A THERMAL INFRARED IMAGING STUDY OF VERY LOW MASS, WIDE-SEPARATION BROWN DWARF COMPANIONS TO UPPER SCORPIUS STARS: CONSTRAINING CIRCUMSTELLAR ENVIRONMENTS^*

The L and M Infrared Camera (LMIRCam) is the 3–5 μm channel of LBTI (Hinz et al. 2008; Skrutskie et al. 2010). The field of view (FOV) with an f/15 beam is 11'' square, with 10.7 mas pixels (calibrated 2011 November by Rodigas et al. 2012). We assume the instrument rotation angle is repeatable to 1°. Field distortion has not been measured observationally; ZEMAX models of the LMIRCam optics predict field distortions of <2% over distances <5''.

The LBT AO system employs a deformable secondary mirror in order to minimize the number of warm surfaces along the optical path. As a result, it benefits from increased sensitivity in the thermal IR (Lloyd-Hart 2000). For system details see Esposito et al. (2011, E11). The LBTI AO hardware design is identical to that used in E11, but the calibration is independent. We corrected 200 aberration modes, rather than the 400 modes E11 used to achieve >80% Strehl ratio (SR) in H band. At the time of these observations, the LMIRCam PSF was degraded due to residual instrumental astigmatism (which has since been mitigated). Nonetheless, because wavefront phase errors decrease with increasing wavelength, our observations approached the diffraction limit. During this run, we achieved an SR on LMIRCam of ∼80% at M' on α CrB (R = 2.2) at air mass ∼1. Data recorded directly from the wavefront sensor itself, not subject to the LMIRCam instrumental astigmatism, indicate the AO system was delivering ∼90% SR during these observations. The science observations were executed under more difficult conditions: low elevation (high air mass), increased winds, and with fainter guide stars, and so were subject to poorer AO performance than the α CrB observations. Thus the sensitivities we derive in this paper are somewhat lower than those which can now be achieved with LBTI on more favorable targets (e.g., Rodigas et al. 2012; Skemer et al. 2012) or with its recently improved AO calibrations.

We observed GSC 06214 and HIP 78530 on UT 2011 May 26. During this LBTI commissioning run, only the AO system on the right LBT aperture was fully operational, so we observed in single aperture mode. The observing strategy for these, and all other LBTI observations, was angular differential imaging (ADI; Marois et al. 2006): the instrument rotation angle was fixed, and the field was allowed to rotate on the detector over the course of the observations. We used a four position nod pattern for taking image data, with nods of 15–3''. Due to the high sky background, we were limited to very short exposures: 87 ms at M' and 146 ms at L'. To limit data volume, 200 individual exposures were coadded per single saved image. We obtained 30 minutes of M' data on GSC 06214 and 15 minutes of L' data on HIP 78530. The FWHM (average of major and minor axes of the astigmatic PSF) for the HIP 78530 (L') and GSC 06214 (M') images were 012 and 018, respectively.

Clio is also a narrow-field 3–5 μm imager (Sivanandam 2006). At the time of these observations it was installed on MMT; it is now at the Magellan Clay telescope. The MMT AO system has an architecture similar to that of LBT AO, with a deformable secondary mirror correcting 52 aberration modes (Brusa et al. 2004). In f/15 mode, the Clio plate scale is 29.9 ± 0.1 mas pixel⁻¹, and the FOV of the entire array is 15'' × 30'' (calibrated 2010 July using the binary system HIP 88817). We assume the instrument rotation angle is repeatable to 01. Plate scale distortion has not been measured, but ZEMAX models predict distortion of <0.2% over 5'' separations.

We observed 1RXS 1609 and GSC 06214 on UT 2011 April 17 and 22, respectively. In order to decrease time lost to detector readout, we used only a subarray with an FOV of 9'' × 30''. Integrations were 10 coadds of 5–6 s at 3.1 μm and 40 coadds of 1 s at 3.3 μm. The data were taken in ADI mode in a 15'' two-position nod, with 1–2'' dithers between nods. These observations suffered from moderate wind from the south that induced vibrations in the telescope and caused smearing in some of the images. The resultant FWHM ranged from 020–035, depending on the severity of the vibrations. On GSC 06214, we obtained approximately 45 and 20 minutes of on-source integration in the 3.1 μm and 3.3 μm narrowband filters, respectively. On 1RXS 1609 we obtained 45 minutes of integration at 3.1 μm and 30 minutes at 3.3 μm.

Each system had been previously observed by MIPS. 1RXS 1609 and HIP 78530 24 μm photometry are published in Carpenter et al. (2009), and we refer the reader to this publication for a description of the observations. For GSC 06214, we analyzed archival MIPS data taken as part of the Spitzer Legacy program: Gould's Belt (PID 30574; PI: L. Allen), which aimed to provide a complete census of star-forming clouds in the solar neighborhood. GSC 06214 was imaged in two AORs: 20000768 and 20001280, obtained with the fast scan rate (3 s per frame). The system was located near the end of the scan map; therefore, only five frames per pointing were obtained near the source position with an equivalent integration time of 15 s at 24 μm. We detected the source only at 24 μm due to the shallowness of the survey; we found no source in the corresponding 70 and 160 μm maps. Therefore, we only discuss the 24 μm result hereafter. Multiple frames were mosaicked based on the image World Coordinate System pointing information and subsampled by a factor of two, resulting in a plate scale of 1245 pixel⁻¹.

Custom MATLAB scripts were used for reducing the LBTI and Clio data. Background, bias, and dark subtraction were achieved by differencing images from adjacent nods. Bad pixels were flagged using a nearest-neighbors median filter and replaced with that median. In LMIRCam, each of the channels had a small time-varying bias level offset as well as variable pattern noise. We subtracted the median value of each channel to remove the offset. Then we subtracted a two-dimensional template of the channel read noise created from the median of all channels. For those calculations we masked out a ∼2'' square region containing the source to avoid biasing the median. Finally, we aligned the images according to their position angle, rotating so north was up.

These observations did not achieve sufficient on-sky rotation to synthesize and subtract the primary's PSF, as per Marois et al. (2006). However, all secondaries were sufficiently separated from their primaries (>2'') that contamination from the primary was negligible. To produce the final image, we mean-combined our set of rotated images together, and 3σ–4σ outliers for each pixel in the image stack were clipped.

Aperture photometry was used to extract the differential flux between primary and secondary. None of the objects were resolved, so we used aperture radii of ∼1 FWHM. The background was estimated using neighboring apertures of the same radius at the same separation from the host star. No aperture correction was necessary, because the same aperture radius was used for both the primary and secondary.

We next calculated sensitivity as a function of radial separation from the primary star. Residual low-spatial frequency offsets were removed by subtracting a smoothed image (Gaussian FWHM = 10λ/D) from the science image. The resulting image was then smoothed with a disk the same diameter as the aperture used for photometry. The noise, σ, as a function of separation was calculated by measuring the standard deviation of 1 pixel wide annuli centered on the primary. From this we calculated the signal-to-noise ratios (S/Ns) of our detections. The uncertainty in our photometry (in magnitudes) is given by 2.5 · (ln (10) · S/N)⁻¹. The noise counts as a function of radius were also converted to contrast (delta magnitudes between the primary star and the measured quantity, noise in this case) by comparing to the peak pixel value of the primary star in the smoothed image.

The primary star photometry in our filters was estimated by interpolating between published photometry in adjacent filters (i.e., between shorter and longer wavelength photometry), because no standard stars were observed. There are no strong spectral features in K7 or B9 stars between 2 μm and 5 μm which would bias this interpolation. The Two Micron All Sky Survey (2MASS) and WISE surveys recorded unresolved photometry for each system. Additionally, Carpenter et al. (2006) conducted a Spitzer IRAC survey which included 1RXS 1609 and HIP 78530. As we will show, the BD contribution to the system flux at 1–5 μm is negligible, so we can treat 2MASS, WISE W1 and W2, and Spitzer IRAC photometry as measurements of the primary star only. Where available, we also used resolved photometry at additional wavelengths from previous work. The data used for interpolation are listed in Tables 2–4.

Table 3. 1RXS 160929.1-210524 System Properties and Apparent Photometry

Property	1RXS 1609A	1RXS 1609B
Distance (pc)a	145 ± 20
Spectral typeb	K7V	L$4^{+1}_{-2}$
Massb	$0.85^{+0.20}_{-0.10}$ M☉	$8^{+4}_{-2}M_{{\rm Jup}}$
Teffb	$4060^{+300}_{-200}$	$1800^{+200}_{-100}$
log (L/L☉)b	−0.37 ± 0.15	−3.55 ± 0.2
Separation ('')c	2.15 ± 0.03
PA (°)c	28.3 ± 0.4
J	9.820 ± 0.027d	17.90 ± 0.12b,e
H	9.121 ± 0.023d	16.87 ± 0.07b,e
K	8.916 ± 0.021d	16.19 ± 0.05b,e
3.1 μmc	8.80 ± 0.05	15.65 ± 0.21
3.3 μmc	8.78 ± 0.05	15.2 ± 0.16
3.4 μmg	8.767 ± 0.023
L'e,f	8.73 ± 0.05	14.8 ± 0.3
4.5 μmh	8.80 ± 0.01
4.6 μmg	8.779 ± 0.021
8.0 μmh	8.735 ± 0.005
12 μmg	8.715 ± 0.027
24 μmi	8.43 ± 0.01

Notes. ^aMean cluster distance (de Zeeuw et al. 1999; Preibisch et al. 2002). ^bLafrenière et al. (2008). ^cThis work. ^d2MASS J/H/K_S survey (Skrutskie et al. 2006). ^eMKO J/H/K'/L'. ^fLafrenière et al. (2010) ^gSystem: WISE survey (Wright et al. 2010). ^hSystem: Spitzer IRAC (Carpenter et al. 2006). ⁱ System: Spitzer MIPS (Carpenter et al. 2009).

Download table as:  ASCIITypeset image

We reduced the raw 24 μm data of GSC 06214 using the MIPS Data Analysis Tool (Gordon et al. 2005) with post-pipeline processing to improve flat fielding and remove instrumental artifacts as detailed in Engelbracht et al. (2007). We extracted the photometry using PSF fitting with the IDL-based StarFinder program (Diolaiti et al. 2000). The input PSF was constructed using observed stars and a smoothed STINYTIM model PSF; it has been tested to ensure the photometry results are consistent with the MIPS calibration (Engelbracht et al. 2007).

The final Clio images at 3.1 μm and 3.3 μm are shown in Figure 2. At 3.1 μm we measured a contrast between "A" and "B" of 5.26 mag, and at 3.3 μm we measured a contrast of 5.30 mag. Using the procedure described in Section 2.3.1, we inferred the 3.1 μm and 3.3 μm magnitudes of the primary to be 9.11 ± 0.05 and 9.10 ± 0.05, respectively. The 5σ sensitivity limits for these two filters were magnitudes 15.82 and 16.14, yielding signal to noise ratios of 18 and 23 for the detections in each wavelength. From this we derive, [3.1]_B = 14.41 ± 0.08 and [3.3]_B = 14.40 ± 0.07 (Table 2).

Zoom In Zoom Out Reset image size

We calculated a separation of 217 ± 002 and a position angle of 1759 ± 01, consistent at the 1.5σ level with astrometry measured by Ireland et al. (2011). Because the companion was not detected in individual frames, we estimated the uncertainty by comparing the values calculated from the final images at each wavelength. Our quoted uncertainty accounts for this comparison as well as our assumed uncertainties in calibration and distortion (see Section 2.2.2).

The final LMIRCam M' image is also presented in Figure 2. The BD was detected at a contrast of 4.65 mag. For the primary, we estimated $M^\prime _A$ = 9.10 ± 0.05. Based on the 5σ sensitivity limit of 13.39 mag, the BD detection has S/N = 3.6, and therefore $M^\prime _B$ = 13.75 ± 0.3. Even though this S/N is lower than that nominally adopted to identify candidate companions in high-contrast imaging data (5σ–6σ; Marois et al. 2008a), we also detect point sources at the same location in the 3.1 μm and 3.3 μm images, justifying a lower detection threshold. The bright spot ∼1'' to the east of the primary does not appear in the more sensitive 3.1 and 3.3 μm images and is almost certainly an artifact.

The 24 μm flux density of the system is 3.3 ± 0.3 mJy (8.34 ± 0.1 mag); the error is dominated by shot noise.

GSC 06214A did not have a published extinction value, so we performed our own estimate. We found A_V ⩽ 0.6 using the following prescription. We compared the primary's flux-calibrated optical spectrum (courtesy Bowler et al. 2011) with Pickles (1998) spectral standards (Figure 3). The observed and template spectra were normalized by their 800–900 nm averages. Interstellar extinction curves with R_V = 3.1 (Cardelli et al. 1989) were used. Based on χ² fitting, both K7V with A_V = 0 and K5V with A_V = 0.6 provide acceptable fits to the data. K4V and A_V = 0.85 is a worse fit to both the continuum and spectral features. In order to make the most conservative estimate of the IR excess in the system, we adopt the maximum allowed extinction value: A_V = 0.6 mag.

Zoom In Zoom Out Reset image size

We placed GSC 06214 in infrared color–color diagrams, compared to several observational brown dwarf and planet samples: field brown dwarfs (Leggett et al. 2010), very young substellar objects (L2010), the young planetary-mass object 2M 1207b (Chauvin et al. 2004; Mohanty et al. 2007; Skemer et al. 2011; Barman et al. 2011), and the planets HR 8799bcd (Marois et al. 2008b, 2010; Currie et al. 2011; Galicher et al. 2011; Skemer et al. 2012). The L2010 sample of M0−L0 dwarfs was derived from dereddened observations of Taurus, Chameleon, and TW Hya, so it provides the most appropriate comparison sample for our similarly young brown dwarfs.

In order to conduct a homogeneous analysis, we transformed all previously published J, H, and K BD photometry to the MKO J/H/K system. For 2MASS J/H/K_S → MKO J/H/K, we used the transformations published in Leggett et al. (2006). For MKO K' → K of stars, we used Wainscoat & Cowie (1992); however, we could not apply this transformation to the BD, because it is only calibrated for H − K < 0.4. Therefore, we calculated the K' → K transformations by integrating over DUSTY evolutionary model spectral energy distributions (SEDs) (Chabrier et al. 2000). We took the SED that most closely matched the color of the BD in our filters of interest (in this case, H−K). We used the same procedure for all color transformations for which empirical relations were not available. We calculated K = 14.75 mag for GSC 06214B. We then dereddened all GSC 06214B photometry, using the extinction value for the primary (A_V = 0.6) discussed in Section 3.1.2. We presumed the extinction toward the primary and secondary was the same, a nominally valid assumption, though counterexamples do exist (e.g., GG Tau, White & Ghez 2001). Figure 4 shows the color–color diagrams H − L'/J − H and H − M'/J − H.

Zoom In Zoom Out Reset image size

GSC 06214B was previously classified using J/H/K spectra as a spectral type of L0 ± 1 (Bowler et al. 2011). Its J−H color is consistent with an L2010 spectral type of M9. The excesses at K, L', and M' compared to an L2010 M9 dwarf photosphere are 0.15 ± 0.05, 0.35 ± 0.07, and 0.4 ± 0.3 mag, respectively, accounting for photometric errors only. Fortuitously, the measured excesses are similar even if we use A_V = 0, because reddening moves the object up and right in color–color space parallel to the line connecting L2010 M9 and L0 spectral types. Therefore, our conclusions about the color excess for this object are independent of the extinction estimate, over the allowed range of extinction values.

The GSC 06214 system was imaged with Spitzer/MIPS at 24 μm. The PSF is broader than the host-companion separation, so it is not possible to measure the primary and secondary objects' emission independently. However, the BD's photosphere does not contribute significantly at this wavelength. Whether based on a Rayleigh–Jeans extrapolation from short wavelength data or using DUSTY model SEDs, the BD's photospheric emission is expected to be less than 1% of the observed 24 μm flux.

We compared the MIPS photometry to the 24 μm flux predicted for the "A" component by an empirical polynomial relation between V − K_S and 24 μm for main sequence stars (Urban et al. 2012). The 24 μm flux prediction is robust to order 0.01 mag between early B and early M spectral types. GSC 06214A is likely in the final stages of contraction onto the main sequence. However, the variation in V − K_S color, and therefore in inferred 24 μm flux, between this object and a main sequence star, is less than the combined observational uncertainties at V, K_S, and 24 μm.

From this comparison, we find evidence of a 24 μm excess in the system. The observed system flux is 3.3 ± 0.3 mJy, while the predicted flux from "A" is 1.73 ± 0.15  mJy. This translates to an excess of 91 ± 24% above the expected photospheric emission from "A." The uncertainty is dominated by the low signal to noise of the 24 μm data. Our detection corroborates the 2σ excess at WISE 22 μm found by Bowler et al. (2011). Therefore, one or both members of the system must retain a warm excess.

No additional objects are detected in our observations. As discussed in Section 3.1.1, the candidate at 1'' separation in M' is likely an artifact. In our collection of images, we could have detected additional companions of mass ≳ 20 M_Jup at separations ≳ 150 AU from the primary, with a limit of ∼10–12 M_Jup beyond 200 AU. Previous aperture masking and direct imaging work have placed deeper limits of 35/15/5 M_Jup at 10/150/300 AU from the primary (Kraus et al. 2008; Ireland et al. 2011). We would have resolved an equally bright binary companion to "B" (i.e., "Bb") if it was >20 AU projected separation from "B."

Figure 5 shows the final 1RXS 1609 narrowband images. We estimated the primary's 3.1 μm and 3.3 μm magnitudes to be 8.80 ± 0.05 and 8.78 ± 0.05, respectively. At 3.1 μm we found a contrast between "A" and "B" of 6.85 mag. The 5σ background sensitivity limit of 15.71 mag yielded S/N = 5.3 for the detection, thus [3.1]_B = 15.65 ± 0.21. At 3.3 μm we found 6.4 mag contrast, a sensitivity limit of 16.14 mag, S/N = 7.5, and [3.3]_B = 15.2 ± 0.16 (see Table 3).

Zoom In Zoom Out Reset image size

We measured a separation of 215 ± 003 at PA = 283 ± 04, consistent with Lafrenière et al. (2010) at the 2σ level. These values represent the average of the astrometry calculated at 3.1 μm and 3.3 μm. Errors account for measurement uncertainties as well as our assumed calibration and distortion uncertainties (see Section 2.2.2).

There is no extinction toward the 1RXS 1609 system (Carpenter et al. 2009). We transformed the K' companion photometry to yield K = 15.99 ± 0.1, following the procedure described in Section 3.1.3. 1RXS 1609B is shown in the color–color diagrams in Figure 4.

Unfortunately, the L2010 sample does not extend past spectral type L0; we therefore compared 1RXS 1609B broadband colors to DUSTY model SEDs with temperatures of 1700 K and 1800 K. A surface gravity of log (g) = 4.0 is predicted by the DUSTY models for these effective temperatures at 5 Myr. We scaled the SEDs so their integrated J band flux matched the observed flux. Both SEDs provide an acceptable fit to the data (Figure 6), consistent with the Lafrenière et al. (2010) effective temperature estimate. Our new photometry confirms the lack of near-IR excess.

Zoom In Zoom Out Reset image size

Following the procedure used in Section 3.1.4, we find evidence for a moderate 24 μm excess in the system. The observed flux is 3.06 ± 0.04 mJy, and the predicted flux from "A" is 2.15 ± 0.07 mJy, an excess of 42% ± 6% above the expected stellar photospheric emission. The error is dominated by the uncertainty in the literature V magnitude. As with GSC 06214, the predicted contribution to the excess from the BD's photosphere is <1%. Neither "A" nor "B" exhibits a near-infrared excess, and the components have not been resolved at wavelengths longer than L'. Thus, though we can detect the presence of a warm excess in the system, we cannot determine whether it is around the primary, BD, or both.

We do not detect any additional objects in the 1RXS 1609 system. Our background limit at 3.3 μm was 15.6 mag, or ∼5 M_Jup, extending out to 500 AU. We do not improve upon previous detection limits of 8/1 M_Jup at 50/450 AU stellocentric distance (Lafrenière et al. 2010). We could have detected an equally luminous binary companion to "B" (i.e., "Bb") at a separation >25 AU from "B."

Figure 7 shows the final L' image of HIP 78530, with the BD to the southeast. For the primary, we inferred $L^\prime _A=6.87\pm 0.05$. HIP 78530B was detected at a contrast between "A" and "B" of 6.93 mag (Table 4). The 5σ sensitivity limit was L' = 16.21, giving S/N = 47 and $L^\prime _B= 13.80\pm 0.06$. Our measured separation and PA are 454 ± 009 and 1407 ± 1°, including plate scale and distortion uncertainties (see Section 2.2.1), consistent with previous observations.

Zoom In Zoom Out Reset image size

The extinction toward HIP 78530 is A_V = 0.5 (Carpenter et al. 2009), and we dereddened the photometry using this value. We calculated a K magnitude for the BD of 14.12 ± 0.06 following the prescription described in Section 3.1.3.

We find the spectral types derived for HIP 78530B from broadband 1–4 μm colors and from near-IR spectra are not consistent. HIP 78530B was classified using J/H/K spectra as M8 ± 1 (Lafrenière et al. 2011), though the authors noted the object was ∼0.2 mag too blue at K for this spectral type. We find that its H − L' color is also ∼0.5 mag too blue compared to an L2010 M8-type BD (Figure 4, left panel). There is a similar mismatch in both H−K and K − L' colors (Figure 8, left panel). With the addition of our new L' data, we find the object's J through L' colors are not consistent with an M8, but instead with an L2010 M3 ± 2. Intriguingly, the object's absolute magnitude is consistent with other USco M8-type BD (Luhman & Mamajek 2012) as shown in Figure 8 (right panel). We discuss the implications of this discrepancy further in Section 4.2. Regardless of whether we adopt spectral class M8 or M3, we find no evidence for K or L' excesses.

Zoom In Zoom Out Reset image size

The observed and predicted fluxes for the HIP 78530 system are 13.10 ± 0.12 mJy and 12.0 ± 0.5 mJy. The excess is 9 ± 4.5%. At 2σ, we do not consider it statistically significant. This, combined with a lack of 1–5 μm excess around both "A" and "B," suggests that neither object retains a massive warm disk.

No additional objects are detected in our data. Figure 9 plots our new contrast and mass limits as a function of projected stellocentric distance; we adopt the Hipparcos system distance of 157 pc (van Leeuwen 2007). We reach a sensitivity of 20 M_Jup beyond 100 AU and ≲ 5 M_Jup beyond 175 AU. In the east half of the system, our FOV extended to 800 AU, while on the west side, the FOV ended at ∼200 AU. An equally luminous binary companion to "B" (i.e., "Bb") is ruled out at >15 AU separation from "B."

Zoom In Zoom Out Reset image size