On-Sky Speckle Nulling Demonstration at Small Angular Separation with SCExAO

The detection of high-contrast features such as disks and companions in diffraction-limited images relies for the most part on the characterization and calibration of diffraction effects, whether static (induced by the geometry of the pupil and the aberrations in the optics), quasi-static (telescope pointing and slowly variable instrument alignment), or dynamic (atmosphere-induced). For ground-based observations, the best performance comes from a combination of adaptive optics (AO), which address the dominant atmospheric term, turning the seeing-limited image into a diffraction-limited one, and post-processing techniques based on differential imaging. These include spectral, polarimetric or angular differential imaging (ADI), the first two relying upon some properties of the target, such as the presence of specific spectral features or a polarimetric signature.

ADI (Marois et al. 2006; Lafrenière et al. 2007) provides the means to calibrate the static and quasi-static features of the PSF by observing a target with an alt-azimuthal telescope, for which the sky rotates relative to the telescope pupil. If enough field rotation occurs over a time that is less than the characteristic quasi-static aberration time-scale, the diversity between the orientation of the sky and the quasi-static point spread function (PSF) allows the calibration of the diffraction pattern, leading to greatly enhanced detection limits.

The calibration requires that the angular rotation induces sufficient local linear displacement of the genuine features, such as faint companions, relative to the quasi-static PSF (Marois et al. 2006) in order to avoid self-subtraction. While highly efficient at angular separations greater than 1'', it is difficult to benefit from this technique at angular separations smaller than 0.5''. For addressing such small angular separations, two approaches are possible: interferometric calibration and additional active wavefront control.

Interferometric calibration of diffraction-limited images is a nascent field that inherits from the ideas and techniques of sparse aperture masking (SAM), a.k.a., nonredundant masking (NRM) interferometry with AO (Tuthill et al. 2006). The fundamental idea behind these techniques is that although the content of images may be corrupted by residual aberrations, it is possible to construct a subset of information that is independent from these aberrations. One possible such subset of information is the kernel-phase (Martinache 2010) which, just like sparse aperture masking using closure phase, goes as far as enabling the direct detection of high contrast features beyond the diffraction limit, a regime referred to as super-resolution (Pope et al. 2013). The fundamental advantage of this approach is that it requires no additional differential "trick" like ADI, although it is for now restricted to well-corrected AO images. Recent improvements of the technique, such as the statistically independent kernel-phase and better calibration procedures described by Ireland (2013), show promise of performance that compares to what is currently achieved by ADI at small angular separations.

The alternative approach to the passive calibration by interferometry is to employ additional wavefront control to modulate the diffraction at small angular separation and create the diversity that enables the separation of genuine structure from diffraction features in the image. This concept, which can be referred to as coherence differential imaging (CDI), is very relevant to the upcoming generation of so called extreme AO (XAO) projects: Palm-3000/P1640 at Palomar (Oppenheimer et al. 2012), GPI at Gemini (Macintosh et al. 2008), SPHERE at VLT (Sauvage et al. 2010), and SCExAO at the Subaru Telescope (Martinache et al. 2011).

Among these projects, the Subaru Coronagraphic Extreme AO (SCExAO) project is unique with its incremental deployment schedule that enabled the on-sky testing of fundamental features of the instrument (Jovanovic et al., in preparation) prior to completion of its closed-loop XAO system (Clergeon et al. 2013). In this paper, we present the first convincing demonstration of on-sky closed-loop diffraction modulation, after an upstream AO correction is performed, providing a Strehl on the order of 20%. While the demonstrated gain is moderate and there is room for improvement, the paper nevertheless demonstrates that it is possible to control the features of a coronagraphic PSF near the inner working angle of the coronagraph using the speckle nulling approach introduced in previous work.

What is usually referred to as an XAO system is a unit capable of producing a very stable and high-quality PSF, characterized by a Strehl ratio on the order of 90%. This very high level of Strehl, corresponding to residual wavefront errors on the order of λ/20 (where λ is the wavelength of observation), is indeed required for the high-contrast technique of coronagraphy, which aims to block the light of an on-axis bright source and reveal its circumstellar environment, in particular disk and/or faint companions.

Particularly in monochromatic light, it is possible to design coronagraphs that provide near-perfect extinction, and an inner working angle (IWA) that approaches the limit of diffraction (λ/D), where D represents the diameter of the telescope aperture. Examples of such coronagraphs include the vortex (Mawet et al. 2010) and the PIAA (Guyon 2003), both implemented on SCExAO. In practice, however, their performance is entirely dictated by the residual wavefront errors. Under poor Strehl conditions, the coronagraph does little more than allow the exposure time to be increased before saturation of the detector.

SCExAO does not currently support a high-order closed-loop AO system, expected to be first commissioned in 2014 fall. In this first phase of the project, SCExAO therefore relies entirely on the upstream Subaru Telescope facility AO system called AO188 (Minowa et al. 2010). AO188 and the coronagraphic imager HiCIAO (Hodapp et al. 2008) form the workhorse of the Strategic Exploration of Exoplanets and Disks with Subaru (SEEDS) observing campaign (Tamura 2009). SCExAO was conceived as a replacement upgrade for the current fore-optics used by HiCIAO, that should include a small IWA PIAA-based coronagraph as well as additional wavefront manipulation capability.

Therefore, while SCExAO does not currently qualify as an XAO system, it nevertheless already implements a 1-k actuator deformable mirror (DM). We have recently shown, using a calibration source in a stable laboratory environment, that the DM and the PIAA coronagraph used together can produce a high-contrast region in the field-of-view, with a 2.2 λ/D IWA (Martinache et al. 2012), with a simple iterative speckle-nulling loop. Using the same approach on-sky, relying on AO188 to produce a 20% Strehl ratio PSF, we demonstrate that it is possible to use speckle nulling to drive down the brightness of the diffraction features at small angular separation. We also show that the DM provides a flexible calibration tool that notably enables astrometric calibration as well as direct measurements of the contrast in images.

Figure 1 provides a schematic overview of the optical layout of the near-IR arm of SCExAO, as it was used on the observing night of 2012 November 12. The upstream AO188 system feeds the instrument with a partially corrected f/14 convergent beam, intercepted by a first tip-tilt controlled mirror located in an image plane and mounted on a focusing stage, and further steered by a second tip-tilt controlled mirror. This combination provides full control of the origin and incidence of the beam entering the coronagraph. The other important feature is the focal plane mask, reflecting the light of any on-axis source onto a camera used to sense the pointing, a subsystem called the coronagraphic low-order wavefront sensor (CLOWFS), as described by Guyon et al. (2009) and Vogt et al. (2011).

Zoom In Zoom Out Reset image size

In addition, the system also comprises the PIAA remapping optics (Guyon et al. 2005) located before the focal plane mask and their inverse counterpart located after the focal plane mask. Recent work (Martinache et al. 2012) showed that when used together, the presence of these two sets of remapping optics can be ignored by the focal plane based wavefront control algorithm. Reflecting upon this conclusion, this discussion will ignore the remapping optics, which for the results presented in this paper, are always in the beam.

One of the two tip-tilt mirrors that steers the beam entering the coronagraph is the 1024-actuator DM itself which, while not exactly located in a pupil plane, can nevertheless be used to introduce/compensate for diffraction features in the coronagraphic image. The square grid geometry of the actuators underneath the DM membrane makes it particularly well-suited to generate pairs of speckles that interfere with already present diffraction features in the coronagraphic image. The total number of actuators across the telescope pupil footprint imposes the size of the region of the coronagraphic image that can be probed by the DM, which will be referred to as the control region. As explained by Martinache et al. (2012), the control region with this current implementation fits within a 27.2 × 24.8 λ/D rectangular box. It is possible to generate a pair of speckles centered around the optical axis within this region by applying a sinusoidal displacement map on the DM.

Each such sinusoidal displacement map on the DM is characterized by four numbers: two cartesian coordinates of spatial frequency (k_x,k_y), one amplitude α (in radians) limited by the stroke of the DM, and a phase φ that can take any value between 0 and 2π. When applying a sinusoidal displacement map, the DM acts like a diffraction grating that creates off-axis copies of the on-axis bright source. Figure 2 illustrates the relation between DM modulation and speckles in the image. The larger the spatial frequency parameters, the faster the sinusoidal modulation of the DM surface and the further the corresponding pair of speckles lies away from the center of the control region. The larger the amplitude of the modulation, the brighter the resulting speckles. The phase does not change the intensity of the speckles unless they interfere with other coherent features in the image. Sinusoidal modulations of distinct properties can be added to the DM as long as the total requested range of displacement does not go beyond the DM stroke, creating in turn multiple pairs of speckles in the image.

Zoom In Zoom Out Reset image size

The ability to produce symmetric pairs of speckles of controllable positions and contrast is a very convenient feature that facilitates otherwise difficult calibration procedures when observing with a coronagraph. Indeed, unlike classical imaging techniques and interferometric methods for which absolute control of the pointing is not a strict requirement (since the central star remains visible), the image of the star needs to be precisely driven and stabilized on the axis of the coronagraph for an efficient suppression.

Sivaramakrishnan & Oppenheimer (2006) astutely proposed to address this issue by inserting a reticulate grid of wires in the pupil plane that produces a predefined periodic diffraction pattern in the image. The grid geometry of deformable mirrors such as the 1k-actuator MEMS DM used in SCExAO can advantageously be used as an adaptive version of this reticulate grid. Just as described by Sivaramakrishnan & Oppenheimer (2006), the added satellite speckles provide a very reliable way to position the star behind the focal plane mask during target acqusition. Once in the ideal position, the low-order wavefront sensor (Guyon et al. 2009) keeps the image in a fixed position relative to the mask. During the post-processing stage, in case noncoherent structures such as companions are identified in the image, the satellite speckles offer a reliable astrometric reference.

In addition to offering astrometric and pointing aids (Hinkley et al. 2013), the generated speckles can be used for contrast calibration purposes. This can efficiently be achieved by an offline predetermination of the satellite speckle contrast relative to the central PSF as a function of spatial frequency and amplitude. Figure 3 shows one example of such on-sky contrast calibrated images, recorded by the coronagraphic imager HiCIAO fed light by SCExAO. The top panel of Figure 3 shows a nonsaturated low-exposure time image with the SCExAO focal plane mask out of the way. A sinusoidal modulation of the SCExAO DM of known characteristics creates a pair of speckles of controlled contrast. The bottom panel shows an image acquired with the same exposure time, this time with the SCExAO focal plane mask placed on-axis. Except at small angular separation, one can observe very little impact of the focal plane mask on the halo PSF features.

Zoom In Zoom Out Reset image size

Martinache et al. (2012) showed, using SCExAO under stable laboratory conditions, that it is possible to actuate the DM before the PIAA coronagraph to create diffraction features that destructively interfere with speckles in the field, creating a uniform high contrast region, often referred to as a "dark hole" in the image. Here, we apply the same iterative speckle nulling algorithm during actual on-sky observations. The major difference with the laboratory scenario is that the static (or quasi-static) features of the PSF probed by the speckle nulling algorithm are buried underneath a very strong, fast-varying dynamic component, due to the incomplete AO correction in the absence of a properly closed XAO loop.

The reported performance of ADI calibration suggests that high-contrast detection limits are set by long-lived aberrations, with a characteristic time-scale ∼1 hr or less. In this experiment, the science detector (HiCIAO) and the speckle-nulling camera cannot simultaneously observe: a retractable mirror sends the light either way. To be useful, the speckle nulling loop must therefore converge within a time-scale on the order of 15–20 minutes, so that sufficient time is left for the acquisition of frames by the (slow) science detector, before the quasi-static aberrations start changing again.

4.1. Technical Constraints

4.2. Speckle Nulling Data

Twelve minutes of the speckle nulling loop running were sufficient to allow us to visually observe a darkening of one side of the field of view. To test the efficiency of the wavefront modulation resulting from the speckle nulling loop, several series of images were acquired with HiCIAO, alternating between a "flat" and an "optimized" SCExAO DM shape. The flat DM shape resulted from an optimization routine maximizing the Strehl in the noncoronagraphic image using Zernike modes: astigmatism, focus, coma, spherical, trefoil, and quadrifoil on an internal calibration source. The optimized shape was the result of speckle nulling corrections applied in addition to the flat. The complete acquisition sequence consisted of four such cycles (to make the analysis robust against slow trend effects), resulting in a total of twenty 5-second exposure frames per DM shape. These HiCIAO images were dark-subtracted and flat-fielded, using standard data reduction procedures established for the reduction of the SEEDS data, assembled into two data cubes.

The images gathered in Figure 4 summarize the statistical properties of these two datacubes, looking at the median (left column) and the standard deviation (right column) for the flat (top row) as well as the optimized (bottom row) DM shape, using common colorbars. For each figure, the extent of the control region is highlighted by a red box, which, for this experiment, is located in the lower part of the HiCIAO image.

Zoom In Zoom Out Reset image size

The combination of exposure time (5 s) and neutral density (ND0.1) for HiCIAO was chosen to acquire nonsaturated images in the H-broadband filter. The low and fast varying seeing experienced during the acquisition and the lack of a yet to come high-order close-loop adaptive correction results in exposures that exhibit no obvious individual speckles in the field, a major difference from what is already achieved by XAO equiped instruments (see, e.g., Fig. 1 of Oppenheimer et al. [2013]). The first bright ring, located at a radius of 2.3 λ/D, is the dominating feature in the images. Note that this radius is very close to the quoted IWA of SCExAO (2.2 λ/D), defined as the angular separation for which the coronagraph throughput reaches 50% (Martinache et al. 2012), and imposed by the focal plane mask. The standard deviation of the DM flat datacube shows that most of the speckle variance is localised onto the diffraction rings.

The speckle nulling algorithm unsurprisingly focused its efforts onto this area of the image. With the optimized DM shape (bottom images in Fig. 4), the median flux for the part of the diffraction ring overlapping with the control region is reduced. The improvement in regards to the median is moderate as the brightest parts of the ring are attenuated by a factor of 2.

4.3. Benefit for High Contrast Imaging

This paper presents the first demonstration of speckle nulling performed on-sky with partial AO correction. In the context of the SCExAO project, using a PIAA coronagraph and a deformable mirror to modulate the focal plane, the results presented here show that it is indeed possible to complement ADI and its variants at small angular separations, even in the absence of a high strehl AO system, and therefore in the presence of a brighter dynamic speckle halo.

Speckle nulling offers improvement at two levels: (1) it removes slow speckles, resulting in images with enhanced contrast before processing, and (2) it reduces the speckle noise amplification by the static diffraction, resulting in enhanced post-processing detection limits. Moreover, it enables this in a regime of angular separation that cannot be addressed by ADI, which remains the reference technique (Brandt et al. 2013) for the calibration of ground based AO PSF. In spite of unfavorable observing conditions, the control loop managed to remain stable and converged to improve the statistical properties of the coronagraphic image.

While consistent, the reported improvement remains modest, and this can in part be attributed to the choices of integration time, constrained by the noise properties of the detector used during the speckle nulling loop. Improving the sensitivity and the dynamical range of the image by using a cooled detector will increase the flexibility of the technique and allow a more efficient use of the observing time. The real performance improvement, however, will come from the use of an actual close-loop XAO system that will improve the raw Strehl of individual images.

Advanced focal plane based wavefront techniques such as electric field conjugation (Give'On 2006; Bordé & Traub 2006) have been used extensively on laboratory high contrast test-beds over the world, including with a PIAA coronagraph like the one used on SCExAO (Guyon et al. 2010) providing access to very high contrast (10^-7 and beyond) detection limits in a stable environment. The fact that, even in difficult observing conditions, a speckle nulling driven wavefront control algorithm originally envisioned for stable test-benches remains operational is very encouraging. This indeed suggests that, with only a few adaptations, all of the high contrast laboratory work that has so far been geared toward space borne applications is relevant to ground based observations, even with partial AO correction.