CHARACTERIZATION OF THE VERY-LOW-MASS SECONDARY IN THE GJ 660.1AB SYSTEM

Our understanding of brown dwarfs and very-low-mass (VLM; M ≲ 0.1 M_⊙) stars has advanced considerably since the first discovery of substellar objects in the 1990s (Nakajima et al. 1995; Rebolo et al. 1995). Subsequent searches in various wide-field, multi-band, infrared imaging surveys have revealed thousands of low-temperature VLM dwarfs, many in the immediate vicinity of the Sun (d ≲ 30 pc; Kirkpatrick 2005). Among these, VLM companions to nearby stars have been particularly valuable because they share the distance, age, and composition of their more massive, and more precisely characterized, primaries. Many of these VLM "benchmarks" have been intensely studied, serving as empirical tests of spectral, evolutionary, and formation models (e.g., Becklin & Zuckerman 1988; Nakajima et al. 1995; Burgasser et al. 2000; Kirkpatrick et al. 2001; Wilson et al. 2001; Chauvin et al. 2005; Metchev & Hillenbrand 2006; Mugrauer et al. 2006; Saumon et al. 2007; Leggett et al. 2008, 2010; Dupuy et al. 2009; Faherty et al. 2010, 2011; Burningham et al. 2011; Bowler et al. 2012b, 2012a, 2013; Gomes et al. 2013).

Deacon et al. (2014) have recently compiled a sample of VLM companions to nearby stars, many identified in the Pan-STARRS survey (Kaiser et al. 2002). One system not included in this compilation is GJ 660.1AB, identified by Schneider et al. (2011) as an M1 dwarf (Gray et al. 2003) with a relatively faint (ΔJ = 4) co-moving secondary at a separation of 608 ± 002 (122 ± 9 AU). Schneider et al. (2011) estimated a spectral type of M9 ± 2 for the companion based on its relative brightness. The system has a parallactic distance of 20.0 ± 1.4 pc, a high proper motion (${\boldsymbol{\mu }}$ = [181 ± 5, −694 ± 3] mas yr⁻¹; van Leeuwen 2007), and a radial velocity of −33 ± 10 km s⁻¹ (Hawley et al. 1996). Because the Galactic velocity of the system, ($U,V,W$) = (0.5 ± 2.1, −52 ± 3, −60 ± 4) km s⁻¹, is inconsistent with any nearby young associations (e.g., Zuckerman & Song 2004; Mamajek 2007; Torres et al. 2008) and indicates old disk kinematics (Leggett 1992), Schneider et al. estimate the age of the system to be older than ∼2 Gyr. No spectral observations of the companion have yet been reported.

In this article, we report separated, near-infrared spectra for GJ 660.1A and B obtained with the NASA Infrared Telescope Facility (IRTF) SpeX spectrograph (Rayner et al. 2003). Section 2 summarizes the observations. Section 3 describes our analysis of the spectra, including the classification of the low-resolution data, the characterization of spectral peculiarities, and the analysis of moderate-resolution spectra of the primary, which yields the system's metallicity. Section 4 presents model fitting of the secondary spectrum and corresponding estimates of the effective temperature (T_eff), surface gravity ($\mathrm{log}g$), metallicity ([M/H]), and other physical properties of this source. We discuss our findings in Section 4, focusing, in particular, on mismatches between the model and observed spectra and their possible origin. For reference, Table 1 summarizes the properties of the GJ 660.1 components based on this and prior work.

Table 1.  Properties of the GJ 660.1AB System

Property	GJ 660.1A	GJ 660.1B	References
Spectral Type	M1	d/sdM7	(1), (2)
Distance (pc)	20.0 ± 1.4		(3)
2MASS J	8.66 ± 0.03	13.05 ± 0.05	(4)
2MASS H	8.07 ± 0.04	12.57 ± 0.02	(4)
2MASS Ks	7.94 ± 0.02	12.23 ± 0.03	(4)
Absolute 2MASS J	7.15 ± 0.15	11.54 ± 0.15	(3), (4)
Absolute 2MASS H	6.56 ± 0.15	11.06 ± 0.15	(3), (4)
Absolute 2MASS Ks	6.43 ± 0.15	10.72 ± 0.15	(3), (4)
Separation ('')	6.08 ± 0.02		(2)
Projected separation (AU)	122 ± 9		(2)
Teff (K)	3800	${2550}_{-90}^{+80}$ a	(1), (5)
$\mathrm{log}g$ (cgs)	4.74	⋯b	(1), (6)
[M/H]	−0.47 ± 0.07c	$-{0.96}_{-0.24}^{+0.19}$ a	(1)
[Fe/H]	−0.63 ± 0.06c	⋯	(1)
Radius (R⊙)	0.53 ± 0.04	0.085 ± 0.014a	(1), (6)
Mass (M⊙)	0.57 ± 0.07	0.084–0.091	(1), (6)

Notes.

^aBased on the BT-Settl spectral model fit; the radius is based on the scale factor between models and data and includes uncertainty in distance; see Section 3.3. ^b $\mathrm{log}g$ value from spectral model fits is unreliable; see Section 4. ^cBased on the analysis of SXD data; see Section 3.2.2 and Table 4.

References.  (1) This paper; (2) Schneider et al. (2011), (3) van Leeuwen (2007), (4) Cutri et al. (2003), (5) Gaidos et al. (2014), (6) Maldonado et al. (2015).

Download table as:  ASCIITypeset image

Observations of the GJ 660.1AB system with SpeX were made during the nights of 2011 March 9 and 2015 April 9 (UT). Conditions were clear on both nights, with 08 and 06 seeing at K band, respectively. Our 2011 data were obtained with the SpeX prism mode and 05 slit, aligned perpendicular to the binary axis (position angle = 315°), which provided 0.7–2.5 μm coverage in a single order for each component separately with a resolution of $\lambda /{\rm{\Delta }}\lambda$ ≈ 120 and a dispersion of 20–30 Å pixel⁻¹. GJ 660.1B was exposed for six images of 90 s each at an airmass of 1.157, while GJ 660.1A was exposed for eight images for 1 s x 4 coadds each at an airmass of 1.139. We also observed the A0 V star calibrator HD 148968 (B = 7.116, V = 6.98) for flux calibration, and obtained flat field and HeNeAr arc lamps. In 2015, we observed the primary only with the SpeX cross dispersed mode and 08 × 15'' slit aligned at the parallactic angle, providing moderate-resolution data ($\lambda /{\rm{\Delta }}\lambda$ ≈ 750) with a dispersion of 3.6 Å pixel⁻¹ covering 0.8–2.4 μm in six orders. Four exposures of 90 s each were obtained at an airmass of 1.139, followed by the same A0 V calibrator star and lamp exposures. Data reduction was performed using the SpeXtool package version 3.4 (Vacca et al. 2003; Cushing et al. 2004), which removes background, bad pixels, and instrumental effects and optimally extracts the spectral data. The wavelength scale was calibrated to air wavelengths using the arc lamp spectrum, and slit and telluric transmission losses were corrected using the A0 V star observation.

The reduced prism spectra of GJ 660.1A and B are displayed in Figure 1, while the reduced SXD spectrum of GJ 660.1A is shown in Figure 2. The spectra of the primary show features consistent with its M1 optical classification, including H₂O (0.92–0.95 μm, 1.08–1.20 μm, 1.325–1.450 μm, and 1.72–2.14 μm) and CO (2.3 μm) absorption, and numerous atomic metal lines in the moderate-resolution data (Leggett et al. 2001; Rayner et al. 2009). We analyze this spectrum in further detail in Section 3.2.2. The secondary spectrum exhibits features consistent with late M/early L dwarfs, including absorption from H₂O, CO, and FeH molecules (1.0 μm and 1.2 μm), and several unresolved alkali lines including K i (1.25 μm) and Na i (1.51 μm; Leggett et al. 2001; Reid et al. 2001). This confirms the photometric classification of Schneider et al. (2011). There are a few clear peculiarities in the spectrum of this source, including its triangle-shaped H-band peak and shallow CO absorption, that are discussed in detail below.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

3.1. Classification of GJ 660.1B

To refine the classification of GJ 660.1B, we first considered the spectral indices and index-SpT relations from Allers et al. (2007). All measured indices had uncertainties estimated from spectral noise through Monte Carlo techniques, and these were propagated into the spectral-type determinations along with each relation's intrinsic scatter. These measurements (Table 2) resulted in a classification of M9.5 ± 0.3, driven primarily by the strong H₂O absorption present in the spectrum of this source.

Table 2.  Index Measurements for GJ 660.1B

Index	Value	Spectral Type
H2O	1.119 ± 0.007	M8.9 ± 0.4
H2O-1	0.604 ± 0.003	L2.8 ± 1.1
H2O-2	0.886 ± 0.009	M9.3 ± 0.6
H2OD	0.987 ± 0.011	[L0.3 ± 0.8]a
Mean		M9.5 ± 0.3

Note. Indices defined in Slesnick et al. (2004), McLean et al. (2003), and Allers et al. (2007).

^aOutside defined range.

Download table as:  ASCIITypeset image

We then compared the spectrum of GJ 660.1B directly to the M and L dwarf spectral standards defined in Kirkpatrick et al. (2010) using a chi-square statistic:

$$\begin{eqnarray}&&{\chi }^{2}(k)=\displaystyle \sum _{\lambda }\ \displaystyle \frac{{({x}_{\lambda }-\alpha {m}_{\lambda }(k))}^{2}}{{\sigma }_{\lambda }^{2}}\end{eqnarray} \tag{ 1 }$$

where x_λ is the measured spectral flux density of GJ 660.1B, σ_λ is the uncertainty of this spectrum, ${m}_{\lambda }(k)$ is the spectrum of standard k, and α is the optimal scaling factor computed as

$$\begin{eqnarray}&&\alpha =\displaystyle \frac{{\displaystyle \sum }_{\lambda }{x}_{\lambda }{m}_{\lambda }(k)/{\sigma }_{\lambda }^{2}}{{\displaystyle \sum }_{\lambda }{m}_{\lambda }^{2}(k)/{\sigma }_{\lambda }^{2}}\end{eqnarray} \tag{ 2 }$$

(see Cushing et al. 2008). Following the procedure described in Kirkpatrick et al. (2010), all sums were performed over the wavelength bins spanning 0.9–1.4 μm. The spectral standard with the lowest ${\chi }^{2}$ was found to be the M7 dwarf VB 8 (van Biesbroeck 1961), but the agreement is poor beyond 1.35 μm (Figure 3). The M8, M9, and L0 spectral standards are also poor matches to the spectrum of GJ 660.1B across the 0.8–2.4 μm near-infrared band.

Zoom In Zoom Out Reset image size

Finally, we compared the spectrum of GJ 660.1B to 911 optically classified M5–L5 spectra in the SpeX Prism Library (SPL; Burgasser 2014) using the same fit statistic as above. Comparisons were made over the wavelength ranges 0.7–1.35 μm, 1.42–1.8 μm, and 1.92–2.45 μm to avoid regions of strong telluric absorption. The best-fit template using the same χ² statistic above is the high proper motion (${\boldsymbol{\mu }}$ = [−124 ± 50, −821 ± 8] mas yr⁻¹) M7.5 brown dwarf GRH 2208-2007 (Tinney et al. 1998; Cruz et al. 2003; Deacon et al. 2005, Figure 4), which fits significantly better than VB 8, although GJ 660.1B has a somewhat bluer overal spectral energy distribution (SED).

Zoom In Zoom Out Reset image size

In summary, we found significant discrepancies in the classifcation of GJ 660.1B between three methods, with results ranging from M7 (standards) to M9.5 (indices). More importantly, the spectrum deviates significantly from the spectral standards in this type range. A robust classification of this source requires more detailed examination of these peculiarities.

3.2. Peculiarities in the Spectrum of GJ 660.1B

3.2.1. Hypothesis 1: A Young Brown Dwarf

The spectrum of GJ 660.1B exhibits an H-band (1.7 μm) peak that resembles those of many low-surface-gravity brown dwarfs in young clusters and associations (e.g., Lucas et al. 2001; Kirkpatrick et al. 2006; Allers et al. 2007; Allers & Liu 2013a). This spectral region in M and L dwarfs is shaped by H₂O and FeH molecular bands and H₂ collision-induced absorption (CIA; Linsky 1969; Borysow et al. 1997; Burgasser et al. 2008b; Allers & Liu 2013a), the last of which is a pressure-sensitive opacity source. Its triangle-shaped H-band peak suggests that GJ 660.1B may have a low surface gravity. To quantify this hypothesis, we used the index-based gravity classification scheme of Allers & Liu (2013a), which examines gravity-sensitive features from VO, FeH, alkali lines, and continuum shape. For a classification of M9.5, the resulting gravity scores (Table 3) yield a classification of INT-G (intermediate low surface gravity), which is consistent with objects of ages ≈100–300 Myr (Allers & Liu 2013a). However, if we assume a spectral type of M7, the scores instead yield an FLD-G (field dwarf) gravity classification.

Table 3.  Gravity Score Measurements for GJ 660.1B

Index	Value	Gravity	Gravity
		Score (M9.5)	Score (M7)
VO-z	1.004 ± 0.004	⋯	⋯
FeH-z	1.189 ± 0.006	0	0
H-cont	0.961 ± 0.006	1	0
KI-J	1.084 ± 0.003	1	0
Mean		INT-G	FLD-G

Note. Indices defined in Allers & Liu (2013a).

Download table as:  ASCIITypeset image

To assess these conflicting results, we compared the spectrum of GJ 660.1B directly to other low- and intermediate-gravity spectra in the SPL. The best-matching case⁶ is the M8γ/M8 VLG 2MASS J05341592−0631401 (Kirkpatrick et al. 2010; Allers & Liu 2013a; Figure 5). While the H-band peaks are similar between these sources, there is a clear mismatch in the 0.9–1.4 μm region. While 2MASS J05341592−0631401 shows weak FeH absorption at 1.0 μm and strong VO at 1.05 μm, the spectrum of GJ 660.1B has strong FeH and no discernable VO band. In addition, the overall SED of GJ 660.1B is significantly bluer ($J-{K}_{s}$ = 0.82 ± 0.06) than comparably classified dwarfs ($\langle J-{K}_{s}\rangle$ = 1.08 ± 0.19 for M7 dwarfs, $\langle J-{K}_{s}\rangle$ = 1.20 ± 0.22 for M9 dwarfs; Faherty et al. 2009), whereas young brown dwarfs tend to be 0.1–0.3 magnitudes redder (Faherty et al. 2012). We therefore conclude that GJ 660.1B is not young, and that its peculiarities arise from another source.

Zoom In Zoom Out Reset image size

3.2.2. Hypothesis 2: A Metal-Poor Dwarf

The best-matching SPL spectra to GJ 660.1B have common physical properties. Two particularly good matches are 2MASS J01151621+3130061, a moderately metal-poor d/sdM8 dwarf (Bowler et al. 2009, Figure 5); and HD 114762B,a d/sdM9±1 which is a companion to a metal-poor star ([Fe/H] = −0.7; Latham et al. 1989; Hale 1995; Bowler et al. 2009). The d/sd "mild subdwarf" classifications of these sources reflect spectral peculiarities arising from their slightly subsolar metallicities (Burgasser et al. 2007). These spectra show the same triangular-shaped H-band peaks, strong FeH and alkali line absorption, and blue SEDs as GJ 660.1B.

To validate this hypothesis, we used the moderate-resolution near-infrared SpeX spectrum of GJ 660.1A (Figure 2) to re-assess its metallicity. As described in Rojas-Ayala et al. (2010, 2012), metal lines in the near-infrared can be used to determine the metallicities of M-type dwarfs based on empirical calibrations from M-dwarf companions to FGK stars.We used the metallicity calibration relations defined by Rojas-Ayala et al. (2012), Terrien et al. (2012), Mann et al. (2013), and Newton et al. (2014), which are based on the equivalent widths of the 1.516 μm K i, 2.205 μm Na i, and 2.263 μm Ca i metal lines and the H₂O spectral indices defined in Covey et al. (2010) and Rojas-Ayala et al. (2012). The resulting estimates of [Fe/H] and [M/H] are listed in Table 4. All of the estimates are in agreement except for those from Terrien et al. (2012), which exhibit a large difference between H- and K-band calibrations. Rejecting these, we determine mean values of [Fe/H] = −0.63 ± 0.06 and [M/H] = −0.47 ± 0.07 for GJ 660.1A. The former is lower than, but statistically consistent with, [Fe/H] = −0.88 ± 0.13 as reported by Gaidos et al. (2014), based on the calibration of optical metal lines given by Mann et al. (2013). We adopt the near-infrared metallicity determination for GJ 660.1A in the remainder of this study.

Table 4.  Metallicity Determinations for GJ 660.1A

Method	[Fe/H]	[M/H]
Rojas-Ayala et al. (2012) K band	−0.57 ± 0.14	−0.40 ± 0.10
Terrien et al. (2012) H band	−0.18 ± 0.12a	⋯
Terrien et al. (2012) K band	−1.08 ± 0.12a	⋯
Mann et al. (2013) J band	−0.60 ± 0.07	−0.32 ± 0.08
Mann et al. (2013) H band	−0.62 ± 0.08	−0.52 ± 0.06
Mann et al. (2013) K band	−0.64 ± 0.06	−0.50 ± 0.05
Newton et al. (2014) K band	−0.80 ± 0.13	⋯
Weighted Mean	−0.63 ± 0.06	−0.47 ± 0.07
Optical (Gaidos et al. 2014)	−0.88 ± 0.13	⋯

Note.

^aNot included in the uncertainty-weighted mean.

Download table as:  ASCIITypeset image

The proximity and co-movement of the two components of GJ 660.1AB strongly indicate coevality, and the subsolar metallicity of the primary supports our second hypothesis that the spectral peculiarities seen in GJ 660.1B likely arise from low-metallicity effects. These peculiarities are not as extreme as those seen in M subdwarfs and extreme M subdwarfs, as illustrated in Figure 6, which shows that the near-infrared spectral morphology of GJ 660.1B is intermediate between those of M7 and sdM7 dwarfs. Following Bowler et al. (2009) and Kirkpatrick et al. (2010), we assign a classification of d/sdM7 for this source based on its similarity to the M7 standard over 0.9–1.4 μm and mild low-metallicity features across the full SED (see also Jao et al. 2008).

Zoom In Zoom Out Reset image size

3.2.3. Hypothesis 3: An Unresolved Binary

We also considered a third hypothesis, that the unusual spectral features of GJ 660.1B arise from the spectral contamination of an unresolved binary (or in this case, a tertiary) companion. This scenario has been used to explain peculiar spectral features in several late M dwarfs, which are suspected or have been confirmed to host T dwarf companions (e.g., Burgasser et al. 2008a; Burgasser et al. 2012, 2015; Bardalez Gagliuffi et al. 2014). To examine this hypothesis, we followed the methods described in Burgasser et al. (2010) and Bardalez Gagliuffi et al. (2014), comparing the spectrum of GJ 660.1B to 487 M6–L2 dwarfs not previously classified as metal-poor, young, or binary in the SPL. We then constructed 279,237 synthetic binary template spectra from these primaries and 574 L0–T6 dwarfs (assuming the secondary spectral type is the same or later than the primary type), using the Dupuy & Liu (2012) M_J versus spectral type relation to scale the individual component spectra. Figure 7 displays the best-fit binary templates resulting from this analysis, which is a combination of GRH 2208-2007 and the T5 pec dwarf WISE J044853.28−193548.6 (Kirkpatrick et al. 2011). While the binary template is a better fit (${\chi }^{2}$ = 44.2 versus 75.2), it fails to surpass the 99% significance threshold in an F-test comparison, and fails to reproduce structure in the 1.0–1.2 μm and 2.2–2.3 μm range. We therefore conclude that unresolved multiplicity is not the most likely source for the observed spectral peculiarities of GJ 660.1B.

Zoom In Zoom Out Reset image size

3.3. Model Fitting

3.3.1. Methodology

To better quantify the physical parameters of GJ 660.1B, we compared its spectrum to two sets of atmosphere models characterized by T_eff, $\mathrm{log}g$, and [M/H]: the BT-Settl models (Allard et al. 2011) and the Drift models (Helling et al. 2008a; Witte et al. 2009, 2011). For the former, we sampled models spanning effective temperatures T_eff = 1400–2900 K, $\mathrm{log}g$ (in cm s⁻²) = 3.5–5.5, and [M/H] = −4.0 to +0.5; for the latter, we sampled T_eff = 1700–3000 K, $\mathrm{log}g$ = 5.0–5.5, and [M/H] = −3.0 to 0.0. In both cases, the models were originally calculated every 100 K in T_eff, 0.5 dex in $\mathrm{log}g$, and 0.5 dex in [M/H], and at much higher resolution than the SpeX data. We smoothed the models to $\lambda /{\rm{\Delta }}\lambda$ = 120 using a Hann filter and interpolated onto a common wavelength scale. As the models are calculated in units of surface flux densities, we calibrated the observed spectrum of GJ 660.1B to absolute flux densities using its absolute 2MASS J-band magnitude (M_J = 11.54 ± 0.16; Schneider et al. 2011). The optimal scaling factor (Equation (2)) is then α = ${F}_{{\rm{spectrum}}}$/${F}_{{\rm{model}}}$ = (R/10 pc)², where R is the radius of GJ 660.1B. These fits, therefore, provide an estimate of the source size (Bowler et al. 2009).

We deployed a Markov Chain Monte Carlo (MCMC) code with a Metropolis–Hastings algorithm (Metropolis et al. 1953; Hastings 1970) to identify the best-fit model parameters and uncertainties for both sets of atmosphere models. We used initial estimates of T_eff = 2650 K based on its d/sdM7 spectral type and the T_eff/spectral type relations of Stephens et al. (2009), Marocco et al. (2013), and Filippazzo et al. (2015). We also assumed $\mathrm{log}g$ = 5.0, which is typical for a field late M dwarf; and [M/H] = 0.0. Two 10⁵-step chains were run in which our three model parameters were alternately updated using step sizes drawn from normal distributions of width ΔT_eff = 50 K, Δ$\mathrm{log}g$ = 0.25, and Δ[M/H] = 0.25. Models at intermediate parameter values were linearly interpolated in logarithmic flux and compared to the observed spectrum using the same χ² statistic as Equation (1). The criterion to adopt a successive parameter ${\theta }_{i}\to {\theta }_{i+1}$ was $U(0,1)\leqslant {e}^{-0.5({\chi }_{(i+1)}^{2}-{\chi }_{(i)}^{2})}$, where $U(0,1)$ is a random number drawn from a uniform distribution between 0 and 1. We eliminated the first 10% of each chain before the evaluation of the model parameters.

3.3.2. Results

The distributions⁷ of model parameters are shown in Figures 8 and 9, and summarized in Table 5. For the BT-Settl models, we find median parameters of T_eff = ${2550}_{-90}^{+80}$ K, $\mathrm{log}g$ = ${4.4}_{-0.5}^{+0.5}$, and [M/H] = $-{0.96}_{-0.24}^{+0.19}$, where uncertainties represent the 16% and 84% quantiles in the marginalized parameter distributions. Both T_eff and [M/H] are well-constrained, the former consistent with the classification of the source. However, the metallicity is 2.4σ lower than the metallicity determined for the GJ 660.1A. Surface gravity constraints are weaker, and the fits primarily limit $\mathrm{log}g$ ≲ 5.0. As discussed below, this is problematic, as it predicts a very-low-mass and age for GJ 660.1B, contrary to the comparative analysis above. There is a strong correlation between [M/H] and $\mathrm{log}g$, with a slope of d[M/H]/d$\mathrm{log}g$ ≈ 0.35. This likely arises from a trade-off in the strength of H₂ CIA, which increases with both increasing surface gravity and decreasing metallicity (e.g., Witte et al. 2009). Solutions with $\mathrm{log}g$ > 5 have mean metallicities of −0.63 ± 0.15, which is more in line with the metallicity of GJ 660.1A. In contrast, the best-fit model (Figure 10) has both a low $\mathrm{log}g$ = 3.85 and a very low [M/H] = −1.20. This model is a reasonable match to the SED of GJ 660.1B beyond 1.4 μm, but poorly reproduces the detailed features in the 0.9–1.4 μm region, most notably FeH absorption at 1.0 and 1.2 μm. Higher surface gravity and higher metallicity models match the <1.4 μm wavelength end of the spectrum better, but poorly reproduce the full SED.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Table 5.  Model Fitting Results

Spectral Models
Spectral Model	[M/H]	Teff (K)	$\mathrm{log}g$ ($\mathrm{cm}\;{{\rm{s}}}^{-2}$)	Radius (${R}_{\odot }$)
BT-Settl	$-{0.96}_{-0.24}^{+0.19}$	${2550}_{-90}^{+80}$	${4.4}_{-0.5}^{+0.5}$	0.085±0.014
Drift	$-{0.20}_{-0.19}^{+0.13}$	${2650}_{-80}^{+90}$	5.0–5.5	0.078 ± 0.013
Spectral and Evolutionary Models
Models	Mass (M⊙)	Age (Gyr)	Luminosity (L⊙)	Radius (R⊙)
BT-Settl & Baraffe	${0.038}_{-0.009}^{+0.018}$	${0.022}_{-0.011}^{+0.081}$	$-{2.8}_{-0.3}^{+0.3}$	${0.23}_{-0.08}^{+0.10}$
BT-Settl & Burrows	${0.031}_{-0.006}^{+0.015}$	${0.027}_{-0.009}^{+0.065}$	$-{2.9}_{-0.3}^{+0.2}$	${0.21}_{-0.07}^{+0.04}$
Drift & Baraffe	${0.088}_{-0.008}^{+0.005}$	${0.6}_{-0.4}^{+0.9}$	$-{3.19}_{-0.10}^{+0.13}$	${0.120}_{-0.008}^{+0.017}$
Drift & Burrows	${0.078}_{-0.009}^{+0.008}$	${0.4}_{-0.2}^{+0.6}$	$-{3.26}_{-0.10}^{+0.13}$	${0.110}_{-0.008}^{+0.015}$

Download table as:  ASCIITypeset image

The radius inferred from the scale factor is 0.085 ± 0.014 R_⊙, where we have incorporated the uncertainty in the distance determination. This is smaller than, but not inconsistent with, radius estimates of 0.10–0.11 R_⊙ based on the evolutionary models (Burrows et al. 2001; Baraffe et al. 2003) for T_eff = 2550 K and an age of τ = 5 Gyr. However, at this age, evolutionary models predict a surface gravity of 5.3–5.4, considerably higher than the surface gravity inferred from the atmosphere model fits. For the median T_eff and $\mathrm{log}g$ from these fits, the evolutionary models of Baraffe et al. (2003) predict a mass of M = ${0.038}_{-0.009}^{+0.018}$M_⊙, τ = ${22}_{-11}^{+81}$ Myr, and R_evol = ${0.23}_{-0.08}^{+0.10}$R_⊙, all consistent with a very young brown dwarf, our rejected hypothesis. The Burrows et al. (2001) evolutionary models yield similar parameters (Table 5). Note that the evolutionary radius is nearly three times larger (1.8σ discrepant) than the scaling factor radius. Bringing these into agreement requires a higher $\mathrm{log}g$ in order to lower R_evol, since the scaling factor is primarily set by the T_eff of the model (which is uncorrelated with the other parameters) and the observed absolute magnitude of GJ 660.1B.

For the Drift models, we find an equivalent T_eff = ${2650}_{-80}^{+90}$ K and R = 0.078 ± 0.013 R_⊙, but a much higher [M/H] = $-{0.20}_{-0.19}^{+0.13}$, although the metallicity distribution has a long tail toward lower values. The Drift models were constrained to have 5.0 $\leqslant$ $\mathrm{log}g$ $\leqslant$ 5.5, and there is no clearly preferred value in that range from the MCMC analysis. The Drift metallicity is inconsistent with that derived from the BT-Settl model (2.8σ discrepant) and considerably lower than metallicity of the primary (1.3σ discrepant), and this may be related to the surface gravity limits on the Drift models. There also appears to be a slight positive correlation between T_eff and [M/H] and a slight negative correlation between T_eff and $\mathrm{log}g$ in these fits. The best-fit Drift model (Figure 11) is a much better (although far from perfect) match to the 0.9–1.4 μm region as compared to the best-fit BT-Settl model, but a poorer match to the full near-infrared SED. For the median parameters, the evolutionary models of Baraffe et al. (2003) predict M = ${0.088}_{-0.008}^{+0.005}$M_⊙, τ = ${0.6}_{-0.4}^{+0.9}$ Gyr, and R_evol = ${0.120}_{-0.007}^{+0.017}$R_⊙, which is biased toward higher masses and older ages by the $\mathrm{log}g$ constraint; again, the Burrows et al. (2001) models yield similar values. Despite the smaller evolutionary model radius, there remains a 2.8σ disagreement between it and the scaling factor radius.

Zoom In Zoom Out Reset image size

Our spectral model fit analysis yields mixed results in elucidating the roles of surface gravity and metallicity in shaping the peculiar spectrum of GJ 660.1B. The models affirm a subsolar metallicity for this source, bracketing the metallicity determined for the primary. However, the BT-Settl models also converge on surface gravities that are appropriate for a very young brown dwarf (≲100 Myr), which disagrees with the kinematics of the system and comparison of the spectrum of GJ 660.1B to young template spectra. In addition, neither of the model sets produce a "best fit" that is an accurate representation of the data. The low-metallicity, low-surface gravity BT-Settl model roughly reproduces the λ > 1.4 μm SED of GJ 660.1B, but fails to match the molecular and atomic features for λ < 1.4 μm. The high-metallicity, high-surface-gravity Drift model has the opposite problem.

Bowler et al. (2009) reported similar disagreements in their analysis of the d/sdM9 companion to the metal-poor F9 star of HD 114762 ([Fe/H] = −0.7). Their Phoenix/GAIA model fits (Brott & Hauschildt 2005) to low-resolution near-infrared spectra of HD 114762B converged to very low surface gravities and retained a high χ². The study concluded that spectral models did not fit the low-resolution spectra of these low-metallicity, low-temperature sources, whereas moderate-resolution spectra yielded better fits. Burgasser et al. (2009), Witte et al. (2011), and Manjavacas et al. (2014) have found comparable problems for late M and L dwarfs and subdwarf low-resolution spectra and photometry.

What is causing the models to fail? We conjecture that the interplay of two opacity sources are primarily responsible: condensate cloud opacity and H₂ CIA. Condensates become an important opacity source in the atmospheres of late M and L dwarfs, producing a scattering haze that suppresses the 0.9–1.4 μm continuum (Ackerman & Marley 2001; Allard et al. 2001; Cooper et al. 2003; Helling et al. 2008b). Burgasser et al. (2003, 2007) and Gizis & Harvin (2006) have conjectured that condensate production is inhibited in metal-poor late M and L subdwarfs, based on their exceptionally blue near-infrared SEDs ($J-K\approx 0$) and persistent TiO optical bands. This has been confirmed in theoretical calculations by Witte et al. (2009), which demonstrated that grain production persists in low-metallicity dwarfs but with an overall opacity that declines with metallicity; and by the absolute J magnitudes of subdwarfs, which are brighter than equivalently classified dwarfs (Schilbach et al. 2009; Faherty et al. 2012). Pressure-sensitive H₂ CIA, on the other hand, strengthens in both higher surface gravity and metal-poor dwarfs ($P\propto g/\kappa$), suppressing K-band flux and producing bluer SEDs. Both of these trends—reduced condensate opacity and stronger H₂ absorption—qualitatively explain the bluer SED tilt and stronger FeH and alkali absorption in the spectrum of GJ 660.1B and other metal-poor dwarfs as compared to field standards (Figure 3).

However, these opacity sources do not appear to be accurately represented in the models. Analysis of field M and L dwarf SPL spectra has shown that the Drift models, even with a prescription for metallicity-dependent condensate grain formation, underpredict condensate opacity in dwarfs with T_eff > 1900 K, resulting in SEDs that are bluer than observed (Witte et al. 2011). Similarly, Allard et al. (2011) have commented that BT-Settl models, and perhaps "all the current cloud models" fail to produce sufficient condensate absorption at the M dwarf/L dwarf transition. Separately, new calculations by Abel et al. (2012) and Frommhold et al. (2010) indicate that current models may be excessively blue in the near-infrared due to incorrect H₂ CIA opacity. Saumon et al. (2012) estimate that revised opacities translate into a 44% decrease in K-band absorption at 2000 K for cloud-free atmosphere models.

We surmise that both insufficient condensate opacity and excessively strong H₂ CIA are present in the Drift and BT-Settl models examined here, which are too blue and therefore biasing fits toward lower surface gravities (BT-Settl) and higher metallicities (Drift). These biases are apparent when we compare our GJ 660.1B spectrum to models for an "expected" parameter set (T_eff = 2600 K, $\mathrm{log}g$ = 5.3 and [M/H] = −0.65), which exhibit excessively strong molecular bands over 1.0–1.2 μm, excess flux at Hband and excess continuum suppression at Kband (Figures 10 and 11). Increasing condensate opacity in the models would mute the molecular gas features, while both grain scattering and weaker H₂ opacities would produce a redder tilt to the SED without the need for lower surface gravities. GJ 660.1B, HD 114762B, and other late M and L dwarf companions to metal-poor stars will be useful benchmarks for testing updated condensate and H₂ CIA opacities in future atmosphere models.

Finally, we comment on the most interesting spectral peculiarity of GJ 660.1B: its triangular-shaped H-band continuum. As discussed above, this feature is typically associated with low-surface-gravity sources and attributed to reduced H₂ CIA in a low-pressure atmosphere. However, it has also been reported in the spectra of high proper motion late M dwarfs with no evidence of youth (Kirkpatrick et al. 2010), and is present in the spectra of both GJ 660.1B and HD 114762B (Bowler et al. 2009). These common features point to subsolar metallicity as the likely source. However, more metal-poor late M and L subdwarfs do not exhibit this feature (Burgasser et al. 2003, 2009; Kirkpatrick et al. 2010, 2014; Figure 6). An additional opacity source that may help shape the H-band continuum of mild subdwarfs is the 1.55–1.60 μm $E{}^{4}{\rm{\Pi }}-A{}^{4}{\rm{\Pi }}$ FeH absorption band (Cushing et al. 2003; Dulick et al. 2003). Like its 1.0 μm and 1.2 μm counterparts, we expect this feature to be enhanced in the spectra of mild subdwarfs, resulting in a sloped, rather than flat, H-band continuum. Stronger H₂O absorption, a consequence of reduced condensate opacity, may also play a role in shaping this feature, while simultaneously skewing index-based near-infrared classifications to later types (Section 3.1). The presence of this feature, and its influence on the gravity indices, counters the claim of Allers & Liu (2013b) that metal-poor dwarfs cannot be misclassified as low gravity by this technique. However, we concur with these authors that "caution should be used when determining" near-infrared spectral types with indices alone. We advocate full-spectum comparisons as a more robust method for characterizing VLM dwarfs with unusual spectral and physical properties.

A.B and J.F were Visiting Astronomers at the Infrared Telescope Facility, which is operated by the University of Hawaii under Cooperative Agreement no. NNX-08AE38A with the National Aeronautics and Space Administration, Science Mission Directorate, Planetary Astronomy Program. The authors thank IRTF operators Paul Sears and Brian Cabreira, and instrument specialist John Rayner, for their assistance with the SpeX observations. We also thank John Johnson for suggesting a metallicity analysis of GJ 660.1A, and our anonymous referee for their rapid review. C.A. acknowledges funding by the University of California Office of the President through the UC-HBCU program. The material presented here is based on work supported by the National Aeronautics and Space Administration under grant No. NNX15AI75G. This research has made use of the SIMBAD database, operated at CDS, Strasbourg, France; NASA's Astrophysics Data System Bibliographic Services; the M, L, T, and Y dwarf compendium housed at http://DwarfArchives.org and the Spex Prism Libraries at http://www.browndwarfs.org/spexprism. The authors recognize and acknowledge the very significant cultural role and reverence that the summit of Mauna Kea has always had within the indigenous Hawaiian community. We are most fortunate and grateful to have the opportunity to conduct observations from this mountain.