closely that used for NIRI.

In total, this paper presents spectra for 43 T-type BDs, of which 37 are new discoveries. Figure 3 presents their H-band spectra, all T-type BDs identified in the CFBDS and followed-up spectroscopically at Gemini and the VLT. Figure 4 presents the 1–2.5 μm spectra of the 14 targets for which we obtained broader NIR spectral coverage. Both figures include previously published CFBDS discoveries.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

All spectroscopic targets have J-band photometry, since they were selected based on their position in an i' − z' versus z' − J diagram. That J photometry was obtained at several observatories. Most was with SOFI at the NTT, with two different J filters (J and J_short). The rest was with NOTCam at the NOT and SQIID at Kitt Peak. More details on the J-band follow-up are given in Delorme et al. (2008b). We therefore used WIRCam (Puget et al. 2004) to obtain further Y, J, H, and/or K_s photometry on the MKO system (Simons & Tokunaga 2002) for several targets, to anchor the flux calibration of the spectra, determine spectroscopic parallaxes, measure proper motions, and identify color trends or peculiarities within our sample. The targets were positioned at the center of the northeast detector (array no. 60) and observed at 5–18 dithered positions within a 60'' radius circle. The WIRCam data were preprocessed with the 'I'iwi pipeline at CFHT and stacked using the Bertin software suite (SExtractor, Bertin & Arnouts 1996; Scamp, Bertin 2006; and Swarp, Bertin et al. 2002). We used the FLUX_AUTO method of SExtractor to extract the photometry, which for J, H, and K_s was anchored to a weighted average of the non-saturated 2MASS stars seen on the full mosaic. The Y photometry was calibrated through observation of a few CALSPEC standards, with no color correction. Table 2 summarizes all photometric measurements, including the MegaCam i'- and z'-band photometry as well as SOFI, NOTCam, and SQIID J-band photometry. The MegaCam photometry and the non-WIRCam J-band photometry were extracted with the point-spread function fitting method of SExtractor, as described in Delorme et al. (2008b).

We use the five spectral indices of the Burgasser et al. (2006) unified classification scheme (H₂O−J, CH₄ − J, H₂O−H, CH₄ − H, and CH₄ − K) to determine spectral types from the full NIR spectra. Additionally, we measure the K/J index, a good gravity probe in late-T dwarfs, the Warren et al. (2007) W_J index, and the Delorme et al. (2008a) NH₃ − H index. The latter probes H-band ammonia absorption beyond spectral type T8. To mitigate the influence of bad pixels in cosmetically poor spectra, we measure the indices after application of a 7 pixel wide median filter. We verified on cosmetically clean spectra (observed with multiple dithers) that this smoothing affects the measured indices by at most 10%. Table 3 presents the spectral indices which could be measured for each spectrum. These indices were converted to decimal subtypes using linear interpolation of Table 5 in Burgasser et al. (2006). We adopt as our best estimate the unweighted mean of the spectral types determined from the individual available indices, after experimenting with weighting schemes and finding that they typically affect the end result by at most a few tenths of a subtype. Upper/lower limits are rejected from the mean. Spectral types are rounded to 0.5 subtypes.

For targets possessing multi-band spectra, indices generally agree within ∼1 subtype. Exceptions are CFBDS 1001+02 (T5.0) for which the H₂O−J (T3.85) index is >1 subtype early; CFBDS 025814 (T1.5) for which the CH₄ − H (<T0) index is >1 subtype early; and SDSS 1511+06 (T2.0) which has discrepant H₂O−H, CH₄ − H indices by >2 subtypes. For targets with only H-band spectra, the average subtype difference (as defined by [CH₄ − H]−[H₂O−H]) is −0.2 with a 1σ scatter of 0.8 subtypes. Two targets have differences of +1.90 and −1.81 (discrepant at the 2σ level), i.e., CFBDS 0258−18 (T2.5) and CFBDS 2141−03 (T1.0), respectively.

After ULAS0034 and CFBDS0059, the two latest BDs isolated in this survey are CFBDS 1028+56 (T8.0) and CFBDS 0133+02 (T8.5). The unrounded average of spectral indices for CFBDS 0133+02 is T8.25 so it is intermediate between T8.0 and T8.5. We use the K/J and W_J indices normalized such that 2M0415 has unity and plot them in Figure 6 of Delorme et al. (2008a) to estimate a temperature and surface gravity of T_eff = 750–800 K and log (g) = 5.0–5.5. For CFBDS 1028+56, the unrounded average of indices is T8.09. A K-band spectrum is not available but very blue photometry (J − K = −0.87 ± 0.10, H − K = −0.47 ± 0.11—refer to Figure 7) indicates weak K-band flux and possibly a small K/J index. We estimate a temperature and surface gravity of T = 700–800 K and log (g) ≈ 5.5. Those estimates should not be taken at face value, for example, models overestimate temperature by 100 K in the case of Wolf 940B (Burningham et al. 2009). Nevertheless, they are indicative that both CFBDS dwarfs have high surface gravity while their thin disk kinematics indicate intermediate age ((0.5–5) × 10⁹ yr). They are thus probably on the massive side.

If we include all the T dwarfs discovered in the CFBDS or re-identified in our survey, 48 have confirmed spectral types, i.e., SDSS 2124−08 of type T5 (Knapp et al. 2004), SDSS 1504+10 (T7) and SDSS 1511+06 (T2) (Chiu et al. 2006), 2MASS 2139+02 (Burgasser et al. 2006) (T1.5), ULAS0034 (Warren et al. 2007) (T9), ULAS 0901−03 (T7.5) (Lodieu et al. 2007), CFBDS0059 (T9) (Delorme et al. 2008a), CFBDS 0959+02 (T3.5), CFBDS 1001+02 (T5.0) and CFBDS 1934−21 (T3.5) (Delorme et al. 2008b), ULAS 1459+08 (T4.5) (Burningham et al. 2010b), and the 37 presented in this paper. The whole 780 deg² survey yielded about 70 T dwarf candidates with i' − z' ⩾ 2.6 and z' − J ⩾ 2.5 of which 48 are spectroscopically confirmed.

The spectral type histogram (Figure 5) shows that a majority of the spectroscopically confirmed BDs (25) have types T2.0 to T4.5 while only five have early-T types. The presence of a peak at T2–T3 holds if we include 10 suspected T dwarfs based on additional NIR photometry. Furthermore, to investigate if the drop at T0–T1 is real or a product of incompletion of our spectroscopic follow-up, we broaden our T dwarf selection box. We add sources with z' − J ⩾ 3.0 and i' − z' ⩾ 2.0 (refer to Figure 1). The z' − J ⩾ 3.0 constraint corresponds roughly to the bluest T dwarf in our diagram and is about the color of a T0 dwarf from our empirical spectral type versus z' − J relation (see below). The i' − z' ⩾ 2.0 constraint represents the color above which our survey is complete (cf. Reylé et al. 2010, Figure 3). This larger selection box brings 37 new candidates. A good fraction is probably T dwarfs with peculiar i' − z' colors (low-metallicity and/or high-gravity) but L dwarf contaminants are likely predominant. We use the strong correlation that exists between spectral type and z' − J based on the 48 spectroscopic T dwarfs of our survey. The empirical relation is Tn = −15.93 + 5.55 × (z' − J) with a 1σ residual to the fit of 1.36 subtype (inset of Figure 5). We then applied the relation to the 37 sources lacking classification. If we assume that 100% of the sources are of T type despite their position within or close to the L dwarf box, the distribution peak at T2–T3 persists but the drop toward T0 is shallower. In reality, L dwarfs likely contribute a significant fraction of these 37 sources. That only strengthens the reality of the peak since L dwarfs are likely to contribute more in the T0 bin than in the T3 bin as they are generally bluer in z' − J than T dwarfs.

Zoom In Zoom Out Reset image size

We conclude that the peak at T2–T3 is not a product of our incomplete spectroscopic follow-up. The cause is instead that our magnitude-limited z'-band imaging survey probes a larger volume for the T2–T3 spectral types than for earlier T types. Indeed, the well-established J-band brightening persists in the z' band at the 0.3 mag level (Reylé et al. 2010) thus our CFBDS samples a volume about 50% bigger for T3 than for T0. That roughly agrees with our measured ratio of $\frac{{\rm T}2+{\rm T}3}{{\rm T}0+{\rm T}1}\approx 2$ from Figure 5.

Table 3 contains the spectrophotometric distances computed using the absolute magnitude versus spectral type relation of Liu et al. (2006). The uncertainties are about equally dominated by the adopted spectral type uncertainty of 0.5 and by the 0.35–0.40 mag scatter in the Liu relation.

It has been argued by Zapatero Osorio et al. (2007) that field BDs belong to a kinematically young population probably because they were discovered in shallow surveys not sensitive to old T dwarfs. They measured velocity dispersions smaller than that of the thin disk and found a high fraction of BDs with Hyades moving group velocities, but used a small sample of only 21 BDs. However, Faherty et al. (2009), using a volume-limited 20 pc sample, conclude that the known field BDs (including T dwarfs) have velocity dispersions consistent with that of the thin disk population. It is informative to know where our CFBDS objects lie.

Our proper motion measurements (Table 4) always use the z'-band MegaCam images as the first epoch; the second epoch image is a WIRCam image (in the band with best-image quality—usually K_s) when available. Except for two exceptions, targets without WIRCam photometry are not included here but will be the subject of a future paper as precision astrometry is more challenging with narrow field-of-view instruments used for the J-band follow-up.

The separation between the two epochs is usually more than two years. We match the astrometric frames of the two instruments by interpolating the MegaCam image to the frame of the near-infrared image, using Scamp (Bertin 2006) to derive a polynomial mapping between bright unsaturated stars of the two images and Swarp (Bertin et al. 2002) to produce the resampled MegaCam images. We then use SExtractor (Bertin & Arnouts 1996) to accurately measure the position of the T dwarfs (and its uncertainty) in both images for each epoch.

The proper motion uncertainties are about equally dominated by the epoch-to-epoch mean position scatter and the differential chromatic refraction (DCR) error. The astrometric mismatch between the two epochs is at the level of 001–002. We adopt the rms of a large number of unsaturated stars as the centroid uncertainty of our target. Due to the very steep slope of T dwarf spectral energy distributions in the z' band, we estimate that T dwarfs have a ∼40 Å redder effective wavelength in the z' filter than M-type field stars which provide the astrometric reference. The T dwarf therefore suffers significantly less refraction than its references, 001–002, depending on the air mass (we used the IDL program by Marchetti found on this Web site: https://www.eso.org/gen-fac/pubs/astclim/lasilla/diffrefr.html). The DCR is neglected in the NIR bands as its effect is an order of magnitude smaller.

We corrected for the parallax motion using the estimated distances of Table 3. For most targets, the motion amounts to <003 and represents a correction of <15% on the quoted proper motions. Two exceptions are CFBDS 1042+58, a nearby slow moving BD, whose parallax motion is 50% of its proper motion; and CFBDS 030226−14, with a 35% effect, because of a half-year observational phase shift.

Our proper motion measurement of ULAS0034 (μα = +11 ± 15 mas yr⁻¹, μδ = −366 ± 24 mas yr⁻¹) agrees with that of Smart et al. (2010) (μα = −20.0 ± 3.7 mas yr⁻¹, μδ = −363.8 ± 4.3 mas yr⁻¹) within 2σ in α and <1σ in δ.

Tangential velocities were derived for the 27 BDs with proper motions and estimated spectrophotometric distances. Most BDs in this paper have tangential speeds below 50 km s⁻¹ with a median of 37 km s⁻¹. This is a bit higher than the 29 km s⁻¹ found for the 20 pc volume-limited field T dwarf sample of Faherty et al. (2009, Figure 7). If the difference were statistically significant, then it could be interpreted as an age difference with our sample being older than the estimated 2.7^+1.0 _{− 0.8} Gyr for the field T dwarfs.

Our survey fields are not uniformly distributed over the sky so there may exist a bias in our median tangential velocity. We attempt to lift this bias by comparing velocities in the UVW local standard of rest, see Figure 6. UVW components were calculated using the IDL implementation (gal_uvw.pro) of Johnson & Soderblom (1987) and corrected for the solar motion. At R ≈500, one element of spectral resolution on our Gemini/VLT spectra corresponds to ∼600 km s⁻¹ yielding a precision no better than 60 km s⁻¹ assuming shifts of a tenth of a resolution element are achieved. That is too large for useful radial velocity measurements so we plot radial velocities of 0 km s⁻¹ and draw lines corresponding to ±25 km s⁻¹ which is roughly the 1σ velocity spread of solar neighborhood stars (Bochanski et al. 2007). Histograms are projected on the axis with the thin/thick disk Bochanski et al. Gaussians for reference.

Zoom In Zoom Out Reset image size

Ellipses represent the 1σ UVW uncertainties and, for clarity, are drawn only for the few kinematically peculiar targets. The ellipse of CFBDS 030225−144125 is typical for that of the rest of the sample. The arbitrary ±25 km s⁻¹ radial velocity is factored into the ellipse.

The kinematics of our sample is more consistent with that of the thin disk than that of the thick disk. Our own Gaussian fits to the histograms yield values of (σ_U, σ_V, σ_W) = (20, 22, 17) km s⁻¹ while that of the thin disk is (σ_U, σ_V, σ_W)_thin = (25.7, 20.9, 14.1) km s⁻¹ and that of the thick disk is (σ_U, σ_V, σ_W)_thick = (34, 29, 31) km s⁻¹. Our numbers have to be taken as lower limits because non-zero radial velocities can only expand the distribution, but likely not enough to mimic a thick disk distribution.

CFBDS J150000−182407 stands out as a particularly high-velocity BD with a tangential velocity of ∼190 km s⁻¹. It is well isolated in the UVW diagram with a V-component in the direction of galactic rotation of ∼146 km s⁻¹ and U and W components of 89 km s⁻¹ and 70 km s⁻¹, consistent with thick disk kinematics. With a spectral type of T4.5, CFBDS J150000−182407 is likely one of the first T-type sub-dwarf candidates with ULAS J141623.94+134836.3B (Burgasser et al. 2010). This is supported by the absence of the K i doublet at 1.25 μm indicative of low metallicity.

Three other BDs show marginally high velocities. CFBDS J223856+034947, a T2.0 dwarf, is a 2σ outlier in the UV plane with U = −71 km s⁻¹ but shows normal photometry and has a spectrum in H-band only.

CFBDS J030225−144125, a T6.0 dwarf at 38 ± 10 pc, is a <2σ outlier in both UV and VW planes. Its photometry is within the color trends so there is no other sign of peculiarity.

Finally, CFBDS J090449+165347, a T4.0 dwarf at 65 ± 12 pc, is a <2σ outlier only in the VW plane but does not show other sign of peculiarity.

It is interesting to compare the colors of our BDs to the literature to identify outliers and perhaps discover new trends. Figure 7 presents the NIR colors of the CFBDS spectroscopic sample as a function of spectral type. Note that several of the 48 BDs do not have full NIR photometric follow-up. Our photometry is plotted as open squares while small plus signs are from Knapp et al. (2004), small crosses are from Chiu et al. (2006), and open triangles are late T8+ dwarfs recently discovered: Wolf 940B (T8.5, Burningham et al. 2009), ULAS 1017 (T8.0), ULAS 1238 (T8.5) and ULAS 1335 (T9.0) (Burningham et al. 2008), CFBDSIR 1458+10 (T8.5, Delorme et al. 2010), ULAS 1302+13 (T8.5, Burningham et al. 2010b), and UGPS 0722−05 (T9/T10, Lucas et al. 2010).

Zoom In Zoom Out Reset image size

The J − H, H − K_S, and J − H plots follow the same general trends as that in the literature. The colors of T dwarfs get bluer with later types until a plateau with larger dispersion is reached at around T6/7. The relation with spectral type is tighter for the J − H color and the spread is larger in the H − K_S plot.

The noteworthy outliers in these diagrams are CFBDS 0226−07, CFBDS 0301−16, CFBDS 0959+02, CFBDS 1042+58, CFBDS 2037−19, and CFBDS 2330−08. In the H − K_S diagram, CFBDS 0226−07, CFBDS 0301−16, and CFBDS 1042+58 all have very red colors for their spectral type, respectively, +0.35, +0.92 ± 0.12, and +0.46 ± 0.06.

The T7 dwarf CFBDS 0301−16 is very peculiar, lying off the mean relation also in J − K_S. According to BT-Settl models, small changes in surface gravity have very significant effects on the H − K_S and J − K_S colors at that spectral type. When gravity goes from log (g) = 4.5 to log (g) = 3.5, H − K_S jumps by +1.5 mag and J − K_S by +3 mag. The surface gravity of CFBDS 0301−16 is possibly small, in the regime of log (g) = 3.5–4.0 if we take the models at face value. Assuming T_eff = 800–1000 K (based on the Stephens et al. (2009) temperature versus spectral type relation), then blindly applying the evolutionary models of Baraffe et al. (2003) suggest a mass and age of ⩽9 M_Jup and ⩽100 Myr. Higher metallicity could also produce similar effects. We also caution that models have notorious difficulties in reproducing BD colors so conclusions are only indicative.

Similar conclusions of lower surface gravity and/or higher metallicity are reached for T6.5 CFBDS 1042+58 and T6.5 CFBDS 0226−07. If we assume log (g) = 4.0–4.5 and T_eff = 800–1100 K, then the mass and age possibly are ⩽15 M_Jup and ⩽500 Myr. In the case of CFBDS 1042+58, youth is also supported by the very small tangential velocity measured at 8 ± 6 km s⁻¹.

CFBDS 0959+02 shows blue colors for a T3.5 dwarf (H − K_S = +0.07 ± 0.16, J − K_S = +0.08 ± 0.19) and is one of the two BDs isolated in the CFHTLS Deep fields with a distance of about 150 ± 31 pc. Again, the H − K_S color is very sensitive to gravity and this dwarf possibly has a large log (g) ≈ 5.0–5.5 or low metallicity. Assuming T_eff = 900–1200 K and using the Baraffe et al. (2003) models at face value yield a large mass and old age of ⩾30 M_Jup and ⩾1.0 Gyr.

CFBDS 2037−19 is 0.5 mag redder in the J − H diagram than any other early-T dwarf in the literature. But the same dwarf follows the normal trends in other color–color diagrams which challenges a model atmosphere-based explanation. Perhaps this L/T transition object is variable?

Finally, we note that one of our two T8 dwarfs, CFBDS J102841+565402, is nearly as blue as CFBDSIR J145829+ 101343 (Delorme et al. 2010), second to only SDSS J141623.94+134836.30B (Burningham et al. 2010a) among the bluest T8/T9 dwarfs for both of the H − K_S and J − K_S color plots. This is an indication of high gravity and/or low metallicity.

Direct imaging surveys (Gizis et al. 2003; Bouy et al. 2003) that probe separations of >2 AU show that ∼10%–15% of field BDs are binaries with separations that peak between 2 and 4 AU. Lower separations probed by spectroscopic surveys (see Joergens 2008, and references therein) contribute a further 7⁺⁵ _{− 3}% (separations <0.3 AU) or 10⁺¹⁸ _{− 8}% (separations <3 AU). The true BD binary fraction is likely between 15% and 25%. For our sample of 48 dwarfs, we therefore expect about 7–11 binary BDs.

A laser guide star adaptive optics (LGSAO) program at Keck is underway (Liu & Dupuy) to follow-up CFBDS candidates and will be the subject of a future publication. What follows is a detailed discussion for three systems for which binarity is suspected or has been constrained by means other than the LGSAO program.

• 1.  
  CFBDS J022644−062522. This T3.0 BD is found ∼40'' from HD 15200, a K0 star, whose Hipparcos parallax indicates a distance of 43 pc (parallax = 23 mas) while its spectrophotometric distance is 50 pc (based on an absolute magnitude of V = +5.5 for K0 dwarfs and an apparent magnitude of V = 8.98). CFBDS 0226−06 is estimated to be at 60 ± 8 pc so their distance is similar within ∼1.5σ. Proper motions alone are consistent within 1σ. The BD proper motion is measured at μ_tot = 46 ± 40 mas yr⁻¹ or (μ_αcos (δ), μ_δ) = (+32 ± 41, −33 ± 39) mas yr⁻¹ while that of the K0 star is μ_tot = 22.3 ± 1.8 mas yr⁻¹ or (μ_αcos (δ), μ_δ) = (+21.8 ± 1.2, +4.6 ± 1.3) mas yr⁻¹ (Høg et al. 2000). Unfortunately, proper motion measurements suffer from a large uncertainty because the second epoch image is a narrow field-of-view SOFI image with poorer astrometry.If the system is physically associated, assuming a distance of 50 pc, the apparent separation is ∼200 AU. The list of star/T dwarf systems is GJ 229B (T8+M1) (Nakajima et al. 1995), GJ 570D (T7.5+K4+M1.5+M3) (Burgasser et al. 2000), Epsilon Indi Bab (T1.5+T6+K5) (Scholz et al. 2003), HN PegB (T2.5+G0) (Mugrauer et al. 2006), HD 3615B (T7.5+K0) (Luhman et al. 2007), Wolf 940B (T8.5+M4) (Burningham et al. 2009), G 204-39B (T6.5+M3) (Faherty et al. 2010), Ross 458C (T7+M0.5+low mass) (Goldman et al. 2010), and Hip 73786B (T6.5+K5) (Scholz 2010). Only HN Peg and Epsilon Indi have an early T companion which makes CFBDS 0226−06 potentially interesting to study the L/T transition. Our search of the literature did not yield interesting information on HD 15200 beside its proper motion and photometry.
• 2.  
  CFBDS J030225−144125 and CFBDS J030226−143719. These two dwarfs, of spectral types T6 and T4, are separated by only 4 arcmin on the sky (mostly along the North–South direction). The WIRCam J-band magnitudes of, respectively, 17.75 ± 0.03 and 17.50 ± 0.03 (Table 2) and absolute J magnitudes of 14.7 and 14.3 for T6 and T4 dwarfs (Knapp et al. 2004, Figure 8) result in spectroscopic distances of 38 ± 10 pc and 51 ± 10 pc. Their difference is well within the dispersion of the magnitude versus spectral type relation, suggesting that they might form a bound pair. A few physical pairs of low-mass stars or BDs are separated by over 1000 AU (Artigau et al. 2007; Caballero 2007), but at 40 pc the separation of this CFBDS pair would be ∼10, 000 AU. The two proper motions however are incompatible, (μ_αcos (δ), μ_δ) = (+ 314 ± 19, −192 ± 13) mas yr⁻¹ versus (μ_αcos (δ), μ_δ) = (+ 87 ± 12, −57 ± 15) mas yr⁻¹, and the pair instead serves as a reminder that objects as rare as T dwarfs (with a density under 1 per 10 deg², to the CFBDS limiting magnitude) can occasionally pair within a few arcminutes.
• 3.  
  CFBDS J102841+565402. This BD is within 15'' of a r = 15.7 and i = 14.6 star, bright enough for use as a reference star for the CFHT PUEO adaptive optics system (Rigaut et al. 1998). A 30 minute H-band integration on 2008 April 17 produced an image with FWHM = 02 and no sign of elongation. For equal luminosity components, we thus conservatively rule out a companion with a separation >2.5 astronomical units.