Temporal Evolution of Coronagraphic Dynamic Range, and Constraints on Companions to Vega

Accepted to ApJ

Temporal Evolution of Coronagraphic Dynamic Range, and
Constraints on Companions to Vega

Sasha Hinkley

O .

O . Department of Astronomy, Columbia University, 550 West 120th Street, New York, NY

10027

Oh.
(D

C/^ ■ shinkleyOastro.columbia.edu
(N

Ben R. Oppenheimer, Remi Soummer^'^, Anand Sivaramakrishnan^'^

, Astrophysics Department, American Museum of Natural History

!>■ ; Central Park West at 79th Street, New York, NY 10022

cn
m

Q\ • Lewis C Roberts Jr.

O

o
o

^ : 96822

The Boeing Company, 535 Lipoa Parkway, Suite 200, Kihei, HI 96753

Jeffrey Kufin

P ! Institute for Astronomy, University of Hawaii, 2680 Woodlawn Drive, Honolulu, Hawaii

Russell B. Makidon^

• 1—1 -
><

;h ' Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218

!

Marsliall D. Perrin^

Department of Astronomy, 601 Campbell Hall, University of California Berkeley, CA 94720

James P. Lloyd

Astronomy Department, 230 Space Sciences Building, Cornell University, Ithaca, NY 14853

Kaitlin Kratter^, Douglas Brenner

Astrophysics Department, American Museum of Natural History
Central Park West at 79th Street, New York, NY 10024

- 2 -

ABSTRACT

The major obstacle to the direct detection of companions to nearby stars is
the overwhelming brightness of the host star. Current instruments employing
the combination of adaptive optics (AO) and coronagraphy can typically detect
objects within 2" of the star that are ~ 10^~^ times fainter. Correlated speckle
noise is one of the biggest obstacles limiting such high-contrast imaging. We have
obtained a series of 284 8 s, AO-corrected, coronagraphically occulted i^-band
images of the star Vega at the 3.63 m AEOS telescope located on Haleakala,
Hawaii. This dataset is unique for studying the temporal behavior of speckle
noise, and represents the first time such a study on highly corrected coronagraphic
AO images has been carried OTit in a quantitative way. We find the speckle
pattern to be highly stable in both position and time in our data. This is due
to the fact that the AO system corrects disturbances to the stellar wave front
at the level where the instrumental wave front errors dominate. Because of
this, we find that our detection limit is not significantly improved simply with
increased exposure time alone. However, we are able to improve our dynamic
range by 1.5-2 magnitudes through subtraction of static/quasi-static speckles
in two rotating frames: the telescope pupil frame and the deformable mirror
frame. The highly stable nature of speckles will exist for any program using a
combination of coronagraphy and high-order AO. and underscores the importance
of calibration of non-common path errors between the wave front sensor and the
image plane. Such calibration is critical for high-contrast AO systems and we
demonstrate this using empirical data. From our data, we are able to constrain
the mass of any purported companion to Vega to be less than ~ 45Mj at 8 AU
and less than ~ SOMj at 16 AU, radii not previously probed at these sensitivities.

Subject headings: instrumentation: adaptive optics — methods: data analysis —
stars: individual (HD172167) techniques: image processing — stars: planetary
systems

^NSF Center for Adaptive Optics.
^Michelson Post-doctoral Fellow.
^Michelson Graduate Student Fellow.

^Current address: Department of Astronomy, University of Toronto, 60 St. George Street, Toronto, ON
M5S 3H8

^ Stony Brook University

-3-

1. Introduction

While the indirect detection of more than 200 planets over the past decade (Mayor et al.
2005; Marcy et al. 2005) has been a significant accomplishment for exoplanetary science, a
current major thrust for the field is the direct detection of these objects through high-
contrast imaging. Very large telescope apertures, although extremely effective for gathering
photons (Lloyd 2002), are not a necessity for direct imaging (Oppenheimer et al. 2003). The
key requirement is the suppression of the host star's overwhelming flux. A "hot Jupiter"
will typically be 10^-10^ times fainter than its host (Baraffe et al. 2003; Burrows 2005) and
often situated within a fraction of an arcsecond of the star. A promising method for direct
imaging of stellar companions involves two techniques working in conjunction: high-order
adaptive optics (AO), providing control and manipulation of the starlight (Angel 1994), and
an optimized Lyot coronagraph (Sivaramakrishnan et al. 2001; Oppenheimer et al. 2004)
to suppress this fight. Together, these two techniques can obtain contrast levels of 10^ or
better within 2". After this level of contrast has been achieved, the major source of noise
in the images is correlated speckle noise. Without a coronagraph, Racine et al. (1999) has
demonstrated that speckle noise will dominate over photon noise by a factor ~ 10^. With a
coronagraph, the contribution from the "pinned" speckle noise can be significantly reduced
(Sivaramakrishnan et al. 2002; Perrin et al. 2003; Aime & Soummer 2004). Such speckle noise
is largely due to non-common path errors (those not measured by the wave front sensor) e.g.
small aberrations in the coronagraphic optics. Other sources include AO correction errors
such as fitting error, aliasing, lag (Jolissaint et al. 2006) and Fresnel wavefront propagation
effects.

The Lyot Project (Oppenheimer et al. 2003; Oppenheimer et al. 2004; Oppenheimer
et al. 2006a, in preparation) employs an optimized, diffraction-limited Lyot coronagraph
(Lyot 1939; Malbet 1996; Sivaramakrishnan et al. 2001). It is deployed at the D — 3.63 m
Advanced Electro-Optical System (AEOS) telescope in Maui, with an AO system equipped
with a 941 actuator deformable mirror (DM). The telescope is an altitude-azimuth design,
with a beam traveling under the dome floor to a coude room containing the AO system
and the DM (Roberts & Neyman 2002). During an observation, the DM frame of reference
is always aligned with the frame of our infrared camera. The frame of the telescope pupil
containing the secondary mirror support struts (the "spiders"), however, rotates with respect
to the camera during an observation. For the purposes of this study, we mention that these
frames contain two identifiable sources of speckle noise in our observations.

In March 2004, the project began a survey of approximately 100 nearby stars in corona-
graphic mode in the J,H, and Ks bands with the goal of detecting faint companions and disks
orbiting the stars. On 2005 May 14 (UTC), we obtained a sequence of 284 AO-corrected,

-4-

coronagraphically occulted images of the bright star Vega (HD172167, AOV, V — 0"*). We
estimate our uncorrected wavefront errors at ~ 150nm, and we observed with adequate see-
ing (ro = 14.1 ± 2.5cm). With atmospheric phase errors well under control by the AO,
any speckle noise should come largely from the AO system or the coronagraph itself. Such
noise largely arises from non-common path errors between the wave front sensor and the
image plane, and can be mitigated through careful calibration (Wallace et al. 2004; Hartung
et al. 2003; Blanc et al. 2003). Although quasi-static speckle noise has been shown to be
the main limitation to earlier instruments (Marois et al. 2005; Boccaletti et al. 2003, 2004)
this dataset is unique for studying the temporal characteristics of correlated speckle noise in
such a highly corrected system coupled to an optimized Lyot coronagraph.

After describing our observations (§2), we will describe our method for quantifying our
sensitivity in the following section. We will show that our sensitivity does not increase
substantially when many images are simply coadded together, but demonstrate that the
dynamic range can be improved by subtracting those speckles that are static in the telescope
pupil frame and also those that are fixed with respect to the DM.

2. Observations

We obtained a sequence of 284 AO-corrected, 8 s images of the star Vega in the if-band
(1.65 urn) with the AEOS 3.63 m telescope on 2005 May 14 (UTC). The AO system's 941
actuator deformable mirror is complemented by a tip/tilt loop capable of running up to
~ 4 kHz, a Shack-Hartmann wave front sensor with a 2.5 kHz frame rate, and a real-time
wave front reconstructor using least-squares calculations performed on dedicated hardware
(Roberts & Neyman 2002). This combination of features has the potential for some of the
highest-order correction in modern AO. For all observations, the AO loop was closed and all
images were fully occulted using our focal plane mask with a 455 mas diameter (4.9A/L').
Although the theoretical if-band full-width at half maximum (FWHM) is 94 mas for AEOS
{D — 3.63 m), the effective FWHM is 121 mas when the size of the Lyot stop is taken into
account {DLyot — 2.81 m). The images were recorded with the Kermit IR Camera (Perrin
et al. 2002) in the H-hand with a 13.5 mas/pixel platescale. During the observations, the
star covered an elevation range between ~ 72° and ~ 67°, and thus suffered a differential
atmospheric refraction offset (Roe 2002) of 60 mas between its iJ-band and its y-band (0.55
/xm) position on the sky. To correct for this, we applied offsets (~ 10 mas each) to the tip-tilt
mirror to obtain the centered i^-band occulting position. In addition, our Lyot coronagraph
has the capacity for simultaneous dual beam polarimetery, with the goal of obtaining Stokes
/, Q, and U images (Oppenheimer et al. 2006b, in preparation). All of the data analysis for

- 5 -

this paper was performed on Stokes / images (the measure of total intensity) . We will save
the in-depth analysis of the polarimetric sensitivity for another work.

3. Analysis &6 Interpretation

The raw data images required a mix of both traditional data reduction steps as well
as some techniques customized for the Kermit Camera. Each of the 284 raw images were
cropped to a region containing the star, dark subtracted, flat field corrected, and cleaned
for bad pixels and cosmic rays through interpolation. Due to cross-talk within the detector
electronics, the raw data also contained easily-removable negative electronic "ghost" images
of the occulted star about 16 times less intense at 128-pixcl intervals from the true position
of the star. Each image was then rotated so that the image y-direction is coincident with
North on the sky. After the basic reduction steps were complete, 284 new coadded images
were formed in which the n*'* coadded image was the mean of the first n individual images.
The image representing the mean of 284 images is shown in the middle of Figure 1.

The dynamic range or contrast at a given position in a two-dimensional image is com-
monly quantified as the faintest companion detectable at that location at the 5a level (Op-
penheimer et al. 2003). We use this convention. In these data, the main noise contribution
is due to the speckles, and has to be estimated locally, from the data itself. In addition,
given that a typical PSF in these data has an oversamphng factor of 7-9 within X/D, the
signal should not be measured from a single point (e.g. the brightest pixel) but rather should
take into account all the pixels within one resolution element. So, at a given location, cal-
culating the dynamic range consists of measuring the signal from a hypothetical companion
by integrating its flux over a patch of size \/D. To remain consistent, this signal needs to
be compared to the noise in the same integrated area if no real PSF exists at that location,
or in the immediate adjacent area of the image. There are alternatives to this particular
method of calculation, but they agree to within 0.3 magnitudes (Soummer et al. 2006).

We constructed a dctcctability map with the same number of pixels and pixel scale as
each science image, indicating the image's 5cr detection limit as a function of position in the
image. The corresponding 5a map for the n — 284 image is shown on the right of Figure 1.
Each pixel in such an image represents five times the root-mean-square variation of the pixel
values in a circular subregion centered on the corresponding pixel in the occulted image.
The subregion was chosen to have a 0.28" diameter (somewhat larger than the if-band point
spread funtion). Each pixel in such a detection limit map represents the minimum flux
necessary for a 5a detection in the co-added image. Thus, these maps show the dynamic
range of the image.

-6-

Due to Vega's brightness, unocculted calibration images of it were not available to
determine apparent magnitudes. Thus, each of the detection limit maps was calibrated to
a fainter, unocculted standard star (HD160346, K3V, V = 6.5™) observed at nearly the
same time and airmass as Vega. The photometric zeropoint was determined by comparing
the calibration star's tabulated i/-band magnitudes with its fiux counts. With this value
in hand, image counts could then be converted to apparent magnitudes. We calculated an
if-band zeropoint of 20.18 magnitudes and use this value for all further analysis in this
paper. In addition, a short sequence of images of this same star was used to derive H-hand
Strehl ratios near 60%. However, during the Vega observations, the AO system was likely
performing somewhat better than this, given Vega's brightness.

In order to determine our detection limits, we studied how the dynamic range and
structure of this map changes with the value of n, the number of images added together.
This is equivalent to studying dynamic range as a function of total exposure time. To do this,
we calculate azimuthally-averaged dynamic range values at many different radial positions
in the image. We plot these values (dashed lines) as a function of exposure time for two
different radii in Figure 2.

It is immediately apparent that these curves are flat. They do not seem to reach substan-
tially fainter magnitudes with longer effective integrations. Any further gain in sensitivity
through subsequent observations seems to be essentially negligible. The gain in dynamic
range over time docs depend on the radial position in the image as shown in Figure 3. At
300 mas, the increase over the course of 2300 s is only ~ 0.2 magnitudes. However, at
1500 mas, the increase over the same time period increases to ~ 0.7 magnitudes.) A certain
degree of the increased dynamic range within this region may be due to the AO control
cutoff, given by \Nact/'2.D (Sivaramakrishnan et al. 2001; Oppenheimer et al. 2003, 2006a,
in preparation; Poyneer & Macintosh 2004), where Nact is the number of hnear actuators
across the pupil (34 for AEOS). For AEOS, this control cutoff is ~ 1500 mas, near where
the solid curve in Figure 3 peaks.

This relative plateau in sensitivity is due to the inherently correlated nature of the
speckle structure from frame to frame, leading to high spatial and temporal stability. Since
the amplitude of this variation in intensity is highly stable across an image (Marois et al.
2003), a number of short exposures added together will not cause the noise to average out
as e.g. Racine et al. (1999) has suggested for the case of atmospheric speckles. Without
subtracting out this source of noise. Figure 2 shows that the fixed speckles dominate the
overall dynamic range.

- 7-

3.1. Subtraction of static speckles

The static and quasi-static speckles that hmit the dynamic range arise primarily from
two sources: those speckles pinned to the diffraction pattern of the secondary support struts
(the "spiders" ) in the telescope pupil, and those that are due to the quasi-static imperfections
in the AO optics. The pupil frame will rotate with respect to the infrared detector, while the
AO frame (DM) is fixed relative to it during an observation. To get a handle on the degree
of speckle stability, we performed two experiments on the data, each an attempt to correct
for the two sources. We performed the experiments on the data after its pre-processing,
but prior to forming the coadded images and carrying out the rest of the dynamic range
calculation. We address the first of these experiments in this section.

In order to quantify the degree to which we can subtract out the longest-lived speckles
due to the spiders, we "de-rotated" each of the 284 processed images by their parallactic
angle offset so that the pupil frame pattern had the same orientation in each image. The
sky thus rotates in each image. These rotated images were median combined into a single
image, and this median image was subtracted from each of its constituent images. This step
is identical to the first of two steps carried out by Marois et al. (2006) to achieve a gain
in dynamic range in high-contrast data. Next, all these constituent images were rotated so
that North/East were ahgned with up/left. With this new data set, the exact same dynamic
range calculation was performed (sequential coadds formed, 5a detection map formed, etc.)
and the results are shown in Figure 2 (dashed-dot curve). An increase of 0.5-1.0 magnitudes
is seen over the case with no speckle subtraction. Moreover, Figure 3 shows that the gain
in dynamic range during the nearly 2300 s observation varies from 0.8 magnitudes nearer to
the star (~ 500 mas) to ~ 1.3 magnitudes at greater separations (~ 1500 mas).

It should be noted, of course, that the measured improvement in dynamic range is
not valid at all radii. The boost in dynamic range is only valid in those regions beyond
which a potential companion in the now rotating sky frame will rotate enough through the
sequence so that it does not become part of the median image and is subsequently subtracted
out (Marois et al. 2006). Given the 40° rotation of the parallactic angle and assuming a
minimum necessary companion separation of 1.5 FWHM over the sequence, this defines an
inner working angle of 250 mas, close to the edge of the occulting mask used.

3.2. Lifetimes of quasi-static speckles

Once these truly static speckles were subtracted out of the image, we studied the lifetime
of those remaining quasi-static speckles. To do this, we performed a pixel-wise temporal

-8-

autocorrelation analysis similar to that in Fitzgerald & Graham (2006). We calculated the
temporal autocorrelation of each pixel over the entire image. The autocorrelation function,
shown in Figure 4 for three sample locations in the image, reveal two distinct time scales. As
a convention, we define these time scales by the half-width of the two regions separated by
the "knee" in the autocorrelation function near 20 s shown schematically in Figure 4. The
first of these, which we denote by Tghort, represents any rapid decorrelation of the speckles
that may exist. This decorrelation does correspond to a slight increase in the dynamic range,
and this can be seen at the very beginning of the curves in Figure 2. The longer lifetime,
denoted in the image by Tiong, gives a reasonable measure of the lifetime of the quasi-static
contribution. A map of both these hfetimes is shown in Figure 5. The maps reveal the
localization of the slowest and fastest speckles: the longest lifetimes of quasi-static speckles
are ~ 400 s and the shortest lifetimes of the short-term speckles are 1 - 5 s. Although
significant work has been done on the lifetimes of atmospheric speckle lifetimes (Roddier
et al. 1982; Macintosh et al. 2005), very little formalism exists for static and quasi-static
speckle lifetimes.

Also in Figure 5, we show radial profiles of the speckle lifetime maps. Although no
particularly strong radial features are evident in the long-term lifetime map, there still is
a notable increase in the lifetimes near 500 mas, reflecting those relatively static speckles
that are pinned just outside the coronagraphic mask. Also, the radial curves reveal why
we achieve slightly more dynamic range towards the outer parts of the images. Specifically,
the average quasi-static speckle lifetime near 740 mas is ~ 210 s, while at 1760 mas it is
~ 150 s. Averaging the shorter lifetime speckles will achieve more dynamic range increase
than averaging longer ones over the same period of time. This is well refiected in the curves
of Figures 2 and 3, where the increase in dynamic range with integration time is ~ 0.8
magnitudes at 740 mas and ~ 1.4 magnitudes at 1760 mas.

In the short-term plot, the lifetimes of those speckles closer to the star are smaller.
Although a definitive physical mechanism underlying this remains unclear, this behavior
is reminiscent of the discussion in Macintosh et al. (2005), in which simulations of pinned
speckles — in our case, residual pinned speckles close to the star — fiuctuate quickly with a
period dictated by the wind speed carrying the wave front error. The type of approach
shown in Figure 5 may serve as a useful tool for future work aiming to further explore the
physical nature behind speckle lifetimes.

-9-

3.3. Subtraction of DM speckles

After the subtraction of the truly static speckles in the frame of the telescope pupil, an
attempt was made to subtract out another contributor to the speckle pattern: those speckles
caused by static aberrations in the adaptive optics system, coronagraph, or science camera.
In a similar manner to that described in Section 3.1, the pupil frame subtracted images
(north aligned up, east aligned left) were de-rotated to the frame in which the placement of
the DM is fixed with respect to the IR array. This is the same frame of reference occupied
by the images when they initially come straight from the infrared camera. Similarly, these
images were median combined into a single image and this image was subtracted from each
of the images in the DM frame. The newly subtracted images were rotated back to their
normal "sky" coordinates (north-up and east-left) and we calculated the dynamic range in
exactly the same way.

The evolution of the dynamic range after both (pupil frame and DM frame) subtractions
is shown in Figure 2 (solid line). Depending on the location in the image plane, the gain
achieved with these two subtractions over simply coadding images (no subtraction at all)
is ~ 1.5-2.0 magnitudes. The increase in sensitivity over different radial positions in the
image is shown more fully in Figure 3. For this second case, we are not allowed as generous
an inner working angle. The transformation between the "sky" frame and the "DM" frame
only spans 12° over the sequence of 284 images. This rather small angle corresponds to an
angular separation of ~ 700 mas outside of which the PSF of a true companion will not be
incorporated into the median image and subsequently subtracted. Hence, improvements to
the dynamic range due to this method are only valid outside this separation. Because of
this, an observing sequence which incorporates the maximum amount of field rotation will
benefit this kind of data reduction.

3.4. Dynamic Range

Figure 7 shows azimuthally-averaged radial plots of our dynamic range incorporating all
images for each of the three cases: no speckle subtraction, pupil frame speckle subtraction,
and both pupil+DM speckles subtracted. It should be noted that for the case of no speckle
subtraction, the dynamic range will still contain some of the highly non-axisymmetric struc-
ture due to the telescope "spiders" . However, the curves for the two speckle-subtracted cases
do not suffer as much from this, and the azimuthally averaged curves are more justified. The
speckle noise will be most prominent close to the star where residual speckle pinning per-
sists due to the insufficient classical Lyot coronagraph starhght suppression (Bloemhof et al.
2001; Sivaramakrishnan et al. 2002; Aime & Soummer 2004). An upgrade using an apodized

-10-

pupil Lyot coronagraph (Soummer 2005) will help the dynamic range in this region. At ~ 2
arcseconds, our H-hand detection limit is approaching AM > 12.5, where the speckle noise
has diminished in strength.

The figure also shows the sensitivities of VLT/NACO observations incorporating the
SDI technique (Biller et al. 2006; Kasper et al. 2005), as well as the Gemini Altair observa-
tions with the Angular Differential Imaging technique (Marois et al. 2006) applied to them.
Although the other projects' curves are for stars other than Vega, and detection diagnostics
in each of these systems will be different and subject to peculiarities associated with each
(non- uniformity in exposure times, different wavelengths and telescopes, etc.), these curves
still place a valuable benchmark for our sensitivity. We feel that these two projects are
strong representations of the current state-of-the-art in high dynamic range imaging. Our
results demonstrate that a 3.63 m telescope with a high-order adaptive optics system and a
coronagraph can provide comparable contrast to that achievable with 8 m telescopes. This
is highly promising for future high-contrast AO systems on 8-10 m telescopes.

However, the relative lack of substantial "halo-clearing" within the AO control cutoff
also arises from several hmiting factors intrinsic to the AO system itself. For one, the AEOS
deformable mirror (DM) was hampered by five malfunctioning actuators. Oppenheimer et

al. (2005) have shown that just a single malfunctioning actuator in a DM can effectively
contribute enough aberration to significantly increase the quasi-static speckle noise, thereby
limiting high-contrast work. Due to the presence of these broken actuators, the DM was
operated with only half of the actuator stroke it was designed for, to ensure that no additional
actuators are damaged. Operating in this mode prevents the system from correcting the full
tilt of the wave front over a given subaperture, especially in poor seeing. Moreover, the AEOS
wave front is not spatially filtered (Poyneer & Macintosh 2004) before wave front sensing,
which may cause its high-frequency spatial components to be interpreted as low-frequency
components by the wave front reconstructor. This results in degraded performance. All
of these effects can lead to the creation of a prominent speckle pattern, which supersedes
the "halo-clearing" effect of the AO control radius. Tests using a Zygo interferometer have
shown that the wave front error induced by the coronagraph alone amounts to no more
than 58 nm RMS (Oppenheimer et al. 2005). By itself, this error will only diminish the
Strehl ratio to 92%. This further suggests that the highly stable noise is arising from other
speckles/scattered light from the telescope/ AO system and not the coronagraphic optics.

- 11 -

4. Limits on Vega's companion

Earlier studies on the inner ~ 20 arcseconds of the Vega system have revealed asym-
metric dust patterns suggesting the system may possess a planet on the order of a few
Jupiter masses (Mj) with 5-10 arcsecond separation on an eccentric orbit (Holland et al.
1998; Wilner et al. 2002). Recent high contrast imaging work by Marois et al. (2006) and
Hinz et al. (2006) use M and 1.58//m imaging to place upper limits of 7 Mj beyond 2.5
arcseconds and 3 Mj beyond 8 arcseconds, respectively. This complements earlier work by
Metchev et al. (2003) and Macintosh et al. (2003) who used H and K-hernd imaging to place
an upper limit of 20 and 10 Mj, respectively, for a hypothetical companion located near 7
arcseconds.

The real value of our current work is our ability to constrain the mass of any unseen
massive companions within 2 arcseconds, corresponding to ~ 16AU for the Vega system.
Figures 2 and 7 show the limiting companion masses at the contrast level achieved. These
were calculated by matching our absolute magnitude detection limits with the models of
Baraffe et al. (2003) assuming an age of 300 Myr for the Vega system. Since we detect
no companions in our images. Figure 7 indicates we can rule out any companions more
massive than about 45 Mj at 1 arcsecond (8AU) and about 30 Mj at 2 arcseconds (16AU).
Figure 6 shows a two-dimensional representation of the upper limits to detectable companion
masses for each of the three methods of coadding: with no speckle subtraction whatsoever,
after the pupil-frame speckle subtraction, and after both the pupil-frame and DM speckle
subtraction. The increase in sensitivity towards lower masses is evident after the subtraction
of the static/quasi-static speckles. Moreover, we compare our mass limits at different angular
separations to that for several other recent imaging studies of Vega's purported companion
in Table 1.

Table 1. Sensitivity limits (in Jupiter masses) at different angular separations for several

selected prior imaging studies of Vega.

Work

Wavelength

Telescope, Technique 0.5"

1" 2"

4"

7"

10"

Macintosh et al. (2003)

K

Keck, AO

10

8

Metchev et al. (2003)

H

Palomar, AO

30

15-20

~ 12

Marois ct al. (2006)

1.58/nm

Gemini, ADI

5

4

3

Hinz ct al. (2006)

M

MMT, AO

26

7

7

7

Itoh et al. (2006)

H

Subaru, AO + coron.

120

7

5-10

This work

H

AEOS, AO + coron. 135

43 27

- 12 -

5. Conclusions

The relative floor in sensitivity shown in Figure 2 is due to the highly correlated and
persistent nature of the speckle noise in the images. In contrast to the way the random,
uncorrelated noise is suppressed through averaging, the data are subject to a static and
quasi-static speckle pattern, placing a limit on the sensitivity of an observing sequence.
Among other factors, such an influx of speckle noise is likely due to a known malfunction
in a subset of the AEOS deformable mirror actuators. In addition, high-order AO systems
in general can suffer from a variety of problems (misalignment, mirror figure errors, etc.),
each of which can be a significant source of speckle noise. We anticipate fixed speckles like
those discussed here to be a major obstacle for other AO+Coronagraphy programs with
similar objectives. But more importantly, this study underscores the primary importance
of non-common path errors between the wave front sensor and the science image plane as
other authors have discussed (Hartung et al. 2003; Blanc et al. 2003). Cahbration of these
errors is critical for AO imaging to exceed the ~ 70% Strehl level. We have demonstrated
this with empirical data.

This paper demonstrates the extent to which fixed speckles can dominate the overall
noise budget of high-contrast AO observations, far exceeding the contributions from residual
atmospheric effects. As we have shown, an increase in sensitivity (1.5-2 magnitudes) can be
obtained by subtracting off individual contributions to the speckle pattern. Another solution
may rely on the wavelength- dependent nature of the speckle noise. Although speckles remain
fixed in space for a given filter, they will reside in different locations in other passbands,
especially if they result from phase errors in or near pupil planes in the optical train. This
key feature will allow a faint companion — with a fixed position in all filters — to be selected
out from the speckle noise. Thus, the addition of multi-wavelength observations will enhance
the efficiency with which companions may be detected (Marois et al. 2000; Sparks & Ford
2002; Biller et al. 2004; Marois et al. 2005). Moreover, an integral field unit (Larkin et al.
2003) will allow the retrieval of spectra across an image, further disentangling companions
from the speckle noise.

The authors would fike to thank Christian Marois for suggesting the use of the ADI-fike
multi-plane rotation subtraction technique, Gilles Chabrier and Isabelle Baraffe for providing
their models, and Michal Simon for helpful discussions about this paper. The Lyot Project is
based upon work supported by the National Science Foundation under Grant Nos. 0334916,
0215793, and 0520822, as well as grant NNG05GJ86G from the National Aeronautics and
Space Administration under the Terrestrial Planet Finder Foundation Science Program. The
Lyot Project grateful acknowledges the support of the US Air Force and NSF in creating

-13-

the special Advanced Technologies and Instrumentation opportunity that provides access to
the AEOS telescope. Eighty percent of the funds for that program are provided by the US
Air Force. This work is based on observations made at the Maui Space Surveillance System,
operated by Detachment 15 of the U.S. Air Force Research Laboratory Directed Energy
Directorate. The Lyot Project is also grateful to the Cordelia Corporation, Hilary and Ethel
Lipsitz, the Vincent Astor Fund, Judy Vale and an anonymous donor, who initiated the
project. L.C.R is funded by AFRL/DE contract number F2901-00-D-0204. R.S and M.R
are supported by the NASA Michelson Fellowship Postdoctoral/Graduate Fellowship under
contract to the Jet Propulsion Laboratory (JPL) funded by NASA. The JPL is managed for
NASA by the Cahfornia Institute of Technology. This work has been partially supported
by the National Science Foundation Science and Technology Center for Adaptive Optics,
managed by the University of California at Santa Cruz under cooperative agreement AST-
9876783.

REFERENCES

Aime, C, & Soummer, R. 2004, ApJ, 612, L85
Angel, J. R. P. 1994, Nature, 368, 203

Baraffe, I., Chabrier, G., Barman, T. S., AUard, F., & Hauschildt, P. H. 2003, A&A, 402,
701

Biller, B. A., Close, L., Lenzen, R., Brandner, W., McCarthy, D. W., Nielsen, E., & Hartung,
M. 2004, in Advancements in Adaptive Optics. Edited by Domenico B. Calia, Brent
L. Ellerbroek, and Roberto Ragazzoni. Proceedings of the SPIE, Volume 5490, pp.
389-397 (2004)., ed. D. Bonaccini Calia, B. L. Ellerbroek, & R. Ragazzoni, 389-397

Biller, B. A., Close, L. M., Masciadri, E., Lenzen, R., Brandner, W., McCarthy, D., Henning,
T., Nielsen, E. L., Hartung, M., Kellner, S., Geissler, K., & Kasper, M. 2006, in Ad-
vances in Adaptive Optics 11. Edited by Brent L. Ellerbroek and Domenico Bonaccini
Caha. Proceedings of the SPIE, Volume 6272, pp. 627229 (2006).

Blanc, A., Fusco, T., Hartung, M., Mugnier, L. M., & Rousset, G. 2003, A&A, 399, 373

Bloemhof, E. E., Dekany, R. G., Troy, M., & Oppenheimer, B. R. 2001, ApJ, 558, L71

Boccaletti, A., Chauvin, G., Lagrange, A.-M., & Marchis, F. 2003, A&A, 410, 283

Boccaletti, A., Riaud, P., Baudoz, P., Baudrand, J., Rouan, D., Gratadour, D., Lacombe,
F., & Lagrange, A.-M. 2004, PASP, 116, 1061

-14-

Burrows, A. 2005, Nature, 433, 261

Fitzgerald, M. P., & Graham, J. R. 2006, ApJ, 637, 541

Hartung, M., Blanc, A., Fusco, T., Lacombe, F., Mugnier, L. M., Rousset, G., & Lenzen, R.
2003, A&A, 399, 385

Hinz, P. M., Heinze, A. N., Sivanandam, S., Miller, D. L., Kenworthy, M. A., Brusa, G.,
Freed, M., & Angel, J. R. P. 2006, ArXiv Astrophysics e-prints

Holland, W. S., Greaves, J. S., Zuckerman, B., Webb, R. A., McCarthy, C., Coulson, I. M.,
Walther, D. M., Dent, W. R. F., Gear, W. K., & Robson, I. 1998, Nature, 392, 788

Itoh, Y., Oasa, Y., & Fukagawa, M. 2006, ArXiv Astrophysics e-prints

Jolissaint, L., Veran, J. -P., & Conan, R. 2006, Journal of the Optical Society of America A,
vol. 23, Issue 2, pp.382-394, 23, 382

Kasper, M., Ageorges, N., Boccaletti, A., Brandner, W., Close, L. M., Davies, R., Finger, G.,
Genzel, R., Hartung, M., Kaufer, A., Kellner, S., Hubin, N., Lenzen, R., Ludman, C,
Monnet, G., Moorwood, A., Ott, T., Riaud, P., Roser, H.-J., Rouan, D., & Spyromilio,
J. 2005, The Messenger, 119, 11

Larkin, J. E., Quirrenbach, A., Krabbe, A., Aliado, T., Barczys, M., Brims, G., Canfield, J.,
Gasaway, T. M., LaFreniere, D., Magnone, N., Skulason, G., Spencer, M., Sprayberry,
D., & Weiss, J. 2003, in Instrument Design and Performance for Optical/Infrared
Ground-based Telescopes. Edited by lye, Masanori; Moorwood, Alan F. M. Pro-
ceedings of the SPIE, Volume 4841, pp. 1600-1610 (2003)., ed. M. lye & A. F. M.
Moorwood, 1600-1610

Lloyd, J. P. 2002, Ph.D. Thesis, University of California, Berkeley.

Lyot, B. 1939, MNRAS, 99, 580

Macintosh, B., Poyneer, L., Sivaramakrishnan, A., & Marois, C. 2005, in Astronomical
Adaptive Optics Systems and Apphcations II. Edited by Tyson, Robert K.; Lloyd-
Hart, Michael. Proceedings of the SPIE, Volume 5903, pp. 170-177 (2005)., ed. R. K.
lyson & M. Lloyd-Hart, 170-177

Macintosh, B. A., Becklin, E. E., Kaisler, D., Konopacky, Q., & Zuckerman, B. 2003, ApJ,

594, 538

Malbet, F. 1996, A&AS, 115, 161

- 15 -

Marcy, G., Butler, R. P., Fischer, D., Vogt, S., Wright, J. T., Tinney, C. G., & Jones,
H. R. A. 2005, Progress of Theoretical Physics Supplement, 158, 24

Marois, C, Doyon, R., Nadeau, D., Racine, R., Riopel, M., Vallee, P., & Lafreniere, D. 2005,
PASP, 117, 745

Marois, C, Doyon, R., Nadeau, D., Racine, R., & Walker, G. A. H. 2003, in EAS Publications
Series, ed. C. Aime & R. Soummer, 233-243

Marois, C., Doyon, R., Racine, R., & Nadeau, D. 2000, PASP, 112, 91

Marois, C, Lafreniere, D., Doyon, R., Macintosh, B., & Nadeau, D. 2006, ApJ, 641, 556

Mayor, M., Pont, F., & Vidal-Madjar, A. 2005, Progress of Theoretical Physics Supplement,
158, 43

Metchev, S. A., Hillenbrand, L. A., & White, R. J. 2003, ApJ, 582, 1102

Oppenheimer, B. R., Digby, A. P., Newburgh, L., Brenner, D., Shara, M., Mey, J., Mandev-
ille, C., Makidon, R. B., Sivaramakrishnan, A., Soummer, R., Graham, J. R., Kalas,
P., Perrin, M. D., Roberts, L. C., Kuhn, J. R., Whitman, K., & Lloyd, J. P. 2004, in
Advancements in Adaptive Optics. Edited by Domenico B. Calia, Brent L. EUerbroek,
and Roberto Ragazzoni. Proceedings of the SPIE, Volume 5490, pp. 433-442 (2004).,
ed. D. Bonaccini Calia, B. L. EUerbroek, & R. Ragazzoni, 433-442

Oppenheimer, B. R., Sivaramakrishnan, A., & Makidon, R. B. 2003, Imaging Exoplanets:
The Role of Small Telescopes (The Future of Small Telescopes In The New Millen-
nium. Volume III - Science in the Shadow of Giants), 155 — h

Oppenheimer et al. 2005, in 2005 AMOS Technical Conference, P. W. Kervin, J. L. Africano;
eds..

Perrin, M. D., Graham, J. R., Trumpis, M., Kuhn, J. R., Whitman, K., Coulter, R., Lloyd,
J. P., & Roberts, L. C. 2002, in 2002 AMOS Technical Conference, P. W. Kervin, J.
L. Africano; eds.

Perrin, M. D., Sivaramakrishnan, A., Makidon, R. B., Oppenheimer, B. R., & Graham, J. R.
2003, ApJ, 596, 702

Poyneer, L. A., & Macintosh, B. 2004, Optical Society of America Journal A, 21, 810
Racine, R., Walker, G. A. H., Nadeau, D., Doyon, R., & Marois, C. 1999, PASP, 111, 587
Roberts, L. C, & Neyman, C. R. 2002, PASP, 114, 1260

-16-

Roddier, F., Gilli, J. M., & Lund, G. 1982, Journal of Optics, 13, 263
Roe, H. G. 2002, PASP, 114, 450

Sivaramakrishnan, A., Koresko, C. D., Makidon, R. B., Berkefeld, T., & Kuchner, M. J.
2001, ApJ, 552, 397

Sivaramakrishnan, A., Lloyd, J. P., Hodge, P. E., & Macintosh, B. A. 2002, ApJ, 581, L59
Soummer, R. 2005, ApJ, 618, L161

Soummer, R., Oppenheimer, B. R., Hinkley, S., Sivaramakrishnan, A., Makidon, R. B.,
Digby, A. P., Brenner, D., Kuhn, J. R., Perrin, M. D., Roberts, L. C., & Kratter, K.
2006, in EAS Publications Series, ed. C. Aime & M. Carbillet

Sparks, W. B., & Ford, H. C. 2002, ApJ, 578, 543

Wallace, J. K., Green, J. J., Shao, M., Troy, M., Lloyd, J. P., & Macintosh, B. 2004, in
Advancements in Adaptive Optics. Edited by Domenico B. Calia, Brent L. Ellerbroek,
and Roberto Ragazzoni. Proceedings of the SPIE, Volume 5490, pp. 370-378 (2004).,
ed. D. Bonaccini Calia, B. L. Ellerbroek, & R. Ragazzoni, 370-378

Wilner, D. J., Holman, M. J., Kuchner, M. J., & Ho, P. T. P. 2002, ApJ, 569, L115

This preprint was prepared with the A AS IM^jX macros v5.2.

-17-

o

CD
CO

o

Single Image Coadded data Sensitivity nnap

-2-10 1 2-2-10 1 2
orcsec orcsec

log counts/sec

log counts/sec

log counts/sec

1.3

2.0

2.7 3.5

4.2

1.6

2.2

2.8 3.5

4.1

2.8 3.5 4.2 5.0

5.7

Fig. 1. — Left: This image shows a single i^-band 8 s exposure of the star Vega. The
occulting coronagraphic mask is in place and the AO control loop is fully closed. The speckle
pattern which limits the dynamic range is evident, and note that residual pinning is present
at regions close to the center, as well as on the telescope "spiders" . Also, the first Airy rings
are not completely removed by the Lyot coronagraph. Middle: This image shows all 284
exposures coadded together. Right: the corresponding 5a detection limit map constructed
from the middle image. Each pixel in this figure represents the minimum fiux necessary for
a 5(7 detection of a companion. For each image, the platescale is 13.5 milliarcseconds/pixel.

-18-

5 -

6 -

13

No subtraction

Static pupil speckle
pattern subtracted

DM + pupil speckle
pattern subtracted

1/Vt,

curve

740mas

1760mas

±

J 200 Mj

100 Mj
60 Mj

_ 50 Mj
_ 35 Mj

_ 25 Mj

500 1000 1500 2000 2500 3000
total exposure time (s)

Fig. 2. — Detection limits as a function of total exposure time for two different radial
positions in the field. Each group of lines represents the azimuthally-averaged 5a detection
limit at the radius listed (in milliarcseconds). By simply co-adding images together (dashed
line), our dynamic range does not improve substantially with exposure time due to to the
presence of highly stable speckles. Subtraction of the truly static portion of the speckle
pattern in the telescope pupil frame will improve the dynamic range by nearly a magnitude
(dashed-dot fine). Finally, another subtraction of those speckles caused by imperfections in
the deformable mirror and AO system increases the dynamic range still a bit more (solid line) .
The dotted line shows the 1/Vtexp curve, the expected contrast gain due to uncorrelated noise
with a Poisson distribution.

-19-

2 -

O

S 1

CO
C\2

5 -

1 -

0.5 -

DM + static pupil
pattern subtracted

— X

V

•>« ^
^ ^

~ Static pupil speckle
pattern subtracted

-

■ ■ ■ ■*

_ No speckle
subtraction

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

1 1 1 1 1

500 1000 1500

radial distance (mas)

2000

Fig. 3. — This plot shows the gain (minimum to maximum) in dynamic range over the full
~ 2300s observing sequence as a function of radial distance from the star. The solid line
shows the increase by simply coadding the images together, and only marginal improvement
is obtained due to the highly static nature of the speckle pattern. When the truly static
contribution in the frame of the telescope pupil is subtracted out (dotted line), the dynamic
range gain improves. Finally, when the speckle pattern due to the imperfections in the
DM are also removed (dashed hne), an increase of ~ 2 magnitudes can be acheived over
the course of the entire observing sequence. This last dynamic range increase is only valid
beyond ~ 700 mas, due to the inner working angle defined by the rotation in this frame.

-20-

time (s)

Fig. 4. — Temporal autocorrelation functions for three different pixel positions located at
0.19" (solid line), 0.10" (dashed line), and .23" (dotted line) from the star in the image
plane. Each of the autocorrelation functions are characterized by two distinct time scales,
here defined as the half-width of the two regions separated by the break near 20s. Each of the
three functions have a nearly identical TsHoh, but three distinct Tiong values. The time axis is
shown in a log scale to better illustrate the shorter timescale, t short- This shorter timescale
corresponds to a quick decorrelation of the speckles and the corresponding improvement in
the dynamic range can be seen in the earliest parts of the curves in Figure 2. The longer
timescale, Tiong gives a measure of the lifetime of the quasi-static component to the speckle
noise.

- 21 -

Long-term speckle lifetimes
2

Short-term speckle lifetimes
2

0.0 0.5 1.0 1.5 2.0
rodiol distonce (orcsec)

0.0 0.5 1.0 1.5 2.0
rodiol distonce (orcsec)

Fig. 5. — Top panels: maps of the derived lifetimes as shown in Figure 4 for the quasi-static
long (riong) and short-lived {rshort) speckles in the image plane. The lower two panels show
azimuthally averaged radial plots of the top panels with Icr error bars. The contours cor-
respond to 100, 150, 200, 300, and 400 seconds for the long-term map and 5, 15, 20, 35,
and 50 second intervals for the short-term map. The increase in the long-term (quasi-static)
lifetimes near 500 mas are due to those speckles that are "pinned" just outside the corona-
graphic mask. Also, the 40% relative difference in quasi-static speckle lifetimes between 740
mas and 1760 mas is likely responsible for the greater increase in dynamic range at these
two locations as shown in Figures 2 and 3 during the 2300 s total exposure time.

-22-

No subtroctions

Pupil speckles removed

Pupil + DM removed

orcsec

orcsec

orcsec

200

150 100

Jupiter Mosses

50

Fig. 6. — Dynamic range maps incorporating all 284 images. Each map shows the upper
limit on companion masses (in units of Jupiter masses) based on the models of Baraffe et al.
(2003). The left map shows the detectable companion masses with no speckle subtraction
whatsoever. The middle plot shows the sensitivity after subtracting the speckle pattern in
the telescope pupil frame, while the right plot shows the limits after both the pupil frame
pattern and those caused by the imperfections in the DM have been removed. The larger
black circle in the last plot shows the inner working angle (~ 700 mas) beyond which the
dynamic range measurements are valid when the DM speckles are removed (see text) .

-23-

radial distance (mas)

Fig. 7. — Azimuthally-averaged radial profiles of our H-hand detection limit (solid lines)
incorporating all 284 images. The curves show our sensitivity with no speckle subtraction
whatsoever, after the static pupil speckle pattern has been subtracted, and after those speck-
les due to the imperfections in the DM have further been subtracted. The shaded region at
left represents the radius of our occulting mask. The dotted line represents the VLT/NACO
sensitivities with the Simultaneous Differential Imaging (SDI) analysis technique (Biller et al.
2006; Kasper et al. 2005). The dashed hne shows the Gemini results from Marois et al. (2006)
using the ADI technique on the star HD97334B. The vertical hne at 1593 mas shows the
XNact/'^D extent of the AEOS AO control radius (Oppenheimer et al. 2003). The corre-
sponding upper limits to companion masses at right are based on models from Baraffe et al.
(2003) and apply only to the Lyot Project results. The other programs may have different
mass limits, given their sensitivities to e.g. methanated companions.