Novel techniques of wavefront sensing for adaptive optics with array telescopes using an artificial neural network.

Abstract

Atmospheric turbulence causes severe degradation of the resolving and signal-to-noise properties of present optical telescopes. Diffraction-limited resolution can be recovered through the use of a deformable ('adaptive') optical element to correct the atmospheric wavefront error. An adaptive optics system operating in the near infrared (1.7 - 3.5 μm) has been developed for use at the Multiple Mirror Telescope (MMT), an array of six co-mounted 1.8 m telescopes, in which six flat mirrors are used to correct the wavefront tilt across each aperture, and the phase differences between apertures. This can reduce the error sufficiently to achieve a diffraction-limited image with a central peak of 0.06 arcseconds full width at half maximum at 2.2 μm wavelength. A number of algorithms are used to drive the adaptive mirror in a closed servo loop, including a trained artificial neural network which deduces the wavefront aberration from a pair of simultaneous in- and out-of-focus images of a star, taken at the combined focal plane of the telescope. Computer simulations have shown that the net is capable of deriving the wavefront for the full six-mirror aperture, and in practice, the net has been demonstrated in the lab to maintain two- and three-aperture diffraction-limited beam profiles in the presence of distorting effects. On the sky, with a real star, the net has successfully restored the diffraction limit for two adjacent MMT segments. High resolution images have been obtained of various objects with a wide-field camera looking in the field around the wavefront reference star. Work has also been carried out to characterise the wavefront aberration at the MMT, which confirms the Kolmogorov model of turbulence. Finally, a new algorithm is discussed which shows great promise for correction of phase errors in array telescopes.