The success of the Hubble Space Telescope (HST) has revolutionized astronomy. Although by modern standards the telescope is small (2.5m in diameter), it has an angular resolution about eight times better than that which could be achieved with a similar ground-based instrument. It will not be possible to make any further repairs to the HST, but astronomers will continue to demand images in visible light wavelengths with comparable quality. However, the much-anticipated James Webb Space Telescope will operate principally in IR wavelengths and will have a poorer angular resolution than the smaller HST.
Measurements with an angular resolution of 1 arcsecond can be typically achieved by the best ground-based observatories. Adaptive optics systems are used to measure and correct for the atmospheric distortion of the wavefront that enters the telescope. These systems work well for near-IR wavelengths, where atmospheric turbulence changes are slow. For visible wavelengths, however, the corrections are difficult for all but the rare, brightest reference stars. Artificial laser guide stars can be used in the adaptive optics systems to help achieve resolutions at visible wavelengths that are close to, but not better than, the HST level.
We have designed a technique with which we can routinely produce HST-resolution images from ground-based telescopes. We use the long-established lucky imaging approach1 to take images at a fast enough rate so that motions due to turbulence are frozen. By combining several of these sharp images we are able to achieve imaging resolutions on HST-sized telescopes that are close to the diffraction limit. It is our goal to extend these methods to significantly larger telescopes, which is essential to realize even higher angular resolutions. The chance of finding a sharp image from a large telescope (>2.5m diameter) is very small and atmospheric turbulence has the greatest effect at these largest scales. We have therefore developed a new type of wavefront sensor that is particularly sensitive to large turbulent scales.
We derive the wavefront curvature with our sensor by measuring the variation in light intensity across a pupil plane as the light propagates through the telescope (see Figure 1).2 Patches in the pupil images that change from bright to dark indicate that the wavefront is diverging and converging. The light propagation is nonlinear and so we measure the intensity on four different planes (two on each side of the pupil), typically at 100Hz. Although our wavefront distortion computations are demanding, they can be made in real time with a high-speed graphics processing unit. We use the derived wavefront errors to drive a deformable mirror in the optical path, which compensates for the errors. By removing the largest scales of turbulence, this process greatly reduces the phase variance across the pupil plane. Thus we increase the size of the patch over which the phase varies by, e.g., 1 radian. As we remove each turbulent scale, the effective cell size increases. Once the cell count across the telescope aperture is sufficiently small, the lucky imaging techniques can be used and high-resolution images can be obtained.
Figure 1. The propagation of light intensity from the pupil plane of a telescope. The pupil has uniform illumination and a phase value to represent the propagation through a simulated turbulent atmosphere. On either side of the pupil, the intensity breaks into speckles. The smaller speckles show the higher order structure and are located nearer to the pupil. The larger structure, which shows the low orders, develops as the propagation distance is increased.
We have used the 5m telescope at the Palomar Observatory in California and a conventional bright-star adaptive optics system, combined with lucky imaging, to demonstrate our technique. Example results from our observations are shown in Figure 2. The images obtained using our adaptation of the lucky imaging technique have significantly better angular resolution than images obtained in the standard manner from the 5m Palomar telescope and from the Hubble Advanced Camera for Surveys. This allows greater detail to be observed for the same regions of the sky. Indeed, the highest resolution image shown in Figure 2 has the highest resolution of any image ever taken at visible or near-IR wavelengths for faint targets.
Figure 2. The core of a globular cluster (Messier object M13) imaged with three different systems. Left: Natural seeing with the Palomar Observatory 5m telescope, angular resolution ∼0.65 arcsecond. Middle: Hubble Advanced Camera for Surveys, angular resolution ∼120 milliarcsecond. Right: Lucky Camera and low-order adaptive optics with the 5m Palomar telescope, angular resolution 35 milliarcsecond. The ability to resolve small, faint objects in the middle and right images illustrates their high resolutions.
We are now building a next-generation instrument for use on telescopes in the Canary Islands.3 This project is funded by the British Science and Technology Facilities Council and is part of a collaboration with groups from Tenerife (Canary Islands), Cartagena (Spain), and Cologne (Germany). This instrument will initially be used on the 4.2m William Herschel Telescope on La Palma and will provide angular resolutions similar to those we have achieved with the Palomar 5m telescope. Our instrument will be subsequently transferred to the much larger Gran Telescopio Canarias—GTC—on La Palma. This 10.5m-diameter telescope, together with our technique, should provide an angular resolution of only 15 milliarcseconds, which is more than a factor of 60 better than that from other ground-based telescopes. With this level of performance we will be able to resolve distant gravitational lenses for the first time, search for binary stars close to the cores of globular clusters, and resolve Milky Way-like galaxies anywhere in the universe.
We have developed an imaging technique that can be used to improve the angular resolution in visible and near-IR wavelengths of ground-based telescopes, by removing the effects of atmospheric turbulence.4 So far, our approach has been used on a 5m-diameter telescope and will soon be used on a telescope that is double the size. We are now working to apply our technique to even larger telescopes and at longer wavelengths. Construction of 30m-class telescopes is now underway, e.g., for the European Extremely Large Telescope in Chile. We believe that we can achieve similar angular resolutions at a wavelength of 2.2μm on a 30m telescope as we can at 700nm on a 10m telescope.
Institute of Astronomy
University of Cambridge
Cambridge, United Kingdom
Craig Mackay is professor of image science and was a member of the Faint Object Camera team for the Hubble Space Telescope. He has been involved with developing instruments for DNA and protein analysis, x-ray and electron beam imaging, as well as high-resolution optical microscopy.
2. P. L. Aisher, J. Crass, C. Mackay, Wavefront phase retrieval with non-linear curvature sensors, Mon. Not. R. Astron. Soc. 429, p. 2019-2031, 2013.
3. C. D. Mackay, R. Rebolo-López, B. Femenia Catellá, J. S. Crass, D. L. King, L. Labadie, P. Aisher, et al., AOLI: Adaptive Optics Lucky Imager: diffraction limited imaging in the visible on large ground-based telescopes, Proc. SPlE
8446, p. 844621, 2006. doi:10.1117/12.925618
4. C. Mackay, High-efficiency lucky imaging, Mon. Not. R. Astron. Soc. 432, p. 702-710, 2013.