of 9
Observing Exoplanets with High-dispersion Coronagraphy. II. Demonstration of an
Active Single-mode Fiber Injection Unit
D. Mawet
1
,
2
, G. Ruane
1
,
3
, W. Xuan
1
, D. Echeverri
1
, N. Klimovich
1
, M. Randolph
1
, J. Fucik
1
, J. K. Wallace
2
, J. Wang
1
, G. Vasisht
2
,
R. Dekany
1
, B. Mennesson
2
, E. Choquet
2
, J.-R. Delorme
1
, and E. Serabyn
2
1
Department of Astronomy, California Institute of Technology, 1200 East California Boulevard, MC 249-17, Pasadena, CA 91125, USA;
dmawet@astro.caltech.edu
2
Jet Propulsion Laboratory, California Institute of Technology, 4800 Oak Grove Drive, Pasadena, CA 91109, USA
Received 2016 October 24; revised 2017 February 12; accepted 2017 February 28; published 2017 March 30
Abstract
High-dispersion coronagraphy
(
HDC
)
optimally combines high-contrast imaging techniques such as adaptive
optics
/
wavefront control plus coronagraphy to high spectral resolution spectroscopy. HDC is a critical pathway
toward fully characterizing exoplanet atmospheres across a broad range of masses from giant gaseous planets down
to Earth-like planets. In addition to determining the molecular composition of exoplanet atmospheres, HDC also
enables Doppler mapping of atmosphere inhomogeneities
(
temperature, clouds, wind
)
, as well as precise
measurements of exoplanet rotational velocities. Here, we demonstrate an innovative concept for injecting the
directly imaged planet light into a single-mode
fi
ber, linking a high-contrast adaptively corrected coronagraph to a
high-resolution spectrograph
(
diffraction-limited or not
)
. Our laboratory demonstration includes three key
milestones: close-to-theoretical injection ef
fi
ciency, accurate pointing and tracking, and on-
fi
ber coherent
modulation and speckle nulling of spurious starlight signal coupling into the
fi
ber. Using the extreme modal
selectivity of single-mode
fi
bers, we also demonstrated speckle suppression gains that outperform conventional
image-based speckle nulling by at least two orders of magnitude.
Key words:
brown dwarfs
instrumentation: adaptive optics
instrumentation: spectrographs
techniques: high
angular resolution
techniques: spectroscopic
Supporting material:
animation
1. Introduction
At the crossroads between planetary science and astronomy,
the
fi
eld of exoplanet studies is undergoing unprecedented
growth. Aided by numerous dedicated ground-based and space-
based facilities and instruments, thousands of new worlds have
been discovered over the past two decades. The vast majority
of detections so far have been through indirect measurements
that take advantage of the gravitational in
fl
uence of planets on
their host star, that of other stars on the spacetime continuum,
or simply the photometric dimming of starlight as the planet
eclipses our line of sight. The techniques exploiting these
effects, namely Doppler radial velocimetry, micro-lensing, and
transit photometry, are now routinely employed for exoplanet
detection and have ushered in a new era in planetary science
called exoplanetology. Exoplanetology has put the Solar
System into a universal perspective, and
fi
nally provides an
opportunity to understand planet formation and evolution in
statistical terms.
Direct detection has eluded the exoplanet community for
many years, mainly due to the stark dif
fi
culty associated with
disentangling the signal of an exoplanet from its host star. The
requirements for directly imaging a planet stretch the limits of
current facilities and instruments in all possible directions:
angular resolution, sensitivity, dynamic range, precision, and
stability. The advent of large ground-based and space-based
telescopes, adaptive optics
(
AO
)
, new infrared and optical
detector technologies, and modern computing have admittedly
done little to overcome these challenges. Coronagraphy, a
niche technology borrowed from solar astronomy, once held
the promise of revolutionizing the
fi
eld, but the long-awaited
breakthrough is slow to unfold.
Coronagraphy was invented in the 1930s by French
astronomer Bernard Lyot
(
1939
)
to observe and characterize
the solar corona without the need for natural eclipses. The
principle of coronagraphy is simple and aims, by way of a
device blocking the glare of the Sun, to reduce the contrast of
the scene to be within the dynamic range of the detectors.
Coronagraphs now come as standard equipment on any high-
contrast imaging instrument, and are paired with wavefront
control systems
(
AO
)
, including deformable mirrors
(
DMs
)
controlled in closed loop via a series of dedicated wavefront
sensors. Downstream from the high-contrast equipment are
classical imaging cameras, and
/
or low spectral resolution
integral
fi
eld spectrographs.
A key strategy for differentiating between planets and
leftover speckles of residual starlight is to modulate the planet
signal against the background of dynamic and quasi-static
speckles. Many differential imaging techniques have been
devised to mitigate speckle noise, such as: angular differential
imaging
(
ADI
)
, spectral
/
simultaneous differential imaging
(
SDI
)
, dual-band differential imaging, reference star differential
imaging, polarization differential imaging, coherent differential
imaging, orbital differential imaging, and binary differential
imaging. ADI and SDI are by far the most successful, but they
present signi
fi
cant challenges at very small inner working
angles owing to signal self-subtraction effects
(
Mawet
et al.
2012
)
.
Here we propose and demonstrate a new concept that
optimally combines high-contrast imaging techniques and
high-resolution spectroscopy, for the sake of simplicity called
high-dispersion coronagraphy
(
HDC
)
. The promise of HDC is
The Astrophysical Journal,
838:92
(
9pp
)
, 2017 April 1
https:
//
doi.org
/
10.3847
/
1538-4357
/
aa647f
© 2017. The American Astronomical Society. All rights reserved.
3
NSF Astronomy and Astrophysics Postdoctoral Fellow.
1
the cumulative gain in the performance offered by each
technique, as
fi
rst suggested by Riaud & Schneider
(
2007
)
and
more recently re
fi
ned by Snellen et al.
(
2015
)
. This is because
high-resolution spectroscopy sidesteps the problem of speckle
noise, since speckle noise has a low spectral resolution
signature
(
Krist et al.
2008
)
and is effectively part of the
continuum at high spectral resolution. Moreover, the planet
signal will be shifted in frequency
(
velocity
)
space with respect
to the star signal due to the Doppler effect induced by the
orbital motion of the planet around its host star, enabling
spectral lines to be disentangled from one another. Thus, HDC
is perhaps the only differential method that will approach the
photon noise limit.
In this paper, we present a new concept for feeding a
fi
ltered
beam of planet light to a high-resolution spectrograph
(
Figure
1
)
.
The framework of this proof-of-concept is the Keck Planet Imager
and Characterizer project
(
KPIC
)
, a planned upgrade to the W.M.
Keck Observatory AO system and high-contrast instrument suite
(
Mawetetal.
2016
)
. KPIC will serve as a path
fi
nder for future
high-contrast spectroscopic instruments for large ground- and
space-based facilities: the Thirty Meter Telescope
(
TMT
)
,the
European-extremely Large Telescope
(
E-ELT
)
,theGiant
Magel-
lan
Telescope, NASA
s Habitability Explorer
(
HabEx
)
,andthe
Large UV Optical InfraRed
(
LUVOIR
)
telescopes.
2. High-contrast High-resolution
Spectroscopy of Exoplanets
Now that thousands of exoplanets have been discovered,
detailed characterization of these planets is the logical next
step. The leading detection methods based on radial velocities
(
RV
)
and transits provide only the mass and
/
or size of the
planet. With these measurements, bulk density and chemical
composition may be inferred with exoplanet internal structure
models. However, this approach suffers from degeneracies,
highlighting the need for directly measuring their chemical
compositions.
Detailed diagnoses of the chemical composition of exoplanet
atmospheres
(
see e.g., Barman et al.
2011
; Konopacky
et al.
2013
; Barman et al.
2015
)
remain a challenge because
of the small angular separation and high contrast between
exoplanets and their host stars. Both constraints are mitigated
by a high-contrast imaging system, which usually consists of an
extreme AO system and a coronagraph. Current state-of-the-art
high-contrast imaging systems such as the Gemini Planet
Imager at the Gemini South telescope
(
Macintosh et al.
2015
)
and SPHERE at the Very Large Telescope
(
Beuzit et al.
2008
)
,
are able to achieve 10
3
10
4
raw starlight suppression levels
at a few tenths of an arcsecond, allowing detections and very
low-resolution spectroscopy
(
spectral resolution
R
50
)
of
gas giant planets and brown dwarfs orbiting nearby young
stars.
Riaud & Schneider
(
2007
)
and Snellen et al.
(
2015
)
suggested that contrast sensitivity may be further improved
by coupling a high-dispersion spectrograph with a high-
contrast imaging system. In this scheme, the high-contrast
imaging system serves as a spatial
fi
lter to separate the light
from the star and the planet, and the high-dispersion
spectrograph serves as a spectral
fi
lter taking advantage of
differences between the stellar and planetary spectra, including
absorption lines and RV
(
see Figure
1
)
.
The use of high-dispersion spectroscopy as a way to
spectrally isolate the planet signal has been successfully
demonstrated by a number of integrated light studies. Indeed,
high-resolution transmission spectroscopy has been used to
detect molecular gas in the atmosphere of transiting planets
(
Snellen et al.
2010
; Birkby et al.
2013
; de Kok et al.
2013
)
.At
a high spectral resolution, resolved molecular lines may be
used to study day-to-night side wind velocity
(
Snellen
et al.
2010
)
and verify 3D exoplanet atmospheric circulation
models
(
Kempton et al.
2014
)
. The spectral lines of a planet
may also be separated from stellar lines with suf
fi
cient
differences in RV
(
>
50
km s
1
)
, breaking the degeneracy
between the true planet mass and orbital inclination
(
Brogi
et al.
2012
,
2013
,
2014
; Lockwood et al.
2014
)
. Moreover,
high-resolution spectroscopy has led to the
fi
rst measurement
of planet rotational velocity
(
Snellen et al.
2014
)
. While not yet
feasible on exoplanets yet, high-resolution spectroscopy has
helped generate the
fi
rst global cloud map of brown dwarf
Luhman 16 B via the Doppler imaging technique
(
Cross
fi
eld
et al.
2014
)
.
High-resolution spectroscopy is poised to become even more
powerful when combined with high-contrast imaging. The
signal-to-noise ratio
(
S
/
N
)
achieved by an HDC instrument is
to
fi
rst order
(
Snellen et al.
2015
)
:
h
sss
=
+++
()
S
SK
N
SN
,
1
planet
star
bg
2
rn
2
dark
2
lines
Figure 1.
High-dispersion coronagraphy
(
HDC
)
concept. A classical high-contrast instrument, with an adaptive optics
(
AO
)
or wavefront control
(
WFC
)
system
followed by a starlight suppressing coronagraph, is linked to a high-resolution spectrograph by a
fi
ber injection unit
(
FIU
)
. The data processing steps are as follows:
the raw data
(
planet
+
residual starlight un
fi
ltered by the high-contrast instrument, and various noise contributors such as photon shot noise, readout noise, background
noise, etc.
)
is cross-correlated with a theoretical template yielding a new observable, called here the cross-correlation function
(
CCF
)
. The CCF pro
fi
le provides
improved dynamic range for detection and molecular characterization. Its broadening with respect to the instrument line pro
fi
le is a direct measure of the planet
s spin
rotation. The variation of the line pro
fi
le morphology over the rotation period enables Doppler imaging.
2
The Astrophysical Journal,
838:92
(
9pp
)
, 2017 April 1
Mawet et al.
where
S
planet
is the planet signal making it to the
spectrograph with ef
fi
ciency
η
,
S
star
is the signal from the star
(
both in units of photo-electrons per pixel
)
,
K
is the suppression
factor of the star at the planet
s position, and
s
bg
2
,
s
rn
2
, and
s
dar
k
2
are the photon shot noise from the sky and telescope
background, the readout noise, and the dark current noise,
respectively.
N
lines
is a multiplication factor that takes into
account the number and strength of the individual planet lines
targeted, which is a de
fi
ning strength of high-resolution
spectroscopy
(
Snellen et al.
2015
)
.
The planet
/
star contrast sensitivity achieved by integrated-
light high-dispersion spectroscopy is currently demonstrated at
the 10
4
level, which corresponds to the stellar photon noise
limit
(
Snellen et al.
2015
)
. When coupled with a state-of-the-art
high-contrast imaging system with a raw starlight suppression
of 10
3
or better, a high-contrast high-dispersion spectroscopy
instrument can potentially exceed 10
7
planet
/
star contrast,
providing a sensitivity superior to a high-contrast imaging
system or a high-dispersion spectrograph alone. This would
allow the physical and chemical processes taking place on an
exoplanet to be studied in unprecedented detail.
It is important to note that high-spectral-resolution observa-
tions of a single spatial resolution element render spatial
speckle variations
(
spatial speckle noise
)
irrelevant. Since the
spectral signature of stellar speckles is a very smooth and a
slowly varying function of wavelength, it becomes part of the
continuum at very high spectral resolutions
(
Krist et al.
2008
)
.
Thus, the dominant limiting factor in low-resolution high-
contrast imaging, spatial speckle noise, is obviated by HDC.
Ground-based HDC observations will enable the detection of
multiple molecular species and their resolved spectral lines in
the J, H, K, L, and M bands
(
Wang et al.
2017
)
. Currently
known directly imaged exoplanets
(
e.g., HR 8799bcde, 51 Eri
b, ROXs 42B b, ROXs 12 b,
β
Pictoris b
)
will be prime targets
for HDC observations. Together with observations from
JWST
in the near-to-mid infrared wavelengths
(
at much larger
wavelengths, e.g.,
>
5
μ
m
)
, ground-based HDC observations
will yield abundances
(
Brogi et al.
2016
)
, remove the
degeneracy of temperature and pressure pro
fi
les, and thus
provide more details on the presence and formation of cloud
/
haze, which is a critical step forward in understanding the
physical and chemical processes in exoplanet atmospheres.
3. Fiber Injection Unit
(
FIU
)
Concept
Here we propose to link the coronagraph instrument and the
high-resolution spectrograph with an FIU, illustrated in
Figure
2
. The purpose of the FIU is to couple planet light
into a single mode
fi
ber
(
SMF
)
and maintain accurate
alignment throughout long-exposure observations
(
up to
several hours
)
. The pointing accuracy and stability is achieved
through active sensing and control of the planet and
fi
ber
positions using a scheme similar to Colavita et al.
(
1999
)
.
An actuated tip-tilt mirror
(
TTM
)
is used to align the planet
image position with the tip of the SMF, the relative locations of
which are determined by simultaneously imaging the scene and
the SMF onto a tracking camera. A beamsplitter
(
BS
)
or
dichroic re
fl
ects part of the science beam to the tracking camera
directly after the TTM. To locate the SMF, a calibration source
(
CAL
)
is retro-fed through the
fi
ber by means of optical
circulators or
Y
-couplers. The BS re
fl
ects light from the SMF
toward a corner cube
(
CC
)
retro
fl
ector, which sends the beam
back through the BS and toward the tracking camera. A beacon
image is formed on the tracking camera at the location of the
Figure 2.
Layout of a typical high-dispersion coronagraph
(
HDC
)
and our laboratory setup, consisting of an AO system with a deformable mirror
(
DM
)
and wavefront
sensor
(
WFS
)
, a coronagraph with a focal plane mask
(
FPM
)
and Lyot stop
(
LS
)
, and a
fi
ber injection unit with a tip-tilt mirror
(
TTM
)
, beamsplitter
(
BS
)
or dichroic,
tracking camera, corner cube
(
CC
)
, single-mode
fi
bers
(
SMF
)
, and optics to image the scene on the camera and inject planet light into the
fi
ber
(
s
)
. The
fi
bers feed light
to the high-resolution spectrograph
(
HRS
)
and an avalanche photodiode
(
APD
)
. A calibration source
(
CAL
)
is used to back propagate light through the SMF for
tracking purposes.
3
The Astrophysical Journal,
838:92
(
9pp
)
, 2017 April 1
Mawet et al.
SMF. The beacon is used to determine the TTM settings to co-
align the object image and the SMF. Alternatively, the CAL
source may feed a separate SMF, creating a beacon nearby to
the spectrograph
fi
ber tip with well-calibrated relative
positions.
The FIU is also designed to provide feedback mechanisms
for starlight suppression using the upstream DM. An optional
low-noise single pixel detector
(
e.g., an avalanche photodiode;
APD
)
may be used to monitor the starlight leaking into the
fi
ber
at high speed
(
>
10 kHz
)
and drive a control loop that
minimizes leaked starlight in real time. We have demonstrated
both the optical alignment procedure and real-time wavefront
control concepts in the laboratory.
4. Laboratory Setup
Our laboratory setup consists of a telescope simulator,
followed by an AOs system, a coronagraph, and the FIU
prototype
(
see Figures
2
and
3
)
. The telescope simulator images
simulated star and off-axis planetary sources, generated by a
Thorlabs 635 nm laser diode and a
fi
ltered NKT Photonics
supercontinuum white light source
(
narrowband
fi
lter centered at
650 nm
)
, respectively. The AO system is made up of a Boston
Micromachines 144-actuator MEMS DM, pellicle BS, and a
Shack
Hartmann wavefront sensor
(
Thorlabs AOK1-UM01
)
.
The AO system is followed by a classical 3-plane
coronagraph with a vortex focal plane mask. The vortex
coronagraph is a phase-based coronagraph enabling high-
contrast imaging at small angular separations, while conserving
high off-axis throughput
(
Mawet et al.
2005
)
. The vortex
coronagraph is currently in operations at Palomar
(
Mawet
et al.
2010
,
2011
; Serabyn et al.
2010
; Bottom
et al.
2015
,
2016
)
, VLT
(
Mawet et al.
2013
)
, Subaru, Keck
(
Absil et al.
2016
; Mawet et al.
2017
; Serabyn et al.
2017
)
, and
at Large Binocular telescopes
(
Defrère et al.
2014
)
. The
particular vortex mask used here has a topological charge of 4,
which applies a phase ramp of the form
q
e
i
4
, where
θ
is the
azimuthal angle in the focal plane. The effective inner working
angle
(
i.e., the angle for 50% off-axis transmission
)
of the
charge-4 vortex coronagraph is
l
~
D
1.7
, where
λ
is the
central wavelength, and
D
is the telescope diameter.
Downstream from the coronagraph
s Lyot stop lies the FIU
described in Section
3
. We use a three-axis TTM from
Newport, actuated by computer-controlled Thorlabs piezo-
actuators. The FIU BS is a 50%
50% BS
(
at the telescope, we
plan to use more optimal splitting ratios and
/
or dichroic BSs
)
.
The tracking camera is a CMOS sensor from Thorlabs. The
SMF is mounted on a Newport Post-mount Singlemode Fiber
Aligner. The CC and other optical elements are off-the-shelf
Newport and Thorlabs products. For the purposes of this
demonstration, we used a Newport Si photodiode power meter
in lieu of the high-resolution spectrograph. The back-end
calibration source is a
fi
ber-coupled 635 nm
laser diode.
5. Results
To validate our new FIU concept, we conducted a series of
experiments. We
fi
rst demonstrated the coupling of starlight
and planet light into the
fi
ber using manual spiral scans. Then,
the co-alignment procedure using the beacon image was
demonstrated by injecting planet light into the
fi
ber in a
reproducible manner. We then validated the wavefront control
procedure to minimize the amount of starlight coupling into the
fi
ber along with the planet light using a technique akin to
speckle nulling
(
Bordé & Traub
2006
; Bottom et al.
2016
)
.
5.1. Planet Injection
Our
fi
rst demonstration consisted of injecting a point source
image into the
fi
ber with an ef
fi
ciency close to the theoretical
limit. A SMF
s fundamental mode is nearly a Gaussian
(
Shaklan & Roddier
1988
)
, while our optical system generates
an Airy function of the form
()
Jr r
1
, which is the result of the
Fourier transform of a circular unobscured input pupil, and thus
the point-spread function
(
PSF
)
of our optical setup. The
monochromatic injection ef
fi
ciency
η
is the modulus squared of
the overlap integral between the incident electric
fi
eld
()
A
xy
,
and the fundamental mode of the SMF
H
E
11
:
*
h
=
∫∫
∫∫∫∫
()()
(
)∣
∣ (
)∣
()
xyA xydxdy
x y dxdy
A x y dxdy
HE ,
,
HE ,
,
.2
11
2
11
22
The theoretical maximum injection ef
fi
ciency is 81% for an
ideal circular, unobstructed pupil
(
Shaklan & Roddier
1988
)
.
Due to aberrations in the system, the impinging
fi
eld is not
exactly an Airy function, and so the overlap integral inevitably
yields lower injection ef
fi
ciencies
(
Wagner & Tomlinson
1982
;
Toyoshima
2006
)
. We note that the theoretical maximum
injection ef
fi
ciency may be increased by apodizing the
impinging beam into a Gaussian function
(
Jovanovic
et al.
2015b
)
.
Figure
4
(
left
)
shows the image on the tracking camera with
the residual starlight concentrated at the center of the image as
well as the simulated planet and FIU beacon. The computer
control of the tip-tilt mirror yielded consistent and reproducible
injection ef
fi
ciencies between 65% and 70%, which was
deemed suf
fi
cient for demonstration purposes
(
Figure
4
,
middle
)
. The difference between our measured ef
fi
ciencies
and the theoretical limit can be traced to optical aberrations and
Fresnel losses at air
/
glass interfaces in the injector and on the
tip of the
fi
ber. The planet injection is very stable over hour
timescales as shown in Figure
4
(
right
)
.
On-sky demonstrations have so far yielded consistent and
reliable results, but nowhere near the theoretical limit due to
residual wavefront error after AO correction
(
Bechter et al.
2016
; Jovanovic et al.
2016
)
. The coupling ef
fi
ciency results
achieved in the lab therefore represent an upper limit on what
can be achieved on-sky. Using Subaru
/
SCExAO, the state-of-
Figure 3.
Top view of the FIU prototype on Caltech
s High Contrast
Spectroscopy Testbed for Segmented telescopes
(
HCST
)
.
4
The Astrophysical Journal,
838:92
(
9pp
)
, 2017 April 1
Mawet et al.
the-art on-sky injection ef
fi
ciency was reported by Jovanovic
et al.
(
2016
)
to be
50%
in the H band.
5.2. Starlight Rejection: Speckle Nulling
A necessity in imaging
let alone characterizing
exopla-
nets is to suppress residual starlight
(
speckles
)
in the
fi
nal
image plane as much as possible. Speckles are caused by
optical aberrations incurred as light travels through Earth
s
atmosphere and imperfect optics in the imaging system.
Speckles are the nemesis of exoplanet imaging, since they
might appear similar to or overwhelm the planet signal,
precluding both discovery and characterization.
In short, speckle nulling is the process of destructively
interfering an intentionally generated anti-speckle with an
existing speckle. We recall that by virtue of the Fourier
transform relationship between the pupil plane and the image
plane, a sinusoid with amplitude
l
h
0
in the pupil plane
will translate into a pair of conjugated anti-speckles in the
image plane
(
Malbet et al.
1995
)
. Speci
fi
cally, we apply a
cosine pattern to the DM surface with height
x
pa
=+
[( ·
)]
()
r
h
h
4
cos 2
,
3
0
where
h
2
0
is the maximum surface height
(
h
0
wavefront
)
,
x
is
the spatial frequency vector and the speckle location in the
image plane,
r
is the position vector in the pupil plane, and
α
is
a constant phase offset. The intensity, position, and phase of the
speckle are controlled by
h
0
,
x
, and
α
, respectively. It is worth
noting that the speckle intensity is
pl
(
)
h
0
2
(
Malbet
et al.
1995
)
. In theory, the limited number of actuators
illuminated on the DM
(
roughly 11
×
11
)
limits the spatial
frequency to
5 cycles per pupil diameter and, therefore, the
range of angular separations at which anti-speckles may be
produced
1-5
l
D
. However, it is possible to use high-spatial
frequency DM surface features, such as print-through and
cupping effects, to increase this range, taking advantage of
harmonics
(
i.e., clone PSFs
)
that appear at integer multiples of
11
l
D
(
Thomas et al.
2015
)
.
To test the viability of this method, we
fi
rst performed linear
searches in
h
0
,
x
, and
α
to
fi
nd the parameters that optimize the
starlight suppression ratio as measured on the tracking camera
at various speckles at
l
~
D
2
,
l
~
D
3
, and
l
~
D
4
from the
star. The optimization of
x
is used as a
fi
ne tuning and might be
super
fl
uous as the
fi
ber location can be known precisely. Initial
searches were done in coarse increments over the full four-
dimensional parameter space, but clearly show that a global
minimum null exists for each speckle. Complete searches,
however, take an implausibly long time
(
20
40 minutes
)
,
which would waste valuable
(
and expensive
)
telescope time.
Hence, we developed an expedited optimization code that
reaches the minimum in
3
5 minutes. This approach relies on
a pattern search optimization in phase and amplitude space,
after a calibration is done to generate the speckle near the
optimal correct location
(
Bottom et al.
2016
)
. However, using
the tracking camera images to probe suppression is not our
fi
nal
goal. Rather, we wish to maximize the S
/
N of the planet signal
in the spectrograph, which depends on the amount of residual
starlight that is injected into the SMF.
5.3. Integration: Speckle Nulling at Fiber-tip Location
The goal of the FIU is to inject as much planet light as
possible while rejecting as much starlight as possible.
Unfortunately, residual aberrations in the system due to
imperfect optics, plus uncorrected atmospheric turbulence in
a ground-based system, will create speckles in the image,
which will also couple into the SMF and propagate into the
spectrograph. Speckle noise would often overwhelm the planet
signal, especially at small angular separations from the star,
where speckle noise is dominant even after the spatial
fi
ltering
from the coronagraph. Preventing starlight from coupling into
the
fi
ber is the best way to increase the S
/
N, and thus ef
fi
ciency
of the observation.
By coupling the other end of the SMF to a photodiode, we
record the total power entering the SMF after applying an
optimal sinusoid pattern to the DM and calculate the
suppression factor
(
the power of speckle with
fl
at DM divided
by the power after nulling
)
of a speckle at the
fi
ber position.
Speci
fi
cally, we recorded monochromatic starlight suppression
factors of
>
1000
with the SMF for a bright speckle located at
roughly
l
D
2
away from the star
(
see Figure
5
)
. We note that
Figure 4.
Left: image from the tracking camera, showing residual starlight after the charge-4 vortex coronagraph, along with our simulated off-axis planet an
d the
beacon calibration lamp. Middle: acquisition sequence showing the quick injection of the off-axis planet signal into the SMF after determining the p
osition of the SMF
via the FIU beacon. Right: injection stability over two hours.
5
The Astrophysical Journal,
838:92
(
9pp
)
, 2017 April 1
Mawet et al.
the measured suppression is limited by the dynamic range and
noise properties of our Si photodiode.
After repeating this experiment multiple times for various
speckles in the image, we
fi
nd that speckle nulling with a SMF
generally improves raw starlight suppression by a factor of
500
1000 beyond the nominal raw starlight suppression level
produced by the wavefront control
/
AO system and corona-
graph. The corresponding gain simultaneously measured on our
tracking camera images is 5
10, which is similar to speckle
nulling gains routinely demonstrated in ground-based imaging
/
spectroscopy systems
(
Bottom et al.
2016
)
. It has been veri
fi
ed
that the presence of the planet signal at the location of the
fi
ber
and speckle does not affect, and is not affected by, the nulling
process
(
Section
5.3.2
)
. The planet signal is indeed an
incoherent background and much fainter than the speckle. It
does not respond to the ripple probes from the DM, and does
not contribute to the sensing of speckle complex amplitude.
5.3.1. Heuristic Explanation for the Suppression Gain
Standard speckle nulling techniques rely on coherent
interference between the speckle and the generated anti-speckle
to minimize the starlight in a particular region on the image
(
Jovanovic et al.
2015b
)
. As noted in Section
5.1
and
Equation
(
2
)
, the injection ef
fi
ciency of the SMF is the
modulus squared of the overlap integral between the incident
E-
fi
eld and the fundamental symmetric HE
11
mode of the
fi
ber.
Therefore, the SMF more ef
fi
ciently uses the existing
degrees of freedom provided by the DM to suppress a speckle,
resulting in a signi
fi
cant improvement in starlight rejection over
traditional speckle nulling using an imaging camera. For
instance, a non-zero incident electric
fi
eld that is antisymmetric
around the center of the
fi
ber tip will be eliminated in the
overlap integral
(
Equation
(
2
))
.We
fi
nd that our optimization
procedure often converges to solution where a node in the
fi
eld
or phase singularity appears at the
fi
ber location, which is
reminiscent of
fi
ber nulling concept presented in Haguenauer &
Serabyn
(
2006
)
.
Generally speaking, a SMF will couple less starlight on
average than a multi-mode
fi
ber or detector resolution element
(
resel
)
with equivalent planet coupling
/
detection capability.
Mathematically, neighboring speckles in a bandlimited com-
plex stellar
fi
eld have opposite parity. Therefore, the overlap
integral between the fundamental mode of the SMF and the
stellar electric
fi
eld is always less than or equal to the
equivalent stellar energy on a resel at the same location when
the resel size is chosen to collect the same planet signal as
the SMF.
In other words, the nulling condition over the resel requires
the anti-speckle
fi
eld
()
A
xy
,
as
to be exactly opposite to the
speckle
fi
eld
(
)
A
xy
,
s
, i.e.,
=-
()
()
A
xy
A xy
,,
as
s
for all spatial
position over the resel
(
x
,
y
)
. The nulling condition through the
SMF from Equation
(
2
)
implies that the resulting complex
fi
eld
+=
() () (
)
A
xy
A xy
Axy
,,,
as
s
is zeroed when projected onto
and integrated over the SMF fundamental mode
()
xy
H
E,
11
,
which is a much less stringent and thus an easier condition
to meet.
5.3.2. Throughput Losses after Nulling
Referring to Equation
(
1
)
, one must be careful that the planet
signal does not suffer throughput losses in the starlight
suppression process. To improve the performance of our
HDC system, we must achieve a starlight suppression ratio that
is better than the square of the planet throughput loss ratio. It is
therefore signi
fi
cant that we have routinely demonstrated a
factor of
>
100
suppression of starlight with no detected loss of
planet throughput, effectively improving the S
/
N by a factor
of
>
10
.
From Equation
(
2
)
, the throughput of the planet signal is
roughly proportional to the Strehl ratio of the planet PSF
described by the
fi
eld
()
A
xy
,
. Using the Marechal approx-
imation
(
Mahajan
1982
)
, we have
h
p
l
μ-
()
h
exp
2
22
,4
0
2
where
h
2
0
is the amplitude of sinusoid we impose on the DM
(
h
0
on the wavefront
)
. To constrain losses to be
<
10%
, the
ripple amplitudes are limited to
l
h
790n
m
0
during the
speckle nulling process. By imposing sinusoidal shapes on the
DM, in particular at low spatial frequencies
(
small angles
)
,we
might introduce slight displacements in the planet and star
beams, which could easily translate into optical aberrations
(
e.g., when traveling through lenses on the bench
)
. In extreme
cases, this could degrade the beam quality and may even cause
lateral beam shifts, which would effect the coupling ef
fi
ciency
and hence the throughput of the planet signal, as SMF coupling
Figure 5.
Experimental results in monochromatic light. Left: stellar PSF with the coronagraph focal plane mask removed. Middle: residual starlight with the f
ocal
plane mask aligned. The
fi
ber and planet position is indicated in both images. The simulated planet is about 1000 times fainter than the peak starlight, and is therefore
invisible in these images. Right: nulling sequence, showing the starlight level as a function of time, while the speckle nulling loop is active. Sharp
peaks in power
appear in the nulling sequence as the optimization algorithm explores the four-dimensional parameter space to
fi
nd a deeper null.
6
The Astrophysical Journal,
838:92
(
9pp
)
, 2017 April 1
Mawet et al.
is sensitive to beam quality, angular offsets, and translational
deviation.
5.3.3. Projected Broadband Performance
To con
fi
rm that this principle is readily extended to
polychromatic light, we performed a series of numerical
simulations that mimic our experimental setup, but allow us
to explore the performance with varying source properties,
fi
ber
positions, and levels of optical aberration. Figure
6
shows the
simulated stellar PSF before and after the speckle nulling
process, as well as the measured spectrum in the spectrograph,
over a 10% passband centered at 632 nm. The simulated optical
system applied
30 nm
rms wavefront error to the beam. The
fi
ber was placed at the location of a bright speckle and the
coupled power was minimized at
fi
ve discrete wavelengths. In
doing so, a suppression factor of
>
100 is achieved across the
full passband
(
see Figure
6
, right
)
. We also veri
fi
ed that the
Figure 6.
Simulation results in polychromatic light. Left:
residual starlight after the coronagraph, with a
fl
at DM surface. Middle:
residual starlight after speckle
nulling at the
fi
ber location
(
2.5
l
D
)
. The optimal DM surface is shown in the inset along with the footprint of the beam
(
yellow circle
)
. Right: spectrum of the
residual stellar power coupled into the SMF before
(
fl
at DM
)
and after speckle nulling, normalized to the total stellar power before the coronagraph.
Figure 7.
(
Online animation
)
Numerical simulation of stellar leakage into a SMF, compared to an equivalent detector pixel, after an AO system. Top row:
(
left
)
the
wavefront at the telescope pupil, given by a Kolmogorov phase screen.
(
middle
)
The wavefront after AO with 250 nm rms error.
(
right
)
The instantaneous stellar PSF
with the control region and
fi
ber location indicated by white circles. Bottom row:
(
left
)
the fraction of planet light coupled into the SMF and sensed by a single pixel.
The size of the pixel was chosen such that these quantities are approximately equal. Time is shown in units of the clearing time
D
v
wind
under the frozen
fl
ow
approximation, where
D
is the telescope diameter and
v
wind
is the wind speed.
(
middle
)
The fraction of starlight coupled into the SMF and sensed by a single pixel. On
average, 3
×
more starlight is sensed by a detector pixel than coupled into a SMF.
(
right
)
The instantaneous spectral responsivity for the star.
(
An animation of this
fi
gure is available.
)
7
The Astrophysical Journal,
838:92
(
9pp
)
, 2017 April 1
Mawet et al.
planet throughput did not decrease due to the speckle
correction. In fact, the planet throughput in this particular case
increased by 0.5%.
5.3.4. Projected On-sky Performance with a Passive FIU
We performed a simulation where the amount of starlight
passively sensed by a SMF and an equivalent single pixel was
monitored over time in the presence of atmospheric turbulence
(
see Figure
7
)
. The effect of the AO was modeled by high pass
fi
ltering a Kolmogorov phase screen up to 8 cycles per pupil
diameter
D
(
equivalent to a DM with 16 actuators
)
giving a
post-AO wavefront error of
250 nm rms. Time lag and
atmospheric chromaticity are ignored and should not affect the
outcome of the numerical experiment. The
fi
ber and pixel were
placed at the same arbitrary location within the AO control
region. We found that over 10 clearing times under the frozen
fl
ow approximation
(
D
v
wind
, where
v
wind
is the wind speed,
typically
1s
)
, an average of 3
×
less starlight was coupled
into the SMF than sensed by the single pixel, without
additional speckle nulling.
5.3.5. Projected On-sky Performance with an Active FIU
We conjecture that implementing a speckle nulling proce-
dure with temporal bandwidth
>
v
D
wind
and predictive control
(
Poyneer et al.
2007
; Riggs et al.
2016
)
may provide an
additional factor of 10
100
×
in starlight suppression
(
Guyon
& Males
2017
)
, taking advantage of the natural speckle
rejection provided by the SMF. We defer further analysis of the
active FIU in the presence of dynamical aberrations to a
forthcoming paper.
6. Perspectives
The FIU described here will be the core of the KPIC instrument
(
Mawetetal.
2016
)
. KPIC is a four-pronged upgrade of the Keck
AO facility. The
fi
rst stage is the addition of a high performance
small inner working angle L-band vortex coronagraph to NIRC2
(
Absiletal.
2016
; Mawet et al.
2017
; Serabyn et al.
2017
)
,
implemented in 2015 and now available to the Keck community
in shared-risk mode. This upgrade not only included a brand new
coronagraph mask, but also a suite of software packages to
entirely script the coronagraph a
cquisition procedure, including
automatic ultra-precise centering
(
Huby et al.
2015
)
, speckle
nulling wavefront control
(
Bottom et al.
2016
)
, and an open
source python-based data reduction package
(
Gomez Gonzalez
et al.
2016
)
.
The second upgrade component is an infrared pyramid
wavefront sensor demonstration and potential facility for the
Keck II AOs system. Near-infrared wavefront sensing is a
critical technology for science with AO on current and future
telescopes. It will enable high-contrast observations of exopla-
nets around low-mass stars and in obscured star-forming regions.
It can be used to extend the performance of natural guide star
AO to redder targets and to increase the sky coverage of laser
guide star AO. Furthermore, it allows the application of optimal
wavefront sensing approaches
(
e.g., pyramid and Zernike
wavefront sensing
)
due to the AO correction at near-infrared
wavelengths. This demonstration is especially relevant because
all of the extremely large telescopes
(
ELTs
)
are planning to use
infrared wavefront sensing as part of their AO facilities.
The third upgrade is a higher-order DM paired with the
infrared pyramid sensor, followed by a new single-stage
coronagraph. The vortex coronagraph installed with the
fi
rst
upgrade component is inside of the NIRC2 cryostat and thus
cannot be used in conjunction with a high-resolution spectro-
graph. Finally, the fourth component of the KPIC is the FIU
discussed in this paper. The second, third, and fourth module
will be integrated within the same optical relay.
KPIC is thus a phased, cost-effective upgrade path for the
Keck II AOs facility. It builds on the lessons learned from
fi
rst-
and second-generation high-contrast AO instruments, meant to
explore new scienti
fi
cally exciting niches and pave the way for
the TMT-Planet Finder Instrument core science, while matur-
ing system-level and critical components for future ground- and
space-based instrumentation, including NASA
s HabEx and
LUVOIR
fl
agship mission concepts.
6.1. Characterization of Known Objects
The FIU concept presented here is amenable to the
characterization of exoplanets discovered by other methods
(
direct imaging, RV, astrometry, etc.
)
where the planet position
is known a priori. Our proposed pointing and tracking system is
accurate enough to offset blindly to the location of a
companion too faint to be visible in acquisition images. To
further improve pointing astrometric accuracy, one could use
the DM to generate a set of satellite spots as routinely used by
VLT
/
SPHERE or Subaru
/
SCExAO
(
Jovanovic et al.
2015a
)
.
6.2. Multiplexing
The HDC technique may also be multiplexed to increase the
effective
fi
eld of view, and so it can be used to detect new
planets, or characterize planets with
positions that are not well
constrained, such as the RV detected Earth-like planet around
Proxima Centauri
(
Lovis et al.
2017
)
. One limitation for spatial
multiplexing is detector real estate. Preliminary design work
has led us to consider a 3
×
3, 9-element multiplexing
capability using a H4RG detector
(
4096
×
4096 pixels
)
. The
sampling in the image plane is done by way of a microlens
array as in Ireland et al.
(
2014
)
, Rains et al.
(
2016
)
, where each
microlens feeds a SMF. The
fi
ber output may be recon
fi
gured
in a pseudo-slit at the entrance of an echelle spectrograph.
Rains et al.
(
2016
)
recently demonstrated a 3
×
3 lenslet-based
9-SMF mini integral
fi
eld unit linking Subaru
/
SCExAO to the
RHEA spectrograph
(
Bento et al.
2016
)
. However, this
fi
rst
attempt was affected by modal noise and cross talk due to the
small spacing between
fi
bers in the output pseudo-slit.
Another multiplexing option currently proposed is to build
as many high-resolution diffraction-limited spectrographs as
there are resolution elements in the search area
fi
eld of view.
Diffraction-limited spectroscopy is economical due to the
conservation of beam etendue and is likely to be the most
realistic implementation of future high-resolution spectro-
graphs on large telescopes
(
Bland-Hawthorn et al.
2004
;
Bland-Hawthorn & Horton
2006
; Bento et al.
2016
)
.
We note that speckle nulling on multiple SMFs should still
work. However, the number of available degrees of freedom
per
fi
ber will be smaller, and likely result in reduced starlight
suppression gains.
7. Conclusion
In this paper, we presented an innovative FIU module
designed to ef
fi
ciently couple a high-contrast imaging system
8
The Astrophysical Journal,
838:92
(
9pp
)
, 2017 April 1
Mawet et al.
(
AO and coronagraph
)
to a high-resolution spectrograph,
enabling HDC of exoplanets.
We built the
fi
rst FIU prototype and performed a series of
laboratory experiments that demonstrated fast off-axis planet
light acquisition, as well as high
(
70%
)
and stable coupling
ef
fi
ciencies.
Using the wavefront control system and a technique akin to
speckle nulling, we achieved high levels of starlight suppression.
Using the extreme modal selectivity of SMFs, we routinely
obtained speckle suppression gains that outperform conventional
image-based speckle nulling by at least two orders of magnitude.
Our FIU demonstrator is a prelude to on-sky scienti
fi
c
demonstrations with the KPIC project, a path
fi
nder to future
HDC instruments on ELTs on the ground and in space.
The authors would like to acknowledge the referee for the
thorough review and constructive comments. The authors
would like to acknowledge the
fi
nancial support of the
Heising
Simons foundation.
Facility:
Caltech
s High-Contrast Spectroscopy Testbed for
Segmented Telescopes
(
HCST
)
.
References
Absil, O., Mawet, D., Karlsson, M., et al. 2016,
Proc. SPIE
,
9908, 99080Q
Barman, T. S., Konopacky, Q. M., Macintosh, B., & Marois, C. 2015,
ApJ
,
804, 61
Barman, T. S., Macintosh, B., Konopacky, Q. M., & Marois, C. 2011,
ApJL
,
735, L39
Bechter, A., Crass, J., Ketterer, R., et al. 2016,
Proc. SPIE
,
9909, 99092X
Bento, J., Feger, T., Ireland, M. J., et al. 2016,
Proc. SPIE
,
9908, 99086K
Beuzit, J.-L., Feldt, M., Dohlen, K., et al. 2008,
Proc. SPIE
,
7014, 701418
Birkby, J. L., de Kok, R. J., Brogi, M., et al. 2013,
MNRAS
,
436, L35
Bland-Hawthorn, J., & Horton, A. 2006,
Proc. SPIE
,
6269, 62690N
Bland-Hawthorn, J., McGrath, A. J., Saunders, W., Haynes, R., &
Gillingham, P. 2004,
Proc. SPIE
,
5492, 242
Bordé, P. J., & Traub, W. A. 2006,
ApJ
,
638, 488
Bottom, M., Femenia, B., Huby, E., et al. 2016,
Proc. SPIE
,
9909, 990955
Bottom, M., Kuhn, J., Mennesson, B., et al. 2015,
ApJ
,
809, 11
Bottom, M., Shelton, J. C., Wallace, J. K., et al. 2016,
PASP
,
128, 075003
Brogi, M., de Kok, R. J., Birkby, J. L., Schwarz, H., & Snellen, I. A. G. 2014,
A&A
,
565, A124
Brogi, M., Line, M., Bean, J., Désert, J.-M., & Schwarz, H. 2016,
ApJ
,
817, 106
Brogi, M., Snellen, I. A. G., de Kok, R. J., et al. 2012,
Natur
,
486, 502
Brogi, M., Snellen, I. A. G., de Kok, R. J., et al. 2013,
ApJ
,
767, 27
Colavita, M. M., Wallace, J. K., Hines, B. E., et al. 1999,
ApJ
,
510, 505
Cross
fi
eld, I. J. M., Biller, B., Schlieder, J. E., et al. 2014,
Natur
,
505, 654
de Kok, R. J., Brogi, M., Snellen, I. A. G., et al. 2013,
A&A
,
554, A82
Defrère, D., Absil, O., Hinz, P., et al. 2014,
Proc. SPIE
,
9148, 91483X
Gomez Gonzalez, C. A., Absil, O., Absil, P.-A., et al. 2016,
A&A
,
589, A54
Guyon, O., & Males, J. R. 2017, ApJ, submitted
Haguenauer, P., & Serabyn, E. 2006,
ApOpt
,
45, 2749
Huby, E., Baudoz, P., Mawet, D., & Absil, O. 2015,
A&A
,
584, A74
Ireland, M., Anthony, A., Burley, G., et al. 2014,
Proc. SPIE
,
9147, 91471J
Jovanovic, N., Cvetojevic, N., Schwab, C., et al. 2016,
Proc. SPIE
,
9908,
99080R
Jovanovic, N., Guyon, O., Martinache, F., et al. 2015a,
ApJL
,
813, L24
Jovanovic, N., Martinache, F., Guyon, O., et al. 2015b,
PASP
,
127, 890
Kempton, E. M. R., Perna, R., & Heng, K. 2014,
ApJ
,
795, 24
Konopacky, Q. M., Barman, T. S., Macintosh, B. A., & Marois, C. 2013,
Sci
,
339, 1398
Krist, J. E., Shaklan, S. B., & Levine, M. B. 2008,
Proc. SPIE
,
7010, 701044
Lockwood, A. C., Johnson, J. A., Bender, C. F., et al. 2014,
ApJL
,
783, L29
Lovis, C., Snellen, I., Mouillet, D., et al. 2017,
A&A
,
599, A16
Lyot, B. 1939,
MNRAS
,
99, 538
Macintosh, B., Graham, J. R., Barman, T., et al. 2015,
Sci
,
350, 64
Mahajan, V. N. 1982,
JOSA
,
72, 1258
Malbet, F., Yu, J. W., & Shao, M. 1995,
PASP
,
107, 386
Mawet, D., Choquet, É., Absil, O., et al. 2017,
AJ
,
153, 44
Mawet, D., Absil, O., Delacroix, C., et al. 2013,
A&A
,
552, L13
Mawet, D., Mennesson, B., Serabyn, E., Stapelfeldt, K., & Absil, O. 2011,
ApJL
,
738, L12
Mawet, D., Pueyo, L., Lawson, P., et al. 2012,
Proc. SPIE
,
8442, 844204
Mawet, D., Riaud, P., Absil, O., & Surdej, J. 2005,
ApJ
,
633, 1191
Mawet, D., Serabyn, E., Liewer, K., et al. 2010,
ApJ
,
709, 53
Mawet, D., Wizinowich, P., Dekany, R., et al. 2016,
Proc. SPIE
,
9909, 99090D
Poyneer, L. A., Macintosh, B. A., & Véran, J.-P. 2007,
JOSAA
,
24, 2645
Rains, A. D., Ireland, M. J., Jovanovic, N., et al. 2016,
Proc. SPIE
,
9908,
990876
Riaud, P., & Schneider, J. 2007,
A&A
,
469, 355
Riggs, A. J. E., Kasdin, N. J., & Groff, T. D. 2016,
JATIS
,
2, 011017
Serabyn, E., Huby, E., Matthews, K., et al. 2017,
AJ
,
153, 43
Serabyn, E., Mawet, D., & Burruss, R. 2010,
Natur
,
464, 1018
Shaklan, S., & Roddier, F. 1988,
ApOpt
,
27, 2334
Snellen, I., de Kok, R., Birkby, J. L., et al. 2015,
A&A
,
576, A59
Snellen, I. A. G., Brandl, B. R., de Kok, R. J., Brogi, M., Birkby, J., &
Schwarz, H. 2014,
Natur
,
509, 63
Snellen, I. A. G., de Kok, R. J., de Mooij, E. J. W., & Albrecht, S. 2010,
Natur
,
465, 1049
Thomas, S., Belikov, R., & Bendek, E. 2015,
ApJ
,
810, 81
Toyoshima, M. 2006,
JOSAA
,
23, 2246
Wagner, R. E., & Tomlinson, W. J. 1982,
ApOpt
,
21, 2671
Wang, J., Mawet, D., Ruane, G., Hu, R., & Benneke, B. 2017,
AJ
, 153, 183
9
The Astrophysical Journal,
838:92
(
9pp
)
, 2017 April 1
Mawet et al.