Документ взят из кэша поисковой машины. Адрес оригинального документа : http://www.stsci.edu/hst/wfc3/documents/ISRs/WFC3-2011-03.pdf Дата изменения: Wed Jan 19 18:36:04 2011 Дата индексирования: Fri Feb 11 10:05:18 2011 Кодировка: Поисковые слова: annular solar eclipse

Instrument Science Report WFC3 2011-03

WFC3 UVIS High-resolution Imaging Performance
R. L. Gilliland and A. Ra jan
gillil@stsci.edu January 2011

Abstract
The sharp and stable point spread function of HST and WFC3 allows for the detection of stellar companions at small offsets from target stars. Deep images obtained through six UVIS filters are used to assess this ability to detect faint companions of brighter targets. Analogous capabilities from the ground make use of Adaptive Optics and Speckle imaging. We contrast the capabilities of HST/WFC3 with several existing AO systems for imaging of V 12 stars for which detection of faint companions as close as possible is desired, and to a delta-magnitude of 10. The observations under comparison are motivated by high-resolution imaging applications for the Kepler Mission to constrain the existence of fainter background stars which, if eclipsing binaries, could mimic transits. The HST/WFC3 observations are found to be superior in the comparisons made to Keck, MMT and Palomar AO and WIYN Speckle systems in terms of delta-magnitude limits outside of about 0.3 arcseconds. In a metric of fraction of phase space that can be probed for delta-magnitude to 10 companions out to 2 arcsecs the UVIS imaging out performs all other approaches by a factor > 5.

1.

Introduction

An important attribute of HST imaging that makes it often superior to any groundbased capabilities to date is the ability to deliver very sharp images over wide fields of view.


­2­ Adaptive optics, at least in the near infrared can be competitive with HST in terms of point spread function (PSF) width, but usually is limited by size of field that can be well corrected compared to HST cameras, and AO generally provides a more variable PSF. The one-orbit HST/WFC3 calibration observations from CAL/WFC3-12354 analyzed here are motivated by a more restrictive science goal that does not require a large field of view, and thus one for which AO may be relatively more competitive. This ISR seeks to establish two things: (1) What stellar companions expressed as delta-magnitudes at varying offset distances to a target star of about V = 12 can be detected in observations that could be obtained in SNAPSHOT mo de (i.e. limited to < 40 minutes per target) with WFC3 UVIS? (2) How do AO and Speckle imaging on commonly available systems compare with HST capabilities? The motivation for this study is relevant for understanding why this comparison uses stars of about 12th magnitude and a delta-magnitude of 10 goal. Rather different results might be expected in significantly different regimes of target brightness. The Kepler mission is observing some 150,000 stars simultaneously for 3.5 years with a goal of detecting planets as small as the Earth orbiting distant stars that are the size of the Sun. The magnitude range for Kepler is typically taken as V = 9 to 15, with 12th magnitude the "design point" at which the number of targets is large enough, and the signal-to-noise still high enough to return go o d statistics on Earth-sized planets. In order to survey such a large number of bright stars Kepler uses 42 CCDs with a pixel scale of 4.0 arcsecs per pixel. See Ko ch et al (2010) for further details on the Kepler Mission. At 12th magnitude the typical photometric aperture will contain a dozen pixels centered on the target star, covering a region of some 14x14 arcseconds. Planets are sought via the transit technique whereby an Earth analog would induce an 85 parts per million drop in the stellar intensity for about 10 hours, once per year when the planet transits the star. A background eclipsing binary (BGEB) anywhere in the 14x14 arcsecond photometric aperture can pro duce a signal mimicking a transit, even to a delta-magnitude of 10 for an Earth analog signal if the binary has intrinsically deep eclipses. Confirming transit-like signals as planets is largely an exercise in eliminating all other possible explanations, and quantifying the relative probability of planet interpretation compared to possible false positive sources. For eliminating, or constraining as much as possible background eclipsing binaries as false positives three factors come into play: (1) Knowing the density of background stars and fraction that are binaries. (2) Using the Kepler data itself with centroid tests across transits to eliminate all possible BGEBs outside of typical radii of 1 ­ 3 arcsecs. (3) Using high resolution imaging to constrain the existence of nearby neighbors within the angle the Kepler data centroiding cannot eliminate.


­3­ In practice this means that high resolution imaging is called upon to determine a census of stellar neighbors over angular separations of 0.1 ­ 3 arcsecs as close to delta-magnitudes of about 10 as possible for targets with typical magnitudes of 12. When high-resolution imaging do es find a nearby companion further investigation would be needed to determine if that may have caused a false positive. If the companion is not an eclipsing binary that when blended induces the transit signal, it is still necessary to estimate the amount of contamination in the Kepler bandpass in order to obtain an undiluted transit signal and draw the right conclusions about the planet.

2.

WFC3 Observations

Since Kepler targets can span a range from 9 - 15th magnitude, and any use of HST for a survey of this type would need to use brief observations, filter pairs have been chosen such that sufficiently high signal-to-noise observations can be obtained with similar exposure time over that full range. For 9th magnitude stars the narrowest filter pair of F487N + F673N would be used, intermediate brightness would be covered by F467M + F680N, and the faintest Kepler targets could still be probed with F555W + F775W. The combination of a blue and red filter provides redundant information and would allow accurate determination of relative magnitudes in the very broad Kepler native bandpass. We have chosen the HST spectroscopic standard P041C (a NICMOS "solar-analog" standard) with V = 12.00 for which high SNR observations in all filter pairs were possible within one HST orbit total. P041C was believed to be a single star based on inspection of previous WFPC2 images, and extensive FGS2r TRANS mo de observations (Gilliland et al 2009). P041C is carried as a standard that will help bring HST and JWST observations to a common calibration. In each filter a total of 5 exposures have been taken, 4 of these in a standard WFC3-UVIS-DITHER-BOX 4-point dither set each of which was aimed at reaching 80% of full well depth in the central pixel (when star centered on a pixel), plus one additional exposure with twice the exposure time of the other 4 summed. The latter saturates the detector, and brings the SNR ratio of the wings up. Exposure times used for the 4-point dither individual exposures are shown in Table 1. The UVIS2-C1K1C subarray nearest a readout amplifier was used allowing all 30 exposures to fit within one orbit (the long exposure for F487N needed to be truncated). If only pairs of filters were used the observations including HST guide star acquisition would span about 15 minutes. It is noted in passing that the exposure level goal of a peak at 56,000 e- was never reached and exposure times 20% larger could have been used. The WFC3 Exposure Time Calculator apparently assumes a slightly to o sharp PSF, or these observations were obtained at less


­4­

Table 1: WFC3 UVIS Observations of P041C Filter F 467M F 487N F 555W F 673N F 680N F 775W t
exp

(s) .3 .3 .9 .1 .5 .8

FWHM (mas) 8 8 8 8 7 7 3 4 6 2 6 7 .2 .0 .0 .0 .0 .6

Peak (e-) 4 3 3 4 3 4 6 4 3 3 9 0 4 2 0 8 7 0 7 4 2 1 6 1 0 6 1 8 9 3

7 28 0 14 4 1

Note. -- texp refers to component dither-box exposures, FWHM is width of PSF core in the multidrizzle combined image, and Peak refers to the greatest value over the four dither-box component exposures.

than optimal fo cus. The 5 exposures in each filter providing cosmic ray elimination a 0.040 arcsecs. Weights were set to to +/-1 along columns beyond the were combined using Multidrizzle by Abhijith Ra jan nd output to a geometrically corrected pixel scale of zero for the pixels saturating in the long exposure, and obvious extent of bleeding.

Figure 1 shows approximately 8в12 arcsec regions for each filter with a direct stretch, while Figure 2 shows the same with a logarithmic stretch. Although selected to be a single star to provide a null test case in these calibrations, inspection of the WFC3 images strongly suggested the presence of a faint companion offset by about 0.57 arcsecs from the target (left by 14, down by 2.5 WFC3 UVIS pixels). This companion is fainter by 7.6 mag at the shortest wavelength filter F467M, about 7.0 magnitudes in F555W, and 5.4 magnitudes in F775W. Experimentation with the WFC3 ETC suggests that an M3 ­ M4 physical companion star to this G0 V target would match both the absolute and wavelength dependent differences. If this were the case, then in H -band the delta-mag would be about 3.5, thus the companion would be down a factor of 25 in direct counts. Subsequent analysis of archival NICMOS imaging in F160W with NIC2 shows a companion with an offset distance exactly matching that determined from WFC3, and down by a factor of 32 in count rate agreeing quite well with the extrapolation based on the WFC3 bandpasses. This means that P041C is actually not a go o d "solar analog" since mo dels of G0 ­ G2 stars would have errors of some 3% already by H , and even larger farther in the IR unless the companion is physically separated from the target. At only 0.57 arcsec separation it


­5­

Fig. 1.-- Combined images in each of the six UVIS filters for P041C. The total counts in each combined image is well over 106 e-, providing SNR of over 1000. FWHM of images is typically about 2.1 pixels, or about 85 mas. is likely the companion and target have been blended in some of the existing STIS and NICMOS calibration spectrophotometry. WFC3 high resolution imaging has been successful in demoting P041C from a high priority "solar analog" calibration target. The existence of the companion has little impact on these results, and its presence is taken into account as needed herein. All images are scaled from e-/s in the drizzled images to e- units for the full exposure time of the combined images to facilitate statistical results. The noise level in the outer parts of all six images shown in Figures 1 and 2 span a narrow range of 4.85 to 4.98 e-. As a quick estimate of companion detectability note that the count level within the central 5x5 pixels in these images averages 2.1в106 e-. Taking into account these noise levels, companions (in the periphery) can be seen at SNR of 5 to a delta-magnitude of 10.46 and at an SNR of 10 to 9.62 ­ very much in line with expectations. At least in the outer regions these images would easily support detection of any stars that if BGEBs could mimic an Earth analog transit signal on the target star.


­6­

Fig. 2.-- Same as Figure 1, but now with a logarithmic stretch providing a larger dynamic range. The faint companion discussed in the text is most evident in F775W to the left and slightly down. It is also interesting to note in passing that these WFC3 images easily detected a 7 magnitude fainter physical companion. For the Kepler Mission a physical companion that is itself an eclipsing binary, or M-dwarf transited by a large planet, can mimic small planet transits. High resolution imaging can thus also restrict the phase space of potential hierarchical triples. P041C at G0 and V = 12.0 would have a distance of 300 parsecs. The companion with observed 0. 57 offset at 300 pc distance would thus be at a minimum physical separation of about 170 parsecs, and could have been seen here down to half this value in the WFC3 UVIS images.

3.

AO and Sp eckle Observations for Comparison

As part of the Kepler follow-up observing program high-resolution imaging is routinely obtained for many high priority planet candidate host stars. From these data available to the Kepler team we have selected what are thought to be representative cases, i.e. neither


­7­

Table 2: AO and Speckle Observations Telescope Palomar 5-m MMT 6.5-m Keck II 10-m WIYN 3.5-m Instrument PHARO ARIES NGS Speckle Target KOIKOIKOIKOI72 85 64 72 V 11.0 11.2 13.3 11.0 Obs Time 20 15 30 15 Band J Ks Ks R FWHM (mas) 66.5 96.4 50.5 66.7

Note. -- `Obs Time' in minutes is how much time at the telescope was required from start of slew to target to start of slew for next target, `Band' is the filter bandpass used in the observations.

the best that can be done, nor po orer than average which would bias the comparison. The observations are of stars with V = 11 - 13, and the amount of observing time used would allow surveying some twenty targets per night. Table 2 provides some basic information about the AO imaging conducted with Keck II, the Palomar 5-m, and the 6.5-m MMT, as well as Speckle imaging with the 3.5-m WIYN telescope. Further details for each observation are provided below. P041C has J 11.0, K have J 10.0. KOI-64 is an AO target is slightly fainter are about one magnitude br 10.7, while KOIs 72 and 85 in Table 2 (all solar type stars) early K dwarf with J = 11.76, and K = 11.23. Thus the Keck by 0.5 magnitude than the HST target, while the other cases ighter.

Figure 3 illustrates the PSFs provided by the several imaging approaches considered in this ISR. All images have been bi-cubically interpolated to a common scale of 25 mas, and normalized to a common central pixel intensity, then displayed with a logarithmic stretch. Tables 1 and 2 include the FWHM for these representative images. For the purposes motivating this calibration the sweet spot is about 0.3 ­ 1.0 arcsecs in radius. Outside of this the Kepler data itself can often discriminate against potential contaminate BGEBs, while inside of 0. 3 arcsecs the small area is less interesting. Within this annulus of primary interest the AO images tend to show more structure compared to HST, and this likely results from speckles that would be different observation to observation. The WIYN Speckle image was acquired with a goal of reaching limiting magnitudes only about half that of the HST and AO imaging, and thus in the forced common stretch


­8­

Fig. 3.-- A 4.2 by 6.4 arcsec region is shown centered on each of the images from all instruments compared in this ISR. All have been adjusted to a common pixel scale and intensity to facilitate easy comparison. The lower-left panel shows the difference of two WFC3 UVIS images to illustrate the potential of PSF subtraction. The lower-right panel shows a reconstructed image from Speckle observations, but caution should be exercised in its interpretation since it is not a true image in the sense of the others. is not shown in a favorable light. The WIYN Speckle imaging is competitive with all of these techniques, if not superior at the smallest working angles, and has an advantage (for Kepler-motivated applications) of being in the optical. The speckle imaging of KOI-72 was obtained on the night of 18 June 2010 UT using the two-color speckle camera on the Kitt Peak 3.5-m WIYN telescope. Speckle images are obtained simultaneously in narrow V and R filters and for KOI-72 consisted of 2000 30 msec EMCCD exposures, thus 60 seconds total which to ok 15 minutes total to acquire the star and take the data before moving on to the next star. These data were reduced and pro cessed to produce a final reconstructed speckle image for each filter. The R-band result is included in Figure 3. Details of the two-color EMCCD speckle camera are presented in Howell et al. (2011) and Horch et al (2011). The speckle imaging is capable of detecting companions from 0.05 arcsec to 1.5 arcsec with delta-magnitudes near 6 in R at the large offset limit. The speckle observations


­9­ are performed simultaneously in V , R or I . The depth derived as al (2011) the depth at significant 6.0 beyond 1 arcsec. The deeper, reported in Table 3 for speckle. two bandpasses each of 40 nm width centered near one of shown in Fig. 9 is about a delta-mag of 4.5. In Batalha et offset for these observations was quoted as having reached published curve has been used for the merit function value

The Palomar Hale 5-m telescope and the PHARO near-infrared camera (Hayward et al. 2001) using the Palomar AO system (Troy et al. 2000) was used on 8 September 2009 UT to image KOI-72 in the J -band. PHARO uses a 1024 в 1024 HgCdTe infrared array, and was utilized in a 25.1 mas/pixel mo de providing a 25 arcsec field of view. These data were collected in a standard 5-point quincunx dither pattern (e.g. dice pattern for number 5) of 5 arcsec steps interlaced with an off-source sky dither pattern. Data were taken at two separate times within the same night ­ 150 frames using 1.4-second and 150 using 2.8-second integrations ­ for a total on source integration time of 10 minutes. The individual frames were reduced with a custom set of IDL routines written for the PHARO camera and were combined into a single final image. The adaptive optics system guided on the primary target itself and pro duced measured Strehl ratios of 0.15 at J. As reported in Batalha et al. (2011) source detection completeness was accomplished by randomly inserting fake sources of various magnitudes in 0.5 mag steps at varying distances in steps of the FWHM from the primary target. Identification of sources was performed by eye and using an IDL version of DAOPhot with detection limits set when a source was not detected by the automated FIND routine or by eye. Within 1 ­ 2 FWHM (0. 1) the eye was better at discerning two sources; beyond this the two metho ds agreed well, and are in go o d agreement with the signal-to-noise based approach adopted for this ISR. The Keck II 10-m telescope and natural guide star AO system on Mauna Kea was used with facility infrared camera NIRC2 and a narrowband 2.271 µm K-cont filter to image the V 13.3, J = 11.76 KOI-64 on 14 July 2010. NIRC2 utilizes a 1024 в 1024 pixel HgCdTe, and the observations were taken with the narrow-field camera, which pro duces a 0. 00994 per pixel scale and 10. 2 field of view. The data were acquired in a standard 5-point quincunx dither pattern with 1 steps. A 30 second integration time per frame and 5 co-additions per position was used for a total integration time of 12.5 minutes. As with the Palomar AO data, the individual frames were reduced and co-added with a custom set of IDL routines. The final coadded image had an FWHM of 0. 06 and a Strehl ratio 0.5. No sources other than the primary target were detected within the 10 field. The MMT 6.5-m telescope with the adaptive secondary AO system in natural guide star mo de was used with the ARIES imager and Ks filter on 04 May 2010 UT to observe the V 11.2, J = 10.06, KOI-85. The ARIES camera was used at a plate scale of 0. 02


­ 10 ­ arcsec/pixel yielding a field of view of 20 arcsec square. Images were acquired in a 2в2 square raster pattern with a 0. 2 arcsec dither. Each exposure was 2 sec, and a total of 16 frames were obtained, totaling 32 sec. The images were reduced using scripts constructed from IRAF. For these AO images the science support did not require reaching a delta-magnitude of 10 at large offsets. The gain in detection depth is expected to improve as square ro ot of time at least in the flat portion at large offset, likely less than this in the inner part of the PSF core.

4.

Qualitative assessment of companion detection limits.

A simple approach to assessing companion detectability is to generate grids of added stars by scaling down, shifting and adding in simulated signals at known delta-magnitude levels and then visually noting ease of recognition. Figure 4 shows the result of doing this for several of the WFC3 UVIS filters, a difference of two UVIS filter images, and the Palomar AO image which is the deepest of the AO images considered here. The simulated companions are added in at a delta-magnitude of 9.0 for Figure 4 on a recti-linear grid. In this griding the simulated companions nearest the target are offset by 0. 39, and the next nearest 0. 61. At the stretch displayed these nearest companions are often discernible to the eye in the HST images, but are lost in the PSF glare for the Palomar AO images. For all images displayed delta-magnitude 9 companions are easy to recognize outside of about 1. 0. Figure 5 is the same as Figure 4, but with simulated images now at a delta-magnitude of 10.0. These remain within range for the HST images at significant distance from the target, but are not reliably seen in the Palomar AO image.

5.

Quantitative companion detection limits.

Setting quantitative companion detection limits is probably established most reliably by inserting simulated stars at random positions and delta-magnitudes, then using a program such as DAOPHOT (DAOFIND) to recover the simulated stars and recording completeness of recovery as a function of radial offsets and delta-magnitude. This approach was used for the Palomar 5-m AO image analyzed here and reported in Batalha et al. (2011) in a figure showing recoverable delta-magnitude plotted against angular separation. We have also performed such simulation experiments followed by DAOPHOT identification for these HST images finding go o d agreement with the approach discussed


­ 11 ­

Fig. 4.-- Similar to earlier figures, but now with a grid of simulated stars added in with a delta-magnitude of 9.0. The simulated stars are easy to recognize at distance from the target star in all cases, but can be discerned at smaller offsets in the WFC3 UVIS images than for the Palomar 5-m AO image. next. A simpler approach also do es an excellent job of repro ducing the results for the Palomar AO image, and facilitates a more concise description and comparison of results across diverse imaging inputs. The simpler approach adopted here establishes what stellar intensity as a function of radial offset would have a SNR of 5.0 given the noise level in the image at this radius. This can then be converted to a delta-magnitude by comparison to the intensity of the target star. The pro cedure is thus: 1. Scale the image to be in electrons so that simple Poisson statistics hold. 2. Set up a series of concentric and contiguous annuli 0. 08 wide with the first one centered at an offset of 0. 1 from the target. 3. Within each annulus estimate the noise level as the standard distribution of all data points with a center contained in the annulus. This will tend to over-estimate the


­ 12 ­

Fig. 5.-- are now r detection visible in

Same as Figure 4, but with ecognizeable at distance in in the Palomar AO image. Figure 2 for the HST imag

the simulated stars at a delta-magnitude of 10.0. These the WFC3 UVIS images, but generally beyond reliable Note that the physical companion discussed in §2 and es has been subtracted out of these images.

noise in regimes where the PSF has a significant gradient over the annular extent, but will under-estimate the impact of infrequent positive spots in the PSF which could mimic stars. 4. Adopt the simple SNR equation equal to the ratio of source counts in an aperture divided by square ro ot of the sum of source counts and total variance on the noise in the aperture. Invert this for source counts assuming a target SNR of 5.0. Adopt an aperture size of 3x3 pixels. 5. Use the noise as a function of radius to obtain required companion counts to yield SNR of 5.0. 6. Evaluate the number of counts for the target in its central 3x3 pixels, and tabulate detectable delta-magnitude as a function of radius. For any science application we would recommend still carrying a full Monte Carlo


­ 13 ­

Fig. 6.-- Shows the delta-magnitude at which the mean noise as a function of radius allows an SNR 5.0 detection relative to the target intensity in a 3x3 pixel aperture. The two broad band WFC3 UVIS filters are contrasted to the Palomar 5-m AO J -band. style simulation and recovery experiment to establish confidence in results. For this cross platform comparison the simpler approach outlined above is preferred. Results contrasting the broad filters, F555W and F775W in Figure 6. These filters would be used to observe faint targ used here the exposure times of about 1 second may have led PSF due to shutter induced jitter (Sabbi 2009), which would recovery potential at the smaller working angles, although the was not a problem for these exposures. used with WFC3 are shown ets. For the V = 12 target to slight degradation of the in turn degrade companion narrow FWHMs suggest this

The intermediate width filters, F467M and F673N are shown in Figure 7 and would be the pair of choice at 12th magnitude. The narrowest pair of filters, F487N and F680N are shown in Figure 8 and would be used to observe stars as bright as 9th magnitude in V . In order to fit all observations in one HST orbit the exposure time was truncated by about a factor of two for F487N explaining the shallower limiting delta-magnitude at large distances.


­ 14 ­

Fig. 7.-- Same as Figure 6 but for the intermediate width filters F467M and F673n. The overall results are shown in Figure 9 which contrasts all of the imaging sources analyzed here. Three curves are shown for the WFC3 UVIS imaging: The curve labeled `WFC3 F467M' is the best result over the six filters used and was shown in Figure 7 as well. The curve labeled `WFC3 UVIS' shows the detection limits for the difference image F673N - F680N ­ the near coincident wavelengths of these two filters allow one to be used as a proxy PSF for the other. In practice the PSF subtraction as shown in Figures 3 ­ 5 was only mo derately successful with residual hot spots from the two filters not controlled by the other, and imperfectly subtracted airy rings. Nonetheless this provides a superior detection limit over 0. 1 to 1. 0. Finally the `UVIS PSF Limit' curve shows what the detection limit would be if the only noise was Poisson from the PSF, i.e. the limit in the domain of being able to subtract a noiseless PSF. In any survey program where many targets per filter pair are all imaged to the high SNR reached here it would be expected that PSF subtraction could support some fraction of this potential gain. A reasonable estimate is that most of the potential could be realized outside of about 0. 3, but that variations in the PSF due to telescope breathing would yield only a fraction of this potential gain for the innermost radial offsets.


­ 15 ­

Fig. 8.-- Same as Figures 6 and 7 but for the narrowest pair of filters ­ F487N and F680N. The results over any of the raw companion observations are made for any of shown in Figure 9 demonstrate an advantage for WFC3 UVIS imaging AO or Speckle observations in the prime 0. 3 to 1. range in terms of delta-magnitude detectability. An additional advantage would follow if desired in the optical where AO is not yet competitive. Gains could be these observations by acquiring more data.

An interesting metric follows from considering what fraction of potential contaminant phase space for background eclipsing binaries that the Kepler data itself cannot discriminate against could be found by the different imaging results as shown in Figure 9. We adopt the assumptions that: (a) the angular domain of interest is 0. 1 ­ 2. , (b) target star has V = 12.0, and (c) delta-magnitudes to 10.0 are of interest. Near the Kepler field center the stellar density at 13.5 is about 3.1в10-5 per square arcsecond increasing to about 10-3 per square arcsec at the faintest relevant magnitude of 21.5. Within the nominal 2 arcsec radius aperture there is a 6% chance of a star having a chance superposition with a background star that if an eclipsing binary could mimic an Earth-sized transit event. The chance of such a contaminant actually being an eclipsing binary is down by another factor of 100, thus only rarely will stars be expected to suffer chance alignment with relevant


­ 16 ­

Fig. 9.-- This provides a comparison of companion detection limits in delta-magnitude relative to the target star for all of the imaging set ups discussed in this ISR, see text for details. background eclipsing binaries, but when surveying 150,000 stars such will happen many times. High resolution imaging can be of value when it can eliminate a significant part of the contaminant phase space. Table 3 shows the fraction of this 6% chance alignment per star with fainter background sources that each curve shown in Figure 9 is still unable to supply. It needs to be emphasized that while this is a relatively unbiased comparison of the several imaging approaches, and that available images were not pre-selected to give results advantageous to HST, it nonetheless may be unfair in that not all of the observations compared had goals of reaching to delta-magnitudes of 10. Additional observing time devoted to the AO and Speckle cases would undoubtedly allow more favorable results, especially in the angular separations beyond 1 arcsec where gains may be expected to improve as square ro ot of the observing effort. Accepting the metric of Table 3 at face value implies a large relative advantage for the WFC3 UVIS imaging for the type of observation motivating this study. This advantage


­ 17 ­

Table 3: Undetected companion fractional residuals. Imaging Case UVIS PSF Limit WFC3 UVIS WFC3 F467M Palomar AO J MMT AO K Keck AO K WIYN Speckle Residual 0.020 0.051 0.101 0.234 0.287 0.437 0.857

can be most simply expressed as the ratio of residual contamination fractions listed in Table 3. The `WFC3 UVIS' line is taken to represent the achieved standard using the HST data. With this approach the advantage of HST over the Palomar 5-m AO is в4.6, and even more for all of the other cases. Furthermore the HST imaging carries the qualitative advantage of both providing a measurement in the optical, and a color for this allowing go o d repro duction of magnitudes in the Kepler bandpass. The AO imaging would leave a significant uncertainty in extrapolated magnitudes.

6.

Summary

We have shown that short HST observations with WFC3 UVIS are capable of returning high resolution imaging for the purpose of detecting faint companions near targets with V = 9 - 15 that are superior to imaging that can be obtained with ground-based AO systems. For the particular application motivating this study HST imaging can eliminate some 95% of residual phase space for possible faint companion contamination ­ a factor of several better than provided by the few AO images compared to. Due to the generally stable operating environment of HST it is expected that PSF subtraction could provide further gains for detection of companions quite near the target. Such an advantage could be tested, if a survey were conducted which obtains 10 individual observations of comparable depth. The Palomar 5-m AO data was provided courtesy of David Ciardi and Nick Gautier, while David also supplied the Keck AO image along with valuable comments. The MMT


­ 18 ­ AO data was provided by Andrea Dupree and Elizabeth Adams. Observations reported here were obtained at the MMT Observatory, a joint facility of the Smithsonian Institution and the University of Arizona. The MMT ARIES PI instrument on the MMT has been developed by Don McCarthy and Craig Kulesa of the University of Arizona. The WIYN Speckle (pseudo-)image was provided by Steve Howell. We am very grateful for this sharing of data and the numerous comments provided. Portions of this work were performed at the California Institute of Technology under contract with the National Aeronautics and Space Administration. Some of the data presented herein were obtained at the W.M. Keck Observatory, which is operated as a scientific partnership among the California Institute of Technology, the University of California and the National Aeronautics and Space Administration. The Observatory was made possible by the generous financial support of the W.M. Keck Foundation. The authors wish to recognize and acknowledge the very significant cultural role and reverence that the summit of Mauna Kea has always had within the indigenous Hawaiian community. We are most fortunate to have the opportunity to conduct observations from this mountain. We thank Paul Kalas for discussions and Linda Dressel for reviewing and commenting on the manuscript.

References
Batalha, N.M. et al, 2011, ApJ, in press. Gilliland, R.L., Bohlin, R.C., McCullough, P.R. & Nelan, E. 2009, Instrument Science Report TEL 2009-01, (Baltimore, MD: STScI) Haward, T.L., et al. 2001, PASP, 113, 105 Horch, E., et al. 2011, AJ, 141, in press Howell, S.B., et al. 2011, in preparation Ko ch, D.G., et al. 2010, ApJ, 713, L79 Sabbi, E. 2009, "WFC3 SMOV Program 11798: UVIS PSF Core Mo dulation," WFC3 ISR 2009-20. (Baltimore: STScI) Troy, M., et al. 2000, Pro c. SPIE, 4007, 31