DIRECT IMAGING DISCOVERY OF A "SUPER-JUPITER" AROUND THE LATE B-TYPE STAR κ And^*

We first detected κ And b using AO188 coupled with HiCIAO on the Subaru Telescope on 2012 January 1, as part of the SEEDS survey (Tamura 2009). The observations used a 20'' × 20'' field of view, 9.5 mas pixels, and an opaque 06 diameter coronagraphic mask, which helped to keep the saturation radius <05. The images were taken in the near-infrared (H band, 1.6 μm), where young substellar objects are expected to be bright with thermal radiation (Baraffe et al. 2003). Pupil tracking was used to enable angular differential imaging (ADI; Marois et al. 2006).

Data reduction of the 46 exposures of 5 s revealed, at 23σ confidence, a point-like source at 107 separation. Follow-up observations in J (1.3 μm; 177 exposures of 10 s), H (1.6 μm; 171 exposures of 8 s), and K_s (2.2 μm; 135 exposures of 10 s), collected on 2012 July 8–9, using the same observing setup, re-detected the source at 6, 28, and 49σ confidence levels, respectively. Unsaturated images of the primary, taken immediately before and after each filter's observing sequence, and using a neutral density filter (0.866% for H, 1.113% for K_s, and 0.590% for J) provided photometric calibration.

To optimize the ADI technique, we first reduced the data using a locally optimized combination of images algorithm (LOCI; Lafrenière et al. 2007). HiCIAO observations of M5, combined with distortion-corrected images obtained with the Advanced Camera for Surveys (ACS) on the Hubble Space Telescope, enabled accurate pixel-scale calibration to within 0.2%; the ACS astrometric calibration was based on van der Marel et al. (2007). Figure 1 (left and middle panels) presents a JHK false-color image and corresponding signal-to-noise ratio (S/N) map after the ADI/LOCI data reduction.

Given the relatively high S/N ratios and the known difficulties in quantifying the impact of LOCI on planet photometry and astrometry, we also performed a classical ADI reduction (Marois et al. 2006) with mean-based point-spread function (PSF) estimation and frame co-adding. Unsharp masking on the spatial scale of 35 pixels (≈ 7 FWHM) was applied to the final image to flatten the residual background. The planet signal was recovered with S/N ratios comparable (within 10%) to the LOCI reduction for all the July data sets. For the somewhat lower-quality January data, the measured S/N reduced from about 23σ to 7σ.

To achieve unbiased photometry and astrometry, we extracted the combined κ And PSF (S/N > 1000) from the neutral density images, and placed it on an empty image frame at the location of κ And b. Applying the same unsharp masking and ADI reduction to this data as we did for the science data, we simulated the parallactic angle evolution, as recorded in the science frames. The resulting processed PSF acted as the photometric and astrometric reference for κ And b. The only nonlinear step in this process was the median-based unsharp masking, but the large spatial scale (≈ 7 FWHM) ensured that the subtraction effects were minimal.

We calibrated the astrometry by cross-correlating the κ And b signal with the processed calibration PSF. We estimated the uncertainty in the κ And b center to be FWHM/(S/N), following Cameron et al. (2008). The uncertainties in the final relative astrometry were dominated by our ability to determine the host star center, which was carried out through Moffat fitting of each individual exposure. We conservatively estimated the uncertainty of the Moffat fit at 0.75 pixels (7 mas). For confirmation, we applied Moffat fitting and peak fitting to unsaturated data of κ And and found that the methods agreed at the 0.5σ level. The photometric uncertainties were calculated as a combination of (1) representative noise in an annulus, centered on the host star, with a radius equal to the companion, (2) photometric variability in the neutral density calibration images, which yielded effective accuracies of 7%–11% for the combined data sets, and (3) uncertainties in the JHK magnitudes of κ And.

On 2012 July 28, we followed the JHK observations with L'-band observations (3.8 μm; 50 exposures of 30 s) using AO188 coupled with the IRCS on the Subaru Telescope. We employed a 105 × 105 field of view, 20.6 mas pixel scale, and no coronagraph. The host star saturated out to ∼01. The dithered observations, carried out in ADI mode, were divided into two identical sequences bracketing observations of the star HR 8799, which provided the photometric calibration (Marois et al. 2008). Observations of a third star, S810-A, were collected before the science observations as a secondary calibration check (Leggett et al. 2003).

We sky-subtracted each image using a median combination of frames taken at the other dither positions. To help maximize the high-contrast sensitivity, we processed the data using an "adaptive" LOCI process (A-LOCI; Currie et al. 2012). We also employed a moving pixel mask, where the LOCI algorithm is prevented from using, in PSF construction, pixels lying within the subtraction zone (see Lafrenière et al. 2007 for details). Figure 1 (right) shows the final image.

To quantify the κ And b throughput, we used fake point sources added to the image and processed with the same algorithm settings. As an additional check on our flux calibration, we determined the relative brightness between the HR 8799 bcd planets (all detected at S/N > 7–10) using identical procedures, and confirmed its agreement with published values (Currie et al. 2011). The independent calibrations all yielded self-consistent results, ensuring confidence in the 22σ detection of κ And b in L'. As a final check, we re-processed the L'-band data using a more classical ADI method, similar to that described for the JHK data set, and achieved consistent results. While the July L' astrometry was consistent with the July JHK results, we refrained from including it in our proper-motion analysis, due to our possession of poorer-quality astrometric calibration.

Located 52.0 pc from the Sun, κ And exhibits a proper motion of 83.5 mas yr⁻¹ (Perryman et al. 1997), enabling an effective test to distinguish bound companions from unrelated background stars. κ And's proper and parallactic motion translates to 76 mas (∼8 HiCIAO pixels) of net movement over the six-month period between epochs. As shown in Figure 2, the companion exhibits common proper motion with the host star, and deviates from the expected background star motion by 7σ. In addition to this 7σ deviation in the magnitude of the motion, the observed direction of the motion and scatter in astrometry are completely inconsistent with that of a background star.

Figure 3 shows that the κ And b colors are most consistent with cloudy L dwarfs and that they overlap with several other benchmark exoplanets and low-mass companions, including HR 8799 bcd, AB Pic b, and 1RXS1609 b. Figure 4 compares κ And b colors and absolute magnitudes with DUSTY and COND evolutionary tracks (Baraffe et al. 2003; Chabrier et al. 2000), as well as low-mass companions around HR 8799 and AB Pic. The plots show κ And b to be well situated between HR 8799 cde and AB Pic b. Its infrared colors are slightly bluer than those of typical field L dwarfs, possibly indicating a low surface gravity (Cruz et al. 2009). However, improved photometry is required to confirm whether or not this color deviation is real.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The estimated temperature of κ And b suggests that its atmospheric properties should align more closely with those of the DUSTY models (see discussions in Chabrier et al. 2000). In deriving a mass estimate from this track, we rely on the July H-band magnitude because (1) alternative J-band and January H-band measurements have higher uncertainties, (2) K_s-band mass estimates are more sensitive to atmospheric composition (see, e.g., Janson et al. 2011b), and (3) L'-band mass estimates have been less thoroughly tested with experimental data, and are more sensitive to age uncertainties for this age and magnitude range (see Chabrier et al. 2000).

Based on the July H-band magnitude of 15.2 ± 0.2, the estimated age of 30⁺²⁰₋₁₀ Myr, a parallax of 19.2 ± 0.7 (Perryman et al. 1997), and the DUSTY evolutionary models, we calculate a mass of 12.8^+2.0_−1.0 M_Jup and a temperature of 1680⁺³⁰₋₂₀ K. As a consistency check, we calculate the predicted JK_sL' magnitudes based on the estimated 1680 K temperature, the 20–50 Myr system age, and the DUSTY evolutionary models. This yields J = 16.5–16.8, K_s = 14.2–14.4, and L' = 13.1–13.2 mag, all of which are in agreement with our measured multiband photometry. Additionally, the two epochs of H-band photometry are in agreement with one another. Table 1 summarizes the complete properties of κ And b.

While the DUSTY models are likely the more relevant, we estimate a possible alternative mass using the COND evolutionary tracks. In this scenario, we determine a mass of 11.5^+2.4_−1.2 M_Jup and a temperature of 1640⁺⁴⁰₋₂₀ K. More recent evolutionary models by Spiegel & Burrows (2012) offer alternative "Warm Start" scenarios that consider formation with lower levels of initial entropy. While these models do not consider combinations of mass and temperature similar to that of κ And b, they do predict generally higher masses than that of the DUSTY and COND models. In the case of κ And b, such models place the most probable mass at a value above the typical deuterium burning limit. While we currently adopt a nominal mass estimate of 12.8^+2.0_−1.0 M_Jup for the analyses in this discovery Letter (based on the DUSTY models), we defer a deeper investigation of the companion mass for a follow-up paper, where we will focus on a more thorough comparison of multiband photometry with synthetic spectra.

We estimate the semimajor axis of κ And b from its observed separation. Assuming a uniform eccentricity distribution of 0 < e < 1, and random viewing angles, Dupuy et al. (2010) compute a median correction factor between projected separation and semimajor axis of 1.1^+0.91_−0.36. Using this relation, we derive a semimajor axis of 61⁺⁵⁰₋₂₀ AU based on its projected separation of 55.2 AU (107) in 2012 January.

The H-band sensitivity levels (see Section 2.1) allow us to rule out secondary companions with temperatures similar to or warmer than that of κ And b, for separations greater than 09 (46 projected AU). For the κ And b separation (11) and beyond, we may rule out secondary companions with masses ⩾11.7 M_Jup, assuming a 30 Myr system age and the DUSTY evolutionary models.