Imaging and characterizing even the brightest extrasolar planets requires million-to-one contrasts near the diffraction limit of giant telescopes.1 Adaptive-optics (AO) systems correct for distortions caused by the Earth's atmosphere, channeling most of the starlight into the telescope's well-behaved diffraction pattern while leaving the rest scattered into a rapidly changing halo of speckles (see Figure 1). The Airy rings of a typical diffraction pattern are only 100–1000 times fainter than the star itself and must be suppressed a further 3–4 orders of magnitude using a coronagraph.2 This does not reduce the residual speckle configuration, however, which remains at least 1000 times more luminous than a bright exoplanet and is very difficult to suppress. Fortunately, atmospheric scatter creates a random intensity distribution, and the rapidly changing speckles should generate a smooth average background that can be subtracted from sufficiently long exposures.
This simple approach fails, however, because of tiny flaws in the telescope optics that cause stationary speckles near the central star image, thus mimicking potential exoplanets. Even nanometer-sized errors in the optical path or a slight coating of dust on any of the mirrors or lenses cause spurious false ‘planets’ to appear, many times brighter than any real exoplanets. These flaws are so small (and vary slowly with telescope orientation and temperature) that they cannot be removed, even if using detailed calibration procedures. In addition, the faint static speckles do not vary rapidly enough to average out over time. As a result, they provide the most serious limit to exoplanet detection.3
Figure 1. Residual adaptive-optics (AO) speckles surround this 5μm series of stellar images, taken at a rate of 15 frames per second.
However, the AO system—which is normally aimed at correcting wind-driven wrinkles in the incoming starlight—can also smooth out the telescope's flaws if we know them sufficiently well. Over a reasonably narrow wavelength range, the entire stellar halo, both diffracted and scattered light, is coherent and can be interfered with itself. Since the problematic static speckles are typically much fainter than the fast AO-corrected speckles, they are only revealed after some integration. The AO system's deformable mirror (DM) can be used to channel starlight into the field of view to probe underlying irregularities, and modulate it to provide the required suppression information as soon as the static speckles can be seen. However, this technique suppresses light from any exoplanets and increases the halo brightness precisely where we want it suppressed. Our technique circumvents problems by using the random atmospheric speckles as operational probes (instead of creating artificial ones). This turns a weakness into a strength.
As the wavefront changes across the telescope pupil due to atmospheric variations, different sections of the incident starlight take slightly different amounts of time to arrive in the focal plane. Over a science camera's few percent bandwidth, this varying delay relative to the bright stellar core appears as a changing phase. The speckles interfere with the diffraction halo and any static speckles, modulating their intensity coherently. The resulting total halo intensity reaches a maximum when the phases match. This modulation is well known but was considered useless given that the fleeting speckle phases were unknown. However, we realized that the latter can be computed from the residual wavefront-sensor (WFS) signal driving the AO system4 (see Figure 2). By post-processing the data to compute the speckle phases we can ‘phase sort’ pixels from short science-camera exposures (see Figure 3) into a set of phase-binned speckle images and compute an interferometric map of the static halo. The technique should be effective even when the WFS and the science camera operate at very different wavelengths.
Figure 2. (a) The wavefront sensor provides us with a low-order estimate of the wavefront errors left behind by the AO system. (b) Using Fourier transformation (inset), we can estimate the phase (color) and amplitude (brightness) of the fast AO speckles and match them up with the science-camera images.
The advantage of this technique lies in the use of information already routinely collected during AO-system operation along with reasonably short science-camera exposures. This means that it can be applied to AO-equipped telescopes as a software enhancement. Once the phase and amplitude of a static speckle have been estimated, adjustments to the DM can suppress it,5 leaving the incoherent light from the exoplanet for detection and further study. We are developing this technique at the MMT Observatory6 using the 5μm Clio science camera.7 Both simulations and initial processing of archived WFS and Clio data are very promising, and we plan to perform the first on-sky demonstration of our phase-sorting interferometry method before the end of 2008. The first on-sky adaptive-halo suppression is expected to be done in 2009. Deep exposures to look for exoplanets will be obtained as soon as the halo-suppression servomechanism is working, with engineering and scientific developments proceeding together.
Figure 3. (a) Science-camera image, (b) computed complex speckle map, and (c) science image color coded by AO speckle phase to sort into a set of interferometric images.
This work is supported by National Science Foundation grant AST-0804586.
University of Arizona
Johanan Codona is senior research scientist at the Center for Astronomical Adaptive Optics. He was previously a distinguished member of technical staff at Bell Labs.
3. Sasha Hinkley, Ben R. Oppenheimer, Rémi Soummer, Anand Sivaramakrishnan, Lewis C. Roberts Jr., Jeffrey Kuhn, Russell B. Makidon, Marshall D. Perrin, James P. Lloyd, Kaitlin Kratter, Douglas Brenner, Temporal evolution of coronagraphic dynamic range and constraints on companions to Vega, Astrophys. J. 654, pp. 633-640, 2007.doi:10.1086/509063