A DISK AROUND THE PLANETARY-MASS COMPANION GSC 06214-00210 b: CLUES ABOUT THE FORMATION OF GAS GIANTS ON WIDE ORBITS^*

We observed GSC 06214-00210 b on 2010 July 9 with the OH-Suppressing Infrared Imaging Spectrograph (OSIRIS; Larkin et al. 2006) integral field unit with natural guide star AO on Keck II. The weather was photometric with good seeing (04–06 according to the Differential Image Motion Monitor (DIMM) on Canada–France–Hawaii Telescope (CFHT)). We observed GSC 06214-00210 b in the Jbb (1.180–1.416 μm) and Hbb (1.473–1.803 μm) bandpasses with the 005 pixel⁻¹ plate scale, resulting in a lenslet geometry of 16×64, a field of view of 08×32, and a resolving power (R ≡ λ/δλ) of ∼3800. The separation of GSC 06214-00210 b from the primary (22) is large compared to the measured FWHM of the companion (011 in Jbb; 013 in Hbb) and the seeing disk (∼05), so contamination from the primary is negligible. We obtained a total of 40 minutes in Jbb (5 minutes/exposure × 4 nodded pairs) and 20 minutes in Hbb (200 seconds/exposure × 3 nodded pairs) with 1'' offsets along the long axis of the field of view. The airmass ranged from 1.35 to 1.64 during the observations. Immediately following our science observations we targeted the nearby A0V star HD 148968 at an airmass of 1.50–1.67. We also acquired sky frames in both filters before GSC 06214-00210 and after HD 148968.

The raw data were flat fielded, sky subtracted, cleaned for cosmic rays and bad pixels, assembled into three-dimensional data cubes using the appropriate rectification matrices, and wavelength calibrated using version 2.3 of the OSIRIS Data Reduction Pipeline. Figure 1 shows an example of a collapsed J-band data cube. The science target and standard star spectra were extracted with aperture radii of 3 and 4 spaxels, respectively. The individual spectra from each band were then scaled to a common level and median combined. Spectral measurement uncertainties were determined by computing the standard errors about the median. The spectra were telluric corrected with the xtellcor_general routine in the Spextool spectroscopic reduction package (Vacca et al. 2003; Cushing et al. 2004). We tested sky subtraction with sky frames (A-sky) and with the nodded science data (A–B) by reducing the data both ways; the influence on the final spectrum was minor. We chose the former (A-sky) for our final spectrum.

Zoom In Zoom Out Reset image size

The J- and H-band spectra were independently flux calibrated using flux ratios for GSC 06214-00210 from Ireland et al. (2011) together with Two Micron All Sky Survey (2MASS) photometry of the primary (Skrutskie et al. 2006). Ireland et al. (2011) report flux ratios in the Mauna Kea Observatory (MKO) system, so we use the 2MASS–MKO conversions from Leggett et al. (2006) to arrive at MKO apparent magnitudes for GSC 06214-00210 b of J = 16.25 ± 0.04 mag and H = 15.55 ± 0.04 mag.

We observed the primary star GSC 06214-00210 with the SpeX spectrograph (Rayner et al. 2003) in short wavelength cross dispersed mode (SXD) at the NASA Infrared Telescope Facility (IRTF) on 2011 April 29 UT. The seeing reported by the CFHT DIMM was between 04 and 06, and there were light cirrus clouds during the observations. We used the 03 slit aligned to the parallactic angle, yielding a resolving power of ∼2000, and obtained a total of 4 minutes of data (30 s/exposure × 4 nodded pairs) by nodding along the slit in an ABBA pattern. After our science observations, we targeted the nearby A0V star HD 144925 and obtained calibration frames at a similar airmass. The data were reduced with version 3.4 of Spextool (Vacca et al. 2003; Cushing et al. 2004).

We also acquired low-resolution (R ∼ 100) spectroscopy of the primary with IRTF/SpeX in prism mode on 2011 May 12 UT. The weather was poor with complete cirrus coverage so we used the opportunity to verify the spectral slope of our SXD spectrum. We used the 08 slit and obtained 160 s of data (20 s/exposure × 4 nodded pairs) at an airmass of 1.38. Immediately afterward we observed HD 144925 for telluric calibration.

We obtained an optical spectrum of GSC 06214-00210 on 2011 May 16 UT under clear conditions using the Super Novae Integral Field Spectrograph (SNIFS) instrument (Lantz et al. 2004) on the University of Hawaii (UH) 2.2 m telescope. SNIFS is an optical integral field spectrograph with R ∼ 1000–1300 that splits the signal with a dichroic mirror into blue (∼3000–5200 Å) and red (∼5200–9500 Å) channels. The images are resampled with microlens arrays, dispersed with grisms, and focused onto blue- and red-sensitive CCDs.

A single 100 s exposure of the science target was sufficient to achieve high signal-to-noise ratio (S/N; ∼200). The processing was performed with the SNIFS data reduction pipeline, which is described in detail in Aldering et al. (2006) and Scalzo et al. (2010). The processing includes dark, bias, and flat-field corrections; assembling the data into red and blue three-dimensional data cubes; cleaning them for cosmic rays and bad pixels; sky subtraction; extracting the spectra using a semi-analytic point-spread function (PSF) model; and wavelength calibrating the spectra with arc lamp exposures taken at the same telescope pointing as the science data. Corrections for instrument response, airmass, and telluric lines are based on observations of the Feige 66 standard star using calibrated observations in Oke (1990). The far ends of the blue and red channels have low QE, so to avoid these regions we trim our final spectrum to 3300–4900 Å and 5170–8700 Å.

3.1.1. Spectral Properties and Classification of GSC 06214-00210 b

Our OSIRIS spectrum of GSC 06214-00210 b is presented in Figure 2. The most striking feature in the data is the strong Paβ emission (EW = −11.4 ± 0.3 Å) at 1.282 μm. The feature is present in each pair of dithered J-band observations of GSC 06214-00210 b and spans ∼9 spectral elements (Figure 2, inset), so it is not a result of a cosmic ray or bad pixel. One of our J-band OSIRIS data sets includes the primary (seen in Figure 1; the rest of data are dithered to avoid the star) and no Paβ emission is observed in that spectrum of the star. Paβ emission is also absent in our medium-resolution SXD spectrum of the primary obtained ∼8.5 months later. We therefore rule out the possibility that the emission observed in the companion is a result of contamination from the primary.

Zoom In Zoom Out Reset image size

Paβ emission in young stars can arise from accretion or outflows (Folha & Emerson 2001; Whelan et al. 2004), both of which imply the presence of a circumplanetary disk around GSC 06214-00210 b. The presence of a disk is bolstered by photometry from Ireland et al. (2011), who found a red K–L' color compared to field M and L dwarfs and suggest that the excess may originate from thermal disk emission. Although other parameters such as metallicity, surface gravity, and dust can also affect the K–L' color of late-type objects, these are unlikely to be the origin of the red color for GSC 06214-00210 b based on the expected near solar metallicity of USco members, a comparison to colors of field objects and giants (Ireland et al. 2011), and the minor influence of dust on K − L' colors at the M/L transition predicted by models (Baraffe et al. 2003; Chabrier et al. 2000). We note that the K–L' color of 1.2 mag is redder than many other young brown dwarfs with excesses attributed to disks (e.g., Liu et al. 2003; Allers et al. 2009). We describe the inferred accretion rate and luminosity in Section 3.1.4 and we discuss the implications of a disk in the context of the formation of GSC 06214-00210 b in Section 4.1.

Other notable features in our spectrum include absorption at ∼1.20 μm from FeH, 1.244 and 1.253 μm K i lines, deep H₂O steam bands at ∼1.4 and ∼1.8 μm, and FeH bandheads in the H band near 1.6 μm (Figure 2). All of these point to a late-M or early-L spectral type (Cushing et al. 2005).

Further classification of GSC 06214-00210 b warrants some care since there is no universally adopted near-infrared classification scheme for young late-type stars and brown dwarfs. The classification of young brown dwarfs in the literature is inconsistent and often a mix of qualitative (comparative) and quantitative (index-based) systems (e.g., Reid et al. 1995; Kirkpatrick et al. 1999; Martín et al. 1999; Reid et al. 2001; Geballe et al. 2002; Riddick et al. 2007); when multiple systems are used the adopted spectral type is usually an (subjectively weighted) average from several schemes.

We classify GSC 06214-00210 b primarily based on comparisons to near-infrared spectra of field objects and late-type free-floating members of Upper Scorpius (Usco; Figure 3). The advantage of using field objects is that they are optically typed and the advantage of using USco members is that they have the same age as GSC 06214-00210 b. Spectra of field objects originate from the IRTF Spectral Library (Cushing et al. 2005; Rayner et al. 2009) and those of USco objects from Lodieu et al. (2008). Many of the USco spectra have modest S/N so we first Gaussian smoothed them from their native resolution of R ∼ 1700 to R ∼ 1000 and cleaned them for strong single pixel outliers probably originating form imperfect bad pixel or cosmic ray removal in the original reduction. For the comparative analysis in Figure 3, the OSIRIS spectrum of GSC 06214-00210 b was smoothed to the appropriate resolving power for each sample.

Zoom In Zoom Out Reset image size

The K i lines in GSC 06214-00210 b are weaker than those in M8–L2 field objects, which is a common signature of youth (Gorlova et al. 2003; McGovern et al. 2004; Kirkpatrick et al. 2006). When normalized to 1.29–1.32 μm, as in Figure 3, the best matches to the J band are M8–L0.5 dwarfs. The J band in GSC 06214-00210 b is significantly bluer compared to the field L1 and L2 dwarfs. Together the joint J+H spectra appear to best fit the L0.5 dwarf, although the 1.6–1.65 μm FeH features are somewhat stronger in the field object.

The best USco matches are M9–L2 objects in the J band and M9–L1 objects in the H band. Interestingly, the 1.6 μm FeH features appear to be slightly stronger and the overall H band less triangular in GSC 06214-00210 b compared to the USco members. The depth of the K i lines are similar for GSC 06214-00210 b and the M8–L2 sequence. Note that the spectral types of some objects from Lodieu et al. (2008) appear to be 1–3 subtypes later than their optical classifications (Herczeg et al. 2009; Biller et al. 2011).

Allers et al. (2007) define a spectral index from 1.50 to 1.57 μm based on the depth of the H₂O absorption that correlates well with optical spectral types from M5 to L0 and is independent of gravity. This index yields a spectral type of M9.5^+1.2 _{− 1.1} for GSC 06214-00210 b, which is consistent with the comparison to field objects and USco members. The errors incorporate spectral measurement uncertainties and intrinsic scatter in the relation. Altogether we assign GSC 06214-00210 b a spectral type of L0 ± 1. Note that this agrees well with the estimate of M8–L4 by Ireland et al. (2011) based on colors.

3.1.2. Effective Temperature and Bolometric Luminosity

We fit the solar metallicity BT-Settl grid of model atmospheres (2010 version; Allard et al. 2010) to our spectrum of GSC 06214-00210 b. The BT-Settl models incorporate updated water opacity sources, revised solar abundances, and a new advanced treatment of dust formation using radiation hydrodynamic simulations. This new grid reproduces the observed colors and spectral energy distributions (SEDs) of late-M and L dwarfs better than the Ames-Cond and Ames-Dusty models of Allard et al. (2001), which examined the limiting cases of atmospheric dust formation. The grid spans effective temperatures between 2000 and 3000 K (ΔT_eff = 100 K) and gravities between 2.5 and 5.5 dex (cgs; Δlog g = 0.5). The fitting procedure relies on χ² minimization as detailed in Cushing et al. (2008) and Bowler et al. (2009). To improve the S/N of the data, we Gaussian smoothed our OSIRIS spectrum from its native resolving power of R ∼ 3800 to R ∼ 2000 and propagated the measurement errors through the convolution. We smoothed the models to R ∼ 2000 and fit them to the data in a Monte Carlo fashion by randomly generating synthetic observations based both on the spectral measurement uncertainties and the J- and H-band photometric uncertainties (Section 2.1) used for flux calibration. For each trial, we determine the best-fitting spectrum and save the χ² values from fitting every model in the grid. Figure 4 displays χ² contour maps averaged over 100 Monte Carlo trials. Since the spectra are flux calibrated and the distance (d) is constrained to 145 ± 14 pc (de Zeeuw et al. 1999; Ireland et al. 2011), the radius (R) can be calculated from the multiplicative factor (R²/d²) used to scale the models to the data (see Bowler et al. 2009). Figure 4 shows the contour plot of the inferred radius for GSC 06214-00210 b, representing averages from the Monte Carlo trials and incorporating the uncertainty in the distance estimate.

Zoom In Zoom Out Reset image size

The J (1.17–1.36 μm), H (1.45–1.83 μm), and J+H (1.17–1.83 μm) spectral segments were fit separately. To avoid the Paβ emission line, the 1.280–1.285 μm region was excluded from the fits. The best-fitting model to the J band is T_eff = 2700 K/log g = 4.5 with R = 1.22 R_Jup.⁶ The models match the depth of the K i lines and 1.35 μm H₂O depth reasonably well, but they fail to reproduce the 1.2 μm FeH feature and the continuum level from 1.27 to 1.31 μm. In the H band the models provide a better match to the data. The best-fit model is T_eff = 2700 K/log g = 4.0 with R = 1.36 R_Jup, which is similar to the result from the J band.

A fit to the combined J+H region produces a different result: T_eff = 3000 K/log g = 2.5 and R = 1.02 R_Jup. The quality of the fit is rather poor as the J-band continuum slope deviates dramatically from the data and the depth of the K i lines are poorly reproduced. The H band, however, is accurately reproduced by the model. In this case, the best-fitting model is at the edge of the model grid so it may not represent the global minimum. The 2700 K/4.0 dex model that matches the individual bands overestimates the J-band flux and underestimates the H-band flux when fit to the entire spectrum.

Another approach to infer effective temperature is to use predictions from evolutionary models based on an object's age and bolometric luminosity. The age of the Upper Scorpius association has long been known to be ∼5 Myr, primarily constrained from fitting isochrones to known members in the H-R diagram (see Preibisch & Mamajek 2008; Slesnick et al. 2008). Inferring the age of a cluster from the H-R diagram can be problematic, however, because pre-main-sequence evolutionary models may have significant systematic errors (Hillenbrand & White 2004; Kraus & Hillenbrand 2009) perhaps caused in part by the influence of episodic accretion (Baraffe et al. 2009; Baraffe & Chabrier 2010). With these uncertainties in mind, we adopt an age of 5 ± 2 Myr for GSC 06214-00210 b.

We calculated the bolometric luminosity by integrating an artificial spectrum constructed from the flux-calibrated OSIRIS J- and H-band spectral segments combined with a model spectrum for wavelengths shorter than the J-band segment (0.001–1.15 μm), between the J and H segments (1.36–1.45 μm), and longer than the H-band segment (1.83–1000 μm). At each model-data interface, the model was scaled to the data to create a continuous spectrum. Uncertainties in the spectra and flux calibration were accounted for in a Monte Carlo fashion. To test the sensitivity of the resulting luminosity on the input atmospheric model temperature, we calculated the luminosity using three BT-Settl models with temperatures of 2000 K, 2500 K, and 3000 K, and a gravity of 4.0 dex. The influence on the resulting luminosity was negligible (∼0.01 dex). Based on 1000 Monte Carlo trials, we derived a bolometric luminosity of −3.1 ± 0.1 dex for GSC 06214-00210 b.

To further verify our luminosity calculation we use the K-band bolometric correction calibrated with field objects from Golimowski et al. (2004). We arrive at a luminosity of −3.04 ± 0.10 dex, where the uncertainty takes into account errors in the distance, spectral type, photometry, and intrinsic scatter in the relation. In addition, we compute the luminosity using parameters predicted from evolutionary models (see below; T_eff = 2200 K/log g = 4.0). We flux calibrated the BT-Settl, Ames-Dusty, and Ames-Cond (Allard et al. 2001) models to the mean of the J-, H-, and K-band flux calibration scaling factors, accounting for photometric uncertainties in a Monte Carlo fashion. The resulting luminosities are nearly identical at −3.11 ± 0.09 dex, only differing by ∼0.01 dex among the models. The radii inferred from the scaling factor and the distance (see above) are also virtually the same at 1.89 ± 0.18 R_Jup.

Figure 5 (top panels) displays interpolated temperature tracks as a function of luminosity and age for various evolutionary models. The left panel shows the Lyon Cond and Dusty models of Baraffe et al. (2003) and Chabrier et al. (2000), which demonstrate the limiting effects of photospheric dust formation (formation and settling versus complete retention; Allard et al. 2001). The right panel shows the evolutionary models of Saumon & Marley (2008) for photospheres with no clouds ("nc") and those with significant amounts of dust ("f_sed = 2").

Zoom In Zoom Out Reset image size

We determined effective temperatures in a Monte Carlo fashion by interpolating the evolutionary models. The Lyon models begin at 1 Myr and the Saumon & Marley models begin at 3 Myr, so we excluded ages younger than those in the analysis. (The resulting probability density functions (PDFs) for the age are truncated Gaussians in linear space). The Cond and Dusty models yielded nearly identical results of 2160 ± 85 K for 10⁴ Monte Carlo draws. The Saumon & Marley models were similar, yielding 2200 ± 100 K for the case with no clouds and 2185 ± 120 K for the cloudy version. These values are substantially lower than the those of the best-fitting model atmospheres.

The temperatures inferred from the atmospheric models (∼2700–3000 K) are warmer than previous determinations of young M/L transition objects in the literature by ∼300–500 K (e.g., Béjar et al. 2008; Kuzuhara et al. 2011; Luhman 2004). A similar result was obtained by Dupuy et al. (2010) in their analysis of resolved late-M binaries in the field with dynamical masses. They found that the most widely used atmospheric models systematically yield effective temperatures ∼250 K higher than the temperatures predicted by evolutionary models, which are thought to be more reliable because they are less sensitive to missing or incomplete opacity sources (Chabrier et al. 2000). Likewise, the radii inferred from the atmospheric model fitting are systematically smaller than the value of ∼1.95 R_Jup predicted by evolutionary models (Section 3.1.4), which is probably a result of overestimated effective temperatures in the fits.

Luhman (1999) developed an effective temperature-spectral-type scale for young M dwarfs intermediate between those of dwarfs and giants. The calibration was defined to ensure that components of the quadruple system GG Tau and members of the young cluster IC 348 were aligned on the same isochrone from the Baraffe et al. (1998) evolutionary models in the H-R diagram. Luhman et al. (2003) revised the scale for M8 and M9 spectral types based on the latest-type members of IC 348 and Taurus. According to Luhman et al., the effective temperature of a young M9 object is ∼2400 K. To estimate temperatures beyond the Luhman et al. scale, Allers et al. (2007) subtract relative offsets taken from the temperature scale of field objects. This amounts to ∼92 ± 175 K for field M9–L0 objects based on the relation of Golimowski et al. (2004) (the uncertainty is from the rms scatter of the SpT–T_eff relation). Assuming approximate uncertainties of 100 K for the Luhman et al. scale, this yields a temperature of ∼2310 ± 200 K, which is consistent with the evolutionary model predictions but disagrees with those from the atmospheric model fitting.

3.1.3. Mass

3.1.4. Accretion

In a study of accretion diagnostics in young brown dwarfs, Natta et al. (2004) found that Paβ luminosity is well correlated with accretion luminosity at low masses. We use their empirical relation to derive the accretion luminosity and mass accretion rate for GSC 06214-00210 b.

The equivalent width (EW) of the Paβ emission line is −11.4 ± 0.3 Å, and the flux from that line is 1.12 ± 0.03 × 10⁻¹⁸ W m⁻². Assuming a distance of 145 ± 14 pc, we find log (L_Paβ/L_☉) = −6.14 ± 0.08 for the Paβ line luminosity. Applying the empirical relationship between Paβ luminosity and accretion luminosity from Natta et al. (2004, Equation 2) yields an accretion luminosity of log (L_acc/L_☉) = −4.4 ± 1.3. The mass accretion rate ($\dot{M}$) and the accretion luminosity are related through $\dot{M}=L_\mathrm{acc}R/GM$, where R is object's radius and M is its mass. We calculate a radius from evolutionary models in the same fashion as in Sections 3.1.2 and 3.1.3 and find R = 0.20 ± 0.01 R_☉. This yields a mass accretion rate of $\log (\dot{M})=-10.7 \pm 1.3$, where $\dot{M}$ is in M_☉ yr⁻¹. The uncertainty is dominated by scatter in the fitted relation from Natta et al. and reduces to 0.14 dex when these are ignored.

Mass accretion rates are observed to depend strongly on stellar mass and roughly follow an $\dot{M}\propto M_* ^2$ empirical relationship (e.g., Muzerolle et al. 2005; Natta et al. 2006; although see Clarke & Pringle 2006). This trend spans two orders of magnitude in mass and six orders of magnitude in mass accretion rate, although there are over two orders of magnitude of intrinsic scatter in the relation. Figure 6 shows the position of GSC 06214-00210 b in the $\log \dot{M}\hbox{--}\log M_*$ diagram relative to members of various star-forming regions. The dependence of accretion rate on mass is clear, and objects with masses ≲30 M_Jup (log (M_*/M_☉) ≲ −1.5) appear to have accretion rates below ∼10⁻¹¹ M_☉ yr⁻¹. GSC 06214-00210 b has a somewhat higher accretion rate than the lowest mass brown dwarfs from Muzerolle et al. (2005) and Herczeg et al. (2009), although it is consistent with the scatter seen at higher masses.

Zoom In Zoom Out Reset image size

While we have assumed that the emission in GSC 06214-00210 b originates from energy released during accretion, we briefly examine whether it could result from chromospheric activity. Short & Doyle (1998) modeled active M dwarf chromospheres and found that weak Paβ emission is possible at some chromospheric pressures. In Figure 7, we show the relative strength of the Paβ EW and line luminosity to objects from the Natta et al. (2006) sample of young stars. The Paβ emission strength for GSC 06214-00210 b is comparable to the strongest accretors from that sample and is larger than the vast majority of comparison stars. We can also examine the correlation between Paβ emission and the Hα 10% line width, which is a widely adopted accretion indicator. Among the sample of young brown dwarfs from Natta et al. (2004), four out of the five that show evidence of accretion based on their Hα 10% widths also exhibit Paβ emission. No objects with Paβ emission were identified as non-accretors, which together with the unusually strong Paβ EW support the notion that the observed emission in GSC 06214-00210 b is a result of accretion.

Zoom In Zoom Out Reset image size

We note that GSC 06214-00210 b is also likely to have other emission lines. Hα is a good candidate since it correlates well with Paβ emission in brown dwarfs. Brγ is observed in some, but not all, accreting objects. For example, Natta et al. (2004) found that only two out of eight brown dwarfs with Paβ emission showed Brγ emission. However, Brγ emission does seem to correlate with the strength of the Paβ line, which increases the likelihood that GSC 06214-00210 b is emitting at that line.

The primary star GSC 06214-00210 is a weak-lined T Tauri member of Upper Sco (see Ireland et al. 2011 for a review of the literature). Medium-resolution optical spectroscopy by Preibisch et al. (1998) revealed weak Hα emission (EW = −1.51 Å) which can be attributed to chromospheric activity. Our 0.8–2.5 μm SXD spectrum is presented in Figure 8; no emission lines are present. Compared to late-K and early-M dwarfs from the IRTF Spectral Library (Rayner et al. 2009), as shown in Figure 9 (right panel), the near-infrared spectrum resembles late-K objects more closely than early-M objects. We verified that our SXD spectrum was properly reduced by obtaining a low-resolution spectrum with SpeX in prism mode during a night of poor weather conditions. The shapes of the spectra are virtually identical (lower left inset of Figure 8), verifying our SXD spectrum and calling into question the nominal spectral type of M1 originally assigned by Preibisch et al. (1998).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Overall, the optical spectrum is best matched by that of a K7 dwarf from the Pickles (1998) Spectral Library from ∼3300 to 8700 Å (Figure 9, left panel).⁷ The shape of the SED and the depth of the TiO absorption bands differ substantially from those of the M1 spectrum. We also infer a spectral type using the spectral indices defined by Reid et al. (1995), which measure the depth of various molecular absorption bands in the optical. Our value for the TiO5 index (0.916) yields a spectral type of K5.5. Since many of the optical absorption bands are gravity-sensitive, a slightly earlier spectral type is not unexpected using this index-based scheme. Similarly, the CaH2 index (0.908) indicates an effective temperature of ∼4200 K using the SpT–T_eff relation from Woolf & Wallerstein (2006). All of these diagnostics suggest a spectral type of late-K rather than early-M, so we revise the spectral type of GSC 06214-00210 from M1 to K7 ± 0.5.

We determine the effective temperature of GSC 06214-00210 by constructing its SED from 0.6 to 22 μm (Figure 10) and comparing it to the grid of solar metallicity Phoenix-Gaia model atmospheres (Brott & Hauschildt 2005). The photometry originate from the Carlsberg Meridian Catalog 14 (CMC14) for the r' band (Evans et al. 2002); the Deep Near-Infrared Southern Sky Survey (DENIS) for i, J, and K_S bands (Epchtein et al. 1997); the 2MASS for J, H, and K_S bands (Skrutskie et al. 2006); and the Wide-Field Infrared Survey Explorer (WISE) for 3.4, 4.6, 12, and 22 μm bands (Wright et al. 2010). Zero-point flux densities are from Stoughton et al. (2002) for CMC14, Fouqué et al. (2000) for DENIS, Rieke et al. (2008) for 2MASS, and Wright et al. (2010) for WISE. The models are flux calibrated using the mean scaling factor from the 2MASS J, H, and K_S bands. We limit the comparison to models with log g = 4.0, which is the approximate surface gravity from evolutionary models for a K7 star at 5 Myr. The r', i, and J bands carry the most weight in constraining the effective temperature because the SED turns over to the Rayleigh–Jeans tail at λ ≳ 1.6 μm (H band). The best match is the 4200 K model, although the 4100 K and 4300 K models provide decent fits. Warmer and cooler model temperatures begin to diverge from the photometry, so we assign an uncertainty of 150 K for the temperature.

Zoom In Zoom Out Reset image size

The WISE 22 μm photometry disagrees with the model in Figure 10, suggesting a slight excess for the primary. In Figure 11, we compare the position of GSC 06214-00210 to field K7 dwarfs and other USco members in the W1–W4 versus J–K_S diagram. To isolate a sample of field K7 dwarfs we used a compilation of the revised Hipparcos catalog (van Leeuwen 2007) generated by E. Mamajek (2011, private communication) with spectral types and V-band magnitudes from the original catalog (Perryman et al. 1997). An absolute magnitude cut of M_V > 5 was imposed to exclude giants. This yielded 272 K7 dwarfs, which we then fed into the WISE Preliminary Release Source Catalog query using a search radius of 10''. (The PSF FWHM for WISE bandpasses range between 6'' and 12''.) Among the resulting detections, we kept only those with the best photometric quality flags in all bands (ph_qual="A" or "B"), the best contamination and confusion flags (cc_flags="0000"), extended source flags consistent with a point source (ext_flg="0"), and stable variable flags (var_flg<5). This produced 41 objects, which are plotted in Figure 11. We performed a similar search of the WISE Preliminary Catalog for known USco members from Preibisch & Mamajek (2008). We queried around 251 USco members with GKM spectral types applying the same flags as above; 67 objects yielded reliable WISE detections. Field stars and some USco members have W1–W4 colors of ∼0.0 ± 0.2 mag, although most USco objects show moderate to large excess in that color. The W1–W4 color of 0.78 ± 0.27 for GSC 06214-00210 indicates a mild (2σ) excess compared to field K7 stars.

Zoom In Zoom Out Reset image size

We calculate the luminosity of the primary by integrating the flux-calibrated 4200 K/log g = 4.0 synthetic spectrum, yielding L_* = 0.38 ± 0.07 L_☉. The error budget is dominated by the uncertainty in the distance (the uncertainty in effective temperature contributes ∼0.01 dex). In Figure 12, we show the position of GSC 06214-00210 in the H-R diagram. The pre-main-sequence evolutionary models of Baraffe et al. (1998) for L_mix = H_p are displayed from 0.02 to 1.4 M_☉ with isochrones from 1 Myr to 1 Gyr (black). We also include the L_mix = 1.9H_p 1 Myr, 5 Myr, 10 Myr, and 1 Gyr isochrones for higher masses (⩾0.6 M_☉). Gray circles show the positions of Upper Scorpius low- and intermediate-mass members in Preibisch & Mamajek (2008; Tables 1 and 2). The position of GSC 06214-00210 is consistent with the scatter exhibited by other Upper Scorpius members, although it sits at a somewhat older isochrone than the nominal 5 Myr age of the complex. For the prescription in Figure 12 with L_mix = H_p, the isochronal age is ∼16 Myr with a 1σ range of 13–25 Myr. For L_mix = 1.9 H_p, which is required to fit the Sun, the inferred age is ∼10 Myr with a range of 6–16 Myr. Note that the isochronal age of low-mass USco members (∼5 Myr) diverges from the cluster age based on the L_mix = H_p prescription (∼10 Myr) and instead is more consistent with the age inferred from the L_mix = 1.9 H_p grid.

Zoom In Zoom Out Reset image size

We estimate the mass of GSC 06214-00210 from its position on the H-R diagram together with predictions from pre-main-sequence evolutionary models. The BCAH98 models imply a mass of ∼1.0 M_☉ (Figure 12). We also compare it to the evolutionary models of D'Antona & Mazzitelli (1994), which yield a mass of ∼0.8 M_☉ and Palla & Stahler (1999), which yield ∼0.9 M_☉. Given these moderate systematic differences, we adopt a mass of 0.9 ± 0.1 M_☉. Note that this is significantly higher than the value of 0.6 ± 0.1 M_☉ that has been cited in previous work using the later M1 spectral type.

As described in Section 1, several explanations can account for the observed population of planetary-mass companions orbiting stars at several hundred AU. Unfortunately, these models make few unique or testable predictions, and there is considerable debate about formation scenarios in the literature. Here we focus on the viability of one explanation, planet–planet scattering, in the context of our results for GSC 06214-00210 b.

The formation of closely packed planetary systems with two or more giant planets naturally leads to dynamical interactions which can significantly alter the orbits of one or more components. These close encounters produce a wide range of outcomes including (but not limited to) collisions between planets, accretion onto the star, rearrangement to new stable or quasi-stable orbits, scattering to close separations, scattering to highly eccentric wide orbits, and complete ejection ("ionization"). There is growing observational evidence that scattering is an important phenomenon in extrasolar planetary systems. It has been invoked to explain the population of planets with high eccentricities found in radial velocity surveys (e.g., Rasio & Ford 1996; Ford & Rasio 2008; Jurić & Tremaine 2008), the distribution of Rossiter–McLaughlin spin–orbit measurements (Morton & Johnson 2011), and, more recently, the possible abundance of unbound gas giants inferred from microlensing surveys (Sumi et al. 2011).

We examine a hypothetical scattering event for GSC 06214-00210 b by assuming it was formed much closer than its present location through conventional means (core accretion or disk instability) and was ejected to a large orbit through the gravitational interaction with another massive body. Monte Carlo simulations of scattering events show that, given unequal planet mass ratios, the lower-mass planets show a much stronger preference for outward scattering than the more massive components (Veras & Armitage 2004; Ford & Rasio 2008) with little dependency on which planet initially had the wider orbit. If GSC 06214-00210 b was ejected to its present location as a result of such an event then it is likely that another object at least as massive (∼14 M_Jup) also formed in that system.

Ejections of planets onto wide orbits are not uncommon outcomes in simulations so the discovery of a planetary-mass object at ∼300 AU is perhaps not surprising. The presence of a disk around GSC 06214-00210 b, however, implies that if it underwent scattering then its disk was not destroyed during point of closest approach with the scatterer. To our knowledge there are no studies that specifically investigate the survivability of circumplanetary disks during scattering events. Here we qualitatively examine the two relevant length scales—the separation at closest approach during a scattering event and the circumplanetary disk radius—to assess the likelihood of a disk surviving such an encounter.

The criterion for orbital stability between two coplanar planets on circular orbits was found by Gladman (1993) to be

$$\begin{equation} \Delta _\mathrm{cr} \ge 2 \sqrt{3} \ R_\mathrm{H,M}, \end{equation} \tag{ 1 }$$

where Δ = a₂ − a₁ is the difference between the initial semimajor axes of the planets and R_{H, M} is the mutual Hill radius (Marchal & Bozis 1982):

$$\begin{equation} {} R_\mathrm{H,M} = \left(\frac{m_1 + m_2}{3 M_*} \right)^{1/3} \frac{a_1 + a_2}{2}. \end{equation} \tag{ 2 }$$

This radius defines the region in which the gravitational force between two bodies is larger than the force on them due to the star. The critical range for stability Δ_cr can be divided into an outer region encompassing weak interactions, which can result in simple rearrangements of the system architectures with new quasi-stable orbits (Ford et al. 2001; Veras & Armitage 2004), and an inner region where strong interactions occur, which is characterized by chaotic events made up of collisions and ejections. This boundary must be found empirically from Monte Carlo simulations (see Gladman 1993 for the empirical relation for equal-mass planets).

To test the critical range for stability Δ_cr, Chambers et al. (1996) ran simulations of two interacting planets with various separations and confirmed that their orbits were stable for at least 10⁷ yr when $\Delta >2\sqrt{3} \ R_\mathrm{H,M}$, and when $\Delta <2\sqrt{3} \ R_\mathrm{H,M}$ they eventually experienced a close encounter with a distance at closest approach of <1 R_{H, M}. Although the planet masses in that study were smaller than mass scales for a GSC 06214-00210-like system, a similar study for ∼Jovian mass planets by Marzari (2002) resulted in similar close approach scales of <1 R_{H, M} and a large frequency of ejections. Most scattering studies focus on various outcomes as a function of the initial planet spacings without discussing the distances of closest approach from the simulations, which is the relevant length scale we are interested in here. The exception is for collisions, defined by various authors to be the outcome when Δ becomes less than about the sum of the planetary radii. Given the relatively large frequency of collisions in some simulations (e.g., Ford et al. 2001 find that collision rates can reach tens of percent depending on the planet radii and semimajor axes), the distances of closest approach appear to be quite small, which is consistent with the separation of <1 R_{H, M} found by Chambers et al. (1996) and Marzari (2002).

Circumplanetary disks are thought to form around young gas giant planets from accretion of circumstellar disk material once a planet has opened a gap in the protoplanetary disk (e.g., Lubow et al. 1999; Ward & Canup 2010). In our solar system, the regular satellites of the giant planets are fossil records of these structures (Canup & Ward 2002) and the excess optical emission from Fomalhaut b has been attributed to scattering from an enormous disk around that extrasolar planet (Kalas et al. 2008). The outer radii of these disks have been investigated by several authors using analytical arguments and hydrodynamical simulations. Quillen & Trilling (1998), Ayliffe & Bate (2009), and Martin & Lubow (2011) obtain similar results of ∼0.3–0.4 R_H, which roughly correspond the centrifugal radius. (Note that the individual Hill radius R_H is the limiting case of the mutual Hill radius R_{H, M} when m₂ goes to zero and a₁ = a₂. If m₂ = c m₁ and a₁ ∼ a₂ then R_{H, M} and R_H differ by a factor of (1 + c)^1/3.)

Since Hill radii depend on the planet masses and semimajor axes, we consider the 1:1, 1:2, and 1:5 cases of planet mass ratios to better understand the relative magnitudes of the scales of disk disruption and scattering closest approach. For an equal-mass coplanar scattering event (m₂ = m₁), the maximum separation at which the disks will not be disrupted is (r_disk 1+r_disk 2)∼0.7 R_H. Assuming a₁ ∼ a₂ during a close encounter, the conversion from mutual Hill radius to the point-particle Hill radius is a factor of 1.26, so the approximate scale of closest approach is ≲1.26 R_H. For the case where m₂ = 2m₁, the disk disruption scale is ∼0.8 R_H and the distance of closest approach is ≲1.4 R_H. Likewise, when m₂ = 5m₁, the disruption scale becomes ∼0.95 R_H and the distance of closest approach is ≲1.8 R_H.

These results suggest that the disruption of circumplanetary disks may be common in scattering events, especially if the distance to closest approach is less than about one-half of a mutual Hill radius. The degree to which a disk is affected by these interactions (i.e., disk truncation versus complete destruction) likely depends on the disk sizes, planet mass ratio, relative disk inclinations, relative planet speeds, and impact parameter. Nevertheless, we interpret the retention of the disk around GSC 06214-00210 b as evidence against a past scattering event, although detailed simulations are needed.

Table 2. Properties of the Companion GSC 06214-00210 b

Property	Value
Observed
SpT	L0 ± 1
Projected separation	22 (320 ± 30 AU)
EW(Paβ)	−11.4 ± 0.3 Å
log (LBol/L☉)	−3.1 ± 0.1
J (MKO)a	16.25 ± 0.04 mag
H (MKO)a	15.55 ± 0.04 mag
K (MKO)a,b	14.94 ± 0.03 mag
L'a	13.75 ± 0.07 mag
Estimated
Mass (Evol.)	14 ± 2 MJup
Teff (Evol.)	2200 ± 100 K
Teff (BT-Settl)	2700 ± 200 K
log (L(Paβ)/L☉)	−6.14 ± 0.08
log (Lacc/L☉)	−4.4 ± 1.3
$\log (\dot{M}/M_{\odot } \,{\rm yr}^{-1})$	−10.7 ± 1.3

Notes. ^aPhotometry from Ireland et al. (2011). ^bWeighted mean and uncertainty of four measurements from Ireland et al. (2011).

Download table as:  ASCIITypeset image

Further insight into the formation of wide, low-mass companions can be gleaned from the architectures of other low-mass ratio systems (see Section 4.2). In a two-planet scattering event involving an ejection, energy conservation requires the remaining bound planet to have a final semimajor axis greater than one-half of its initial semimajor axis (e.g., Ford & Rasio 2008). Since none of the wide planetary-mass companions discovered so far have an observed scatterer present in the system, the scattering hypothesis appears unlikely, although the dearth of published detection limits for additional companions in these systems hampers a more quantitative assessment. The notable exception is deep AO imaging of the 1RXS J1609–2105 system by Lafrenière et al. (2010) and somewhat shallower AO imaging of the GSC 06214-00210 system by Ireland et al. (2011).

Finally, we note that if GSC 06214-00210 b was scattered to a large orbit then we can constrain the final location of the hypothetical scatterer using the aperture masking detection limits from Kraus et al. (2008) combined with the direct imaging detection limits from Ireland et al. (2011). As shown in Figure 13, the detection limits exclude objects with the same mass as GSC 06214-00210 b at projected separations ≳170 AU, 30 M_Jup objects (roughly twice the mass of GSC 06214-00210 b) ≳60 AU, and 40 M_Jup objects ≳5 AU.

Zoom In Zoom Out Reset image size

The discovery of strong accretion in GSC 06214-00210 b prompts the broader question of how common this phenomenon is in other objects. Although the census of young planetary-mass companions has been steadily growing over the past several years, the moderate-resolution spectra needed to detect infrared emission lines have only been obtained for a fraction of these objects.

Six companions with masses ≲20 M_Jup and ages ≲10 Myr are known: SR 12 C (Kuzuhara et al. 2011), CHXR 73 B (Luhman et al. 2006), 1RXS J1609–2105 b (Lafrenière et al. 2008), DH Tau b (Itoh et al. 2005), CT Cha B (Schmidt et al. 2008), and GSC 06214-00210 b (note that the uncertainties in masses and ages can be rather large). We have ignored companions to young brown dwarfs (e.g., 2M1207 Ab) since the formation of those systems was probably different from that of stars and wide companions. Low-resolution spectra have been published for most of these companions, but moderate-resolution spectra have only been acquired for 1RXS J1609–2105 b (Lafrenière et al. 2008; Lafrenière et al. 2010), CT Cha B (Schmidt et al. 2008), and GSC 06214-00210 b (this work). CT Cha B and GSC 06214-00210 b both show Paβ emission, but there is no evidence of accretion or thermal disk emission out to 4 μm for 1RXS J1609–2105 b.

Two out of three very low mass companions therefore show evidence of a disk. The sample size is minute, but this could hint that a large fraction of these objects are accreting. Moderate-resolution near-infrared spectroscopy of the remaining companions would help address this matter. Intriguingly, the low-mass (∼25 M_Jup) brown dwarf companion GQ Lup B also shows Paβ emission (Seifahrt et al. 2007; Schmidt et al. 2008; Lavigne et al. 2009), which may support this notion.

With a large enough sample size, the presence of accretion or thermal disk emission from young planetary-mass companions might be used to learn about their formation mechanisms. If this population formed like free-floating planetary-mass objects then they should both share similar physical properties such as disk frequencies. Such a result would be a sign of a common formation mechanism. In the future, a similar analysis might be applied to extrasolar planets found at close separations (e.g., <100 AU) and those at wide separations (>100 AU) to test formation mechanisms, perhaps in the context of core accretion and disk instability.