Integral Field Spectroscopy of the Low-Mass Companion HD984B with the Gemini Planet Imager

TL;DR

This paper reports detailed observations and orbit fitting of the low-mass companion HD 984 B using the Gemini Planet Imager, revealing its spectral type, temperature, luminosity, and mass estimates, and introduces a spectral analysis method.

Contribution

It presents the first orbit determination of HD 984 B and a novel spectral covariance analysis method for better spectral characterization.

Findings

Orbit of HD 984 B is approximately 18 AU with a 70-year period.

Spectral type of the companion is M6.5+/-1.5 with a temperature of about 2730 K.

Mass estimates range from 34 to 95 Jupiter masses depending on age.

Abstract

We present new observations of the low-mass companion to HD 984 taken with the Gemini Planet Imager as a part of the Gemini Planet Imager Exoplanet Survey campaign. Images of HD 984 B were obtained in the J (1.12--1.3 micron) and H (1.50--1.80 micron) bands. Combined with archival epochs from 2012 and 2014, we fit the first orbit to the companion to find an 18 AU (70 year) orbit with a 68% confidence interval between 14 and 28 AU, an eccentricity of 0.18 with a 68% confidence interval between 0.05 and 0.47, and an inclination of 119 degrees with a 68% confidence interval between 114 degrees and 125 degrees. To address considerable spectral covariance in both spectra, we present a method of splitting the spectra into low and high frequencies to analyze the spectral structure at different spatial frequencies with the proper spectral noise correlation. Using the split spectra, we compare to known spectral types using field brown dwarf and low-mass star spectra and find a best fit match of a field gravity M6.5+/-1.5 spectral type with a corresponding temperature of 2730+120 K. Photometry of the companion yields a luminosity of log(L_bol/L_sun) = -2.88+/-0.07 dex, using DUSTY models. Mass estimates, again from DUSTY models, find an age-dependent mass of 34+/-1 to 95+/-4 M_Jup. These results are consistent with previous measurements of the object.

Tables

Table 1. Table 1 Astrometry

Band	Date	Separation (mas)	PA($°$)	Instrument	Note
$L^{\prime }$	2012 July 18	$190\pm 20$	$108.8\pm 3.0$	NaCo	APP data
$L^{\prime }$	2012 July 20	$208\pm 23$	$108.9\pm 3.1$	NaCo	direct imaging
$HK$	2014 Sep 9	$201.6\pm 0.4$	$92.2\pm 0.5$	SINFONI
$H$	2015 Aug 29	$216.3\pm 1.0$	$83.3\pm 0.3$	GPI
$J$	2015 Aug 29	$217.9\pm 0.7$	$83.6\pm 0.2$	GPI

Table 2. Table 2 System Properties

Property	unit	HD 984	HD 984 Ba	HD 984 Bb	References
Distance	pc	$47.1\pm 1.4$			1
Age	Myr	30–200			2
mH	–	$6.170\pm 0.038$	$12.58\pm 0.05$	$12.60\pm 0.05$	3
mJ	–	$6.402\pm 0.023$		$13.28\pm 0.06$	3
Spectral Type	–	F7V	M$6.0\pm 0.5$	M$6.5\pm 1.5$	4
Temperature	K	$6315\pm 89$	${2777}_{-130}^{+127}$	${2730}_{-180}^{+120}$	5
$\log \left(L_{\mathrm{bol}}/L_{☉}\right)$	dex	$0.346\pm 0.027$	$-2.815\pm 0.024$	$-2.88\pm 0.07$	2
Mass	–	$\sim$1.2$M_{☉}$	$33\pm 6$ to $94\pm 10$ MJup	$34\pm 1$ to $94\pm 4$ MJup	2
Semi Major Axisc	au			${18}_{-4}^{+10}$
Periodc	yrs			${70}_{-25}^{+69}$
Inclinationc	deg			${119}_{-5}^{+6}$
Eccentricityc	–			${0.18}_{-0.13}^{+0.29}$

Full text

1. Introduction

The search for exoplanets through direct imaging has led to many serendipitous detections of brown dwarfs and low-mass stellar companions (e.g. Chauvin et al., 2005; Biller et al., 2010; Nielsen et al., 2012; Mawet et al., 2015; Konopacky et al., 2016). These surveys tend to target young, bright stars whose potential companions would still be warm and bright in the infrared (e.g., Burrows, 1997). Brown dwarfs, having higher temperatures than planetary-mass companions of the same age, are significantly brighter and therefore easier to detect. While brown dwarf companions are not the primary focus of direct imaging searches, they are useful in their own right for a better understanding of substellar atmospheres and for comparing competing formation models (e.g. Perets & Kouwenhoven, 2012; Bodenheimer et al., 2013; Boley & Durisen, 2010).

Meshkat et al. (2015) reports the discovery of a bound low-mass companion to HD 984, a bright, nearby ($47.1\pm 1.4$ pc) F7V star of mass $\sim$1.2 M☉ (van Leeuwen, 2007; Meshkat et al., 2015) and a temperature of $6315\pm 89$ K (White et al., 2007; Casagrande et al., 2011). With an age estimate of 30–200 Myr ($115\pm 85$ Myr at a 95% confidence level) derived from isochronal age, X-ray emission and rotation (Meshkat et al., 2015), and consistent with previous age estimates (Wright et al., 2004; Torres et al., 2008), HD 984 is ideal for direct imaging campaigns to search for young substellar objects.

The results presented in Meshkat et al. (2015) finds HD 984 B at a separation of $0\farcs 19\pm 0\farcs 02$ ($9.0\pm 1.0$ au) based on the $L^{\prime}$ observations with NaCo (Lenzen et al., 2003; Rousset et al., 2003) in 2012 July, and the $H+K$ band observations with SINFONI (Eisenhauer et al., 2003; Bonnet et al., 2004) in 2014 September, both mounted on UT4 on the Very Large Telescope (VLT) at Cerro Paranal, Chile. Comparing the SINFONI spectrum to field brown dwarfs and low-mass star in the NASA Infrared Telescope Facility (IRTF) library, Meshkat et al. (2015) conclude the companion to be a M$6.0\pm 0.5$ object (Cushing et al., 2005; Rayner et al., 2009). The paper reports an estimated mass of $33-96$ MJup, overlapping the brown dwarf and low-mass star regime. This mass range corresponds to a mass-ratio of $q=0.03-0.08$. Meshkat et al. (2015) note that future observations in the $J$ band could provide additional constraints on the surface gravity of the companion, which could help place the system in the lower or higher range of the age estimate of $115\pm 85$ Myr.

Competing theories of companion formation are reliant upon observational studies to verify their models. Systems like HD 984 B, with low mass-ratios are particularly interesting for understanding brown dwarf and low-mass companion formation. Kraus et al. (2011) find a flat mass-ratio distribution for solar type stars, which they state to be consistent with formation via fragmentation on small scales, however their sample does not extend below $q=0.1$. Furthermore, the long baseline of observations of HD 984 B allows for better determination of the companion’s orbit. This is necessary for understanding formation scenarios, given their dependence on separation, as some suggest disk instability is responsible for companion formation at large ($\gtrapprox 50$ au) separations and protostellar core fragmentation at small ones ($\lessapprox 50$ au) (e.g. Boss, 2001; Bodenheimer & Burkert, 2001; Clarke, 2009). Finally, characterizing young binary star systems like HD 984 is important for future observations with JWST and$/$or 30-m class telescopes that could discoverer lower mass exoplanets, giving insight of planet formation in binary systems.

HD 984 was observed as one of the targets of the Gemini Planet Imager Exoplanet Survey (GPIES, Macintosh et al., 2014), an ongoing survey of over 600 nearby young stars with nearly 900 hours dedicated time on the 8-meter Gemini South Telescope using the Gemini Planet Imager (GPI, Macintosh et al., 2014), a coronagraphic adaptive optics integral field spectrograph and imaging polarimeter. Using near-infrared imaging spectroscopy (0.9–2.4 µm) and advanced imaging and post-processing techniques, GPI detects thermal emission from exoplanets and brown dwarfs at angular separations of $0\farcs 2$–$1\farcs 0$ from their parent star.

We present new spectroscopic observations of HD 984 B at $J$ (1.12–1.35 µm, $R\sim 34$) and $H$ (1.50–1.80 µm, $R\sim 45$) bands with GPI, from which we derive the first orbit estimate and update luminosity, magnitude and mass measurements. In Section 2 we describe the GPI observations. Basic reductions are explained in Section 3. Section 4 details the point spread function (PSF) subtraction technique. Astrometry is discussed in Section 5, orbital fitting in Section 6 and spectral and photometric analyses are presented in Section 7. Finally, we conclude in Section 8.

2. Observations

HD 984 was observed with the GPI integral field spectrograph (IFS, Chilcote et al., 2012; Larkin et al., 2014) on 2015 August 30 UT during the GPIES campaign (program GPIES-2015B-01, Gemini observation ID: GS-2015B-Q-500-982) at Gemini South. The GPI IFS has a field-of-view (FOV) of $2.8\times 2.8$ arcsec2 with a plate scale of $14.166\pm 0.007$ miliarcseconds/pixel and a position angle offset of $-0\fdg 10\pm 0\fdg 13$ (Macintosh et al., 2014; De Rosa et al., 2015). Coronographic images were taken in spectral mode in the $J$ and $H$ bands. Observations were performed when the star was close to the meridian at an average airmass of 1.1 so as to maximize FOV rotation for angular differential imaging (ADI, Marois et al., 2006) and minimize the airmass during observations. Twenty-three exposures of 60 seconds of one coadd each were taken in the H band and 23 exposures, also of 60s and one coadd, were followed up in the $J$ band; the $J$ band data was acquired with the $H$ band apodizer due an apodizer wheel mechanical issue. Two $H$ band exposures and five $J$ band exposures were rejected due to unusable data quality. Total FOV rotation for H band was $15\fdg 2$ and a total rotation of $11\fdg 9$ was acquired with J band. Average DIMM seeing for $H$ and $J$ band sequences was $1\farcs 14$ and $0\farcs 82$ respectively, higher than the median seeing of $0\farcs 65$. The windspeed averages for $H$ and $J$ bands were 2.5 m s*-1* and 1.9 m s*-1*. Images for wavelength calibration were taken during the daytime at zenith and short exposure arc images were acquired just before the sequences at the target elevation to correct for instrument flexure (Wolff et al., 2014).

3. Reductions

The images were reduced using the GPI Data Reduction Pipeline (Perrin et al., 2014) v1.3.0111http://docs.planetimager.org/pipeline. Using primitives in the pipeline, raw images were dark subtracted, argon arc image comparisons were used to compensate for instrument flexure (Wolff et al., 2014), the spectral data cube was extracted from the 2D images (Maire et al., 2014), bad pixels were interpolated in the cube and distortion corrections were applied (Konopacky et al., 2014). A wavelength solution was obtained using arc lamp images taken during the day prior to data acquisition. Four satellite spots, PSF replicas of the star generated by the pupil-plane diffraction grating (Marois et al., 2006; Sivaramakrishnan & Oppenheimer, 2006), were used to measure the location of the star behind the coronagraphic mask for image registration at a common centre, and to calibrate the object flux to star flux (Wang et al., 2014).

4. PSF Subtraction

After the initial data reduction, each slice of each data cube, which were each flux normalized using the average maximum of a Gaussian fit on the four calibration spots, were spatially magnified to align diffraction-induced speckles using the pipeline-derived spot positions to determine the star position. The images were then unsharp masked using a $11\times 11$ pixel kernel to remove the seeing halo and background flux, and were PSF subtracted using the TLOCI (Template Locally Optimized Combination of Images) algorithm (Marois et al., 2014). TLOCI uses a priori planet spectral information to optimize the least squares-based subtraction algorithm to maximize planet detection when using data acquired with the simultaneous spectral differential imaging (SSDI, Racine et al., 1999; Marois et al., 2000) and the ADI (Marois et al., 2006) techniques. T8 and L0 templates created from the Brown Dwarf Spectroscopic Survey (McLean et al., 2003) were used for these reductions. The TLOCI pipeline was run in a 6 pixel (85 mas) width annulus of increasing size with an inner gap to avoid the coronagraph focal plane mask. The IDL invert.pro algorithm, which uses a Gaussian elimination method, was used to invert the correlation matrix with the single value decomposition cutoff algorithm which shows limited gains in the signal-to-noise ratio (SNR) for GPI data. Two annuli, one with a width of 5 pixels interior to the subtraction zone inner annuli, and another of a 10 pixel width exterior to the subtraction zone outer annuli, are used for the reference pixels to derive the correlation matrix (pixels in the subtraction zones are not included in the correlation matrix to avoid the algorithm to fit the planet). Once all the data cubes had been PSF subtracted, a final 2D image was obtained by performing a weighted-mean of the 37 slices, using the input template spectrum and image noise to maximize the object’s SNR. Final images for $J$ and $H$ bands shown in Figure 1. While the initial discovery was obtained by performing both an SSDI and ADI subtractions, to avoid spectral cross-talk bias, only a less aggressive ADI-only subtraction (reference images are selected if they have less than 30% of the substellar object flux in a 1.5 $\lambda/D$ diameter aperture centred at the object position) was used for spectral and astrometry extractions.

5. Astrometry and Spectral Extraction

Using the TLOCI ADI-only subtracted final combined data cube, the flux and position of the companion were measured relative to the star. As TLOCI uses a training zone that differs from the subtraction zone, the companion is not fitted by the least-squares, thus removing one bias. To further take into account the self-subtraction bias for both spectral and astrometry extraction, the companion’s signal was fitted using a forward model derived from the median-average PSF of the four calibration spots for the entire sequence. For each slice of each data cube, a noiseless image was created with the calibration spot PSF at the approximate location of the companion. These simulated images were then processed using the same steps as the science images to produce the companion forward model. The forward model was then iterated in flux and position to minimize the local residual post subtraction in a 1.5 $\lambda/D$ circular aperture, where $\lambda$ is the wavelength and $D$ is the telescope diameter. Error bars were derived by adding and extracting the forward model flux at the same separation as the companion, but at nine different position angles. The standard deviation in flux and position of the nine simulated companions are adopted as the spectral and astrometric errors. A correction is applied to account for forward model errors. Instead of adding the simulated companions having the same flux as the recovered flux, they are added after normalizing the companion signal by the ratio of the local residual noise inside a 1.5 $\lambda/D$ aperture after the forward model best subtraction relative to the noise at the same angular separation calculated away from the companion.

In the $H$ band, the separation was $216.3\pm 1.0$ mas, and in the $J$ band, $217.9\pm 0.7$ mas. The position angles for $H$ and $J$ bands were $83\fdg 3\pm 0\fdg 3$ and $83\fdg 6\pm 0\fdg 2$, respectively. The uncertainty is a combination of measurement error added in quadrature with plate scale, north angle error, and star position error (0.05 pixels, Wang et al., 2014). The astrometry is summarized alongside the previous measurements from Meshkat et al. (2015) in Table 1.

Since TLOCI does not directly yield the spectrum of the companion, but rather the ratio of the companion to the spectrum of the primary star, the stellar spectra therefore needs to be divided out.This calibration uses the known stellar spectral shape to recover the true shape of the object’s spectra and was done using a custom IDL program and the Pickles stellar spectral flux library (Pickles, 1998). First, spectra for both F5V and F8V models in $J$ and $H$ bands were degraded to GPI spectral resolution and interpolated at the same wavelengths as the GPI wavelength channels. The zero points were computed with a Vega spectrum222http://www.stsci.edu/hst/observatory/crds/calspec.html and the $J$ and $H$ filter transmission curves and used to calibrate the Pickles template to the $J$ and $H$ magnitudes of HD 984. For each band, the normalized F5V and F8V spectra were interpolated to compute a F7V since the Pickles library did not contain a model spectrum for an F7V star. Finally, the calibrated F7V model spectrum was multiplied by the planet-to-star spectra for each band extracted from the GPI data to obtain the final J and H band spectra of HD 984 B.

6. Orbital Fitting

We determined the orbital parameters consistent with the full astrometric record of HD 984 using the rejection sampling method previously presented in De Rosa et al. (2016) and Rameau et al. (2016). Orbital parameters were drawn from priors that are uniform for eccentricity ($e$), argument of periastron ($\omega$), and epoch of periastron passage ($T_{0}$), and uniform in $\cos i$ for inclination angle. Semimajor axis ($a$) and position angle of nodes ($\Omega$) are assigned initial values of $a=1$ au and $\Omega=0$ degrees. These values are then adjusted so that the orbit matches one of the epochs of data which has the effect of imposing a prior uniform in position angle of nodes and uniform in $\log a$. Observational errors are taken into account by scaling to Gaussian distributions in separation and position angle, centered on the measurements from the reference epoch with the standard deviations of the Gaussians equal to the errors. Period is not fit independently but is instead derived from Kepler’s third law and the mass of the star of 1.2 $M_{\sun}$.

The fitting gives a median semimajor axis of 18 au (70 year) orbit, with a 68% confidence interval between 14 and 28 au and an eccentricity of 0.18 with a 68% confidence interval between 0.05 and 0.47 and inclination of 119∘ with a 68% confidence interval between 114∘ and 125∘. The highest probability orbit has parameters of $a=23.17$ au, $e=1.74$, $i=115\fdg 26$, $\omega=49\fdg 92$, $\Omega=208\fdg 48$, $T_{0}=2105.80$, $P=101.80$ yrs, with $\chi^{2}=9.80$. Fits to the orbit are shown in Figure 2 and posterior probability distributions and covariances of the orbital parameters are shown in Figure 3. Meshkat et al. (2015) do not perform an orbital fit in their analysis, but from their two epochs believe the system to have a non-zero inclination, which we have confirmed here.

7. Photometric and Spectroscopic Analysis

We first calculate the absolute magnitudes for HD 984 B by integrating the companion-to-star spectra, and correcting for the GPI filter transmission profile and Vega zero points (De Rosa et al., 2016). The $J$ and $H$ band apparent magnitudes were calculated to be $13.28\pm 0.06$ and $12.60\pm 0.05$, respectively. Assuming a distance to the star of $47.1\pm 1.4$ pc (van Leeuwen, 2007), the absolute magnitudes of the object in $J$ and $H$ bands are $9.92\pm 0.09$ and $9.23\pm 0.08$, respectively. The $H$ band magnitude is consistent with the $H$ band magnitude reported in Meshkat et al. (2015). A rudimentary spectral type can be ascertained using the $J$ and $H$ band magnitudes as compared to other brown dwarfs and low mass stars via a colour-magnitude diagram (see Figure 4). When compared with literature brown dwarfs and low-mass star from Dupuy & Liu (2012), these magnitudes further corroborate the Meshkat et al. (2015) result of a late M-type object.

For further characterization of the companion, we discuss the spectral analysis here. A more detailed analysis of the spectral type requires attention to spectral noise covariance, which arises from the coupling of neighbouring wavelength channels in the spectra and is a result of the finite resolution of GPI. This type of correlation correction is necessary for proper error calculation. IFS instruments observations often produce spectral noise covariance (e.g. Greco & Brandt, 2016), and GPI data cubes are also known to suffer from this effect, especially at small separations close to the focal plane mask. Before any comparisons to field objects can be made, this spectral noise covariance needs to be characterized to avoid biasing any analysis with improper error calculations. Given the high SNR of detections, it may be possible to fit higher frequency structures (e.g. spectral lines) in the spectrum independently to the low frequency envelope (the overall shape of the spectra). The noise characteristics, especially the spectral noise correlation, may differ with spectral frequencies, with the low frequencies mostly limited by highly correlated speckle noise slowly moving over the object as a function of wavelength, while higher frequencies could be mainly limited by read or background noises, thus being weakly correlated between wavelength channels.

The $J$ and $H$ band spectra of HD 984 B were split into low and high frequencies by taking the Fourier transform of the spectra (see Figure 5). Any frequencies between -2 and 2 cycles per bandwidth are considered low frequencies, and anything outside of that range is designated as high frequencies. This range was selected to incorporate the bulk of the flux in the low frequency range. The high and low spatial frequency errors were propagated by taking the spectra at nine different position angles, splitting their high and low components and taking the spread in the noise of each component as the error. The split high and low frequency spectra are shown in Figure 6. Testing the spectral noise correlation for the high frequency spectra separately found correlations covering just three wavelength channels for both bands as expected from read and background noises (see Figure 7); this spectral correlation is consistent with the GPI pipeline spectral over-sampling. The low frequency spectra were still highly correlated to fifteen and eight wavelength channels for $J$ and $H$ bands respectively (correlation reaching 50%).

Since the spectral channels at different wavelengths are correlated, each spectrum can be binned accordingly to avoid biasing comparisons to the spectra of other objects. High frequency spectra were binned by averaging three adjacent channels into twelve groups, with the last group averaging four channels to incorporate the leftover channel. $H$ band low frequency spectra were binned into two groups of twelve channels each and one group of thirteen channels, and $J$ band low frequency spectra were binned into five groups, three of seven channels each and two of eight channels each.

The extracted GPI spectra of HD 984 B, and its low and high-frequency components, and the SINFONI K band spectrum from Meshkat et al. (2015) were compared to a library of 1600 M, L, and T dwarf near-infrared spectra compiled from the SpeX Prism Library (Burgasser, 2014), the IRTF Spectral Library Cushing et al. (2005), the Montreal Spectral Library (e.g., Gagné et al., 2015; Robert et al., 2016), and the sample of young ultracool dwarfs presented in Allers & Liu (2013). The spectrum of each object was convolved with a Gaussian to degrade it to the same resolution as the GPI data. For the comparison with the low and high-frequency components, the library spectra were processed through the same Fourier filtering and binning procedure as for the HD 984 B spectrum. For the fits to the individual bands, the scaling factor between the HD 984 B spectrum and each comparison object was found analytically by evaluating the derivative of the $\chi^{2}$ equation (e.g., Burgasser, 2016). The SINFONI K band spectrum was fit using a similar procedure, with the spectrum of HD 984 B and the comparison objects degraded to a resolution of $R\sim 120$, similar to that of the majority of the objects within the spectral library. The SINFONI K band spectrum was not processed through the same Fourier filtering and binning steps as the GPI data.

The best fit to the three combined JHK spectra – one each with the unfiltered JH spectrum, the low frequency JH spectrum (JHlow), and the high frequency JH spectrum (JHhigh) – was found by minimizing the goodness of fit statistic $G$ (Cushing et al., 2008), similar to the $\chi^{2}$ statistic but with each term of the summation weighted proportionally to the bandwidth of each channel within the spectrum (e.g., Burgasser et al., 2010; Gagné et al., 2015). This weighting was particularly important for the fit to the JHlow spectrum which consisted of eight spectral channels, compared to the 160 within the K band spectrum. For each fit the same K band spectrum was used. Each band was fit with an independent scale factor, which accounts for uncertainties in the absolute flux calibration of the GPI data (Maire et al., 2014), the differences in flux calibration between the GPI and SINFONI data, and for the spread in near-infrared colors seen for young brown dwarfs (Cruz et al. submitted).

The goodness of fit statistic for each comparison object is plotted as a function of spectral type for each fit for the unfiltered spectrum and the low and high frequency components in Figure 8. The spectral type of HD 984 B was estimated by calculating the goodness of fit statistic (either $\chi^{2}$ or $G$) of the spectrum of HD 984 B and a template spectrum from the library created by averaging all objects of a given spectral type. The weighted average of the spectral type of each template was adopted as the spectral type, with the individual weights drawn from the F-test probability distribution function (e.g., Burgasser et al., 2010). A systematic uncertainty of 1 subclass was assumed for all objects within the spectral library, and incorporated into the uncertainty on the spectral types given in Figure 8. The spectral types estimated from the fit to both the JHK and the JHlowK spectra are consistent, both M$6.5\pm 1.5$ when rounded to the nearest half subtype, suggesting that the covariance between neighbouring wavelength channels within the GPI data did not strongly bias the fit. While the spectral type from the fit to the JHhighK spectrum was also consistent at M$7\pm 2$, this was almost entirely driven by the K band fit as neither the J nor H band GPI spectra of HD 984 exhibited strong spectral features, as demonstrated by the very large uncertainty on the spectral type from the JHhigh fit (L$4\pm 6$). While the high frequency spectrum did not help constrain the spectral type of HD 984 B, a similar analysis may prove more useful for K band spectra which cover the CO band heads at 2.29$\mu$m and 2.32$\mu$m (e.g., Konopacky et al., 2013).

The five objects within the spectral library which best fit the unfiltered J and H band spectra of HD 984 B are shown in Figure 9, and are consistent with the spectral type estimate from each band given in Figure 8. For the J band fit, two of the objects are known members of young moving groups: DENIS J004135.3-562112 (catalog ) (M7.5$\beta$, Tucana–Horologium 20–40 Myr; Reiners & Basri, 2009) and 2MASSI J0019262+461407 (catalog ) (M8$\beta$, AB Doradus 30–50 Myr; Schlieder et al., 2012). Additionally, 2MASS J20491972-1944324 (catalog ) was listed by Gagné et al. (2014) as having indications of youth based on both a red color and fits to model atmospheres. For the H band fit, the best fit object has a no indication of low surface gravity (Gagné et al., 2015), and of the five best fit only one is a candidate member of a young moving group, 2MASS J12271545-0636458 (catalog ) (M8.5$\beta$, TWA 5–15 Myr; Gagné et al., 2015). The five best fit objects to the JHK spectrum are shown in Figure 10. The best fit object was 2MASS J0019262+461407 (catalog ) (M$8\beta$), a likely member of the AB Doradus moving group (Schlieder et al., 2012), consistent with the $115\pm 85$ Myr age estimate for HD 984 (Meshkat et al., 2015). Of the remaining four, two are classified as having low surface gravity and are likely members of the Tucana–Horologium (2MASS J225511530-6811216 (catalog ), M$5\gamma$) and Columba (2MASS J05123569-3041067 (catalog ), M6.5$\gamma$) moving groups (Gagné et al., 2015), one has intermediate surface gravity and is a likely member of the AB Doradus moving group (SIPS J2039-1126 (catalog ), M$7\beta$; Gagné et al., 2015), and one has a field surface gravity classification (LP 759-17 (catalog ), M$7\alpha$; Gagné et al., 2015).

Based on the spectral fitting alone there is not a conclusive result in terms of the surface gravity classification (e.g. Cruz et al., 2009) for HD 984 B. While the global minima of $G$ for both the JHK and JHlowK spectra are populated by field or intermediate gravity objects, the low gravity objects appear to form a sequence displaced from field gravity objects by approximately one to two spectral subtypes hotter. Performing the same spectral typing procedure outlined previously on the JHK spectrum of HD 984 B using the subset of field gravity objects (gravity classification of $\alpha$ or FLD-G) yields a spectral type of M$7\pm 1$, compared with M$6\pm 1$ when using the low gravity subset ($\gamma$/$\delta$/vl-g). This effect is most pronounced for the H band where surface gravity has a significant effect on spectral morphology, with the same analysis yielding a spectral type of M$8.5\pm 1.5$ using field gravity objects, and of M$6.5\pm 1.0$ using low gravity objects. As the difference between the minimum goodness of fit statistic of the two subsets is small ($G=0.70$ and $G=0.81$ respectively), it is not possible to assign either a field or low surface gravity classification to HD 984 B based on this analysis.

Photometric and other spectral indicators can also be used to classify the gravity of late M to early L type objects (Allers & Liu, 2013; Filippazzo et al., 2015). We test the H-cont index described in Allers & Liu (2013) which uses the blue end slope of the H band continuum. A higher index ($\sim 1.0$) indicates a straighter slope which is seen in lower gravity objects. HD 984 B, when compared to other objects, has a comparable H-cont index ($0.95\pm 0.01$) for its determined spectral type, and there is no indication it is of low gravity. Furthermore, comparisons of HD 984 B’s $J$ band magnitude versus spectral type (see Filippazzo et al., 2015) is consistent with objects from the BDNYC catalogue (doi:10.5281/zenodo.45169) and does not show low gravity signatures. Similarly, a CMD of $J$ vs $J-K$ magnitudes shows HD 984 B at a similar location as other late M types, instead of shifted redder which would indicate low gravity. From these indicators, there is no conclusive evidence showing HD 985 B has low surface gravity which suggests that the true age for the system may be at the high end of the age estimate of $115\pm 85$ Myr (Meshkat et al., 2015).

We note a significant difference between the GPI and SINFONI spectra towards the blue end of the $H$ band. The cause for this difference is unknown but may arise from differences in data processing and speckle subtraction artifacts. A complete analysis of these differences is beyond the scope of this paper.

In addition to spectral type matching, the photometry can also be analyzed to determine the object’s luminosity, temperature and mass. These calculations from the new GPI data can be used both to compare to the earlier SINFONI estimates of luminosity, temperature and mass, and to provide better estimates with the additional bands. Since Meshkat et al. (2015) find no significant difference between evolutionary models, we use DUSTY isochrone models (Chabrier et al., 2000). As these models are highly dependent on age since low-mass objects slowly cool with time, it is necessary to have an accurate age estimate. Using the detailed analysis from Meshkat et al. (2015), who made a comprehensive age estimate, we adopt the same age range, 30–200 Myr, for HD 984 B. Taking the DUSTY isochrone models, which provide tables of the luminosity and magnitudes for a range of planet masses at particular ages, we first interpolate between model ages to generate new finer grid tables for 10 to 500Myr. This is then used to derive the predicted luminosities at 30 and 200 Myr that correspond to the measured values of J and H. Luminosity and mass uncertainties are propagated from uncertainties in the absolute magnitudes. Although the age of the system is inconsequential when computing luminosity (see Figure 11), it is highly influential when estimating mass (see Figure 12). The luminosity, accounting for the age range and both bands is $\log(L_{\mathrm{bol}}/L_{\sun})=-2.88\pm 0.07$ dex, in agreement with Meshkat et al. (2015). We use the same technique to find the mass, and we find the $H$ band absolute magnitude corresponds to a range of masses from $39\pm 2$ MJup at 30 Myr to $94\pm 4$ MJup at 200 Myr. The $J$ band yields masses of $34\pm 1$ MJup and $84\pm 4$ MJup for the same ages. Temperature analysis, conducted in the same manner as the luminosity models and using the same DUSTY models, found object temperatures of $2458\pm 32$ K to $2800\pm 37$ K for $J$ band over the same age range. The $H$ band magnitude give temperatures of $2545\pm 28$ K to $2896\pm 31$ K. The low photometric uncertainties correspond to the low quoted temperature errors, but much larger systematic errors likely are present due to model uncertainties. Using the spectral type estimated previously (M$6.5\pm 1.5$) and the spectral type-to-temperature conversion from Stephens et al. (2009), we derive an effective temperature of $T_{\rm eff}=2730^{+120}_{-180}$ K, consistent with the results of Meshkat et al. (2015), and intermediate to the temperature estimates from the J and H band absolute magnitudes.

Due to extensive spectral covariance, more rigorous characterization estimation through spectroscopy is unfeasible. Instead, we further empirically compare HD 984 B to other objects fitted by colour, which are independent of models. For this, we use field objects from Filippazzo et al. (2015). Using HD 984 B’s best matched specturm type, M6.5, we estimate $L_{\mathrm{bol}}=-3.1^{+0.1}_{-0.2}$.

8. Conclusion

Our new observations of HD 984 B with the Gemini Planet Imager have built upon the results presented by Meshkat et al. (2015) and provided a more comprehensive understanding of this young substellar companion. Here we summarize the characterization we have achieved in our analysis. All numerical results can be found in Table 2.

With a three year baseline between the first epoch of Meshkat et al. (2015) astrometry and the new GPI astrometry presented in this work, we derive the first constraints on all of the orbital parameters for this system using a rejection sampling technique. Continued astrometric monitoring of HD 984 B with GPI will help to further constrain these orbital parameters, and reduce the effects of any systematic biases between measurements obtained from different instruments.

The GPI observations were also used to investigate the photometric and spectroscopic properties of HD 984 B, which were compared to predictions of evolutionary models and to other substellar objects spanning a range of spectral types and ages. From the integrated J and H band spectra, and knowing the distance to the object, we measure an absolute magnitude which is used with DUSTY evolutionary models to derive a luminosity, mass and temperature.

Complementing our GPI JH spectrum with the SINFONI K band spectrum presented in Meshkat et al. (2015), we estimated the spectral type of HD 984 B from spectral templates constructed from a large number of near-infrared spectra of low-mass stars and brown dwarfs. To account for spectral covariances, the spectra were split into high and low spatial frequencies and binned according to the correlation length measured within each spectrum. The best spectrum match was further used to calculate an effective temperature using empirical spectral type-to-temperature relations. While the results from the fit to the unfiltered and low frequency spectra were consistent in terms of the derived spectral type, the high frequency component did not help constrain the spectral type due to the lack of sharp spectral features. This method however, may prove useful to match spectral features with $K$ band data, where narrow spectral features, such as CO, can be identified and fitted. Splitting the spectra in these cases will allow for better noise statistics and improved $\chi^{2}$ analysis.

The surface gravity of HD 984 B was also investigated, as a low surface gravity could have been used to provide a further constraint on the age and mass of the system. The five best fitting objects to the JHK spectrum of HD 984 B had a mix of gravity classifications. With this ambiguous result, several photometric and spectral indicators were also computed to look for additional evidence of low surface gravity. None of these indicators suggested a low surface gravity for HD 984 B, and we therefore do not assign a gravity classification to the spectral type.

HD 984 B is one of the latest of a growing number of brown dwarf and low-mass companions discovered via direct imaging, often serendipitously during searches for exoplanets. As demonstrated for this object, substellar companions discovered in these campaigns can be rapidly characterized with an integral field spectrograph, and preliminary constraints on orbital parameters can be derived with a relatively short baseline between epochs. With the many ongoing surveys using extreme adaptive optics instruments such as GPI, SPHERE, and SCExAO, there is ample opportunity to discover and characterize new substellar objects in the near future. Although one of object alone cannot prove a rule, continued identification and characterization of this class of objects will undoubtedly further our understanding of their formation.

Acknowledgments

We would like to thank Dr. Tiffany Meshkat for providing the SINFONI $K$ band spectra. This research has benefited from the SpeX Prism Library and SpeX Prism Library Analysis Toolkit, maintained by Adam Burgasser at http://www.browndwarfs.org/spexprism, the Montreal Brown Dwarf and Exoplanet Spectral Library, maintained by Jonathan Gagné, and the SIMBAD database, operated at CDS, Strasbourg, France. This paper is based on observations obtained at the Gemini Observatory, which is operated by the Association of Universities for Research in Astronomy, Inc., under a cooperative agreement with the NSF on behalf of the Gemini partnership: the National Science Foundation (United States), the National Research Council (Canada), CONICYT (Chile), Ministerio de Ciencia, Tecnología e Innovación Productiva (Argentina), and Ministério da Ciência, Tecnologia e Inovação (Brazil). This publication additionally makes use of data products from the Two Micron All Sky Survey, which is a joint project of the University of Massachusetts and the Infrared Processing and Analysis Center/California Institute of Technology, funded by the National Aeronautics and Space Administration and the National Science Foundation. Portions of this work were performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344. Supported by NASA grants NNX11AD21G and NNX15AD95/NEXSS (R.J.D.R., J.R.G, J.J.W, T.M.E, P.G.K.) and NX14AJ80G (E.L.N., S.B, F.M.), and NSF grants AST-0909188 and AST-1313718 (R.J.D.R., J.R.G., J.J.W., T.M.E., P.G.K.)