KNOW THE STAR, KNOW THE PLANET. IV. A STELLAR COMPANION TO THE HOST STAR OF THE ECCENTRIC EXOPLANET HD 8673b

We used the Keck II telescope, its AO system (Wizinowich et al. 2000), and the NIRC2 instrument to observe the system at three epochs in 2011 and 2012. Each time, we collected data in multiple filters with multiple coadds. Since the existence of the companion was still questionable, the multiple filters allowed us to determine if the object was an astronomical object or a quasi-static speckle whose position is wavelength dependent. The 2012 Brγ image of HD 8673 and its companion is shown in Figure 1.

After reducing each image, we measured the astrometry and photometry using the fitstars algorithm on each image (ten Brummelaar et al. 1996, 2000). Photometric error bars were set equal to the standard deviation of the measurements from all the images. We computed the weighted mean of the astrometry from the images taken in different filters on the same night, with the weight being the number of coadds for each image. The error bar was set to the standard deviation of the results. That covers much of the random errors, but does not cover the systematic errors. We expect those to be smaller than those of the Advanced Electro-Optical System (AEOS; Section 2.3) or Hale data (Section 2.2) because of the smaller point-spread function (PSF) arising from the larger aperture size, but we do not have a way of quantifying those errors. The astrometry is given in Table 1; it lists the Besselian date of the observations, the position angle and separation of the companion relative to the primary and the telescope used to collect the measurements. The photometric measurements are listed in Table 2. The table lists the measured magnitude difference in each filter, the central wavelength of those filters, and the telescope used to take the observations.

We observed HD 8673 on 2011 August 22 UT and 2013 September 28 UT with the Palomar Observatory Hale 5 m telescope using the PALM 3000 AO system (Dekany et al. 2013) and the PHARO near-IR camera (Hayward et al. 2001). HD 8673 was centered in the detector's 25'' field of view. In 2011, we collected 10 frames of the star in ${\rm Br}\gamma$ filter and after reducing the data, the frames were coadded. In 2013, we collected 50 frames in the K_s filter. After the individual frames were calibrated, we created five images by coadding 10 frames into each image, allowing us to analyze the precision of the measurements. The fitstars algorithm was used to measure the astrometry and photometry of the objects. Photometric error bars were assigned using the technique described in Roberts et al. (2005). The resulting astrometry and photometry are listed in Tables 1 and 2. Figure 2 shows an image from the 2013 data collection.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Following the Keck II and Hale resolutions, archival observations obtained with the AEOS 3.67 m telescope (Vigil et al. 1996) were re-examined for a comparable companion detection. On 2004 October 14 UT the star was observed in an effort to detect the bright companion detected by Mason et al. (2011). The observations used the telescope's AO system and the Visible Imager camera (Roberts & Neyman 2002). We collected 1000 frames of the star with the Bessel I filter. After debiasing, dark subtraction, and flat fielding, the resulting frames were coadded with each frame weighted by its peak pixel to emphasize the best quality data. The result is shown in Figure 3. The companion is harder to see than in the Keck II or Hale images, due to a lower image quality because of the shorter wavelength and because the system has a larger dynamic range in the I filter.

Zoom In Zoom Out Reset image size

Again, we measured the astrometry and photometry with the iterative blind deconvolution algorithm, fitstars. Photometric error bars were assigned using the technique described in Roberts et al. (2005). Tables 1 and 2 list the resulting astrometry and photometry measurements.

In 2012 and 2013, we observed HD 8673 four times with the Robo-AO system (Baranec et al. 2013, 2014) on the Palomar 1.5 m telescope. On 2012 July 17 UT and 2012 July 18 UT, we used the SDSS r', i', z' filters, on 2012 October 6 UT we used the SDSS i' and g' filters, and on 2013 January 20 UT we used the SDSS r', i', z' filters. In 2012, the observations were collected as part of the automated Robo-AO target queue, where the telescope is pointed at the object robotically with a 44'' wide square field of view. For the 2013 January observations, the star was acquired manually which allowed the star to be positioned at the exact center of camera's field of view, and a 55 sub window of the image to be read out at 50.5 Hz, which is 5 times faster than the standard rate for the full image. This improves the image quality by reducing the uncompensated intra-exposure stellar image motion. Capturing data in this manner is done infrequently, as it requires human intervention and drastically slows the acquisition rate.

Each data set was reduced using the standard Robo-AO lucky imaging pipeline to perform the image alignment (Terziev et al. 2013; Law et al. 2014). The pipeline produces four images, with each image selecting either the best 1, 10, 50 or 100% of the images. Each image was examined visually for companions and no companions were detected. We computed the dynamic range of the image via the method of Turner et al. (2008) and determined that the images have insufficient dynamic range to detect the companion. We also examined the wider field of view image taken in 2012 and were able to rule out any additional companions to a limit of 10 magnitudes fainter than the primary.

Between 2011 and 2012 the measurements from Keck II show the companion moved at a rate of 16.47 mas yr⁻¹ compared to the measured proper motion of HD 8673 of 251.8 mas yr⁻¹ (van Leeuwen 2007). The low relative motion of the companion indicates that it shares common proper motion with the primary star and is not a background object. The Keck II data were used to measure the relative proper motion since it is a collection of data taken with the same instrument. The AEOS and Hale data are also consistent with it being a bound object.

There is very little difference in the short term proper motion (α = 237.19 ± 0.33 mas yr⁻¹, δ = −84.64 ± 0.25 mas yr⁻¹; van Leeuwen 2007) and the long term proper motion $(\alpha =237.9\pm 1.0\;{\rm mas}\;{\rm y}{{{\rm r}}^{-1}}$, $\delta =-87.2\pm 1.1\;{\rm mas}\;{\rm y}{{{\rm r}}^{-1}}$, Høg et al. 2000). Significant differences in these values can indicate the presence of companions (Makarov & Kaplan 2005), but their absence is not surprising, especially in the case of a pair with a large differential magnitude and hence large mass ratio.

The two observations of the companion reported by Mason et al. (2011) were both optained with visible speckle interferometry using an intensified CCD camera. The observations were in the Strömgren y filter (550 ± 24 nm) and the Δm from the 2007 measure is estimated as 2.3 ± 0.5. The quality of the 2001 measure was insufficient for an estimation of the differential magnitude. Using the same speckle interferometry camera as Mason et al. (2011), Hartkopf & Mason (2009) did not detect the companion in observations using the Mt. Wilson 2.5 m telescope on 2007.8179, only a short time after the 2007.6049 detection in Mason et al. (2011). The observations had a limiting resolution of 0054. This would have been sufficient to detect the 0109 separation detected by Mason et al. (2011) measured just a few months earlier. In addition, no companions were detected by Ginski et al. (2012) using the 2.2 m Calar Alto telescope and the AstraLux lucky imaging camera in the SDSS i' filter on 2011 January 14.

Our observations using AO have a higher dynamic range than speckle interferometry and would have detected any companions detected with that technique, even if those companions had considerable motion. The astrometry in Table 1 is inconsistent with the companion reported in Mason et al. (2011). Our non-detection of the candidate companion from Mason et al. (2011) and the non-detection in Hartkopf & Mason (2009) and Ginski et al. (2012) lead us to conclude that the Mason et al. (2011) detection was spurious. This is not the only known case of false doubles in speckle interferometry; see McAlister et al. (1993) and Tokovinin (2012) for examples and further discussion.

The companion reported in this paper is a redder object than the purported object in Mason et al. (2011). Visible speckle interferometry using intensified CCD cameras can only detect companions with differential magnitudes less than 3.5 (Mason 1996). This is too low to detect the companion reported in this paper and this explains why it was not seen by Hartkopf & Mason (2009), or Mason et al. (2011). The observations of Ginski et al. (2012) also did not have sufficient dynamic range to detect the companion.

Based on the distance (van Leeuwen 2007) to HD 8673 AB and the I-band apparent magnitude (Monet et al. 2003), we computed the absolute magnitude of HD 8673 B in I band of 9.5 ± 0.94. Using the 2MASS apparent magnitudes in J, H, and K (Skrutskie et al. 2006), we also computed the absolute magnitudes of companion in the J, H and K filters. Our data were taken in Jc, Hc, and Ks and we do not have a transformation between these filters into J, H, and K. Instead we use the values from narrow band filters and realize that there is an unestimated error in the magnitudes. We compared the resulting absolute magnitudes to the values listed on the Dartmouth Stellar Evolution Database¹³ (Dotter et al. 2008) for a 3 Gyr old star (Saffe et al. 2005) with an Fe/H of −0.01 (Nordström et al. 2004). The J, H, Ks absolute magnitudes all fall in the same bin. This bin has a mass of 0.33–0.45 M$_{\odot }$, T_eff of 3520–3690 K, and ${\rm log} G$ of 4.94–4.85. The I absolute magnitude has a much larger error bar and corresponds to a mass range of 0.17–0.44 M$_{\odot }$. There is some overlap between the infrared and the I data. This variance in the I-band result is not surprising. The I data is the most error prone of all the measurements. As seen in Figure 3 the secondary PSF is in the halo of the primary; it also has the largest dynamic range.