1: %Paper on Vega and Eps Eri based on 2006 Clio data.
2:
3: \documentclass[12pt,preprint]{aastex}
4:
5: \slugcomment{To be submitted to \apj .}
6:
7: \shorttitle{Searching for Planets Around Vega and $\epsilon$ Eri}
8: \shortauthors{Heinze and Hinz}
9:
10: \begin{document}
11:
12: \title{Deep L' and M-band Imaging for Planets Around Vega and $\epsilon$
13: Eridani\altaffilmark{1}}
14:
15: \author{A. N. Heinze}
16: \affil{Swarthmore College, 500 College Avenue, Swarthmore, PA 19081}
17: \email{aheinze1@swarthmore.edu}
18:
19: \author{Philip M. Hinz}
20: \affil{Steward Observatory, University of Arizona, 933 N Cherry Avenue, Tucson, AZ 85721-0065}
21: \email{phinz@as.arizona.edu}
22:
23: \author{Matthew Kenworthy}
24: \affil{Steward Observatory, University of Arizona, 933 N Cherry Avenue, Tucson, AZ 85721-0065}
25: \email{mkenworthy@as.arizona.edu}
26:
27: \author{Douglas Miller}
28: \affil{Steward Observatory, University of Arizona, 933 N Cherry Avenue, Tucson, AZ 85721-0065}
29: \email{dlmiller@as.arizona.edu}
30:
31: \author{Suresh Sivanandam}
32: \affil{Steward Observatory, University of Arizona, 933 N Cherry Avenue, Tucson, AZ 85721-0065}
33: \email{suresh@as.arizona.edu}
34:
35: \altaffiltext{1}{Observations reported here were obtained at the MMT
36: Observatory, a joint facility of the University of Arizona and the
37: Smithsonian Institution.}
38:
39:
40:
41: \begin{abstract}
42: We have obtained deep Adaptive Optics (AO) images of
43: Vega and $\epsilon$ Eri to search for planetary-mass
44: companions. We observed at the MMT
45: in the $L'$ (3.8 $\mu$m) and $M$ (4.8 $\mu$m) bands using Clio,
46: a recently commissioned imager optimized for these wavelengths.
47: Observing at these long wavelengths represents a departure from
48: the $H$ band (1.65 $\mu$m) more commonly used for AO imaging
49: searches for extrasolar planets. The long wavelengths offer
50: better predicted planet/star flux ratios and cleaner (higher Strehl) AO images,
51: at the cost of lower diffraction limited resolution and higher sky background.
52: We have not detected any planets or planet candidates around Vega or
53: $\epsilon$ Eri. We report the sensitivities obtained around
54: both stars, which correspond to upper limits on any planetary companions
55: which may exist. The sensitivities of our $L'$ and $M$ band
56: observations are comparable to those of the best $H$-regime observations
57: of these stars. For $\epsilon$ Eri our $M$ band
58: observations deliver considerably better sensitivity to close-in planets
59: than any previously published results, and we show that the $M$ band
60: is by far the best wavelength choice for attempts at ground-based
61: AO imaging of the known planet $\epsilon$ Eri b. The Clio camera
62: itself with MMTAO may be capable of detecting $\epsilon$ Eri b
63: at its 2010 apastron, given a multi-night observing campaign. Clio appears
64: to be the only currently existing AO imager that has a realistic
65: possibility of detecting $\epsilon$ Eri b.
66: \end{abstract}
67:
68: \keywords{planetary systems, debris disks,
69: techniques: IR imaging, stars: individual: \objectname
70: {Vega}, \objectname{GJ 144}, \objectname{$\epsilon$ Eri}}
71:
72: \section{Introduction}
73: Early space based observations with the IRAS satellite identified four bright,
74: nearby stars with strong IR excesses: $\beta$ Pic, Vega, Fomalhaut, and
75: $\epsilon$ Eri \citep{IRASv1,IRASa1,IRASa2,IRASa3}. The only reasonable
76: explanation for these excesses is that
77: the systems contain substantial dust, which is warmed by starlight until
78: it radiates brightly in the IR because of the large total surface
79: area of its numerous small grains (see for example \citet{backman, HD4796, deller}).
80:
81: The dust in these systems cannot be primordial but must be
82: continually generated by the grinding down of larger bodies such as
83: asteroids \citep{backman, HD4796, deller}.
84: The stars are therefore said to have `debris disks'. The clear implication
85: is that each of these stars has at least an asteroid belt, and probably
86: a more extensive planetary system, because it is unlikely that an asteroid belt
87: would form without planets also forming, or that it would continue
88: to grind down without ongoing gravitational stirring due to planets.
89:
90: Theoretical models (e.g. \citet{bur} and \citet{bar}) predict that
91: it should be possible to make direct
92: images of giant planets orbiting nearby, young stars,
93: using the current generation of large ground-based telescopes with
94: adaptive optics (AO). These observations are only possible at near infrared
95: wavelengths from about 1-5 $\mu$m, where giant planets are self-luminous due
96: to the gravitational energy converted to internal heat in their formation
97: and subsequent slow contraction. Because giant planets radiate this
98: energy away over time, they become cooler and fainter as they age.
99: The youngest nearby stars are therefore the most promising targets for AO
100: surveys attempting to image self-luminous giant planets.
101:
102: Each of the four debris-disk stars discovered using IRAS is relatively
103: young, so orbiting giant planets might be detectable if any exist. We
104: have imaged the two stars most easily observable from Northern Hemisphere
105: sites: Vega and $\epsilon$ Eridani. Vega's age is about 0.3 Gyr
106: \citep{agevega}, while the age of $\epsilon$ Eri is about
107: 0.56 Gyr \citep{fischer}.
108: Besides the dust-dispersion timescale
109: argument mentioned above for the existence of planetary systems around
110: these stars, asymmetries in the dust distributions around each have led to hypotheses
111: that the dust is being gravitationally sculpted by giant planets orbiting
112: at large distances \citep{ozernoy,quillen,wyatt,wilner,deller,marsh}.
113: In the case of Vega there are suggestions that the dust may reveal the mass
114: and approximate position of a giant planet \citep{wilner,deller}. For
115: $\epsilon$ Eridani, in addition to evidence for a planet in a distant
116: orbit that may be sculpting the dust \citep{deller,benedict}, there is the
117: the radial velocity and astrometric detection of the closer-in planet
118: $\epsilon$ Eri b \citep{benedict}. The combination of radial
119: velocity and astrometry
120: observations permits a full orbital solution yielding ephemerides
121: for the separation and position angle of $\epsilon$ Eri b \citep{benedict},
122: making this the most promising case yet where attempts to image a
123: known extrasolar planet can target a specific location.
124:
125: Most imaging searches for extrasolar planets to date have used either the
126: $H$ band (1.5 - 1.8 $\mu$m) or other filters in the same wavelength regime (see for
127: example \citet{hsttwa7a, masciadri, sdi1, epsindia, biller, GDPS}).
128: The magnitude vs. mass tables of \citet{bar} and the theoretical
129: spectra of \citet{bur} show clearly why the $H$ band is usually chosen:
130: giant planets are predicted to be very bright at these wavelengths, much brighter than black bodies
131: at their effective temperatures. Detector formats are large, technology well
132: developed, and sky backgrounds faint at the $H$ band relative to longer wavelengths.
133:
134: However, theoretical models indicate that planet/star flux ratios are much more
135: favorable at the longer wavelength $L'$ and $M$ bands (3.4-4.1 $\mu$m and 4.5-5.0 $\mu$m, respectively).
136: For planets at sufficiently large separations, or planets orbiting faint stars, the
137: planet/star flux ratio is not relevant. Rather, it is the planet's brightness relative
138: to the sky background and/or detector read noise that matters. In this regime
139: the very high sky background in the $L'$ and $M$ bands prevents them from being
140: as sensitive as the $H$ band regime. However, close to very bright
141: stars the background becomes irrelevant and only the planet/star flux ratio
142: matters. Under these circumstances using the longer wavelengths makes sense.
143:
144: Vega is a magnitude 0.0 standard star and is among the brightest stars in the sky at
145: almost any wavelength. $\epsilon$ Eri, while not impressive at visible wavelengths, is
146: a very bright magnitude 1.9 at $H$ band. The stars are therefore excellent targets
147: for Clio, an $L'$ and $M$ band optimized AO camera that had its first light
148: on the MMT in June 2005 \citep{oldvega}. We have made deep $\sim$1 hour
149: integrations in both the $L'$ and $M$ bands on both stars. Our $M$-band
150: observations are the deepest ground-based images yet published in this band.
151:
152: In Section \ref{obs} we present our observations and data analysis strategy.
153: In Section \ref{sensanal}, we describe our methods of analyzing
154: our sensitivity, and present our sensitivity results. Blind sensitivity tests
155: in which simulated planet images were inserted directly into the raw data show that
156: we have obtained 100\% completeness for sources at 10$\sigma$ signficance, 77\% completeness for 7$\sigma$ sources, and 41\%
157: completeness for 5$\sigma$ sources, where $\sigma$ is an estimate of the
158: RMS noise amplitude in the image at the spatial scale of the PSF core (the
159: relevant scale for detection of faint point sources). We note
160: that no other planet-imaging papers to date present such careful blind tests
161: in their sensitivity analyses, and that the fact that our tests result in somewhat
162: lower completeness values at each significance level than might have been expected
163: suggests such tests should always
164: be attempted and may result in a need to revise some sensitivity estimates to more
165: conservative values.
166:
167: In Section \ref{vega} we compare the sensitivity we have obtained around
168: Vega to that of other deep
169: observations of Vega, and to the expected brightness of planets that have been hypothesized to explain the
170: dust distribution. In Section \ref{eri}, we present the same comparisons for
171: $\epsilon$ Eri, and in Section \ref{conclusion} we present the conclusions of our
172: study.
173:
174: \section{Observations and Data Analysis} \label{obs}
175:
176: \subsection{The Instrument} \label{clio_ins}
177:
178: The Clio instrument we used for our observations has been well described
179: elsewhere (\citet{freed},
180: \citet{suresh}, and \citet{oldvega}). We present only a brief overview here.
181:
182: The MMT AO system delivers a lower thermal background than other AO systems because it
183: uses the world's first deformable secondary mirror, thereby avoiding the multiple warm-mirror
184: reflections (each adding to the thermal background) that are needed in AO systems
185: where the deformable mirror is not the secondary. This unique property makes the
186: MMT ideal for AO observations in wavelengths such as the $L'$ and $M$ bands
187: that are strongly affected by thermal glow. Clio was developed to take advantage of
188: this to search for planets in these bands. It saw first light as a simple
189: imager offering F/20 and F/35 modes. The design allowed for coronagrapic
190: capability, which has since been developed \citep{phaseplate} but was not fully operational
191: at the time of our Vega and $\epsilon$ Eri observations. In the F/20 mode
192: we used for all the observations of Vega and $\epsilon$ Eri, Clio's
193: field of view is 15.5$\times$12.4 arcseconds. Its plate scale
194: is $0.04857 \pm 0.00003$ arcseconds per pixel, which gives finer than Nyquist
195: sampling of the diffraction-limited PSF of the MMT in the $L'$ and $M$ bands.
196:
197: \subsection{Observing Strategy} \label{obsstrat}
198: We carry out $L'$ and $M$ band imaging with Clio using
199: the technique of nod-subtraction, in which we take images of our
200: target star in two different
201: telescope positions offset typically by about 5.5 arcsec, and then
202: subtract the images taken in one position from those taken in the
203: other to remove artifacts from the bright sky background and
204: detector imperfections. Since the star is present on images taken
205: in both positions, both provide useful
206: science data. Nod-subtraction does result in a dark negative image of the star
207: reducing the sensitivity in part of each image, but the area affected is
208: fractionally small, far (5.5 arcsec) from the star, where planets are less
209: likely to be found, and can be placed away from objects of potential interest
210: by a good choice of the nod direction. We also have alternative ways of
211: processing nodded data that do away with the dark images entirely.
212:
213: We typically nod the telescope every 2-5 minutes, which appears to be fast enough
214: that variations in the sky background are sampled well enough to be
215: essentially removed. We take 5 or 10 images in each nod position,
216: each of which typically represents
217: about 20 seconds worth of data. A full data set consists of 100-500
218: such images.
219:
220: We choose the exposure for most of the images so that the sky background level
221: is about 70\% of the detector full-well. At such exposure times the cores
222: of bright stars such as Vega and $\epsilon$ Eri are saturated, but
223: optimal sensitivity is obtained to faint point sources beyond the
224: saturation radii. When possible, we interleave a few nod cycles of
225: shorter exposures yielding unsaturated star images into the sequence
226: of longer exposure images. This allows us to measure the unsaturated
227: PSF under the exact conditions of a particular observing sequence.
228: We achieved the PSF measurement with $\epsilon$ Eri, but
229: Vega proved too bright for us reasonably to obtain unsaturated images. We
230: used other stars observed close in time to our Vega observations to
231: provide a reference PSF for the Vega data.
232:
233: Tables \ref{tab:vobs01} and \ref{tab:vobs02} give details of our observations.
234: The June 2006 $M$ band Vega observations had
235: far higher sky noise than the April 2006 data, possibly because of the higher
236: thermal background during warm summer weather, and therefore were not used in
237: calculating the final sensitivity.
238:
239: \clearpage
240: \begin{deluxetable}{llcrrr}
241: \tablewidth{0pt}
242: \tablecolumns{6}
243: \tablecaption{Observations of Science Targets: Basic Parameters \label{tab:vobs01}}
244: \tablehead{\colhead{Star} & \colhead{Date Obs} & \colhead{Band} & \colhead{Clio int(msec)} & \colhead{Coadds} & \colhead{\# Images}}
245: \startdata
246: Vega & April 12, 2006 & $L'$ & 2000 & 10 & 160 \\
247: Vega & April 13, 2006 & $M$ & 200 & 90 & 110 \\
248: Vega & June 10, 2006 & $M$ & 100 & 50 & 558 \\
249: Vega & June 11, 2006 & $M$ & 120 & 100 & 180 \\
250: $\epsilon$ Eri & September 09, 2006 & $M$ & 130 & 100 & 180\\
251: $\epsilon$ Eri & September 11, 2006 & $L'$ & 1500 & 15 & 184 \\
252: \enddata
253: \tablecomments{Clio int(msec) refers to the nominal single-frame
254: exposure time in Clio. The integrate-while-reading mode used
255: in high efficiency science imaging causes the true single-frame
256: exposure time to be about 59.6 msec longer than the nominal exposures
257: listed here. Coadds is the number of frames internally coadded
258: by Clio to produce a single 2-D FITS image.}
259: \end{deluxetable}
260:
261: \begin{deluxetable}{lccccr}
262: \tablewidth{0pc}
263: \tablecolumns{5}
264: \tablecaption{Observations of Science Targets: Data Acquired \label{tab:vobs02}}
265: \tablehead{\colhead{Star} & \colhead{Date} & \colhead{Band} & \colhead{Exposure(sec)} & \colhead{Mean Airmass} & \colhead{Rotation}}
266: \startdata
267: Vega & April 12, 2006 & L' & 3295.4 & 1.018 & $80.63^{\circ}$ \\
268: Vega & April 13, 2006 & M & 2570.0 & 1.026 & $36.39^{\circ}$ \\
269: Vega & June 10, 2006 & M & 4452.8 & 1.034 & $72.36^{\circ}$ \\
270: Vega & June 11, 2006 & M & 3232.8 & 1.054 & $25.53^{\circ}$ \\
271: $\epsilon$ Eri & September 09, 2006 & M & 3412.8 & 1.334 & $23.41^{\circ}$ \\
272: $\epsilon$ Eri & September 11, 2006 & L' & 4304.5 & 1.342 & $36.92^{\circ}$ \\
273: \enddata
274: \tablecomments{The observations in June were plagued with high
275: sky noise, which may have been due to the higher thermal background
276: during warm summer weather. Adding them to the
277: April M-band data on Vega did not significantly increase the
278: sensitivity to objects far from the star, though in the speckle
279: dominated regime near the star, the sensitivity did increase
280: by about 40 \%.}
281: \end{deluxetable}
282: \clearpage
283:
284: \subsection{Data Analysis} \label{dataan}
285: Our Clio image processing pipeline
286: will be described in more detail in a future paper.
287: Here we briefly state that our baseline processing involves dark
288: subtraction; flat fielding; nod subtraction; several iterations
289: of different types of deviant (`hot') pixel removal; a pattern
290: noise correction (Figure \ref{fig:tileim} Panel B shows
291: an example image at this stage); shifting, rotation, and
292: zeropadding in a single
293: bicubic spline operation; final stacking; and then unsharp masking
294: of the stacked image using a gaussian kernel 3-4 times wider than
295: the PSF.
296:
297: For the final image stacks we use a creeping mean algorithm with
298: 20\% rejection. This algorithm works by finding the
299: mean of all values for a given pixel, rejecting the most deviant one,
300: finding the new mean, rejecting the new maximally deviant value,
301: etc, until the specified rejection fraction is reached. For data sets
302: where ghosts or other artifacts can render a large fraction of the
303: data at a given location deviant, the creeping mean produces a cleaner
304: final stack than the median. Figure \ref{fig:tileim} Panel A shows
305: an example of a raw image; Panel B shows a partially processed version
306: of the same image just before shifting and rotation;
307: and Panels C and D show examples of final stacked images after
308: unsharp masking.
309:
310: In addition to the image made using our baseline processing,
311: we make images using two types of more advanced processing,
312: one that avoids the negative star
313: images from standard nod subtraction at the cost of slightly
314: increased noise, and one that includes
315: subtraction of the stellar PSF using a technique similar
316: to the angular differential imaging (ADI) described by \citet{marois}.
317: We use all three images when we search for companions, since
318: the detection of a faint companion on images processed in more
319: than one way increases the likelihood that it is real. We also
320: construct a separate sensitivity map for each of the three
321: differently-processed
322: master images, and then combine them into a single master sensitivity
323: map. Since the different processing methods obtain optimal sensitivity
324: at different locations, we set the sensitivity at a given location
325: on the master map to the best sensitivity obtained at that location
326: on any of the three separate maps. Details on how the separate
327: sensitivity maps themselves are made may be found in
328: Section \ref{sensanal}.
329:
330: \clearpage
331: \begin{figure*}
332: \plotone{f1.eps}
333: \caption[Examples of our Vega $M$ Band Images] {(A)
334: Raw single $M$ band image of Vega. (B) Nod subtracted,
335: processed version of the same image just before shift and rotation.
336: Contrast stretched 100 times more than in Panel A. (C) Final master
337: $M$ band image of Vega, consisting of 110 Panel B-like
338: images, shifted, rotated, and coadded. Contrast stretched 10 times more
339: than in Panel B. (D) Like Panel C, but with fake planets added
340: to the raw data. The field shown in all panels is about 15.5$\times$ 12.5
341: arcseconds. The unsharp masking which has removed the bright
342: stellar halo in Panels C and D is responsible for
343: the black spaces between the inner diffraction rings.
344: The noisy region at bottom right in these panels is due
345: to the negative stellar images from nod subtraction.
346: }
347: \label{fig:tileim}
348: \end{figure*}
349:
350:
351: Intensive image processing such as we describe here is often
352: used for AO planet search data, where high contrast
353: is required and artifacts must be aggressively removed.
354: Such processing can remove flux from the faint point sources
355: whose detection it is intended to facilitate. Careful tests of
356: our processing methods, however, indicate that the flux loss
357: from a faint PSF is no more than about 10\%, and appears to be close
358: to zero in most cases.
359:
360: \section{Sensitivity Measurements and Source Detection Tests} \label{sensanal}
361:
362: \subsection{Sensitivity Estimation}
363: We create a sensitivity map from each stacked master image produced
364: by the processing outlined above. We are careful to measure the
365: noise at the relevant spatial scale -- that is, the scale of the
366: PSF. Our method requires an unsaturated star image taken under
367: similar conditions to the science data, and therefore representing
368: a good estimate of the PSF. We perform a two-parameter least
369: square fit centered on each pixel in the master science image
370: in turn, with the two parameters being the amplitude of a PSF
371: centered on that pixel, and a constant background value. This
372: fit is performed within a disk of six pixel radius about each given
373: pixel. The best-fit PSF amplitude from the fit
374: centered on each pixel of the master science image becomes the
375: value of the corresponding pixel of a new image:
376: the PSF amplitude map. This PSF amplitude
377: map image is essentially the result of PSF-fitting photometry
378: centered in turn on every pixel in the original master science image.
379: This PSF-fitting has, of course, mostly measured simply noise -- the
380: point is that it has measured the noise at the spatial scale of the PSF.
381:
382: Our method may be expected to produce results somewhat similar to
383: the `matched filter' technique (see for example \citet{matchfilt}),
384: although the least-square fitting that we use is mathematically
385: more sophisticated than the straightforward convolution used in
386: a matched filter. The most obvious advantage of our method is that
387: it automatically fits and removes any slowly-varying background (since
388: our least square fit determines a separate background value within
389: the disk centered on each pixel), while an ordinary matched filter
390: requires the separate construction of a background model.
391:
392: The noise in the amplitude map constructed by our PSF-fitting
393: accurately reflects the PSF-scale noise in the original image -- that
394: is, the noise at the spatial frequencies relevant for the
395: detection of real point sources. We calculate the
396: sensitivity at every point in the original image by computing the RMS
397: in an 8 pixel radius aperture about that point on the PSF amplitude map
398: (for regions too close to the star, where a circular aperture would not produce
399: accurate results, we use a 45-pixel long arc at constant radius from the
400: star instead). Note that since the data are only slightly oversampled,
401: both the 8 pixel radius disk and the 45-pixel long arc span many resolution
402: elements or speckles. Calculating the RMS on the PSF amplitude map
403: rather than the original master image takes into account
404: spatial correlations between pixels (that is, the fact that the
405: noise in adjacent pixels is not independent). This is a
406: large effect in the case of speckle noise.
407:
408: We note that many previous planet-imaging papers have not
409: used a sensitivity estimator mathematically able to account
410: for correlated noise in speckles --- or, at least, have not devoted
411: sufficient space to the description of their sensitivity
412: estimator to make it clear whether or not it properly
413: measures correlated noise. Estimators that contain an implicit
414: mathematical assumption that the noise is independent from
415: one pixel to the next can significantly overestimate the
416: sensitivity in speckle-dominated regions close to the star.
417: The careful design, description, and testing of sensitivity estimators is
418: an important task, because in the case of a non-detection all
419: the science rests on upper limits set through sensitivity estimation.
420: The only observational planet-imaging paper we are aware of prior to
421: this work in which a sensitivity estimator able to account for
422: correlated noise is clearly described is \citet{GDPS}. (However,
423: we may safely assume that \citet{marois} used the same estimator
424: as \citet{GDPS}. \citet{hinkley} also used, and carefully
425: described, such an estimator in their paper to set limits on
426: brown dwarfs in close orbits around Vega.)
427:
428: \subsection{Testing the Sensitivity Estimator}
429:
430: To test the accuracy of our sensitivity estimator, we conducted blind
431: tests in which fake planets were inserted into the raw
432: data. The altered images were then processed in exactly the same way
433: as the original raw data, and the `planets' were
434: detected using both automatic and manual methods by an experimenter
435: who knew neither their positions nor their number. These planets were inserted
436: at fixed nominal significance levels of 10$\sigma$, 7$\sigma$, and 5$\sigma$ based
437: on the sensitivity map. We conducted such tests for each of
438: our four data sets (the $L'$ and $M$ band data sets for each of the two
439: stars). The final result of each test was that every inserted planet
440: was classified as `Confirmed', `Noticed', or `Unnoticed'.
441: `Confirmed' means the source was confidently detected and
442: would certainly be worthy of long-exposure followup observations at the MMT.
443: If a source is detected with this confidence level in an unaltered data set,
444: there is no significant doubt it is a real object. In calculating our
445: completeness, we count only confirmed sources as true detections.
446: `Noticed' means the source was flagged by our automatic detection algorithm,
447: or noticed as a possible real object during the purely manual phase
448: of planet-searching, but could not be confirmed beyond reasonable doubt.
449: Many spurious sources are `Noticed' whereas the false-positive rate for
450: `Confirmed' detections is extremely low, with none for any of the data
451: sets discussed here. `Unnoticed' means a fake planet was not automatically
452: flagged or noticed manually.
453:
454: The end result of the four blind sensitivity tests was that at 10$\sigma$, 50 of
455: 50 total inserted planets were confirmed, giving us 100\% completeness
456: to the limits of the statistical accuracy of the test. At 7$\sigma$,
457: 23 of 30 total inserted sources were confirmed, giving us 77\% completeness,
458: and 29 of the 30 sources were at least noticed.
459: At 5$\sigma$, 11 of 27 total inserted planets were confirmed, for 41\%
460: completeness, and 23 of the 27 sources were at least noticed.
461: In addition to the completeness levels for confirmed sources, the percentages
462: of fake planets that were at least noticed is of potential
463: interest for setting limits: 100\% of 10$\sigma$ sources, 97\% of 7$\sigma$
464: sources, and 85\% of 5$\sigma$ sources were at least noticed.
465: We note that if we had quoted 5$\sigma$ sensitivities without conducting a
466: blind sensitivity test we would have significantly overestimated our true
467: high-completeness sensitivity. Most papers in the field of
468: planet-imaging surveys do in
469: fact quote 5$\sigma$ limits, and do not verify their validity by
470: a blind test.
471:
472: In our sensitivity experiments there were no false positives among
473: the `Confirmed' sources. Many spurious sources were classified as
474: `Noticed', which is why we do not count `Noticed' sources as detections
475: for completeness purposes. The conclusion of our fake planet experiments
476: is that our detection strategy has an extremely low
477: false alarm probability, and delivers the completeness values given above.
478: The fact that a large majority of low significance sources were noticed,
479: even if not confirmed, indicates that upper limits stronger than those
480: implied by our formal completeness values may be set on planets in clean
481: regions of an image where no spurious sources were noticed.
482:
483: \subsection{Final Sensitivity Results} \label{finsens}
484: We have converted the master sensitivity maps described above into magnitude
485: contour images, with 10$\sigma$ sensitivity values shown. We quote
486: sensitivities in apparent magnitudes based on observations of photometric
487: standard stars (from \citet{leggett}), rather than giving $\Delta$-magnitudes
488: relative to the primary. We present our $L'$ and $M$ band Vega results in
489: Figures \ref{fig:vegamap}
490: and \ref{fig:vegammap}, with the approximate position of the hypothetical planet from
491: \citet{wilner} marked with a white `X'. Figures \ref{fig:erilmap} and
492: \ref{fig:erimmap} present the analogous results for $\epsilon$ Eri.
493: Our Vega $M$ band observation is the deepest ground-based $M$
494: band observation yet published.
495:
496: \clearpage
497: \begin{figure*}
498: \plotone{f2.eps}
499: \caption[Sensitivity contour map for the Vega $L'$ observations] {
500: 10$\sigma$ sensitivity contour map for our Vega $L'$ observations in magnitudes.
501: The grid squares superposed on the figure for astrometric reference are 2x2 arcsec.
502: The approximate location of the hypothetical planet from \citet{wilner}
503: is marked with a white `X'. The best areas in this image give sensitivity
504: to objects fainter than $L'$ = 15.5. The numbers at the top of the colorbar
505: give the sensitivity of each contour in magnitudes, while the numbers at
506: the bottom give the equivalent value in MJ, where applicable, based on the
507: models of \citet{bur} with the age set to 0.3 Gyr.
508: }
509: \label{fig:vegamap}
510: \end{figure*}
511:
512: \begin{figure*}
513: \plotone{f3.eps}
514: \caption[Sensitivity contour map for the Vega $M$ band observations] {
515: 10$\sigma$ sensitivity contour map for our Vega $M$ observations in magnitudes.
516: The grid squares superposed on the figure for astrometric reference are 2x2 arcsec.
517: The approximate location of the hypothetical planet from \citet{wilner}
518: is marked with a white `X'. The numbers at the top of the colorbar
519: give the sensitivity of each contour in magnitudes, while the numbers at
520: the bottom give the equivalent value in MJ, where applicable,
521: based on the models of \citet{bur} with the age set to 0.3 Gyr. }
522: \label{fig:vegammap}
523: \end{figure*}
524:
525: \begin{figure*}
526: \plotone{f4.eps}
527: \caption[Sensitivity contour map for our $\epsilon$ Eri $L'$ band observations] {
528: 10$\sigma$ sensitivity contour map for our $\epsilon$ Eri $L'$ observations in
529: magnitudes.The best areas in this image give sensitivity
530: to objects fainter than $L'$ = 15.5.
531: The grid squares superposed on the figure for astrometric reference are 2x2 arcsec. The numbers at the top of the colorbar
532: give the sensitivity of each contour in magnitudes, while the numbers at
533: the bottom give the equivalent value in MJ, where applicable,
534: based on the models of \citet{bur} with the age set to 0.56 Gyr.}
535: \label{fig:erilmap}
536: \end{figure*}
537:
538: \begin{figure*}
539: \plotone{f5.eps}
540: \caption[Sensitivity contour map for our $\epsilon$ Eri $M$ band observations] {
541: 10$\sigma$ sensitivity contour map for our $\epsilon$ Eri $M$ observations in
542: magnitudes. The grid squares superposed on the figure for astrometric reference are 2x2 arcsec. The numbers at the top of the colorbar
543: give the sensitivity of each contour in magnitudes, while the numbers at
544: the bottom give the equivalent value in MJ, where applicable,
545: based on the models of \citet{bur} with the age set to 0.56 Gyr.}
546: \label{fig:erimmap}
547: \end{figure*}
548: \clearpage
549:
550: We have further translated our master sensitivity
551: map for each data set into 10$\sigma$ sensitivity curves, and plotted them
552: in Figures \ref{fig:vegal_mag} through \ref{fig:erim_mass}. Our sensitivity
553: varies azimuthally as well as radially due to the negative nod subtraction
554: images, ghosts, and the different distances to the edge of the valid
555: data region in different directions. Therefore we have
556: computed both the 50th and 90th percentile sensitivities at each radius.
557: Both are shown, with the 50th percentile, of course, indicating
558: our median sensitivity and the 90th percentile indicating our sensitivity
559: in the cleanest 10 \% of the image at a given radius from the star.
560: We have also indicated the `Confirmed',
561: `Noticed', and `Unnoticed' planets from our sensitivity tests with
562: appropriate symbols. The sensitivity in these plots increases
563: with separation from the star as one would expect, but
564: then decreases again as the edge of the good data region (ie, useful field on
565: the master stacked images) is reached. The noise goes up at
566: the edge of the useful field because, due to the shifts and rotations
567: required to register the images, the coverage (number of images
568: supplying data to a given pixel) goes down near the edge of the field.
569:
570: Figures \ref{fig:vegal_mag} and \ref{fig:vegal_mass} show the sensitivity we
571: obtained in our $L'$ observations of Vega, first in `observational' units
572: of sensitivity in magnitudes vs. separation in arcsec, and then in `physical'
573: units of MJ (based on the \citet{bur} models, and adopting an 0.3 Gyr age
574: for Vega \citep{agevega}) vs. projected separation in AU.
575: Vega has approximately magnitude 0.0 at every band, so the magnitudes in
576: Figure \ref{fig:vegal_mag} correspond approximately to $\Delta$-magnitude values.
577: Figures \ref{fig:vegam_mag} and \ref{fig:vegam_mass} show the sensitivity
578: obtained in our $M$ band observations of Vega, following exactly the same
579: conventions as the $L'$ figures that precede them.
580:
581: Comparison of \ref{fig:vegal_mass} and \ref{fig:vegam_mass} shows that the
582: $L'$ and $M$ band results provided similar sensitivity to planets around
583: Vega. The $M$ band results are slightly better, especially at
584: smaller separations. This is not surprising, because the predicted
585: planet/star flux ratio is even more favorable at $M$ band than at $L'$. Also,
586: MMTAO, like all AO systems, delivers better Strehl ratios at longer
587: wavelengths, so the PSF subtraction is more effective at $M$ band than
588: at $L'$.
589:
590: Figures \ref{fig:eril_mag} through \ref{fig:erim_mass} show
591: the sensitivity of our $L'$ and $M$ band observations of $\epsilon$ Eri,
592: following the same conventions as the Vega figures that precede them.
593: For $\epsilon$ Eri we have adopted an age of 0.56 Gyr \citep{fischer}.
594: Note that the magnitudes in Figures \ref{fig:eril_mag} and \ref{fig:erim_mag}
595: may be converted to $\Delta$-magnitudes by subtracting the $L'$ magnitude
596: of $\epsilon$ Eri, which is about 1.72 (the $L'$ - $M$ color of the star
597: is near zero).
598:
599: Comparing Figure \ref{fig:eril_mass} with
600: \ref{fig:erim_mass} shows that for $\epsilon$ Eri the advantage of the
601: $M$ band over $L'$ is considerably more than for Vega. The fundamental reason
602: for this is that $\epsilon$ Eri is closer to us than Vega. This is
603: an important point we will refer back to later: the smaller the
604: distance to a star system, the more favorably long wavelength planet
605: search observations of the system will compare to short wavelength ones.
606: There are several logical links in the explaination of this observational fact.
607: First, intrinsically low-luminosity planets can be detected only
608: in the nearest systems. Second, low-luminosity planets have low
609: $\mathrm{T_{eff}}$. Third, low $\mathrm{T_{eff}}$ planets have
610: red $L' - M$ colors. Therefore, the faintest detectable planets
611: will be more red in nearby systems than in distant ones, and it
612: follows that longer
613: wavelength observations (i.e., $M$ band) will perform best relative to shorter
614: wavelength ones (i.e., $L'$) on the very nearest stars. This conclusion is
615: most obvious when one considers background-limited regions
616: of images at large separations from the star, but it applies
617: in the contrast limited regime as well.
618:
619: \clearpage
620: \begin{figure*}
621: \plotone{f6.eps}
622: \caption[Sensitivity of the Vega $L'$ Band Observations in Magnitudes] {
623: 10$\sigma$ sensitivity of our Vega $L'$ band observations in magnitudes, plotted
624: against separation in arcseconds. The 50th and 90th percentile
625: sensitivity curves are shown, along with fake planets from the
626: blind sensitivity test. The star symbols are fake planets that were
627: confidently detected; the diamonds are those that were suspected
628: but not confirmed, and the tiny triangle represents the only
629: fake planet that was not at least suspected.}
630: \label{fig:vegal_mag}
631: \end{figure*}
632:
633: \begin{figure*}
634: \plotone{f7.eps}
635: \caption[Sensitivity of the Vega $L'$ Band Observations in MJ] {
636: Sensitivity of our Vega $L'$ band observations in terms of the minimum mass
637: for a planet detectable at the 10 $\sigma$ level in MJ, plotted
638: against projected separation in AU.
639: The magnitude-mass conversion was done using the \citet{bur} models for
640: an age of 0.3 Gyr.
641: The 50th and 90th percentile sensitivity curves are shown, along with fake planets from the
642: blind sensitivity test. The star symbols are fake planets that were
643: confidently detected; the diamonds are those that were suspected
644: but not confirmed, and the tiny triangle represents the only
645: fake planet that was not at least suspected.}
646: \label{fig:vegal_mass}
647: \end{figure*}
648:
649: \begin{figure*}
650: \plotone{f8.eps}
651: \caption[Sensitivity of the Vega $M$ Band Observations in Magnitudes] {
652: 10$\sigma$ sensitivity of our Vega $M$ band observations in magnitudes, plotted
653: against separation in arcseconds. The 50th and 90th percentile
654: sensitivity curves are shown, along with fake planets from the
655: blind sensitivity test. The star symbols are fake planets that were
656: confidently detected; the diamonds are those that were suspected
657: but not confirmed, and the tiny triangles are those that were
658: not suspected.}
659: \label{fig:vegam_mag}
660: \end{figure*}
661:
662: \begin{figure*}
663: \plotone{f9.eps}
664: \caption[Sensitivity of the Vega $M$ Band Observations in MJ] {
665: Sensitivity of our Vega $M$ band observations in terms of the minimum mass
666: for a planet detectable at the 10 $\sigma$ level in MJ, plotted
667: against projected separation in AU.
668: The magnitude-mass conversion was done using the \citet{bur} models for
669: an age of 0.3 Gyr.
670: The 50th and 90th percentile sensitivity curves are shown, along with fake planets from the
671: blind sensitivity test. The star symbols are fake planets that were
672: confidently detected; the diamonds are those that were suspected
673: but not confirmed, and the tiny triangles are those that were
674: not suspected.}
675: \label{fig:vegam_mass}
676: \end{figure*}
677:
678:
679: \begin{figure*}
680: \plotone{f10.eps}
681: \caption[Sensitivity of the $\epsilon$ Eri $L'$ Band Observations in Magnitudes] {
682: 10$\sigma$ sensitivity of our $\epsilon$ Eri $L'$ band observations in magnitudes, plotted
683: against separation in arcseconds. The 50th and 90th percentile
684: sensitivity curves are shown, along with simulated planets from the
685: blind sensitivity test. The star symbols are fake planets that were
686: confidently detected; the diamonds are those that were suspected
687: but not confirmed, and the tiny triangle represents the only
688: fake planet that was not at least suspected.}
689: \label{fig:eril_mag}
690: \end{figure*}
691:
692: \begin{figure*}
693: \plotone{f11.eps}
694: \caption[Sensitivity of the $\epsilon$ Eri $L'$ Band Observations in Magnitudes] {
695: Sensitivity of our $\epsilon$ Eri $L'$ band observations in terms of the minimum mass
696: for a planet detectable at the 10 $\sigma$ level in MJ, plotted
697: against projected separation in AU. The magnitude-mass conversion
698: was done using the \citet{bur} models for an age of 0.56 Gyr.
699: The 50th and 90th percentile
700: sensitivity curves are shown, along with fake planets from the blind
701: sensitivity test. The star symbols are fake planets that were
702: confidently detected; the diamonds are those that were suspected
703: but not confirmed, and the tiny triangle represents the only
704: fake planet that was not at least suspected.}
705: \label{fig:eril_mass}
706: \end{figure*}
707:
708: \begin{figure*}
709: \plotone{f12.eps}
710: \caption[Sensitivity of the $\epsilon$ Eri $M$ band Observations in Magnitudes] {
711: 10$\sigma$ sensitivity of our $\epsilon$ Eri $M$ band observations in magnitudes, plotted
712: against separation in arcseconds.
713: The 50th and 90th percentile
714: sensitivity curves are shown, along with simulated planets from the
715: blind sensitivity test. The star symbols are fake planets that were
716: confidently detected; the diamonds are those that were suspected
717: but not confirmed. In the sensitivity test for this data set all
718: of the fake planets were at least suspected.}
719: \label{fig:erim_mag}
720: \end{figure*}
721:
722: \begin{figure*}
723: \plotone{f13.eps}
724: \caption[Sensitivity of the $\epsilon$ Eri $M$ band Observations in MJ] {
725: 10$\sigma$ sensitivity of our $\epsilon$ Eri $M$ band observations in terms
726: of minimum detectable planet mass in MJ, plotted
727: against projected separation in AU. The magnitude-mass conversion
728: was done using the \citet{bur} models for an age of 0.56 Gyr.
729: The 50th and 90th percentile
730: sensitivity curves are shown, along with fake planets from the blind
731: sensitivity test. The star symbols are fake planets that were
732: confidently detected; the diamonds are those that were suspected
733: but not confirmed. In the sensitivity test for this data set all of
734: the fake planets were at least suspected.}
735: \label{fig:erim_mass}
736: \end{figure*}
737: \clearpage
738:
739: \section{Vega: Comparison with Other Studies, and Upper Limits for Hypothetical Planets} \label{vega}
740:
741: \subsection{Comparing Our Sensitivity with Other Studies}
742: We have not attempted to compare our Vega results with an exhaustive list
743: of all previous attempts to image planets or other faint objects around
744: Vega. Instead, we have chosen two of the best previous results.
745: First, the $H$ band imaging results of \citet{yoichi}, and second,
746: the narrow band, $H$-regime images of \citet{marois}. The latter
747: presents the most sensitive images yet published
748: for substellar companions at 3-10 arcsecond separations from Vega.
749:
750: Before comparing our sensitivities with these other observations a
751: brief discussion about the different sensitivity estimation techniques
752: used by the respective observers is in order. As described above, in this
753: work we have used an estimator able to account for correlated noise,
754: we have performed blind tests of our sensitivity estimator, and we
755: have quoted 10$\sigma$ limits.
756:
757: \citet{yoichi} did not calculate sensitivity limits in terms of $\sigma$.
758: Instead, they calculated their sensitivities by performing numerous
759: tests in which they placed 4 planets into their data at a fixed separation
760: and $\Delta$-magnitude with respect to the primary. These tests differ
761: from our own blind sensitivity tests in that the locations of the
762: \citet{yoichi} fake planets were known, and fixed from one test
763: to the next. \citet{yoichi} set their sensitivity at each
764: separation to the faintest $\Delta$-magnitude at which at least
765: 3 of the 4 planets were recovered by their automatic
766: detection algorithm. Therefore the \citet{yoichi} sensitivities
767: correspond to planet brightness
768: values at which they had at least 75\% completeness, with an unknown
769: false-positive rate. Although it appears the completeness level corresponding
770: to the \citet{yoichi} sensitivities corresponds better to our 7$\sigma$ level,
771: we have conservatively chosen to compare the \citet{yoichi} sensitivity
772: values to our own 10$\sigma$ results without alteration.
773:
774: \citet{marois} do not explain how their quoted 5$\sigma$ sensitivity
775: limits are obtained. We assume, however, that they used the same method
776: as \citet{GDPS}, another planet imaging survey by a very similar set
777: of authors, presenting observations made with the same telescope, instrument,
778: and observing and analysis stategies. \citet{GDPS} set $\sigma$
779: limits using a sensitivity estimator carefully designed to
780: account for correlated noise. They also carefully account for
781: processing losses, but they do not present blind
782: sensitivity tests. Assuming that \citet{marois} used the same
783: good estimator and careful correction of processing losses, we
784: conservatively choose to consider their quoted 5$\sigma$ limits to
785: be comparable with our 7$\sigma$ limits. Based on this assumption we
786: transform them to 10$\sigma$ limits for comparison with out own.
787: We also adjust their limits
788: by a factor of 2 (0.753 mag) in the direction of greater sensitivity,
789: to scale from the planet-optimized narrowband filter they used to
790: the broadband $H$ filter. (\citet{GDPS} estimate this correction
791: at a factor between 1.5 and 2.5; we have used the mean value of 2.0.)
792:
793: Figure \ref{fig:vhbur} shows the sensitivities of our Vega $L'$ and $M$
794: band observations compared to those of \citet{yoichi} and \citet{marois}.
795: The magnitude limits, adjusted as described above, have been
796: converted to planet masses using the theoretical planet models of \citet{bur},
797: adopting the \citet{agevega} age of 0.3 Gyr.
798: We plot our 90th percentile 10$\sigma$ sensitivity values because the 90th
799: percentile curves are smoother and easier to interpret, and because
800: sensitivity at least this good can be obtained at a position angle of
801: choice by a well-tuned observing strategy. Although our observations are
802: more sensitive to planetary-mass objects
803: around Vega than the observations of \citet{yoichi}, the carefully processed
804: narrowband observations of \citet{marois} are more sensitive than ours by
805: 1.5-3 MJ at all separations beyond 3 arcseconds, which was their approximate saturation
806: radius. Inward of 3 arcsec our images are sensitive mainly to brown dwarfs
807: and and the most massive planets, while the other plotted observations
808: are saturated or very insensitive. However, in
809: the regime of higher masses and smaller separations than
810: covered by our figure, we note that LYOT project
811: $H$ band observations of Vega \citep{hinkley}
812: obtain sensitivity to massive brown dwarfs inward to about
813: 0.7 arcsec. Their observations appear to be sensitive
814: to lower mass brown dwarfs than ours inside of 1.5 arcsec,
815: while ours are more sensitive at 2 arcsec and farther out.
816:
817:
818: \clearpage
819: \begin{figure*}
820: \plotone{f14.eps}
821: \caption[Comparison of sensitivities obtained around Vega with different techniques (using the \citet{bur} models)]{
822: Comparison of the sensitivities obtained around Vega with different
823: techniques. Magnitude
824: sensitivities have been converted to planet mass limits in MJ using
825: the theoretical
826: models of \citet{bur} for an age of 0.3 Gyr. The dashed line is
827: the narrowband $H$-regime
828: result from \citet{marois}; the dotted line is the $H$ band result
829: from \citet{yoichi}, the gray continuous line is our 90th percentile
830: $L'$ result, and the black continuous line is our
831: 90th percentile $M$ band result.}
832: \label{fig:vhbur}
833: \end{figure*}
834: \clearpage
835:
836: It is interesting to note that the Figure \ref{fig:vhbur}
837: would look very different if we plotted $\Delta$-mag rather than
838: minimum detectable planet mass. The sensitivity of the $H$ regime
839: results of \citet{marois} would surpass the sensitivity of our
840: observations by a far greater
841: margin in $\Delta$-mag terms. At the $L'$ and $M$ bands the sky
842: background is far brighter than in the $H$ regime. Also, diffraction
843: limited resolution is several times lower, and the Airy pattern is
844: correspondingly larger in angular terms. The result is that
845: despite the cleaner, higher Strehl images offered by AO systems at
846: longer wavelengths, the $\Delta$-mag vs. angular separation curves at $L'$
847: and $M$ band are typically considerably less good than those in the $H$
848: regime. Because the planet/star flux ratios are so much better
849: in the $L'$ and $M$ bands, however, when we convert from
850: $\Delta$-magnitudes to planet masses the sensitivity gap closes
851: considerably, and in fact (as will be seen below in the case of
852: $\epsilon$ Eri) the longer wavelengths may turn out
853: to be more sensitive.
854:
855:
856: In terms of planet mass the \citet{marois} $H$-band regime observations
857: were more sensitive than our $L'$ and $M$ band results beyond 3 arcseconds,
858: but not by a huge margin. Theoretical
859: planet models are still somewhat uncertain because of the dearth of
860: observational constraints. $L'$ and $M$ band observations of bright stars
861: such as Vega make sense to diversify the investment of planet-imaging
862: effort and hedge the overall results against the possibility that
863: unexpected atmospheric chemistry, clouds, or evolutionary effects
864: (see for example \citet{faintJup}) cause planets to appear fainter in
865: $H$ band than current models predict. It is also possible that planets could be
866: fainter than predicted at the longer wavelengths, specifically $M$
867: \citep{L07}. However, the supression of $M$ band flux observed by
868: \citet{L07} applied only to objects with $\mathrm{T_{eff}}$ from 700-1300 K.
869: The situation for objects cooler than 700 K is unknown.
870: According to the \citet{bur} models, our Vega $M$ band observations
871: were sensitive to planets with $\mathrm{T_{eff}}$ below 400 K. Such objects may be too
872: cold to have the enhanced concentrations of $\mathrm{CO}$ to which \citet{L07}
873: attributed the $M$ band flux supression (see \citet{NCE}).
874:
875: Because they offer better
876: flux ratios relative to the primary star than shorter wavelengths,
877: the $L'$ and $M$ bands we have used are
878: optimal for detecting massive planets and low mass brown dwarfs
879: at small separations from Vega and other very bright stars.
880:
881: \subsection{Upper Limits at the Locations of Hypothetical Planets}
882: \citet{wilner} presents high-resolution submillimeter observations of
883: Vega which show two bright clumps arranged asymmetrically
884: relative to the star. He states that is very unlikely the clumps
885: could be background galaxies, and is essentially certain that they
886: are concentrations of dust in the Vega system. Further, the dust
887: could represent the remains of two different planetesimal collisions in the
888: system, but the collisions would have to have happened fairly recently or
889: the dust would have dispersed. \citet{wilner} therefore concludes the
890: most reasonable assumption is that the clumps are dust concentrations
891: resulting from resonant interactions between the dust and a massive planet.
892: He shows that the observations could be explained by a 3 MJ planet in
893: a large, eccentric orbit, which would currently be near apastron and located
894: about 7.1 arcsec NW of the star (though the submillimeter observations
895: were carried out
896: a few years before our imaging, a planet near apastron in such a large orbit
897: would not move appreciably over that interval).
898:
899: We chose the target position and nod direction for our Vega observations
900: to obtain good sensitivity at the location of this hypothetical
901: planet. The planet's location is marked on our sensitivity contour
902: plots (Figures \ref{fig:vegamap} and \ref{fig:vegammap}). We do not
903: detect the planet, so our observations place upper limits on its
904: mass.
905:
906: At an approximate separation of 7.1 arcsec, PA 315 degrees (due NW), our $L'$
907: images of Vega give a 10$\sigma$ sensitivity of $L'$ = 15.21,
908: or a 7$\sigma$ sensitivity of $L'$ = 15.60. Translating these
909: magnitudes to masses using the \citet{bur} models for an age
910: of 0.3 Gyr, and using the results
911: of our blind sensitivity tests, we can rule out a planet at this location
912: with a mass above 6.02 MJ with near 100\% confidence, and one more massive than
913: 4.30 MJ with 77\% confidence. If the images were very clean at this
914: location, showing no suspected sources, we could set stronger limits.
915: However, there was a suspected source within about 0.4 arcsec of this
916: location. Careful records of the manual examination of the images make
917: it clear that the suspected source can be identified as spurious
918: with high confidence, and should by no means be considered a candidate
919: detection of the \citet{wilner} planet. Its appearence simply means
920: the images are not very clean at this location, and the stronger
921: limits possible in regions without suspected sources do not apply.
922:
923: At the same location on our $M$ band images, we
924: obtained 10, 7, and 5$\sigma$ limits of $M$ = 13.39, $M$ = 13.78,
925: and $M$ = 14.14, respectively. Using the \citet{bur} models for
926: an age of 0.3 Gyr, these
927: magnitude limits correspond to planets of 5.14 MJ, 3.76 MJ, and 2.86 MJ,
928: respectively. Records from our automatic and manual examination
929: of the images show no suspected source within
930: 1.5 arcseconds of this location. Since in the sensitivity tests
931: 97\% of 7$\sigma$ planets and 85\% of 5$\sigma$ planets were at
932: least suspected, we can rule out a planet above 3.76 MJ at this
933: location at the 97\% confidence level, and one above 2.86 MJ at
934: the 85\% confidence level. The excellent sensitivity obtained at
935: this location is due in part to the fact that our
936: observing strategy was optimized to give good sensitivity near
937: the position of the \citet{wilner} hypothetical planet.
938:
939: % here
940:
941: We can set limits on the hypothetical planet of \citet{wilner}
942: close to, or perhaps even below, the proposed mass of 3 MJ. It would
943: appear from Figure \ref{fig:vhbur} that \citet{marois} set
944: similar or slightly lower limits, though the exact sensitivity
945: of their observations at the position angle of the
946: \citet{wilner} planet cannot be explicitly
947: analyzed because they present their sensitivity only in a radially
948: averaged sense. Observations at the $H$ and $M$ bands have thus consistently
949: set upper limits near the predicted mass of 3 MJ. A 3 MJ planet at
950: the 0.3 Gyr age we have adopted for Vega would have $\mathrm{T_{eff}}$
951: between 300 and 400 K. No objects in this temperature range have
952: yet been observed, so model fluxes are not observationally constrained
953: at any wavelength. Where an upper limit from a single band would
954: be tentative because of the uncertainties of the models, the consistent
955: results from a range of wavelengths allow us to conclude that it is
956: probable no 3 MJ planet exists at this location.
957:
958: \citet{wilner} makes it clear that other models besides his hypothetical 3 MJ planet
959: might explain the observed dust distribution, and
960: that further modeling is needed to see what range of planetary orbits
961: and masses might be capable of producing the resonant dust concentrations
962: seen in the submillimeter. \citet{marsh}, for example, explain the
963: distribution of dust they observe around Vega at 350-450 $\mu$m wavelengths
964: (vs 850 $\mu$m for \citet{wilner})
965: by a Neptune-mass planet in a 65 AU orbit. It is not entirely clear whether
966: their model also explains the \citet{wilner} images; however, \citet{wyatt}
967: presents a model of a migrating Neptune-mass planet that does match the 850 $\mu$m
968: images. In contrast to \citet{wyatt} and \citet{marsh}, \citet{deller} present
969: a model that explains
970: the 850 $\mu$m images by a 3 MJ planet in a considerably larger orbit than
971: that suggested by \citet{wilner}. It would have the same current PA
972: as the \citet{wilner} planet (NW of the star, near PA $315^{\circ}$), but
973: it would be 12-13 arcsec from Vega as opposed to 7 arcsec. Our Clio observations
974: do not obtain good sensitivity at these larger separations, though new, differently
975: targeted Clio images could.
976:
977: No current observational technique can image Neptune-mass extrasolar planets in distant
978: orbits. The non-detections of our survey and that of \citet{marois} lend
979: some support to models explaining the Vega dust distribution using such
980: planets rather than the model of \citet{wilner} in which the planet
981: has a mass a few times that of Jupiter. However, we cannot rule out a 3 MJ
982: planet in the more distant orbit suggested by \citet{deller}, simply
983: because our observing strategy was not designed to give good sensitivity
984: at such a large separation.
985:
986: Theoretical planet models indicate that observations at $L'$, $M$ band, and
987: the narrowband $H$-regime filter of \citet{marois} can
988: detect planets down to 3 MJ in the Vega system. Further work
989: at all three bands would either detect such a planet or rule out the
990: existence of one at large separation with very high confidence.
991: Consistent results at a variety of wavelengths will ensure that conclusions
992: are less vulnerable to model uncertainties at any particular wavelength.
993: More submillimeter work and orbital modeling of the
994: Vega system is also desirable, because if models explaining the dust
995: distribution without a massive planet can be ruled out, deep
996: targeted AO observations to detect the planet could be strongly
997: prioritized, and success could be anticipated with confidence.
998:
999:
1000:
1001:
1002: \section{$\epsilon$ Eri: Comparison with Other Studies, and Upper Limits for Hypothetical Planets} \label{eri}
1003:
1004: \subsection{Comparing Our Sensitivity with Other Studies} \label{compsens}
1005: As with Vega we do not attempt to compare our $\epsilon$ Eri
1006: results with an exhaustive list of other studies, but only
1007: with a few that obtained the best sensitivity results.
1008: We have chosen \citet{yoichi}, \citet{biller}, and \citet{GDPS}.
1009: Figure \ref{fig:ehbur} shows the results of the comparison, with
1010: again, the 90th percentile 10$\sigma$ sensitivity curves for our observations
1011: plotted.
1012:
1013: Of the other studies, the sensitivity methods of \citet{yoichi} have
1014: already been discussed in
1015: Section \ref{vega} above, as have those of \citet{GDPS} because we
1016: assumed \citet{marois} used the same methods for their Vega data.
1017: It only remains to consider the methods of \citet{biller}.
1018: They use a sensitivity estimator which is based on the single-pixel
1019: RMS in 6 pixel (0.05 arcsec, or 1.2 $\lambda / D$) square
1020: boxes on the images, and they quote 5$\sigma$ limits. It is not
1021: clear whether they take processing losses into account in their
1022: sensitivity calculation. In general we expect sensitivity estimators
1023: involving the single-pixel RMS to overestimate the sensitivity, as
1024: they assume independence of noise in adjacent pixels. This assumption
1025: is always violated in the speckle-dominated regions on AO images (that
1026: is, speckle noise is always spatially correlated, though the extent
1027: of the correlation depends on the details of the raw images
1028: and the type of PSF subtraction used).
1029:
1030: The above would seem to imply that the \citet{biller}
1031: 5$\sigma$ sensitivity results are comparable to our 5$\sigma$
1032: limits, and that we should adjust them by a factor of 2 (0.753 mag) toward
1033: decreased sensitivity in order to compare them properly against
1034: our 10$\sigma$ limits. This would not include any correction for
1035: the possible overestimation of sensitivity in the presence of correlated
1036: noise.
1037:
1038: However, several characteristics of the \citet{biller} data suggest
1039: their sensitivity should be rated higher than this. First, they
1040: use a `roll subtraction' technique which effectively
1041: creates both a positive and a negative image of any
1042: real companion, separated by $33^{\circ}$ of rotation about the primary
1043: star, and the presence of both can be used to evaluate
1044: the reality of potential sources. This doubles the
1045: data and the sensitivity should accordingly go up by
1046: $\sqrt{2}$.
1047:
1048: Second, their simultaneous differential
1049: imaging (SDI) technique involves two independent spectral differences.
1050: They are not necessarily equally sensitive, but in the best case
1051: this again doubles the data available for planet detection.
1052: With, potentially, four equal-brightness images of any
1053: real object in their data (two independent spectral difference images
1054: at each of two `roll angles'), the
1055: sensitivity of the \citet{biller} observations should in
1056: principle go up by a factor of as much as 2 (i.e. $\sqrt{4}$)
1057: over their nominal values.
1058:
1059: Finally, Beth Biller has explained
1060: to us that the \citet{biller} 5$\sigma$ point-source sensitivities
1061: were calculated by comparing the single-pixel RMS noise
1062: to the brightness of the peak pixel of a PSF. This method is
1063: conservative for well-sampled data such as that of \citet{biller},
1064: since it does not take into account the fact that bright pixels
1065: surrounding the peak of a PSF allow it to be detected with additional
1066: confidence. The single-pixel method also does not overestimate
1067: the sensitivity in the presence of correlated noise (provided the
1068: RMS noise is calculated over a large enough region).
1069:
1070: The above might indicate we should compare the \citet{biller}
1071: nominal 5$\sigma$ sensitivities directly to our 10$\sigma$ sensitivities
1072: (since obtaining 4 separate images of any real source could
1073: in principle raise the sensitivity to twice its nominal value).
1074: However, since the two spectral difference images do not neccesarily
1075: have equal sensitivity, we have scaled the \citet{biller}
1076: nominal 5$\sigma$ limits down in sensitivity by about a factor of
1077: $\sqrt{2}$ (0.38 mag) to compare them with our 10$\sigma$
1078: limits. This is equivalent to taking into account the \citet{biller}
1079: sensitivity gain only from the fact that an image is obtained
1080: in each of two `roll angles', and not from the additional fact
1081: that at each roll angle two independent spectral difference images
1082: are produced. The reader should keep this in mind when examining
1083: Figure \ref{fig:ehbur}: we may have underestimated the relative
1084: sensitivity of the \citet{biller} obsevations by a factor of
1085: around $\sqrt{2}$ (0.38 mag). This rather small correction would
1086: not affect our conclusions.
1087:
1088: As with the \citet{marois} Vega data (and also the \citet{GDPS} $\epsilon$
1089: Eri data), we have adjusted the \citet{biller}
1090: sensitivities toward greater sensitivity to convert
1091: magnitudes from narrowband filters tuned to a predicted peak
1092: in giant planet spectra to broadband $H$ magnitudes. For \citet{biller},
1093: the correction factor we applied was 0.84 magnitudes. This is an
1094: approximate value based on the SDI observers' analysis of their own filters.
1095:
1096:
1097: \clearpage
1098: \begin{figure*}
1099: \plotone{f15.eps}
1100: \caption[Comparison of sensitivities obtained around $\epsilon$ Eri with different
1101: techniques (using the \citet{bur} models)]{
1102: Comparison of the sensitivities obtained around $\epsilon$ Eri with different
1103: techniques. Magnitude
1104: sensitivities have been converted to planet mass limits in MJ using
1105: the theoretical
1106: models of \citet{bur} for an age of 0.56 Gyr. The long-dashed line
1107: is the narrowband $H$-regime
1108: result from \citet{GDPS}, the dot-dashed line at small separations
1109: is the SDI result from \citet{biller}, the dotted line is the $H$ band result
1110: from \citet{yoichi}, the gray continuous line is our 90th percentile
1111: $L'$ result, and the black continuous line is our
1112: 90th percentile $M$ band result.}
1113: \label{fig:ehbur}
1114: \end{figure*}
1115: \clearpage
1116:
1117: Figure \ref{fig:ehbur} makes it clear that though the best $H$-regime
1118: results for Vega delivered better sensitivity than our $L'$ and $M$ band
1119: observations, the sensitivity of our $M$ band observation of $\epsilon$ Eri
1120: is better than that of all previous observations out
1121: to a separation of at least 7 arcseconds from the
1122: star. Within three arcseconds of the star the sensitivity advantage
1123: of the longer wavelength observation is especially great. We note
1124: that this applies only to our $M$ band result: the SDI method of
1125: \citet{biller}, which is designed to give excellent sensitivity
1126: close to bright stars, does give results comparable those of
1127: our $L'$ observation. The good performance of the
1128: $M$ band is due to the fact that the planet/star flux ratio
1129: is much more favorable at $M$ band than even in the most
1130: optimized intervals of the $H$ band.
1131:
1132: In closing this section on comparitive sensitivities, we note
1133: that although we have used mainly the theoretical planet
1134: models of \citet{bur} to calculate sensitivities in this work, those
1135: presented in \citet{bar} are a good complement and comparison
1136: to the former. However, the filter set over which \citet{bar}
1137: integrated their theoretical spectra is slightly different from
1138: the Clio filter set, over which we integrated the \citet{bur} models.
1139: We have, however, also done tests in which we performed
1140: magnitude-mass conversions using the original mass/mag/age
1141: tables presented in \citet{bar}. In general, the \citet{bar} models
1142: give us somewhat better sensitivity in $L'$ than those of \citet{bur},
1143: with a typical disagreement of 1-2 MJ. The $M$ band predictions of the
1144: two model sets are very close. At present we cannot say for sure if
1145: the $L'$ band discrepancy is inherent in the different models or is
1146: a pure artifact of the filter set. In any case the two model sets
1147: are broadly in agreement, except for very old, cool planets, where the
1148: differences become very large and it appears clear that slightly different
1149: filter sets cannot be the whole explanation (see the discussion of the $H$ band
1150: flux of $\epsilon$ Eri b below).
1151:
1152:
1153: \subsection{Upper Limits at the Locations of Hypothetical Planets}
1154:
1155: $\epsilon$ Eri has the extremely important distinction of being
1156: one of only a few stars around which a single planet has been detected with both RV and
1157: astrometric methods \citep{hatzes, benedict}. This means that a complete, unique
1158: solution for the size, eccentricity, and orientation of the orbit is possible,
1159: as is a solution for the mass of the planet. \citet{benedict} present
1160: such an orbit solution, and give the mass of the planet as 1.55 MJ.
1161:
1162: At the time of our observations the \citet{benedict} orbit predicts a separation
1163: of about 0.684 arcsec. Our observations do not set limits in the
1164: planetary mass regime this close to the star. We note, however, that our observations
1165: were not timed with the idea of obtaining good sensitivity to this
1166: planet. If we had observed the planet near its apastron, at which
1167: point the separation is about 1.7 arcsec, our $M$ band observations
1168: in particular would have been in the range to set possibly interesting
1169: limits, though still above the \citet{benedict} mass of 1.55 MJ.
1170: The median 10 and 7$\sigma$ sensitivities of our $M$ band observation at
1171: 1.7 arcseconds are 5.3 and 4.2 MJ, respectively, and our 5$\sigma$ limit
1172: is 3.9 MJ. These are good sensitivities at a very small separation
1173: from a bright star, but, of course, the planet would still not have been
1174: detected, unless it is far more massive than the
1175: \citet{benedict} orbital solution indicates.
1176:
1177: Could any current-technology telescope detect this planet, and if so
1178: what would be the best method?
1179:
1180: \citet{janson} applied the same SDI methodology used by \citet{biller}
1181: to observe $\epsilon$ Eri at several different epochs.
1182: The data from their
1183: second epoch gave them the best limit on the planet, with a 3$\sigma$
1184: sensitivity of $\Delta$-magnitude 13.1 at the expected location
1185: of the planet based on \citet{benedict}. As discussed above, the
1186: \citet{biller} observations using the SDI method had two independent
1187: roll angles and two independent spectral differences for each observation,
1188: and the sensitivity estimation method they used was conservative.
1189: Assuming that \citet{janson} used the same methodology, we will compare
1190: their 3$\sigma$ limits directly to our 10$\sigma$ limits. Note that
1191: even considering all the issues mentioned in Section \ref{compsens}
1192: above this results in a conservative estimation of our
1193: sensitivities relative to those of \citet{janson}.
1194:
1195: We can adjust the \citet{janson} 3$\sigma$
1196: sensitivity of $\Delta$-magnitude 13.1 by the 0.84 mag
1197: value used before and add the $H$ = 1.88 magnitude of the
1198: star itself to get an equivalent sensitivity of $H$ = 15.8;
1199: the equivalent masses are 9.6 and 9.1 MJ according to the \citet{bur}
1200: and \citet{bar} models, respectively, with the age set to 0.56 Gyr
1201: in both cases.
1202:
1203: However, \citet{janson} mention that the correction from the
1204: narrowband SDI filters to $H$ band would actually be much greater
1205: than 0.84 magnitudes for a very cool object such as $\epsilon$ Eri b. According to
1206: their Figure 5, the correction is about 2.2 magnitudes for
1207: the appropriate filter difference at our adopted age of 0.56 Gyr
1208: for $\epsilon$ Eri. This different correction does not
1209: change the upper limit of 9.6 MJ quoted above, because the
1210: larger correction applies only to a planet with the 1.55 MJ
1211: mass determined by \citet{benedict}, which would have been
1212: far too faint for \citet{janson} to detect. However, the 2.2 magnitude
1213: correction is appropriate for estimating by what factor
1214: the \citet{janson} observations missed the planet -- that is, how
1215: much their sensitivity would have to be increased in order to
1216: detect it.
1217:
1218: The sensitivity of the \citet{janson} observations
1219: in their narrowband filter was about
1220: 13.1 + 1.88 = 14.98 mag, assuming that the magnitude of $\epsilon$ Eri A
1221: is the same in the narrowband filter as in broadband $H$. According
1222: to the models of \citet{bur}, an 0.56 Gyr-old planet of mass 1.55 MJ
1223: located 3.27 pc away has an $H$ band magnitude of about 28.5. We subtract
1224: the 2.2 magnitude correction to obtain a narrowband magnitude of 26.3,
1225: and difference the result with the \citet{janson} sensitivity of
1226: 14.98 mag. The conclusion is that the \citet{janson} sensitivity
1227: was insufficient to detect the planet by 11.3 magnitudes (a factor of 34,000) under the
1228: \citet{bur} models. As noted in Section \ref{compsens} above, the \citet{bar} models
1229: disagree with the \citet{bur} ones on the brightness of $\epsilon$ Eri b:
1230: the former indicate that the \citet{janson} miss factor is only
1231: about 1000, rather than 34,000. The reason for this discrepancy
1232: is not clear. It does not affect our conclusions about the best
1233: band at which to search for $\epsilon$ Eri b. The large
1234: discrepancy for two model sets that are in close agreement for warmer
1235: objects does suggest that theoretical $H$ band magnitudes for
1236: objects with temperatures as low as $\epsilon$ Eri b have a large uncertainty,
1237: and therefore any constraints based on them will be tentative.
1238: The $M$ band brightnesses predicted by the two
1239: model sets for $\epsilon$ Eri b are discrepant by a much smaller
1240: factor, about 1.7 rather than 34 (see below).
1241:
1242: The miss factors calculated above indicate the SDI sensitivity
1243: would have to be increased at least a thousandfold to
1244: detect the planet. Assuming we
1245: had observed the planet at apastron, by what factor would we
1246: have failed to detect it? We will consider only our $M$ band
1247: results, as they are more sensitive than our $L'$ observations
1248: to low mass planets close to the star.
1249: Our median 10$\sigma$ sensitivity at the apastron separation
1250: of 1.7 arcsec was $M$ = 12.02. The models of \citet{bur} give
1251: the brightness of the planet as $M$ = 14.7. This means we
1252: would have come short of a 10$\sigma$ detection by
1253: 2.68 mag, or a factor of 11.8, according to the \citet{bur} models
1254: (the \citet{bar} models give a higher but not enormously discrepant
1255: miss factor of 20.5; as in the paper up to this point we focus on
1256: the \citet{bur} models in the discussion that follows).
1257: Our blind sensitivity tests indicate about 40\% completeness
1258: at 5$\sigma$, with 85\% of sources at least noticed.
1259: Thus if we could increase our sensitivity
1260: by only a factor of 5.9 (that is, 11.8 divided by 2 to change
1261: from 10 to 5$\sigma$), we would have some chance of confidently
1262: detecting the planet, with a greater likelihood of at least noticing it.
1263:
1264: These lower miss factors suggest that $\epsilon$ Eri b
1265: might actually be detectable near apastron with ground based
1266: $M$ band imaging. It is almost certain that $\epsilon$ Eri b
1267: is at too low a $\mathrm{T_{eff}}$ for its $M$ band flux to
1268: be dimmed by the above-LTE $\mathrm{CO}$ concentrations suggested
1269: by \citet{L07} and \citet{reid} to account for the supressed
1270: $M$ band flux observed for much hotter objects (see \citet{NCE}
1271: for an analysis of how the effects of non-equilibrium $\mathrm{CO}$
1272: concentrations diminish with decreasing $\mathrm{T_{eff}}$).
1273: We note also that even if the supression of $M$ band flux remained, $M$ would
1274: still be better for the detection than the $H$-regime.
1275:
1276: The next apastron of $\epsilon$ Eri b is in 2010, and this
1277: would be the best time to attempt to image it with a very
1278: deep $M$ band observation. We have observed that
1279: our sensitivity in both speckle-limited and background-limited
1280: regimes increases roughly as the square root of the integration
1281: time, as we would expect. Therefore, barring further
1282: improvements in Clio or MMTAO, an exposure 35 (or $5.9^2$)
1283: times as long as our $M$ band integration would be
1284: required to have a 40\% chance of making a confirmed
1285: detection of the planet using Clio at the MMT.
1286: This means 35 hours of observing, or about seven
1287: good nights. Improvements to the Clio instrument,
1288: MMTAO, and our processing methods might bring the detection
1289: in range with a shorter exposure, perhaps only two nights.
1290: We note that at 1.7 arcseconds from $\epsilon$ Eri our
1291: current images are speckle-limited -- the background
1292: limit is still a factor of about 3 lower. Alterations
1293: to the instrument, or improved PSF subtraction methods
1294: in post-processing, may in future obtain near-background limited
1295: performance at this separation. The planet might then
1296: be detectable with only one or two nights worth of integration,
1297: though four to six nights would still be preferred to ensure
1298: that an interesting upper limit could be set in the event of a non-detection.
1299: Clio has already been used with a phase plate coronagraph \citep{phaseplate}
1300: which improves the close-in sensitivity.
1301:
1302: As far as we know Clio, when used with the adaptive
1303: secondary AO system of the MMT, is the only
1304: currently operating AO imager able to make the
1305: deep, high-efficiency integrations in the broad $M$ band required
1306: to detect $\epsilon$ Eri b. Other AO imagers exist that
1307: can use the $M'$ band, where the narrower bandpass reduces the
1308: intensity of the thermal background. However at $M'$ the sensitivity
1309: to planets is also reduced and the project becomes unfeasible.
1310: Given a multi-night $M$ band integration with
1311: Clio, the goal of obtaining the first direct image of a
1312: mature extrasolar planet appears to be within reach.
1313:
1314: Detection and characterization of $\epsilon$ Eri b should be
1315: quite straightforward with new large telescopes such as the
1316: LBT (which might be used instead of the MMT to make the first
1317: detection), GMT, TMT, or E-ELT, provided the latter two are equipped
1318: with the adaptive secondary AO systems necessary to reduce thermal
1319: background and make deep $M$ band
1320: observations feasible. Space-based observations are likely to be
1321: useful as well. The planet might be studied at $L'$,
1322: $M$ band, or longer wavelengths using JWST, or it could be detected
1323: in reflected light at visible wavelengths by a sensitive space-based
1324: coronagraph. However, the first detection may come well ahead
1325: of JWST and the next generation of giant telescopes --- it may
1326: be achieved in the $M$ band with the MMT during the 2010 apastron.
1327:
1328: \citet{ozernoy} and \citet{quillen} suggest that the dust disk
1329: of $\epsilon$ Eri has been sculpted by a planet of 0.1-0.2 MJ in
1330: an orbit between 40 and 65 AU in radius. \citet{deller} agree,
1331: and prefer the model of \citet{quillen}. Such a planet
1332: would be far too faint to detect with any telescope in the
1333: near future. However, \citet{deller} state
1334: that an additional, $\sim$ 1 MJ planet in a closer-in orbit is
1335: likely required to produce the observed clearing of the dust inside
1336: of about 30 AU \citep{greaves}. The RV/astrometric planet of \citet{hatzes} and
1337: \citet{benedict} has too small an orbit to account for this dust clearing;
1338: \citet{deller} suggest a larger orbital radius between 10 and 18 AU for the planet
1339: responsible for clearing the dust. \citet{benedict} mention a long-term
1340: trend in RV measurements for $\epsilon$ Eri A that might indicate just
1341: such a planet: a $\sim$ 1 MJ object
1342: orbiting with a period longer than 50 years. Since such a planet
1343: would probably appear at least 3-4 arcsec from the star, we would likely
1344: have detected it if it had a mass of 4-5 MJ or
1345: greater, as would the \citet{GDPS} observation. Since the mass
1346: is expected to be closer to 1 MJ, it is not surprising the planet has not
1347: yet been detected. It might be imaged serendipitously in the course
1348: of a very long exposure intended to detect the known RV/astrometric planet.
1349:
1350: \section{Conclusions} \label{conclusion}
1351: We have taken very deep $L'$ and $M$ band images of the interesting
1352: debris disk stars Vega and $\epsilon$ Eri to search each system
1353: for orbiting planets and brown dwarfs. For both stars we
1354: obtained better sensitivity than shorter-wavelength observations
1355: at small separations from the star. The sensitivity of our observations
1356: compared more favorably to the sensitivity of $H$-regime observations in the case
1357: of $\epsilon$ Eri than in the case of Vega. For $\epsilon$ Eri, our
1358: $M$ band observation appears to set the best upper limits yet for
1359: planets out to a separation of about 7 arcseconds, beyond which the
1360: sensitivity of the \citet{GDPS} $H$-regime observations becomes very
1361: slightly superior.
1362:
1363: %At the location of the
1364: %\citet{wilner} hypothetical planet of Vega, the $H$-regime observation of
1365: %\citet{marois} obtained considerably better sensitivity than ours. However,
1366: %at the 1.7 arcsecond apastron separation of the known planet of $\epsilon$ Eri, the
1367: %sensitivity of our observations is much closer to what is required to
1368: %detect the planet than the sensitivity of the \citet{janson} observations.
1369: %The \citet{GDPS} observations of $\epsilon$ Eri are saturated
1370:
1371: The reason our $\epsilon$ Eri observations have a
1372: greater sensitivity advantage over $H$ regime observations than
1373: do our images of Vega is the smaller distance to the $\epsilon$ Eri system.
1374: This is another instance of the same physical reality we discussed
1375: in Section \ref{finsens} above, when explaining why our $M$ band
1376: sensitivity is much better than our $L'$ results on $\epsilon$ Eri
1377: but not on Vega. As we stated above, for $\epsilon$ Eri,
1378: the sensitivity of a given observation at any wavelength extends
1379: down to less luminous, lower $\mathrm{T_{eff}}$ planets than for
1380: Vega. The $H-L'$ and $H-M$ colors, as well as the $L' - M$ color, of
1381: low $\mathrm{T_{eff}}$ giant planets are more red than those of
1382: hotter ones. Therefore the faintest detectable objects in the
1383: $\epsilon$ Eri system would be more red than those
1384: in the more distant Vega system, and longer wavelength observations
1385: are most useful for the nearer system. This is a general and
1386: important principle for planning optimal planet search strategies:
1387: the faintest detectable planets will be more red, and therefore
1388: the relative advantage of long wavelengths over short ones will be higher,
1389: for the nearest stars. For distant stars where only hot objects
1390: with blue infrared colors can be detected, long wavelengths observations
1391: are not as useful. For very nearby stars such as $\epsilon$ Eri,
1392: where very interesting, extremely low-mass, low $\mathrm{T_{eff}}$
1393: planets can be detected, the long wavelengths are very useful because
1394: the planets being sought have such red colors.
1395:
1396: Planet-search observations at the $L'$ and $M$ bands have a considerable
1397: advantage over those in the more commonly used $H$ band regime for
1398: $\epsilon$ Eri and a handful of other bright, very nearby stars. For
1399: more distant bright stars such as Vega, $L'$ and $M$ band observations
1400: give markedly better results only at separations inside about 3 arcsec,
1401: and in this regime no currently employed method gives sensitivity
1402: to any but the highest mass planets. Observations in the bands we
1403: have employed are still useful on Vega, but their use tends toward
1404: a diversification of planet-search effort in case theoretical models
1405: are overpredicting planets' $H$ band brightnesses. For nearer
1406: systems such as $\epsilon$ Eri, by contrast, $L'$ and $M$ band observations
1407: clearly provide the best sensitivity at the most interesting separations,
1408: and it is the $H$ regime images that naturally take the role of
1409: diversifying effort under the supposition that the models may
1410: overpredict planet brightness at longer wavelengths.
1411:
1412: We have set a limit on the Vega planet hypothesized by \citet{wilner}
1413: that is close to the 3 MJ mass he suggested for it. It appears that
1414: \citet{marois} could set a similar limit. The evidence seems fairly
1415: strong that no 3 MJ planet exists at this location. This favors alternative
1416: models involving smaller planets, such as those of \citet{marsh} and
1417: \citet{wyatt}, or a 3 MJ planet in a larger orbit, such as that of
1418: \citet{deller}. Since a 3 MJ planet around Vega could be imaged in
1419: multiple wavelength regimes with current technology, more
1420: submillimeter observations and
1421: further modeling to determine if such a planet is required to
1422: explain the observed dust distribution is very desirable. If this
1423: does turn out to be the case, deep AO observations to detect the
1424: planet could be strongly prioritized, and a sucessful detection in
1425: one or more wavelength bands would be very likely.
1426:
1427: Our $\epsilon$ Eri observation was not timed to catch the
1428: known planet $\epsilon$ Eri b at a large separation, and
1429: therefore our current data do not allow us to set an interesting
1430: limit on its mass. \citet{janson} observed $\epsilon$ Eri at several
1431: epochs of more promising separation using SDI, and set limits
1432: in the 9-10 MJ range.
1433:
1434: We have set a limit of 4-5 MJ for additional planets in
1435: more distant orbits around $\epsilon$ Eri. The existence
1436: of a planet in such an orbit may be indicated by a long term RV trend
1437: \citep{benedict} and by a clearing of dust from the inner disk \citep{deller}.
1438: \citet{benedict} and \citet{deller} suggest a mass of around 1 MJ
1439: for this hypothetical outer planet, so our non-detection is not
1440: surprising.
1441:
1442: We have explored the question of whether SDI imaging \citep{janson, biller}
1443: or $L'$ and $M$ band imaging (this work) is the method most
1444: likely ultimately to detect $\epsilon$ Eri b.
1445: Our $M$ band images were much more sensitive at small separations than
1446: our $L'$ results, so we have not considered the latter. We find
1447: that the sensitivity of the \citet{janson} observations
1448: at the best epoch, where
1449: the planet was near the optimal separation for SDI imaging, was
1450: still insufficiently sensitive to detect the planet by a factor
1451: of at least a thousand. By contrast our observations, if carried
1452: out at apastron, would have missed the planet by a factor of only about 12.
1453:
1454: This striking difference suggests that it is at $M$ band that the planet
1455: $\epsilon$ Eri b will first be imaged. A several-night observing
1456: campaign using Clio at the MMT might detect it during the 2010
1457: apastron passage, since we have observed the sensitivity in the
1458: speckle-dominated regions of $M$ band
1459: images does go up approximately as the square root of the exposure
1460: time. More advanced PSF subtraction, or coronagraphic capability in
1461: Clio \citep{phaseplate}, might reduce the required exposure time to detect the
1462: planet to as little as 2 nights.
1463: At present, we believe Clio with MMTAO
1464: is the only system capable of deep planet imaging
1465: integrations in the $M$ band. Spitzer, despite its enormously
1466: lower background and correspondingly excellent sensitivity,
1467: does not have sufficient resolution to detect objects at the separations
1468: expected for orbiting planets.
1469:
1470: $\epsilon$ Eri b could be studied in more detail using
1471: new giant telescopes such as the LBT and GMT with planned
1472: adaptive secondary AO systems. The $M$ band will remain
1473: the best wavelength choice for observations using these
1474: larger telescopes, so adaptive secondaries will remain
1475: essential: conventional AO systems even on giant telescopes
1476: will likely still have too high a thermal background for
1477: efficient, deep $M$ band images. An $L'$ and $M$ band
1478: imager called LMIRCam is planned for the LBT \citep{elmer}.
1479: When JWST is launched, it should also deliver interesting
1480: scientific results on $\epsilon$ Eri b. However, it is possible
1481: that the first image of this planet --- the first direct
1482: image of any mature extrasolar planet --- will be obtained using
1483: Clio at the MMT in 2010.
1484:
1485: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1486:
1487:
1488:
1489: \acknowledgments
1490: This research has made use of the SIMBAD database,
1491: operated at CDS, Strasbourg, France.
1492:
1493: This research has made extensive use of information and code
1494: from \citet{nrc}.
1495:
1496: We thank I. Baraffe for kindly suppling us with theoretical
1497: spectrum files corresponding to the planet models used in \citet{bar}.
1498:
1499: We thank the referee, Remi Soummer, for helpful comments.
1500:
1501: Facilities: \facility{MMT}
1502:
1503:
1504: \begin{thebibliography}{}
1505: \bibitem[Aumann et al.(1984)]{IRASv1} Aumann, H., Gillett, F., Beichman, A.,
1506: de Jong, T., Houck, J., Low, F., Neugebauer, G., Walker, R., \& Wesselius, P.
1507: 1984, \apj, 278, L23
1508: \bibitem[Aumann(1985)]{IRASa3} Aumann, H. 1985, \pasp, 97, 885
1509: \bibitem[Backman \& Paresce(1993)]{paresce} Backman, D. \& Paresce, F. 1993, in Protostars
1510: and Planets III, ed. E. H. Levy \& J. I. Lunine (Tucson: Univ. Arizona Press),
1511: 1253
1512: \bibitem[Backman(1996)]{backman} Backman, D. 1996, AAS/DPS 28.0122B
1513: \bibitem[Baraffe et al.(2003)]{bar} Baraffe, I., Chabrier, G., Barman, T. S.,
1514: Allard, F., \& Hauschildt, P. H. 2003, \aap, 402, 701
1515: %\bibitem[Barrado y Navascu\'{e}s(1998)]{BYN98} Barrado y Navascu\'{e}s, D.
1516: %1998, \aap, 339, 831
1517: %\bibitem[Barrado y Navascu\'{e}s et al.(1999)]{BYN99} Barrado y Navascu\'{e}s,
1518: %D., Stauffer, J., Song, I., \& Caillault, J-P. 1999, \apj, 520, L123
1519: \bibitem[Benedict et al.(2006)]{benedict} Benedict, G., McArthur, B.,
1520: Gatewood, G., Nelan, E., Cochran, W., Hatzes, A., Endl, M.,
1521: Wittenmyer, R., Baliunas, S., Walker, G., Yang, S., K\"{u}rster, M.,
1522: Els, S., \& Paulson, D. 2006, \aj, 132, 2206
1523: %\bibitem[Bessell \& Brett(1988)]{bessell} Bessell, M. \& Brett, J. 1988,
1524: %\pasp, 100, 1134
1525: \bibitem[Biller et al.(2006)]{sdi1} Biller, Beth A.; Close, Laird M.; Masciadri, Elena; Lenzen, Rainer; Brandner, Wolfgang; McCarthy, Donald; Henning, Thomas; Nielsen, Eric L.; Hartung, Markus; Kellner, Stephan; Geissler, Kerstin; \& Kasper, Markus 2006, \procspie, 6272E, 74B
1526: %\bibitem[Biller et al.(2006b)]{SCR1845} Biller, B., Kasper, M., Close, L., Brandner, W.,
1527: %\& Kellner, S. 2006, ArXiv e-prints, astro-ph/0601.440
1528: \bibitem[Biller et al.(2007)]{biller} Biller, B., Close, L., Masciadri, E., Nielsen, E., Lenzen, R., Brandner, W.,
1529: McCarthy, D., Hartung, M., Kellner, S., Mamajek, E., Henning, T., Miller, D., Kenworthy, M., \& Kulesa, C. 2007, ArXiv e-prints, astro-ph/0705.0066
1530: %\bibitem[Bryden et al.(2006)]{bryden} Bryden G., Beichman, C., Trilling, D.,
1531: %Rieke, G., Holmes, E., Lawler, S., Stapelfeldt, K., Werner, M.,
1532: %Gautier, T., Blaylock, M., Gordon, K., Stansberry, J., \& Su, K. 2006,
1533: %\apj, 636, 1098
1534: \bibitem[Burrows et al.(2003)]{bur} Burrows, A., Sudarsky, D., \& Lunine, J. I. 2003, \apj, 596, 578B
1535: %\bibitem[Chauvin et al.(2005)]{ABPic} Chauvin, G., Lagrange, A., Zuckerman, B.,
1536: %Dumas, C., Mouillet, D., Song. L., Beuzit, J., Lowrance, P.,
1537: %\& Bessel, M. 2005, \aap, 438, L29
1538: %\bibitem[Cohen et al.(1992)]{mkofilt} Cohen, M., Walker, R. G., Michael, J. B., \& Deacon, J. R. 2003, \aj, 104, 1650
1539: %\bibitem[Cox(2000)]{aaq} Cox, A. N. 2000, Allen's Astrophysical Quantities (Fourth Edition; New York, NY: Springer-Verlag New York, Inc.)
1540: %\bibitem[Cutri et al.(2003)]{2MASS} Cutri, R., Skrutskie, M., van Dyk, S.,
1541: %Beichman, C., Carpenter, J., Chester, T., Cambresy, L., Evans, T., Fowler, J.,
1542: %Gizis, J., Howard, E., Huchra, J., Jarett, T., Kopan, E., Kirkpatrick, J.,
1543: %Light, R., Marsh, K., McCallon, H., Schneider, S., Stiening, R.,
1544: %Sykes, M., Weinberg, M., Wheaton, W., Wheelock, S., \& Zacarias, N. 2003,
1545: %2MASSS All Sky Catlog of point sources (The IRSA 2MASS All-Sky
1546: %Point Source Catalog, NASA/IPAC Infrared Science Archive.
1547: %\url{http://irsa.ipac.caltech.edu/applications/Gator/})
1548: \bibitem[Deller \& Maddison(2005)]{deller} Deller, A. \& Maddison, S. 2005, \apj, 625, 398
1549: %\bibitem[Favata et al.(1998)]{favata} Favata, F., Micela, G., Sciortino, S.,
1550: %\& D'Antona, F. 1998, \aap, 335, 218
1551: \bibitem[Fischer(1998)]{fischer} Fischer, D. 1998, PhD Thesis entitled
1552: \textbf{Lithium Abundances in Field K Dwarfs}, University of
1553: California Santa Cruz
1554: %\bibitem[Fischer \& Valenti(2005)]{carnp} Fischer, D. A. \& Valenti, J. 2005,
1555: %\apj, 622, 1102
1556: \bibitem[Freed et al.(2004)]{freed} Freed, M., Hinz, P., Meyer, M.,
1557: Milton, N., \& Lloyd-Hart, M. 2004, \procspie, 5492, 1561
1558: \bibitem[Gei\ss ler et al.(2007)]{epsindia} Gei\ss ler, K.; Kellner, S.; Brandner, W.; Masciadri, E.; Hartung, M.; Henning, T.; Lenzen, R.; Close, L.; Endl, M.; \& K\"{u}rster, M. 2007, \aap, 461, 665G
1559: \bibitem[Gillett et al.(1984)]{IRASa1} Gillett, F., Aumann, H., \& Low, F.
1560: 1984, paper presented at the Metting on Protostars and Planets,
1561: University of Arizona, Tucson, January 1984
1562: \bibitem[Gillett et al(1985)]{IRASa2} Gillett, F., Aumann, H., Neugebauer, G.,
1563: Low, F., \& Walters, R. 1985, in Proceedings of the First International
1564: IRAS symposium, June 10, 1985, Noordwijk, The Netherlands.
1565: \bibitem[Greaves et al.(1998)]{greaves} Greaves, J., Holland, W., Moriarty-Schieven, G.,
1566: Jenness, T., Dent, W., Zuckerman, B., McCarthy, C., Webb, R., Butner, H., Gear, W.,
1567: \& Walker, H. 1998, \apj, 506, L133
1568: \bibitem[Hatzes et al.(2000)]{hatzes} Hatzes, A., Cochran, W., McArthur, B., Baliunas, S.,
1569: Walker, G., Campbell, B., Irwin, A., Stephenson, Y., K\"{u}rster, M.,
1570: Endl, M., Els, S., Butler, R., \& Marcy, G. 2000, \apj, 544, 145
1571: \bibitem[Hinkley et al.(2007)]{hinkley} Hinkley, S., Oppenheimer, B.,
1572: Soummer, R., Sivaramakrishnan, A., Roberts, L., Kuhn, J., Makidon, R., Perrin,
1573: M., Lloyd, J., Kratter, K., \& Brenner, D. 2007, \apj, 654, 633
1574: \bibitem[Hinz et al.(2006)]{oldvega} Hinz, P., Heinze, A., Sivanandam, S.,
1575: Miller, D., Kenworthy, M., Brusa, G., Freed, M., \& Angel, J. 2006, \apj,
1576: 653, 1486.
1577: %\bibitem[Holland et al.(1998)]{holland} Holland, W. et al. 1998, Nature, 392, 788
1578: %\bibitem[H\"{u}nsch et al.(1998)]{hunsch} H\"{u}nsch, M., Schmitt, J.,
1579: %Sterzik, M. \& Voges, W. 1998, Astron. Astrophys. Suppl. Ser. 135, 319
1580: \bibitem[Hubeny \& Burrows(2007)]{NCE} Hubeny, I. \& Burrows, A. 2007, \apj, 668,1248
1581: \bibitem[Janson et al.(2007)]{janson} Janson, M., Brandner, W., Henning, T.,
1582: Lenzen, R., McArthur, B., Benedict, G., Reffert, S., Nielsen, E., Close, L.,
1583: Biller, B., Kellner, S., G\"{u}ther, E., Hatzes, A., Masciadri, E.,
1584: Geissler, K., \& Hartung, M. 2007, \aj, 133, 2442
1585: %\bibitem[Johnson et al.(2007)]{johnson} Johnson, J., Butler, R.,
1586: %Marcy, G., Fischer, D., Vogt, S., Wright, J., \& Peek, K. 2007,
1587: %ArXiv e-prints, astro-ph/0707.2409
1588: %\bibitem[Juric \& Tremaine (2007)]{juric} Juric, M. \& Tremaine, S. 2007, ArXiv e-prints, astro-ph/0703.0160
1589: %\bibitem[Karatas et al.(2005)]{karatas} Karatas, Y., Bilir, S., \& Schuster, W.
1590: %2005, \mnras, 360, 1345
1591: %\bibitem[Kasper et al.(2007)]{kasper} Kasper, M., Apai, D., Janson, M., \& Brandner, W. 2007, ArXiv e-prints, astro-ph/0706.0095
1592: \bibitem[Kenworthy et al.(2007)]{phaseplate} Kenworthy, M., Codona, J., Johanan, L.,
1593: Hinz, P., Angel, J., Heinze, A., \& Sivanandam, S. 2007, \apj, 660, 762
1594: %\bibitem[King et al.(2003)]{king03} King, J., Villarreal, A., Soderblom, D.,
1595: %Gulliver, A., \& Adelman, S. 2003, \apj, 125,1980
1596: %\bibitem[Koerner et al.(2001)]{koerner} Koerner, C., Sargent, A., Ostroff, N. 2001, \apj, 560, L181
1597: %\bibitem[Lachaume et al.(1999)]{taucetiage} Lachaume, R., Dominik, C.,
1598: %Lanz, T., \& Habing, H. 1999, \aap, 348, 897
1599: %\bibitem[Lafreni\`{e}re et al.(2007a)]{LOCI} Lafreni\`{e}re, D., Doyon, R., Marois, C., Doyon, R., Nadeau, D.,
1600: %\& Artigau, E. 2007, \apj, 660, 770
1601: \bibitem[Lafreni\`{e}re et al.(2007)]{GDPS} Lafreni\`{e}re, D., Doyon, R., Marois, C., Nadeau, D., Oppenheimer, B.,
1602: Roche, P., Rigaut, F., Graham, J., Jayawardhana, R., Johnstone, D., Kalas, P., Macintosh, B., \& Racine, R. 2007,
1603: ArXiv e-prints, astro-ph/0705.4290
1604: %\bibitem[Latham et al.(1989)]{planet1} Latham, D., Stefanik, R., Tsevi, M.,
1605: %Mayor, M., \& Gilbert, B. 1989, \nat, 339, 38.
1606: \bibitem[Leggett et al.(2003)]{leggett} Leggett, S., Hawarden, T., Currie, M., Adamson, A., Carrol, T.,
1607: Kerr, T., Kuhn, O., Seigar, M., Varricatt, W., \& Wold, T. 2003, \mnras, 345,144
1608: \bibitem[Leggett et al.(2007)]{L07} Leggett, S., Saumon, D., Marley, M.,
1609: Geballe, T., Golimowksi, D., Stephens, D., \& Fan, X. 2007, \apj, 655, 1079
1610: %\bibitem[Gomez et al.(2005)]{levison} Gomez, R., Levison, H.,
1611: %Tsiganis, K., \& Morbidelli, A. 2005 \nat, 435, 466
1612: \bibitem[Li \& Lunine(2003)]{HD4796} Li, A. \& Lunine, J. 2003, \apj, 590, 368
1613: %\bibitem[Lopez-Santiago et al.(2006)]{lopez} Lopez-Santiago, J.,
1614: %Montes, D., Crespo-Chacon, I., \& Fernandez-Figueroa, M. 2006, \apj,
1615: %643, 1160
1616: %\bibitem[Lowrance et al.(2005)]{GJ3541A} Lowrance, P., Becklin, E.,
1617: %Schneider, G., Kirkpatrick, J., Weinberger, A., Zuckerman, B., Dumas, C.,
1618: %Beuzit, J., Plait, P., Malumuth, E., Heap, S., Terrile, R., \& Hines, D. 2005,
1619: %\aj, 130, 1845
1620: %\bibitem[Marcy \& Butler(1992)]{iodine} Marcy, G. \& Butler, R. 1992, \pasp,
1621: %104, 270
1622: \bibitem[Marley et al.(2007)]{faintJup} Marley, M. S., Fortney, J. J., Hubickyj, O., Bodenheimer, P., \& Lissauer, J. J. 2007, \apj, 655, 541
1623: \bibitem[Marois et al.(2006)]{marois} Marois, C., Lafreni\`{e}re, D., Doyon, R., Macintosh, B., \& Nadeau, D. 2006, \apj, 641, 556.
1624: \bibitem[Marsh et al.(2006)]{marsh} Marsh, K., Dowell, C., Velusamy, T.,
1625: Grogan, K., Beichman, C. 2006 \apj, 646, 77.
1626: \bibitem[Masciadri et al.(2005)]{masciadri} Masciadri, E.; Mundt, R.; Henning, Th.; Alvarez, C.; \& Barrado y Navascu\'{e}s, D 2005, \apj, 625, 1004M
1627: %\bibitem[Mayor \& Queloz(1995)]{51peg} Mayor, M. \& Queloz, D. 1995,
1628: %\nat, 378, 355
1629: %\bibitem[Montes et al.(2001)]{montes01} Montes, D., L\'{o}pez-Santiago, J.,
1630: %Ferna\'{a}ndez-Figueroa, M., \& G\'{a}lvez, M. 2001, \aap, 379, 976
1631: %\bibitem[Nielson et al.(2007)]{nielson} Nielson, E., Close, L., \& Biller, B.
1632: %2007, ArXiv e-prints, astro-ph/0706.4331
1633: \bibitem[Neuh\"{a}user et al.(2000)]{hsttwa7a} Neuh\"{a}user, R.; Brandner, W.; Eckart, A.; Guenther, E.; Alves, J.; Ott, T.; Hu\'{e}lamo, N.; \& Fern\'{a}ndez, M 2000, \aap, 354L, 9N
1634: %\bibitem[Neuh\"{a}user et al.(2005)]{GQLup} Neuh\"{a}user, R., Guenther, E.,
1635: %Wuchterl, G., Mugrauer, M., Bedalov, A., \& Hauschildt, P. 2005, \aap, 435, L13
1636: \bibitem[Ozernoy et al.(2000)]{ozernoy} Ozernoy, L., Gorkavyi, N., Mather, J., \& Taidakova, T. 2000, \apj,
1637: 537, L147
1638: \bibitem[Perryman et al.(1997)]{hip} Perryman, M., Lindegren, L., Kovalevsky, J., Hog, E.,
1639: Bastian, U., Bernacca, P., Creze, M., Donati, F., Grenon, M., Grewing, M.,
1640: Van Leeuwen, F., Van Der Marel, H., Mignard, F., Murray, C., Le Poole, R., Schrijver, H.,
1641: Turon, C., Arenou, F., Froeschle, M., \& Peterson, C. 1997, \aap, 323, L49
1642: %\bibitem[Potter et al.(2003)]{potter} Potter, D., Cushing, M., \&
1643: %Neuhauser, R. 2003, csss, 12, 689
1644: \bibitem[Press et al.(1992)]{nrc} Press, W. H., Teukolsky, S.A., Vetterling, W. T., \& Flannery, B. P. 1992, Numerical Recipes in C (Second Edition; New York, NY: Cambridge University Press)
1645: \bibitem[Quillen \& Thorndike(2002)]{quillen} Quillen, A. \& Thorndike, S. 2002, \apj, 578, L149
1646: \bibitem[Reid \& Cruz(2002)]{reid} Reid, I. \& Cruz, K. 2002, \aj, 123, 466
1647: \bibitem[Sivanandam et al.(2006)]{suresh} Sivanandam, S., Hinz, P., Heinze, A.,
1648: \& Freed, M. 2006, Proc. SPIE, 6269, 27
1649: \bibitem[Song et al.(2001)]{agevega} Song, I., Caillault, J.-P., Barrado
1650: y Navascu\'{e}s, D., \& Stauffer, J. 2001, \apj, 546, 352
1651: %\bibitem[Song et al.(2003)]{song03} Song, I., Zuckerman, B., \& Bessell, M.
1652: %2003, \apj, 599, 342
1653: %\bibitem[Song et al.(2006)]{2M1207} Song, I., Schneider, G., Zuckerman, B.,
1654: %Farihi, J., Becklin, E., Bessell, M., Lowrance, P., \& Macintosh, B.
1655: %2006, \apj, 652, 724
1656: %\bibitem[Wichmann et al.(2003)]{wichmann03a} Wichmann, R., Schmitt, J.,
1657: %\& Hubrig, S. 2003, \aap, 399, 983
1658: %\bibitem[Wichmann \& Schmitt(2003)]{wichmann03b} Wichmann, R. \& Schmitt, J.
1659: %2003, \mnras, 342, 1021
1660: \bibitem[Wilner(2004)]{wilner} Wilner, D. 2004, ASP Conference Series,
1661: Vol. 324.
1662: \bibitem[Wilson et al.(2007)]{elmer} Wilson, J., Hinz, P., Kenworthy, M., Skrutskie, M., Jones, T.,
1663: Nelson, M., Woodward, C., \& Garnavich, P. 2007, Proceedings of the conference In the Spirit of
1664: Bernard Lyot: The Direct Detection of Planets and Circumstellar Disks in the 21st Century. June 04 - 08, 2007. University of California, Berkeley, CA, USA. Edited by Paul Kalas
1665: %\bibitem[Wolszczan \& Frail(1992)]{pulsar} Wolszczan, A. \& Frail, D. 1992,
1666: %\nat, 355, 145
1667: \bibitem[Wyatt(2003)]{wyatt} Wyatt, M. 2003, \apj, 598, 1321
1668: \bibitem[Vikhlinin et al.(1995)]{matchfilt} Vikhlinin, A., Forman, W., Jones, C., \&
1669: Murray, S. 1995, \apj, 451, 542
1670: \bibitem[Yoichi et al.(2006)]{yoichi} Yoichi, I., Oasa, Y., \& Fukagawa, M.
1671: 2006, \apj, 652, 1729
1672: %\bibitem[Zuckerman et al.(2001)]{zuckerman01} Zuckerman, B., Song, I.,
1673: %Bessell, M., \& Webb, R. 2001, \apj, 562, L87
1674: \end{thebibliography}
1675:
1676: \end{document}
1677:
1678:
1679: %To test the accuracy of our sensitivity estimator, we conducted blind
1680: %tests in which fake planets were inserted into the raw
1681: %data. The altered images were then processed in exactly the same way
1682: %as the original raw data, and the `planets' were
1683: %detected using both automatic and manual methods by an experimenter
1684: %who knew neither their positions nor their number. These planets were inserted
1685: %at fixed nominal significance levels of 10$\sigma$, 7$\sigma$, and 5$\sigma$ based
1686: %on the sensitivity map. We conducted such tests for each of
1687: %our four data sets (the $L'$ and $M$ band data sets for each of the two
1688: %stars). The results of the tests
1689: %are given in Tables \ref{tab:vegalsimv}, \ref{tab:vegamsimv},
1690: %\ref{tab:erilsimv}, and \ref{tab:erimsimv}.
1691:
1692: %Fake planets in these tables are classified as 'Confirmed', 'Noticed', or
1693: %'Unnoticed'. 'Confirmed' means the source was confidently detected and
1694: %would certainly be worthy of long-exposure followup observations at the MMT.
1695: %If a source is detected with this confidence level in an unaltered data set,
1696: %there is no significant doubt it is a real object. In calculating our
1697: %completeness, we count only confirmed sources as true detections.
1698: %'Noticed' means the source was flagged by our automatic detection algorithm,
1699: %or noticed as a possible real object during the purely manual phase
1700: %of planet-searching, but could not be confirmed beyond reasonable doubt.
1701: %Many spurious sources are 'Noticed' whereas the false-positive rate for
1702: %'Confirmed' detections is extremely low, with none for any of the data
1703: %sets discussed here. 'Unnoticed' means a fake planet was not automatically
1704: %flagged or noticed manually.
1705:
1706: %\begin{deluxetable}{rrrcc}
1707: %\tabletypesize{\small}
1708: %\tablecaption{Vega L'-band fake planet experiment.\label{tab:vegalsimv}}
1709: %\tablewidth{0pt}
1710: %\tablehead{\colhead{Sep} & & \colhead{Mass} & \colhead{Input} & \colhead{Detection}\\
1711: %\colhead{(asec)} & \colhead{$L'$ Mag} & \colhead{(MJ)} & \colhead{Significance} & \colhead{Status}}
1712: %\startdata
1713: %1.68 & 10.83 & $>$20 & 10$\sigma$ & Confirmed \\
1714: %1.97 & 12.00 & $>$20 & 10$\sigma$ & Confirmed \\
1715: %2.48 & 12.22 & $>$20 & 10$\sigma$ & Confirmed \\
1716: %2.53 & 12.60 & 18.05 & 10$\sigma$ & Confirmed \\
1717: %3.42 & 13.34 & 12.62 & 10$\sigma$ & Confirmed \\
1718: %4.41 & 13.53 & 11.45 & 10$\sigma$ & Confirmed \\
1719: %4.67 & 14.28 & 8.32 & 10$\sigma$ & Confirmed \\
1720: %5.63 & 14.82 & 6.87 & 10$\sigma$ & Confirmed \\
1721: %6.40 & 14.61 & 7.40 & 10$\sigma$ & Confirmed \\
1722: %8.65 & 15.03 & 6.42 & 10$\sigma$ & Confirmed \\
1723: %8.69 & 15.06 & 6.35 & 10$\sigma$ & Confirmed \\
1724: %9.21 & 15.31 & 5.84 & 10$\sigma$ & Confirmed \\
1725: %10.69 & 15.22 & 6.00 & 10$\sigma$ & Confirmed \\
1726: %11.21 & 14.88 & 6.74 & 10$\sigma$ & Confirmed \\
1727: %1.75 & 11.61 & $>$20 & 7$\sigma$ & Confirmed \\
1728: %2.24 & 12.60 & 18.05 & 7$\sigma$ & Confirmed \\
1729: %2.37 & 12.51 & 18.79 & 7$\sigma$ & Confirmed \\
1730: %2.86 & 12.92 & 15.60 & 7$\sigma$ & Confirmed \\
1731: %3.35 & 12.79 & 16.58 & 7$\sigma$ & Confirmed \\
1732: %4.08 & 13.88 & 9.73 & 7$\sigma$ & Confirmed \\
1733: %6.57 & 14.92 & 6.65 & 7$\sigma$ & Confirmed \\
1734: %8.26 & 15.04 & 6.39 & 7$\sigma$ & Noticed \\
1735: %2.32 & 12.88 & 15.90 & 5$\sigma$ & Confirmed \\
1736: %3.68 & 13.42 & 12.10 & 5$\sigma$ & Confirmed \\
1737: %5.22 & 15.12 & 6.22 & 5$\sigma$ & Unnoticed \\
1738: %5.70 & 15.22 & 6.00 & 5$\sigma$ & Noticed \\
1739: %8.11 & 16.29 & 4.15 & 5$\sigma$ & Noticed \\
1740: %10.16 & 16.06 & 4.52 & 5$\sigma$ & Noticed \\
1741: %\enddata
1742: %\tablecomments{Planets confirmed: 14/14 at 10$\sigma$; 7/8 at 7$\sigma$;
1743: %3/6 at 5$\sigma$. Planets noticed: 14/14 at 10$\sigma$; 8/8 at 7$\sigma$;
1744: %5/6 at 5$\sigma$. Planet magnitude to mass conversion carried out
1745: %using the models of \citet{bur}.}
1746: %\end{deluxetable}
1747:
1748:
1749: %\begin{deluxetable}{rrrcc}
1750: %\tabletypesize{\small}
1751: %\tablecaption{Vega M-band fake planet experiment.\label{tab:vegamsimv}}
1752: %\tablewidth{0pt}
1753: %\tablehead{\colhead{Sep} & & \colhead{Mass} & \colhead{Input} & \colhead{Detection}\\
1754: %\colhead{(asec)} & \colhead{$M$ Mag} & \colhead{(MJ)} & \colhead{Significance} & \colhead{Status}}
1755: %\startdata
1756: %0.49 & 6.66 & $>$20 & 10$\sigma$ & Confirmed \\
1757: %0.93 & 7.77 & $>$20 & 10$\sigma$ & Confirmed \\
1758: %0.97 & 8.07 & $>$20 & 10$\sigma$ & Confirmed \\
1759: %1.41 & 9.70 & $>$20 & 10$\sigma$ & Confirmed \\
1760: %1.69 & 10.80 & $>$20 & 10$\sigma$ & Confirmed \\
1761: %2.93 & 12.29 & 11.97 & 10$\sigma$ & Confirmed \\
1762: %3.60 & 12.71 & 8.54 & 10$\sigma$ & Confirmed \\
1763: %5.53 & 12.72 & 8.47 & 10$\sigma$ & Confirmed \\
1764: %7.80 & 12.31 & 11.78 & 10$\sigma$ & Confirmed \\
1765: %7.93 & 13.16 & 6.11 & 10$\sigma$ & Confirmed \\
1766: %8.42 & 12.80 & 7.96 & 10$\sigma$ & Confirmed \\
1767: %8.74 & 12.55 & 9.67 & 10$\sigma$ & Confirmed \\
1768: %11.09 & 12.69 & 8.67 & 10$\sigma$ & Confirmed \\
1769: %0.50 & 12.79 & $>$20 & 7$\sigma$ & Noticed \\
1770: %2.54 & 13.02 & 8.02 & 7$\sigma$ & Noticed \\
1771: %2.59 & 12.87 & 9.29 & 7$\sigma$ & Unnoticed \\
1772: %3.26 & 13.17 & 7.57 & 7$\sigma$ & Confirmed \\
1773: %4.57 & 13.42 & 7.96 & 7$\sigma$ & Confirmed \\
1774: %4.71 & 13.12 & 6.77 & 7$\sigma$ & Confirmed \\
1775: %8.61 & 13.26 & 5.68 & 7$\sigma$ & Confirmed \\
1776: %8.65 & 13.00 & 6.30 & 7$\sigma$ & Confirmed \\
1777: %9.73 & 12.80 & 6.06 & 7$\sigma$ & Noticed \\
1778: %2.46 & 13.38 & 6.68 & 5$\sigma$ & Unnoticed \\
1779: %4.62 & 13.04 & 4.35 & 5$\sigma$ & Noticed \\
1780: %6.90 & 12.60 & 6.87 & 5$\sigma$ & Confirmed \\
1781: %7.93 & 13.59 & 3.31 & 5$\sigma$ & Noticed \\
1782: %8.42 & 7.08 & 4.39 & 5$\sigma$ & Noticed \\
1783: %9.53 & 13.60 & 5.02 & 5$\sigma$ & Confirmed \\
1784: %9.73 & 13.94 & 5.18 & 5$\sigma$ & Unnoticed \\
1785: %\enddata
1786: %\tablecomments{Planets confirmed: 13/13 at 10$\sigma$; 5/9 at 7$\sigma$;
1787: %2/7 at 5$\sigma$. Planets noticed: 13/13 at 10$\sigma$; 8/9 at 7$\sigma$;
1788: %5/7 at 5$\sigma$. Planet magnitude to mass conversion carried out
1789: %using the models of \citet{bur}.}
1790: %\end{deluxetable}
1791:
1792:
1793: %\begin{deluxetable}{rrrcc}
1794: %\tabletypesize{\small}
1795: %\tablecaption{$\epsilon$ Eri L'-band fake planet experiment.\label{tab:erilsimv}}
1796: %\tablewidth{0pt}
1797: %\tablehead{\colhead{Sep} & & \colhead{Mass} & \colhead{Input} & \colhead{Detection}\\
1798: %\colhead{(asec)} & \colhead{$L'$ Mag} & \colhead{(MJ)} & \colhead{Significance} & \colhead{Status}}
1799: %\startdata
1800: %0.87 & 10.15 & $>$20 & 10$\sigma$ & Confirmed \\
1801: %0.88 & 10.18 & $>$20 & 10$\sigma$ & Confirmed \\
1802: %1.41 & 12.20 & 12.54 & 10$\sigma$ & Confirmed \\
1803: %1.76 & 12.87 & 9.96 & 10$\sigma$ & Confirmed \\
1804: %1.80 & 13.02 & 9.46 & 10$\sigma$ & Confirmed \\
1805: %2.71 & 13.71 & 7.40 & 10$\sigma$ & Confirmed \\
1806: %4.62 & 14.95 & 4.97 & 10$\sigma$ & Confirmed \\
1807: %5.06 & 15.26 & 4.59 & 10$\sigma$ & Confirmed \\
1808: %5.46 & 14.72 & 5.35 & 10$\sigma$ & Confirmed \\
1809: %6.11 & 14.51 & 5.70 & 10$\sigma$ & Confirmed \\
1810: %7.11 & 14.54 & 5.65 & 10$\sigma$ & Confirmed \\
1811: %7.24 & 14.99 & 4.92 & 10$\sigma$ & Confirmed \\
1812: %9.88 & 15.00 & 4.91 & 10$\sigma$ & Confirmed \\
1813: %2.48 & 14.46 & 5.78 & 7$\sigma$ & Noticed \\
1814: %3.11 & 14.47 & 5.76 & 7$\sigma$ & Confirmed \\
1815: %5.16 & 15.32 & 4.52 & 7$\sigma$ & Confirmed \\
1816: %9.35 & 14.94 & 4.99 & 7$\sigma$ & Confirmed \\
1817: %10.56 & 14.89 & 5.06 & 7$\sigma$ & Confirmed \\
1818: %2.07 & 13.43 & 8.16 & 5$\sigma$ & Noticed \\
1819: %3.89 & 15.63 & 4.14 & 5$\sigma$ & Noticed \\
1820: %5.66 & 15.99 & 3.79 & 5$\sigma$ & Confirmed \\
1821: %7.38 & 14.76 & 5.28 & 5$\sigma$ & Confirmed \\
1822: %7.97 & 15.40 & 4.42 & 5$\sigma$ & Noticed \\
1823: %8.15 & 15.73 & 4.02 & 5$\sigma$ & Confirmed \\
1824: %10.87 & 14.49 & 5.73 & 5$\sigma$ & Unnoticed \\
1825: %\enddata
1826: %\tablecomments{Planets confirmed: 13/13 at 10$\sigma$; 4/5 at 7$\sigma$;
1827: %3/7 at 5$\sigma$. Planets noticed: 13/13 at 10$\sigma$; 5/5 at 7$\sigma$;
1828: %6/7 at 5$\sigma$. Planet magnitude to mass conversion carried out
1829: %using the models of \citet{bur}.}
1830: %\end{deluxetable}
1831:
1832: %\begin{deluxetable}{rrrcc}
1833: %\tabletypesize{\small}
1834: %\tablecaption{$\epsilon$ Eri M-band fake planet experiment.\label{tab:erimsimv}}
1835: %\tablewidth{0pt}
1836: %\tablehead{\colhead{Sep} & & \colhead{Mass} & \colhead{Input} & \colhead{Detection}\\
1837: %\colhead{(asec)} & \colhead{$M'$ Mag} & \colhead{(MJ)} & \colhead{Significance} & \colhead{Status}}
1838: %\startdata
1839: %0.59 & 6.20 & $>$20 & 10$\sigma$ & Confirmed \\
1840: %0.77 & 9.92 & $>$20 & 10$\sigma$ & Confirmed \\
1841: %1.15 & 11.20 & 9.61 & 10$\sigma$ & Confirmed \\
1842: %1.64 & 11.93 & 5.60 & 10$\sigma$ & Confirmed \\
1843: %1.73 & 12.48 & 4.05 & 10$\sigma$ & Confirmed \\
1844: %3.50 & 12.69 & 3.65 & 10$\sigma$ & Confirmed \\
1845: %4.00 & 12.69 & 3.65 & 10$\sigma$ & Confirmed \\
1846: %5.25 & 13.02 & 3.03 & 10$\sigma$ & Confirmed \\
1847: %6.58 & 12.63 & 3.76 & 10$\sigma$ & Confirmed \\
1848: %9.91 & 12.69 & 3.65 & 10$\sigma$ & Confirmed \\
1849: %0.62 & 6.89 & $>$20 & 7$\sigma$ & Confirmed \\
1850: %2.46 & 13.19 & 2.83 & 7$\sigma$ & Confirmed \\
1851: %2.85 & 13.10 & 2.93 & 7$\sigma$ & Noticed \\
1852: %3.43 & 13.39 & 2.60 & 7$\sigma$ & Confirmed \\
1853: %3.64 & 13.49 & 2.48 & 7$\sigma$ & Confirmed \\
1854: %4.05 & 13.10 & 2.93 & 7$\sigma$ & Confirmed \\
1855: %5.23 & 13.49 & 2.48 & 7$\sigma$ & Confirmed \\
1856: %5.87 & 12.85 & 3.35 & 7$\sigma$ & Confirmed \\
1857: %1.79 & 12.72 & 3.59 & 5$\sigma$ & Confirmed \\
1858: %3.77 & 13.77 & 2.16 & 5$\sigma$ & Noticed \\
1859: %4.15 & 13.30 & 2.70 & 5$\sigma$ & Noticed \\
1860: %4.72 & 13.85 & 2.07 & 5$\sigma$ & Noticed \\
1861: %6.00 & 13.73 & 2.21 & 5$\sigma$ & Confirmed \\
1862: %6.53 & 13.52 & 2.45 & 5$\sigma$ & Noticed \\
1863: %6.65 & 13.66 & 2.29 & 5$\sigma$ & Confirmed \\
1864: %\enddata
1865: %\tablecomments{Planets confirmed: 10/10 at 10$\sigma$; 7/8 at 7$\sigma$;
1866: %3/7 at 5$\sigma$. Planets noticed: 10/10 at 10$\sigma$; 8/8 at 7$\sigma$;
1867: %7/7 at 5$\sigma$. Planet magnitude to mass conversion carried out
1868: %using the models of \citet{bur}.}
1869: %\end{deluxetable}
1870:
1871:
1872: %The masses in Tables \ref{tab:vegalsimv} through \ref{tab:erimsimv} are calculated
1873: %by converting our sensitivity values
1874: %to $L'$ and $M$ band magnitudes and then converting the magnitudes to planet
1875: %masses in Jupiter Masses (MJ) by interpolating the planet models of \citet{bur}
1876: %for the appropriate mass, age, and distance of the two stars.
1877: %We have adopted an age of 0.56 Gyr \citep{fischer} and a distance of
1878: %3.27 pc \citep{hip} for $\epsilon$ Eri;
1879: %and an age of 0.3 Gyr \citep{agevega} and a distance of 7.756 pc \citep{hip} for Vega.
1880:
1881: %The end result of the three blind sensitivity tests was that at 10$\sigma$, 50 of
1882: %50 total inserted planets were confirmed. At 7$\sigma$,
1883: %23 of 30 total inserted sources were confirmed and 29 of the 30 were at least noticed.
1884: %At 5$\sigma$, 11 of 27 total inserted planets were confirmed, and 23 of the 27
1885: %were at least noticed. These statistics give us 100\% completeness at 10$\sigma$,
1886: %77\% completeness at 7$\sigma$, and a 41\% completeness at 5$\sigma$.
1887: %The percentages of fake planets that were at least noticed is of potential
1888: %interest for setting limits: 100\% of 10$\sigma$ sources, 97\% of 7$\sigma$
1889: %sources, and 85\% of 5$\sigma$ sources were at least noticed.
1890: %We note that if we had quoted 5$\sigma$ sensitivities without conducting a
1891: %blind sensitivity test we would have significantly overestimated our true
1892: %sensitivity. Most papers in the field of planet-imaging surveys do in
1893: %fact quote 5$\sigma$ limits, and do not verify their validity by
1894: %a blind test.
1895:
1896:
1897: %Letter to Phil and Matt about this version:
1898: %Hello Phil and Matt,
1899: % I've attached the Vega paper. Phil, I think I've attended to all
1900: %your edits without exception, and additionally did a lot of editing of
1901: %my own. This was mostly polishing and shortening -- the fundamental
1902: %content of the paper hasn't changed. I removed the tables giving the
1903: %masses and separations of the fake planets. I clarified some
1904: %parts, fixed errors of spelling, grammar, and unclear phrasing,
1905: %and stated the possibility of detecting eps Eri b in 2010 with Clio
1906: %more strongly as one of the most important results of the paper.
1907: %I also added Suresh as a coauthor. I think he deserves it, especially
1908: %since the Vega obs. were obtained in April 2006, the first really
1909: %successful run with Clio -- which worked partly because of all his
1910: %effort. Phil, if you agree please email Suresh the paper so he can
1911: %look it over as well. Please send it to Doug also.
1912: % I think the paper is essentially ready to submit. I won't submit
1913: %it without the OK from both of you and from Doug and Suresh, but
1914: %I'd encourage you to get back to me as soon as possible -- please pass this
1915: %encouragement on to Suresh and Doug. If you only have time to
1916: %read the abstract and the conclusions, and don't find anything there
1917: %that worries you, the paper is probably OK: I put the most potentially
1918: %dangerous stuff --- the strong claims about Clio's unique ability to
1919: %image eps Eri b in 2010 --- in the abstract and the conclusion.
1920:
1921: %Clear skies,
1922: %Ari
1923:
1924: %Original names of figures
1925:
1926: %vega_errgrid01.eps -> f1.eps
1927: %vegam_errgrid01.eps -> f2.eps
1928: %GJ144S_errgrid01.eps -> f3.eps
1929: %GJ144M_errgrid01.eps -> f4.eps
1930: %VegaL_mag01.eps -> f5.eps
1931: %VegaL_mass01.eps -> f6.eps
1932: %VegaM_mag02.eps -> f7.eps
1933: %VegaM_mass01.eps -> f8.eps
1934: %EriL_mag01.eps -> f9.eps
1935: %EriL_mass01.eps -> f10.eps
1936: %EriM_mag01.eps -> f11.eps
1937: %EriM_mass01.eps -> f12.eps
1938: %VegaComp01.eps -> f13.eps
1939: %EriComp01.eps -> f14.eps
1940:
1941: %Later, all of these figure numbers were increased by one to permit
1942: %the addition of a new figure1, which was originally
1943: %called vegamtilefig02.eps.