A High-Contrast Imaging Survey of SIM Lite Planet Search Targets

For the last decade, planet search efforts around nearby stars have been led by radial-velocity surveys that are sensitive to Jupiter-mass planets with close-in (< 5 AU) orbits (Butler et al. 2006). High-contrast imaging surveys have recently culminated with the direct detection of Jupiter-mass planets at wider separations (>20 AU, Marois et al. 2008; Kalas et al. 2008). While the radial-velocity surveys are approaching sensitivities capable of detecting hot super-Earth mass planets around solar-mass stars and even habitable terrestrial-mass planets around some low-mass stars, Earth analogs remain out of reach, due to intrinsic stellar jitter. Space Interferometry Mission Lite (SIM Lite, formerly called SIM PlanetQuest, Unwin et al. 2008) will be the first instrument capable of detecting Earth analogs in the habitable zone of nearby stars. SIM Lite is a planned space-based interferometer with two 50 cm telescopes separated by 6 m. It is capable of achieving a single-measurement positional accuracy of 1 μas, enough to detect terrestrial planets in the habitable zone of nearby solar-type stars.

There are two SIM Key projects chosen to conduct surveys capable of detecting terrestrial-mass planets around nearby main-sequence stars—Discovery of Planetary Systems (Geoff Marcy, PI) and the Extrasolar Planets Interferometric Survey (EPIcS, Mike Shao, PI). These projects will carry out an astrometric search for rocky planets around ∼200 stars located within 30 pc of the Sun.

There are many precursor programs currently being conducted to characterize and vet all potential targets for all the SIM Key projects (Unwin et al. 2008). These programs include radial-velocity studies of the reference stars utilized in the narrow-angle observations, accurate spectral type determinations through optical spectroscopy, and photometry studies of the stars in the young-star planet search program. Some of these programs have resulted in planet detections themselves (Niedzielski et al. 2007) or data that can be applied to upcoming planet surveys (Tanner et al. 2007). Here, we present the results of a companion survey of stars included in both of the terrestrial planet SIM Lite surveys. The observational goal of this companion survey is to identify bright nearby companions to the target stars. A companion with ΔV < 4 and within 1'' can bias the position of the photocenter of the star, thus reducing the astrometric accuracy. While this survey was designed as a precursor program for SIM Lite, the observations presented here also serve as reconnaissance data for upcoming high-contrast imaging surveys such as the Gemini Planet Imager (GPI) (Macintosh et al. 2006), the VLT SPHERE coronagraph (Beuzit et al. 2008), and P1640, the recently commissioned Lyot adaptive optics (AO) coronagraph on the Palomar 200 inch (5 m) telescope (Hinkley et al. 2008). As a result of the high-contrast images collected for this survey, for most of the stars in the sample, we are sensitive to brown dwarf mass objects at separations of >50 AU. Because this program is sensitive to brown dwarf companions at wide separations, the results serve as an additional probe of the brown dwarf desert.

Section 2 describes the criteria by which we created the sample for the survey, § 3 details our observations, § 4 describes the data reduction and analysis, § 5 summarizes the results of the survey and our achieved sensitivities, and in § 6 we compare our results to previous surveys and discuss the importance of publishing nondetections.

The two SIM Lite terrestrial planet search programs coordinated their target selection so as to create unique lists for each team. Since one of the projects is primarily aimed at finding terrestrial planets in the habitable zone ("Extrasolar Planets Interferometric Survey [EPIcS]," Shao, PI) and the other is aimed at detecting all planets around stars within 8 pc ("Discovery of Planetary Systems," Marcy, PI), it was not difficult to agree on the two samples. Target stars were selected based on spectral type, heliocentric distance, brightness, companion separation (if applicable), habitable zone location, and the orbital period at the habitable zone.

Since these lists were created, the EPIcS program completed a study on how to further optimize the list. An additional sample list of 240 stars has been created for the purpose of performing Monte Carlo simulations to predict the potential yield of a SIM Lite terrestrial planet search program (Catanzarite et al. 2006). This SIM-optimized target list is derived from an initial list of 2350 stars taken from the Hipparcos catalog, with distances of less than 30 pc (Turnbull & Tarter 2003). They excluded stars with luminosity greater than 25 times solar, thereby eliminating giants from our sample. To eliminate the possibility of fringe contamination from a binary companion, we applied the following: stars with a known companion closer than 0.4'' and ΔV < 4 were excluded. If the target-star candidate had a wide binary companion that was separated by more than 1.5'', the companion was added to the list of target-star candidates. Further improvements on the target ranking will include considerations of stellar metallicity and the number and orientation of the reference stars with respect to the target. In the end, these stars will represent the best targets for a microarcsecond astrometric planet search.

After two years of observations, we have collected observations for a subset of 84 nearby main-sequence (V) stars taken from both the original EPIcS sample and the one created for Catanzarite et al. (2006). The properties of each of the stars in the sample are given in Table 1. The stars in our sample have distances of 3.5–30 pc, spectral types of M4-A3V, and V magnitudes of 0.77–11.35. Most, if not all, of the stars in the sample are mature stars with ages of 1–10 Gyr. We made sure not to reobserve those stars with published AO coronagraphic or Hubble Space Telescope observations (Carson et al. 2006; Lowrance et al. 2005; Metchev et al. 2005).

Observations were obtained with the Palomar Observatory Hale 5 m Telescope using the Palomar High Angular Resolution Observer (PHARO) near-IR camera (Hayward et al. 2001) behind the Palomar adaptive optics (PALAO) system (Troy et al. 2000). Observing dates and sky conditions are given in Table 2. We used a 25 mas pixel^-1 scale camera (25'' field of view) and the 0.45'' radius occulting spot. Each star was observed in the K_s filter (2.16 μm) with integration times of 1–30 s, depending on the brightness of the star. Stacks of 20–100 short-exposure images were collected for each star, resulting in effective integration times of 100–1200 s.⁴ Stacks of sky images were taken adjacent to each set of target images by offsetting 30'' from the target in the four cardinal directions and turning off the AO system. For flux calibration, observations of the target stars were taken by inserting the neutral-density filter, offsetting the star from the coronagraph and placing it in a five-point dither pattern to allow for adequate sky subtraction.⁵ These offset images allowed us to determine that the FWHM of the AO-corrected observations varied from 0.2 to 1.3'' with Strehl ratios of 0.20–0.74. The seeing quality varies considerably for the different observing runs, from good (0.5'') to bad (>3^'') based on the FWHM of the point-spread function (PSF) with no AO correction (see Table 2). The variable seeing affected our ability to achieve uniform sensitivities for all the target stars in the survey.

To aid in the suppression of speckle noise, we observed a set of PSF stars in conjunction with about half of our targets. The PSF stars were selected to have colors (i.e., spectral types), brightness, and air masses similar to those of the targets so as to produce a speckle pattern as close as possible to the science targets. Finally, sets of known binary stars with high-quality orbital solutions were also observed during the 2005 November and 2004 October runs to provide an accurate determination of the plate scale and image orientation. During the observation, each binary was placed in multiple positions over the field of view of the camera. In order to observe as many targets as possible while also reaching image sensitivities necessary to detect brown dwarf companions, our observing goal was to complete all observations associated with each target within a half-hour.

The individual AO images were sky-subtracted, flat-fielded, and corrected for bad pixels, with the final image created from the median of the stack of reduced individual images. For those images with additional stars visible in the field of view, we used them to determine the offset of each image in the stack with respect to a reference image and shifted them accordingly prior to taking the median of the stack. This helped mitigate smearing from movement of the field during the observations due to telescope flexure or small losses in telescope tracking. A subset of the target images has corresponding observations of PSF stars with similar brightness and color. The relative positions of the target and PSF star were estimated using the Poisson spot present in the middle of the coronagraphic spot. As with previous studies (Metchev et al. 2004; Tanner et al. 2007), we are able to register the star positions to 0.7 pixels using the centroid position of the spot (Tanner et al. 2007). The scaling for the PSF is the multiplicative factor that minimizes the residuals remaining after the subtraction. Comparisons of the standard deviations of the speckle noise of the stars before and after PSF subtraction suggest that the subtraction reduced the noise within the halo by ∼25%. Figure 1 shows the difference image of GL 876 and its PSF star HD 216789.

Zoom In Zoom Out Reset image size

The coronagraphic images were flux-calibrated using the images taken with the primary star offset from the coronagraph while accounting for a difference in integration time and the well-defined neutral-density filter used for the off-spot images (Metchev et al. 2004). The magnitudes for both the primary stars in the off-spot images and the companions in the coronagraphic images are estimated from aperture photometry with an aperture of 0.8'' and sky annulus of 1.–1.25''. The K_s-band magnitudes of the primaries were taken from the Two Micron All-Sky Survey (2MASS, Cutri et al. 2006). The uncertainties for the photometry were estimated from the errors given for the 2MASS magnitude and an assumption of a 5% calibration error determined by comparing the photometry of the companion to GJ 105a to published values (Golimowski et al. 2000).

The images of the calibration binaries were reduced in the standard manner. For the 2004 data, we assume a plate scale of 25.11 ± 0.04 mas pixel^-1 estimated from the known orbital solutions of three different binary stars (WDS 09006 + 4147, WDS 18055 + 0230, and WDS 20467 + 1607) that were observed very close to our October observations using the same instrument (2004 October 4–5, Metchev 2005, see their Table 4.1). For the 2005 data we estimate a plate scale of 25.21 ± 0.36 mas pixel^-1 using the average and standard deviations of the measured pixel separation of one binary (WDS 09006 + 4147, Hartkopf et al. 1996), compared to its predicted orbital separation in arcseconds. This binary, which was observed in 2005 November, was placed in multiple positions across the field of view after correction for the known distortion in the camera.

A thorough visual inspection of both the median-averaged coronagraphic images and the difference images between the targets and the PSF identified all stars in the 25'' field of view. Those target stars with no additional stars visible in the field of view are listed in Table 3, and the companion candidates identified in these images are listed in Table 4, along with their distance from the target star, position angle, and magnitude difference compared to the primary, when available. To accurately determine the position of the star behind the coronagraphic spot, we utilized the static waffle pattern, which has a distinct set of four speckles framing the PSF (see Fig.  2). As determined in Tanner et al. (2007), by using the centroid position of each of the four spots in the waffle pattern, we can use the intersection of the lines crisscrossing the pattern to determine the star's position to 2.3 pixels. This positional accuracy and the error in the plate scale are then propagated when determining the uncertainties for the offset and position angle.

Zoom In Zoom Out Reset image size

4.1. Common Proper Motion Determination

For 13 of the stars with companion candidates, we collected second epoch observations at least a year later during AO observing runs dedicated for other projects. The additional image allows us to determine whether the companion is bound to the star through their common proper motion. This positional accuracies given in Table 4 are sufficient to conclude whether the companion candidates are bound to their stars given the high proper motions (>100 mas yr^-1) of all the stars in the sample and the minimum of a full year between observations. In most cases, simply blinking between the images shows the movement of the background star once the target star is held fixed. We determined the separation and position angles of the candidates in both images and compared them to what would be expected given the star's proper motion. After this analysis, we find two stars with confirmed common proper motion companions—GJ 454 and GJ 1020. The companions to these stars are new detections having not been mentioned in previous publications.

Figure 3 plots the offset in right ascension and declination between GL 454 and its companion 1'' away, and Table 5 lists the astrometry. The solid line denotes the expected motion of the companion if it were a stationary background object. The fact that the offset of the companion star estimated for two different epochs does not change significantly given the associated errors (10–30 mas) confirms this object as a bona fide companion. Additional observations of GJ 454 at 1.25 μm were made with the same instrument during the 2004 June run and were reduced in the same way as the K_s data. Given the K_s and J magnitudes of the companion (K_s = 7.75 and J = 9.07) and the distance to the star (12.91 pc), it is most likely an M3 dwarf (Baraffe & Chabrier 1996). Despite its proximity to GJ 454 and its relative brightness, ΔK_s > 6, it should not pose a problem for SIM Lite. Figure 3 also shows the astrometry for the companion to GJ 1020, which was originally observed as a PSF star to GJ 10. This confirmed proper motion companion has a K_s magnitude of 10.4. With an absolute K magnitude of >10, this companion is most likely a low-mass star.

Zoom In Zoom Out Reset image size

For some of the targets with no second epoch observations in our survey, we utilized images from the HST,⁶ ESO,⁷ and Gemini⁸ archives. We were able to use these archive data sets to confirm that companion candidates around GJ 726, GJ 722, and GJ 892 are background stars. There are 17 companion candidates that remain unconfirmed (see Table 4 and Fig. 4).

Zoom In Zoom Out Reset image size

Out of 84 stars, we found two unpublished common proper motion companions to GJ 454 and GJ 1020, neither of which will cause problems for SIM Lite observations, since it is much fainter than the target star. None of the companions have absolute magnitudes or colors consistent with brown dwarfs. This is not unexpected, given the observed paucity of brown dwarf companions to solar-mass main-sequence stars (McCarthy & Zuckerman 2004; Butler et al. 2006).

5.1. Image Sensitivities

To estimate the sensitivities of all of our target fields as a function of distance from the star, we employ "PSF planting," in which a PSF corresponding to an object of known brightness is inserted into the image. The PSF extracted from the off-spot flux-calibration image of each target is sky-subtracted, normalized, multiplied by an array of contrast values (ΔK_s = 7.7–15.1 mag), and placed at a range (0–10'') of distances from the target at random position angles. We completed 10⁴ iterations of the PSF planting algorithm to fill out the parameter space of contrast and distance from the primary star. We make sure that the same number of planted PSFs occur in each radius bin. To determine whether the planted star is detected, the image is cross-correlated with a flux-normalized PSF. For each distance bin we estimate the minimum PSF intensity for which at least 90% of the fake sources had a correlation value of 0.75 or higher. This correlation value was determined by visually inspecting an image with inserted scaled PSFs. Previous studies have shown that the eye is often the best natural detector when determining the presence of a faint star in an image (Metchev et al. 2005).

The intensities are converted into magnitudes using the flux calibration from the off-spot image and the 2MASS K_s magnitude of the star. Figure 5 plots the largest K_s magnitude difference between the target star and planted PSF as a function of distance from the star for all targets with calibration data.

Zoom In Zoom Out Reset image size

Table 6 lists the values of the faintest detectable K_s magnitudes at 0.5, 1, 2, and 5". We were able to detect sources with a magnitude contrast of ΔK_S ∼ 4–6.5 mag at 0.5'', ΔK_S ∼ 5.5–8 mag at 2'', and ΔK_S ∼ 6.5–9.5 at 5'' with ∼90% completeness. The range of image contrasts is due primarily to variations in seeing conditions throughout the night. Unfortunately, the brightest objects suffer from two additional components of degradation in image sensitivity: 1) a smudge in the optics that creates a long streak across the spot at a position angle of 45° and 2) a thick offset ring due to a reflection off the array, entrance windows, or the coronographic slide that is illuminating the Lyot stop (see Fig. 6). These image artifacts reduce the sensitivity in these areas of the image by 15–20%.

Zoom In Zoom Out Reset image size

5.2. Mass Sensitivities

By assuming a set of values relating the absolute magnitude of any brown dwarf companion to its mass and age based on theoretical evolution models (Burrows et al. 1997; Baraffe et al. 2003), we can estimate an lower limit to the mass sensitivity around those target stars with flux calibrations. Figure 7 plots the lower limits of the mass sensitivity at a separation of 1'' from each star as a function of separation in AU based on its Hipparcos distance. The upper limits assume both 1 Gyr (crosses) and 5 Gyr (asterisks) ages for the primary star. This plot shows that most of our observations are reaching into the brown dwarf regime (∼75 M_J) at separations of 1'' and beyond.

Zoom In Zoom Out Reset image size

6.1. How Does This Survey Compare with Previous Work?

6.2. The Value of Publishing Nondetections

We have completed a high-contrast imaging survey of a sample of stars slated to be potential targets for the SIM Lite astrometric space telescope. While many of our observations were sensitive enough to detect brown dwarf companions at separations of >1^'', our survey found two unpublished confirmed common proper motion stellar companions around GJ 454 and GJ 1020. This survey will serve as a resource for both SIM Lite and direct imaging surveys such as GPI and SPHERE, as all the images will be given to the NStED database. These images can be used to vet future targets and aid in common proper motion determinations. Additional SIM Lite targets will be observed with high-contrast imaging if they have not all ready been observed with other programs.

We would like to thank our anonymous referee for their valuable insights into how to improve our manuscript. We would also like to thank Shri Kulkarni and Mike Shao for invaluable discussions regarding the selection of the EPIcS targets. Based on observations obtained at the Hale Telescope, Palomar Observatory, as part of a continuing collaboration between the California Institute of Technology, NASA/JPL, and Cornell University. The research described in this publication was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration. This publication 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.