Recent developments in adaptive optics

The detail that astronomical telescopes can resolve is limited by the effects of turbulence in the Earth's atmosphere. In fact, the resolving power of the largest telescopes in the world equals that obtained with 10-20cm amateur telescopes. The goal of adaptive optics (AO), originally described in the 1950s, is to measure and correct the wavefront perturbations in real time^1,2 using a bright reference source. It was not until the 1970s that the first AO systems were developed, and then the context was military. AO systems using natural reference sources have been installed on most of the main astronomical telescopes in the world.³ Most of them operate in the near infrared, since the wavefront correction is easier as wavelength increases.

However, bright natural reference sources are relatively rare. The solution of using artificial guide stars, generated using laser beams, was proposed in the 1980s by astronomers,⁴ but had already been tested by the military. Systems suitable for large telescopes are based on illuminating a layer of sodium ions that exists at an altitude of about 90km. It has been difficult to obtain systems that provide enough power and are robust enough to operate reliably in the harsh environment of mountaintop observatories. Dye laser solutions have been implemented, for example, at the Lick and Keck telescopes, and significant astrophysical results have been obtained.⁵ Currently there is significant development in the area of solid state solutions, including fiber lasers. These have the advantage of providing both beam generation and transport to launch in a single system.⁶

In addition to the rarity of natural reference sources, another problem is the smallness of the field over which the wavefront aberrations are the same (the isoplanatic patch). These two factors severely limit the sky coverage of AO, which turns out to be less than 1%. Attempts are underway to increase the AO field of view. In this case, wavefront sensing has to be carried out using several reference sources. If one deformable mirror (DM) is available, this mirror can be controlled to correct turbulence near the pupil, which will provide moderate correction over the full field of view. This is called ground-layer AO, and the concept has recently been demonstrated by a group from the University of Arizona.⁷ Alternatively, the DM can be controlled to provide enhanced correction in one particular direction. If two or more DMs are available, they can be controlled to correct over a significant field. This approach has recently been demonstrated on the sky by the European Southern Observatory.⁸

Another AO challenge lies in developing techniques to directly detect extrasolar planets. Since the planet will be very close to the parent star, and only 10⁻⁶ to 10⁻⁹ times as bright, this requires wavefront correction to the level of a few nanometers, and very efficient rejection of light from the parent star. This high-order AO, referred to as extreme AO or XAO, is another area of rapid development. Various coronographic and pupil masking schemes have been devised to reject the parent starlight. The limit to exoplanet detection is usually set by residual, quasi-static speckles associated with the optics of the telescope and the AO system itself. Differential imaging strategies have been developed to minimize this effect.⁹

Figure 1. The multiconjugate adaptive-optics laboratory setup. Click image to enlarge.

Figure 2. The receiver operating characteristic (ROC) curves compare the performance of planet detection algorithms as a function of the threshold employed as obtained from numerical simulations. The relative intensity of the star to the planet is 10⁵, with a separation of six times the diffraction limit. The dot-dashed curve corresponds to subtracting the star point spread function and searching for pixels above the threshold, while the solid line corresponds to applying a Hotelling observer, which is the optimal linear algorithm. (Click to enlarge graph.)

At the Applied Optics group at the National University of Ireland, Galway, (NUIG) we are working to address several of the current challenges in astronomical and retinal adaptive optics. For instance, we have developed a test bench for multiconjugate adaptive optics (see Figure 1), including a very compact multi-object wavefront sensor.¹⁰ The bench is being upgraded to test a concept called virtual wavefront sensing.¹¹ This technique aims to solve some problems inherent to the use of laser guide stars on future extremely large telescopes (ELTs). The laser guide star images will be severely aberrated due to the finite distance to the sodium layer, and it will be very difficult to relay the light from the lasers through the adaptive optics. In the virtual wavefront sensor, the laser light is extracted to wavefront sensors before it enters the AO system. The DMs are sensed independently, e.g., by wavefront sensing using test sources at the focus before the AO system. The outputs of the laser and test wavefront sensors are combined to give an output that approximates what would have been measured by propagating the lasers through the whole system.

Another problem that will be particularly relevant for ELTs is the correction of atmospheric dispersion. This will blur images with a finite bandpass by more than the diffraction limit for even modest zenith angles. The usual solution is to correct dispersion using counter-rotating prisms. An alternative solution is to use two wedges that can be separated by different amounts; we have recently published a design solution for ELTs based on this concept.¹²

In collaboration with Prof. Harrison Barrett and co-workers at the University of Arizona, we are also developing approaches to the analysis of data from AO systems.¹³ These include optimal techniques for the detection of exoplanets.¹⁴ The usual ad hoc approach is to look for pixel intensities that are some number of standard deviations above the background noise. It is possible, using techniques developed for medical imaging, to do much better. The optimal linear approach is the Hotelling observer, which corresponds to a prewhitening matched filter. This approach is being tested using numerical simulations (see Figure 2) and real data obtained with the AO system at the 3m Shane telescope at the Lick Observatory, California.

While astronomical AO has finally come of age, there are many ongoing, exciting developments. The future will see AO play a key role in ground-based astronomy, from the provision of modest, wide-field correction for general-purpose astronomy, to extreme correction as required for the direct imaging of exoplanets. Ultimately, astrophysical breakthroughs will be obtained using multi-laser guide star systems on ELTs. The technical challenges are formidable, but the return in scientific discovery should make the effort well worthwhile.