A FIRST-LOOK ATMOSPHERIC MODELING STUDY OF THE YOUNG DIRECTLY IMAGED PLANET-MASS COMPANION, ROXS 42Bb

2.1.1. Observations and Image Processing

We reduced J, K_s and L' data for ROXs 42B obtained from the Keck Observatory Archive (KOA), taken on 2011 June 22 with the NIRC2 "narrow" camera (9.952 mas pixel⁻¹; Yelda et al. 2010; PID HD242N2; P.I. A. Kraus). All new data were taken with good adaptive optics corrections in classical imaging, and in dither/nod patterns to remove the sky background. The field of view was ≈10 arcsec on a side. Basic image processing steps were identical to those carried out for previously published ROXs 42Bb photometry (Currie et al. 2014). Briefly, we first sky-subtracted each image from a median combination of images obtained at other dither positions. Then, we identified and removed hot/cold/bad pixels, corrected each image for distortion using the nirc2dewarp.pro IDL routine, copied each to larger blank image, and registered the images to a common center (Currie et al. 2012a, 2012b). We then subtracted off a two-dimensional radial profile of the primary to remove the halo light from each image. After these steps, we median-combined together our set of images and rotated the combined image to the north-up position (P.A._north = −0252; Yelda et al. 2010).

Figure 1 displays the combined J (left), K_s (middle), and L' images, revealing ROXs 42Bb at an angular separation of ≈117 and position angle of ≈270°. Additionally, each data set reveals the second candidate companion, "ROXs 42B c," at roughly 05 separation, although the L' detection is noticeably weaker than detections at J and K_s. ROXs 42B is defined as a close binary (Δθ ≈ 005) identified from lunar occultation data (e.g., Simon et al. 1995); however, we fail to resolve the two components (see also Currie et al. 2014; Kraus et al. 2014).

Zoom In Zoom Out Reset image size

2.1.2. Photometry

To these data we add Very Large Telescope/SINFONI spectra and H-band photometry previously published in Currie et al. (2014). We measure the primary-to-companion contrast for ROXs 42Bb as ΔH = 6.86 ± 0.05. Given the 2MASS H-band photometric measurement of m_H = 9.017 ± 0.020, we derive an apparent magnitude of 15.88 ± 0.06 and a dereddened absolute magnitude of 9.87 ± 0.14.

As noted in Currie et al. (2014), the putative "ROXs 42Bc" is of comparable brightness in H band. Quantitatively, we derive a contrast of ΔH = 6.94 ± 0.09 and an apparent magnitude of 15.96 ± 0.09. These values significantly disagree with the contrast estimate by Kraus et al. (2014), who claim ΔH = 6.20. With the background signal largely removed from a radial profile subtraction or explicitly measured in a median-combined frame, "ROXs 42Bc" clearly cannot be approximately twice as bright as ROXs 42Bb.⁵

Table 1 summarizes our photometric measurements for both ROXs 42Bb and "ROXs 42Bc." While both companions have similar H-band contrasts and apparent magnitudes, their observed colors are quite different. In particular, ROXs 42Bb has a J–L' color of 2.86, significantly redder than that of "ROXs 42Bc" (1.42), a difference largely owing to the latter's significantly fainter L' flux density.

Additional arguments show that "ROXs 42Bc" is likely a reddened background object. Besides being significantly fainter at L', it is about 0.45 mag fainter at K_s band and ∼0.31 mag brighter at J. If the line-of-sight extinction between us and "ROXs 42Bc" is in the A_V = 4–6 range consistent with it being behind the filamentary structure containing ROXs 42B, then the object dereddens to the colors expected for an early-type star. For example, assuming A_V = 4.5 yields m(J) = 15.33 ± 0.12, m(H) = 15.11 ± 0.09, m(K_s) = 14.94 ± 0.07 and m(L') = 14.93 ± 0.20, consistent with the colors for an A0 to mid F star (see Pecaut et al. 2013; Currie et al. 2010). Thus, as previously suggested in Currie et al. (2014) and argued independently by Kraus et al. (2014), our analysis shows that "ROXs 42Bc" is likely not a second planetary-mass companion orbiting ROXs 42B.

To compare the near-to-mid IR photometric properties of ROXs 42Bb with those for other substellar objects, we follow similar analysis in Currie et al. (2013), using the Leggett et al. (2010) sample of field brown dwarfs with spectral classes between M7 and T5. To these objects, we add the directly imaged planetary-mass companions/candidates listed in their Table 4 using photometric measurements listed in their Table 5. For clarity, we do not include young objects with inferred masses above the deuterium-burning limit: these tend to be much more luminous in each infrared filter than ROXs 42Bb.

Figure 2 compares the J/J–K_s and H/H–L' color–magnitude position of ROXs 42Bb (blue diamond) to those positions for older, field brown dwarfs (black/grey dots) and other planetary-mass companions (aqua squares). Currie et al. (2014) found that the near-IR H/H–K_s color–magnitude diagram position for ROXs 42Bb, like that for directly imaged planets β Pic b and HR 8799 bcde (Marois et al. 2008; Currie et al. 2011, 2013), was discrepant compared to the field MLT dwarf sequence. The same trends persist when we consider J and L' data, as ROXs 42Bb is ∼0.5 mag redder than the field sequence in both J–K_s and H/H–L'.

Zoom In Zoom Out Reset image size

Compared to the planet-mass companions plotted, ROXs 42Bb's J/J–K_s, H/H–L', and H/H–K_s (Currie et al. 2014) color–magnitude diagram position bears the strongest resemblance to 6–16 M_J M/L transition objects GSC 06214B and USco CTIO 108 B (Ireland et al. 2011; Béjar et al. 2008). These objects have masses of ∼6–16 M_J, and spectral types of M9.5–L0. Given their slightly older but qualitatively similar ages of ≈5–10 Myr (Pecaut et al. 2012; Bowler et al. 2011), we expect they should serve as good indicators for ROXs 42Bb's spectral type and inferred mass.

In Currie et al. (2014), we focused on simple, empirical comparisons to the ROXs 42Bb spectral shape to estimate its spectral type and assess evidence for low surface gravity (see also Bowler et al. 2014). Here, we further analyze the spectrum (Figure 3) to estimate its spectral type from gravity independent (or weakly dependent) indices (Allers and Liu 2013; Bonnefoy et al. 2014), identify chemical species present, and corroborate evidence for low surface gravity from alternate diagnostics (Canty et al. 2013).

Zoom In Zoom Out Reset image size

3.2.1. Spectral Type

3.2.2. Chemical Composition and Gravity-sensitive Features

Table 3 summarizes our model-fitting results. Figure 4 displays the χ² distributions for our photometric model fits while Figure 5 shows the best-fit model spectrum and predicted photometry (magenta lines) with the photometric data overplotted. The dust-free COND models fail to reproduce the photometry (Figures 4 and 5, top left panels) at a statistically significant level ($\chi ^{2}_{{\rm min}}$ ∼ 35.9 ≫ $\chi ^{2}_{95\%}$ ∼ 9.5). The BT-Settl models (same figures, bottom-left panels) perform better but include only one model (T_eff = 1800, log(g) = 3.5) consistent with the data at the 95% confidence limit.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Models incorporating substantial atmospheric dust and/or thick clouds best reproduce ROXs 42Bb's photometry. The DUSTY models quantitatively yield the better fits than either COND or BT-Settl (top right panel), with a χ² minima of ∼2.47 (T_eff = 1900 K, log(g) = 4) and models spanning T_eff ∼ 1800–1900 K consistent within the 95% confidence limit. The Burrows models reproduce the data even better (bottom right panel) for T_eff = 1800–2000 K and log(g) = 3.6–4 and T_eff = 1900–2000 K with log(g) = 3.4 ($\chi ^{2}_{{\rm min}}$ = 0.57). While empirical comparisons to ROXs 42Bb's spectrum allowed temperatures of 1800–2600 K, modeling the object's photometry strongly favors the lower part of this range.

Figures 6 and 7 display the χ² distributions for our spectroscopic model fits and best-fit model with the K-band spectrum overplotted. The COND and DUSTY model spectra yield significant fits only at high temperatures (T_eff = 2600–2800 K). The BT-Settl models fit the spectrum well at 1700–1900 K/log(g) = 4–5 and 2400 K/log(g) = 3.5, accurately reproducing the shape of most of the K-band peak and its location. While the fit from the Burrows models tends to predict too flat of a K-band spectrum compared to BT-Settl, models with T_eff = 1950–2100 K and log(g) = 3.4–3.8 are consistent with the spectrum to within the 95% confidence limit.⁸

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figures 8 and 9 and the righthand columns of Table 3 clarify which model parameter space matches the ROXs 42Bb photometry and K-band spectrum simultaneously. Although some COND and DUSTY models fit the K-band spectrum, the same models fail to reproduce the ROXs 42Bb photometry. Similarly, the COND and DUSTY models best fitting the photometry fail to reproduce the K-band spectral shape. Thus, neither of the model limiting cases can simultaneously fit photometry and spectra.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The inferred temperatures for BT-Settl models that best fit the spectrum and the photometry are similar. However, as with the COND and DUSTY models, the specific BT-Settl models yielding fits to within the 95% confidence limit for the spectra fail to match the photometry (Figure 8) and vice-versa (Figure 9), owing to the model predicting too narrow of a 2.2 μm peak. In contrast to all other models, the Burrows models are able to yield some parameter space that fits both the ROXs 42Bb photometry and spectroscopy at the 95% confidence limit. For instance, the T_eff = 2000 K, log(g) =3.6 model that best fits the spectrum also agrees with the photometric data to within the 95% confidence limit (Figure 8, bottom right panel).

To derive best-estimated atmospheric parameters, we focus on the Burrows thick cloud models that simultaneously fit the photometry and K-band spectrum: T_eff = 1950–2000 K, log(g) = 3.4–3.8. Thus, spectral analysis (Section 3.2) and atmospheric modeling (Section 4.1–4.2) combined shows ROXs 42Bb to have a very dusty/cloudy, low surface gravity L0 dwarf atmosphere with T_eff = 1950–2000 K.

The models matching ROXs 42Bb's photometry and spectroscopy allow us to derive an estimated radius for the object. Compared to the nominal radii for a given T_eff and log(g) assumed from the Burrows et al. (1997) evolutionary models, our fits require radii larger by factors of 1.25–1.53 for models with T_eff = 1950 K and log(g) = 3.4–3.8. For the T_eff = 2000 K and log(g) = 3.4–3.8 models, the scaling factors are smaller: 1.2–1.45. In each case, the scaling factors that best fit the photometry match those that best fit the spectrum to within ∼2.5%. We obtain best-fit radii of R/R_J ∼ 2.55 ± 0.20 for the T_eff = 1950 K models and R/R_J ∼ 2.43 ± 0.18 for the T_eff = 2000 K models, where the errors include consider the dispersion in radius scaling estimates among good-fitting models and the distance uncertainty of ∼5%.

ROXs 42Bb's luminosity-estimated mass, derived from the temperature and radius of our fits, is consistent with the sub-deuterium burning, 9$^{+6}_{-3}$ M_J estimate from Currie et al. (2014). Luminosity estimates for the T_eff = 1950 K and 2000 K models yield identical values: log(L/L_☉) = −3.07 ± 0.07. Given an estimate age of 2.5–3 Myr for ROXs 42B (Currie et al. 2014), both the Baraffe et al. (2003) and Burrows et al. (1997) "hot start" evolutionary models nominally predict a mass of 10$^{+2}_{-1}$ M_J. Displacing the evolutionary tracks by a characteristic protostar lifetime of τ_Class0/I ∼ 0.5 Myr (Evans et al. 2009) lowers the masses to 9$^{+1.5}_{-1}$ M_J. If instead ROXs 42B is 1 Myr old like much of the ρ Oph complex, ROXs 42Bb's mass is ≈6 M_J.

Considering these uncertainties, we estimate a mass of 9$^{+3}_{-3}$ M_J for ROXs 42Bb. We derive consistent (slightly lower, much higher) mass estimates if we take the surface gravity and radii of our best-fit models at face value: for log(g) = 3.6 (3.4, 3.8) we derive M ≈ 9.1–10.4 M_J (6–7.3 M_J, 14.3–15.8 M_J).