SEARCHING FOR PLANETS IN HOLEY DEBRIS DISKS WITH THE APODIZING PHASE PLATE^*

The APP data were obtained in 2012 and 2013 (090.C-0148(A) and 091.C-0457(A), PI: Tiffany Meshkat) at the VLT/UT4 with NaCo. The APP was used for additional diffraction suppression from 02 to 10, increasing the chance of detecting faint companions close to the target stars. The infrared wavefront sensor was used with the target stars themselves as the natural guide star. Data were acquired with the L27 camera (27 mas/pix) and the L'-band filter (λ = 3.80 μm and Δλ = 0.62 μm). We used pupil tracking mode for Angular Differential Imaging (ADI; Marois et al. 2006) and intentionally saturated the stellar PSF core to increase the signal-to-noise ratio (S/N) from possible faint companions. We also obtained unsaturated data to calibrate the photometry relative to the central star and determine the sensitivity achieved in each data set.

The APP suppresses diffraction over a 180° hemisphere on one side of the target star. Thus, two data sets need to be acquired, with different initial position angles (P.A.s) for full coverage around the target star. We observed six targets with the APP (see Table 1), however only four of these have complete APP hemisphere coverage around all P.A.s. One of the targets (HD 134888) has 270° coverage with only one APP hemisphere (from 135° to −135°), the final target (HD 110058) does not have sufficient coverage for a detailed analysis.

Data were obtained in cube mode. Each data cube contains 200 frames, with an integration time between 0.1 s to 0.3 s per frame, depending on the L' mag of the star. Details about total integration time and field rotation per target are contained in Table 1. A three-point dither pattern (with an amplitude of 4'') was used on the detector to subtract the sky background systematics from each data set, detailed in Kenworthy et al. (2013).

Data cubes closest in time at different dither positions were subtracted from each other and centroided. The average is taken over the subtracted cube in order to decrease the full data set size. The final averaged frames cover a rectangular area of 31 by 15 centered on the star. The reason for this asymmetry is that the APP can only observe on the upper quartile portion of the CCD, due to the wedge, deliberately introduced in the optics to avoid ghost images (Kenworthy et al. 2010). Thus, we have complete coverage around the star out to 15 and incomplete coverage beyond that radius.

Optimized principal component analysis (PCA) was run on each target APP hemisphere independently, following Meshkat et al. (2014). PCA processed frames were derotated, averaged over, and combined with the other hemisphere to generate the final image with North facing up. If two regions of APP data overlap, those regions were combined by average. We generated final PCA processed frames for a range of principal components (PCs) from 3 up to 20. This first test was to determine if there were any companion candidates in our data, and if so, how robust they were to the number of PCs used in the image reduction.

We next fixed the number of PCs (approximately 10% of cubes in the data set). We injected artificial planets in the data cubes and ran the extraction pipeline to generate contrast curves of point sources sensitivity for 5σ. The artificial planets were generated from the data with unsaturated PSF cores. The unsaturated data was scaled to the same exposure time as the saturated data. Since the APP is in the pupil plane, it affects the PSF of every source in the field of view (FOV) in the same way, thus we can use the unsaturated star itself to generate artificial planets. The artificial planets were added to the data before PCA. The planets were added at different angular separations from 018 to 30 in steps of 015 and at different contrasts from 5 to 12 mag in steps of 1 mag. This was repeated for two different P.A.s, in order to place a fake planet in each APP hemisphere. We smoothed the final PCA image by a 1 λ/D aperture (Bailey et al. 2013). We measured the S/N of the injected planet and decreased the flux of the injected planet until it reached a S/N of 5. The S/N was determined by dividing the flux in one pixel at the center of the injected planet by the noise in a 2 pixel wide ring around the star at the planet separation (not including the planet). We then interpolated between the contrasts to determine the 5σ contrast limit.

Our targets were selected based on the presence of a bright, unresolved infrared excess indicative of a dusty debris disk. Most of them have well covered SEDs using Spitzer broad-band and spectroscopic data and Herschel broad-band photometry. We collected all the published broad-band photometry from the literature, and performed our own photometry measurements if the data were not published. Spitzer MIPS 24 and 70 μm photometry is part of the Spitzer Debris Disk Master Catalog (K. Y. L. Su et al., in preparation), where data reduction and photometry extraction were briefly summarized in Sierchio et al. (2014). Herschel PACS data reduction and photometry extraction were performed following the procedure published by Balog et al. (2014) for calibration stars, except that, as detailed in the following section, smaller apertures were used for photometry measurements to minimize background contamination. All our debris disk targets have existing Spitzer IRS low-resolution spectra. We downloaded the extracted spectra using the CASSIS database (Lebouteiller et al. 2011). Three of the targets have MIPS-SED low-resolution spectra covering 55 to 95 μm with a slit width of ∼20''. We reduced and calibrated the MIPS-SED data as described by Lu et al. (2008).

Herschel photometry of HD 95086 has been published by Moór et al. (2013) and Su et al. (2014). Here we briefly summarize the Hershel photometry results for HD 134888, HD 28355 and HD17848. The PACS 70 and 160 μm observation for HD 134888 was obtained under Program OT1_dpadget_1. The source appears to be point-like surrounded by background cirrus structure apparent on the 160 μm data, therefore we used an aperture size of 6'' at 70 μm and 11'' at 160 μm to measure the photometry. Including the absolute calibration errors (7%, Balog et al. 2014), the PACS photometry for HD 134888 is: 117 ± 8.4 mJy and 87 ± 11 mJy at 70 and 160 μm, respectively, and the PACS 70 μm photometry agrees with the MIPS 70 μm photometry (119 ± 12 mJy) very well.

For HD 28355, PACS 100 and 160 μm data were obtained under Program OT2_fmorales_3. The source appears to be point-like, located near a bright source ∼28'' away. To avoid contamination from the nearby bright source which is extended at 160 μm, we used an aperture size of 6'' to measure the photometry at both bands. Our adopted photometry for the HD 28355 system is 127 ± 12 mJy and 195 mJy ± 20 mJy at 100 and 160 μm, respectively. The contamination from the nearby bright source explains the up-turn in the MIPS-SED data for wavelengths longward of ∼70 μm (see Figure 1).

Zoom In Zoom Out Reset image size

The PACS and SPIRE data for HD 17848 were obtained under program OT1_pabraham_2 using all the available photometry bands. The source appears to be elongated, suggesting a close to edge-on orientation to the disk. We used an aperture size of 22'' to measure photometry at all three bands, at which radius the encircled flux reached the maximum and flattened afterward. The final PACS fluxes are: 201 ± 14 mJy, 213 ± 15 mJy, and 148 ± 11 mJy at 70, 100 and 160 μm, respectively. The PACS 70/100 μm photometry agrees well with the MIPS 70 μm and MIPS-SED data (see Figure 1). For the SPIRE data, we used the level 2 product for point sources provided by the Herschel Science Center (HIPE ver. 11). The source was detected at ≳3σ at both 250 and 350 μm, but not detected at 500 μm. The submillimeter fluxes are: 57 ± 14 mJy, 30 ± 12 mJy and <14 mJy (1σ) at 250, 350 and 500 μm, respectively. These values are consistent with the Herschel photometry recently published by Pawellek et al. (2014) within uncertainties.

The final disk SEDs were constructed by subtracting the best-fit Kurucz ATLAS9 stellar atmosphere models (Castelli & Kurucz 2004) that fit the optical and near-infrared data. The uncertainties in the excess fluxes also included 2% photospheric extrapolation errors. Figure 1 shows the disk SEDs for our debris targets. We have excluded the disk SED for HD 95086, which will be published in Su et al. (2014), and for HD 110058, on which we did not obtain sufficient APP sky coverage for analysis.

To estimate the likely debris location, we performed basic SED modeling. We started with the simplest blackbody fitting for the disk SEDs (with a typical error of  ± 5 K) and used these temperatures to guide a more complicated SED model with adopted grain properties. Without spatial information, SED modeling is degenerate; therefore, our strategy is to minimize model parameters with some reasonable assumptions.

Similar to the SED model for HD 95086 (Su et al. 2014), we adopted icy silicates as our grain properties. The particle size distribution was assumed to be a power-law form, ∼a^−q, where a is the grain radius with a minimum a_min and maximum a_max cutoffs. We adopted q = 3.5 for the power index of the particle distribution and a_max = 1000 μm for all the models. The minimum grain size is set to be close to the radiation blowout size estimated based on the grain density and stellar luminosity and mass. We assumed that the debris is uniformly distributed (constant surface density) from the inner radius (R_in) to the outer radius (R_out), and computed the thermal dust emission under optically thin conditions where the star is the only heating source. We then derived the best-fit inner and outer boundaries of the cold disk component, along with the total cold disk dust mass, quantifying the goodness of fit with reduced χ². As detailed in the following sections, we excluded any long wavelength data which was contaminated by nearby background galaxies, as well as any short wavelength data which might include a contribution from a warm disk component.

In some cases, weak warm excess shortward of ∼20 μm was present. When the warm excess signal was greater than the uncertainty in photospheric subtraction, we also performed a blackbody fit to the short wavelength data to derive the warm component's approximate temperature. Because the location of the tentative warm disks (≲2 to 10 AU) is less than the inner working angle of our high contrast observations, we do not perform detailed SED modeling for the warm excesses. We comment on the derived warm dust temperatures, but their nature will be discussed in a separate paper.

HD 17848 (ν Hor) is an A2V (Houk & Cowley 1975) field star at distance d = 50.5 ± 0.5 pc (adopting parallax ϖ = 19.82 ± 0.18 mas from van Leeuwen 2007). We estimate the effective temperature to be T_eff = 8470 ± 130 K based on multiple photometric T_eff estimates: using UBVK_s photometry with the color/T_eff table of Pecaut & Mamajek (2013), and the tight cluster of reported T_eff estimates in the literature (Allende Prieto & Lambert 1999; Paunzen et al. 2006; Rhee et al. 2007; Patel et al. 2014). Our adopted T_eff is systematically cooler than that reported by Chen et al. (2014) (9000 K), which is an outlier compared to the other estimates. Adopting the V magnitude from Mermilliod (1994) (V = 5.254 ± 0.005), van Leeuwen (2007) parallax, and adopted T_eff-appropriate bolometric correction BC_V = −0.040 ± 0.034 (Pecaut & Mamajek 2013), we estimate ν Hor's luminosity to be log(L/L_☉) = 1.222 ± 0.016 dex. Based on our HR diagram position for ν Hor, we use the evolutionary tracks of Bertelli et al. (2009) to infer an isochronal age of ∼540 Myr and mass 1.90 M_☉. Sampling a reasonable range of protostellar chemical compositions (Y = 0.26–0.27, Z = 0.014–0.017), we estimate isochronal age 1σ uncertainties of ±90 Myr (statistical) and ±60 Myr (systematic), and mass 1σ uncertainties of ±0.02 M_☉ (statistical) and ±0.04 M_☉ (systematic). Hence, we do not find support for the young age of ∼100 Myr proposed by Rhee et al. (2007), and estimate ν Hor to be ∼5× older than previously thought.

Its IRS spectrum was first analyzed by Ballering et al. (2013), and suggested the system is a two-component disk with dust temperatures of 164 K and 50 K. However, Chen et al. (2014) suggested a much warmer temperature, 353$^{+4}_{-8}$ K, while the cold temperature (57 ± 5 K) was consistent with the early result. The discrepancy is most likely due to how the excess flux in the mid-infrared range was determined (especially how the different modules of IRS spectra joined and pinned down to the photosphere). Therefore, the nature and amount of warm excess is still in debate. Using the observed dust temperature as gauges, the warm excess characterized by Ballering et al. (2013) would be asteroid-like with orbital distances of 9–12 AU, while the value from Chen et al. (2014) would be terrestrial-like with orbital distances of 2–3 AU. The new far-infrared photometry from Herschel and the MIPS-SED data confirm that the cold disk has a typical dust temperature of ∼56 K (see Figure 1). We excluded data points shortward of ∼30 μm in our SED model, resulting in an inner cold disk radius of 96$^{+9}_{-37}$ AU and an outer cold radius of 410$^{+24}_{-96}$ AU with a total dust mass of (1.3 ± 0.7) × 10⁻³M_⊕.

No bright point sources were detected around HD 17848 (see Figure 3). The overlapping region between two APP hemispheres is noisy due to the edge of the dark APP hole. Therefore, the bright point (at 07 and P.A. = −74°) on the reduced image is probably a noise spike. We compared any possible faint point sources in our data with archival data obtained with NaCo/VLT from both 2011 (087.C-0142(A)) and 2009 (084.C-0396(A)) and verified that none of these faint point sources are present in all three data sets.

Based on the disk structure, we would expect to find companions inside 19 if the inner edge of the cold disk is maintained by unseen planet(s). Our high-contrast observations rule out the presence of a low-mass star down to the brown dwarf regime in the range of 05–3''(∼25–150 AU). Bear in mind, as shown in Figure 3, beyond 15 we do not have complete sky coverage around the star. As mentioned in Section 3.1, the HD 17848 disk may be an edge-on. This limits our ability to detect companions unless they are fortuitously at a projected separation outward of 05.

HD 28355 is an A7V (Abt & Morrell 1995) member of the 650 Myr-old Hyades cluster (Su et al. 2006; De Gennaro et al. 2009) at a distance of 48.8 ± 0.6 pc away (van Leeuwen 2007). Its IRS spectrum along with MIPS 70 μm photometry has been published by several papers with different derived dust temperatures. Morales et al. (2011) suggest the disk SED is best described by two temperatures of 128 K and 60 K, which are consistent with the values (120 K+54 K) published by Ballering et al. (2013) within errors. Analysis performed by Chen et al. (2014) gives the warm temperature of 176$^{+7}_{-8}$ K and the cold temperature of 69$^{+5}_{-6}$ K. As discussed in Section 3, HD 28355 is near a bright infrared source, which is extended at 160 μm, as a result its far-infrared fluxes are contaminated. The overall disk SED can be fit with two temperatures: ∼120 K and ∼55 K (see Figure 1). However, the contamination-free part of the SED (data shortward of ∼80 μm) can also be described by one single temperature of ∼80 K (within a few σ). Therefore, we excluded the Herschel 160 μm data and the MIPS-SED points longward of 85 μm in our one-component SED model. The inner boundary is estimated to be 46 ± 12 AU, and the outer is 130 ± 30 AU with a total dust mass of (1.8 ± 0.7)× 10⁻³M_⊕.

No bright point sources were detected around HD 28355 (see Figure 3). The structure inward of 15 is due to residuals from the APP PSF. We were not sensitive to planetary mass companions due to the older age of the star. We were able to rule out substellar companions from 05–3''(∼24–147 AU). Our single component SED model places the inner boundary of the disk at 09. Thus, if the cold disk is shaped by a companion, it must be less than ∼50 M_Jup.

HD 37484 is a F3V (Houk 1982) star at a distance of d = 56.8 ± 2.0 pc (adopting parallax ϖ = 17.61 ± 0.62 mas from van Leeuwen 2007). Its B − V versus M_V position places it on the zero age main sequence (ZAMS), thus it must be >27 Myr in order to allow enough time for a ∼1.36 M_☉ star to reach the main sequence. Its position is commensurate with that of the Pleiades and IC2391 clusters, making it <100 Myr. It has a brighter and hotter (∼A5V) common proper motion companion 0.34 deg away (HR 1915). Both stars are considered Columba members by Malo et al. (2013), who list 10-40 Myr for the group. The companion appears to be on or below the ZAMS, consistent with an age of >19 Myr. HD 37484's HR diagram position is consistent with the previous assigned age of ∼30 Myr. Given its Columba membership and consistency with the age of other Columba members, we adopt the age of 30 ± 10 Myr.

The disk SED is composed of data from Spitzer MIPS 24 and 70 μm photometry as well as IRS 10–35 μm spectroscopy. The disk SED is best fit with a blackbody temperature of ∼88 K. The best fitting disk model has an inner radius of $12^{+20}_{-4}$ AU, outer radius of $100^{+100}_{-20}$ AU and dust mass of (2 ± 1)× 10⁻³ M_⊕. This source does not have longer wavelength Herschel data, which makes the outer disk radius and dust mass fairly unconstrained.

We initially detected a faint point source around HD 37484 at 097 (55 AU from the star), 9.0 magnitudes fainter than the host star at a P.A. of 103° in our APP data (see Figure 3). However, this point source was not detected in archival data with equivalent sensitivity taken in 2011 (088.C-0085(A)), allowing us to conclude that it is unlikely to be a real source. This false detection demonstrates the importance of archival data, which can be used to quickly confirm or deny the physical nature of a point source.

HD 95086 is a 17 Myr old A8 LCC member star at a distance of 90.4 ± 3.3 pc (Meshkat et al. 2013). The system was also resolved by Herschel at 70 and 100 μm (Moór et al. 2013) with an estimated inclination of 25° from face-on. Re-analysis of the Herschel resolved images combined with detailed SED modeling reveal that the extended part of the images arises from a disk halo (Su et al. 2014), similar to the disk halo found around HR 8799 (Su et al. 2009; Matthews et al. 2014). In the three-component disk model presented by Su et al. (2014), the inner belt is located from ∼7 AU to ∼10 AU, and the cold planetesimal disk likely ranges from 63 ± 6 AU to 189 ± 13 AU, and are surrounded by an extended halo up to ∼800 AU.

A 5 M_Jup planet was discovered around HD 95086 by Rameau et al. (2013b). Our H-band NICI/Gemini non-detection provided a strict color lower limit for the planet of H − L' > 3.1 ± 0.5 mag (Meshkat et al. 2013). The subsequent H-band Gemini Planet Imager (GPI) detection of HD 95086 b by (Galicher et al. 2014) provides a red planet color of H − L' = 3.6 ± 1.0. We also observed the system with the APP, but did not detect it due to decreased AO quality at high airmass (>1.4) during the observation (APP hemisphere 2; see Figure 3). Due to lack of sky rotation, we are missing about 20° of coverage. Figure 2 shows that we achieve a sensitivity of ∼10 M_Jup at the separation of the planet, however the detected planet is 5 ± 2 M_Jup (Rameau et al. 2013a).

HD 134888 is an F4V star, located 90$^{+9}_{-8}$ pc away (adopting parallax ϖ = 11.12 ± 1.05 mas from van Leeuwen 2007). We adopt an age of 16 Myr based on its membership in the Lower Centaurus-Crux (LCC) association, commensurate with isochronal age estimates for the star (16-25 Myr; Pecaut et al. 2012). Chen et al. (2014) conclude it has a two-component disk with the warm temperature of 387$^{+6}_{-7}$ K and the cold temperature of 72 ± 5 K. The far-IR excesses along with the IRS excess longward of ∼30 μm are consistent with a blackbody temperature of 75 K (see Figure 1). However, we found the warm excess is better fit with a dust temperature of ∼175 K (i.e., asteroid-like). The derived temperature in the warm component is highly dependent on the amount and shape of the excess emission in the 10–20 μm region. Future mid-IR observations are needed to better characterize the warm component. We excluded data points shortward of ∼30 μm in our SED fitting of the cold component. The best-fit one cold-component disk ranges from 60 ± 11 AU to 135 ± 29 AU with a dust mass of 3.25 ± 1.4 × 10⁻² M_⊕.

Based on our inferred disk structure, the ideal place for detecting the potential low-mass companion that sculpts the disk is interior of 60 AU (066) or exterior of 135 AU (15). A possible interesting point source appeared at 02 from the star, but after comparing with archival data (087.C-0142(A)), the point source appears to be an artifact of the data reduction. No believable point sources are detected in our reduced images around the star at any projected separation, however we cannot conclusively say there are no companions around HD 134888. From a P.A. of 135° to −135° (clock-wise, passing through North; see Figure 3), we do not detect any companions with an average mass limit of 5–7 M_Jup.

HD 110058 is a A0V star (Houk 1978) and member of the LCC subgroup of Sco-Cen (de Zeeuw et al. 1999; Rizzuto et al. 2011). The revised Hipparcos parallax from van Leeuwen (2007) (ϖ = 9.31 ± 0.78 mas) translates to a distance of 107$^{+10}_{-8}$ pc. The star is situated near ℓ, b = 301°, +136, in the northern part of LCC, which appears to be the oldest part of the subgroup (Preibisch & Mamajek 2008; Pecaut et al. 2012), so we follow Chen et al. (2014) and adopt an age of 17 Myr.

We did not achieve full 360° APP coverage around this star. Only one APP hemisphere was observed and it had poor sky rotation. Without full sky coverage, the contrast curve for HD 110058 (Figure 2) is only valid from a P.A. of ∼100° to −90° (clock-wise, passing through North; see Figure 3). We are sensitive to 5–8 M_Jup from 05–3''. We detect no point sources in this limited region around the star.