In this paper we present the Adaptive Optics Lucky Imager (AOLI), a state-of-the-art instrument which makes use of two well proved techniques for extremely high spatial resolution with ground-based telescopes: Lucky Imaging (LI) and Adaptive Optics (
AO). AOLI comprises an AO system, including a low order non-linear curvature wavefront sensor together with a 241 actuators deformable mirror, a science array of four 1024x1024 EMCCDs, allowing a 120x120 down to 36x36 arcseconds field of view, a calibration subsystem and a powerful LI software. Thanks to the revolutionary WFS, AOLI shall have the capability of using faint reference stars ({it I/} $sim$ 16.5-17.5), enabling it to be used over a much wider part of the sky than with common Shack-Hartmann AO systems. This instrument saw first light in September 2013 at William Herschel Telescope. Although the instrument was not complete, these commissioning demonstrated its feasibility, obtaining a FWHM for the best PSF of 0.151$pm$0.005 arcsec and a plate scale of 55.0$pm$0.3 mas/pixel. Those observations served us to prove some characteristics of the interesting multiple T Tauri system LkH$alpha$ 262-263, finding it to be gravitationally bounded. This interesting multiple system mixes the presence of proto-planetary discs, one proved to be double, and the first-time optically resolved pair LkH$alpha$ 263AB (0.42 arcsec separation).
Increasing interest in astronomical applications of non-linear curvature wavefront sensors for turbulence detection and correction makes it important to understand how best to handle the data they produce, particularly at low light levels. Algorithms
for wavefront phase-retrieval from a four-plane curvature wavefront sensor are developed and compared, with a view to their use for low order phase compensation in instruments combining adaptive optics and Lucky Imaging. The convergence speed and quality of iterative algorithms is compared to their step-size and techniques for phase retrieval at low photon counts are explored. Computer simulations show that at low light levels, preprocessing by convolution of the measured signal with a gaussian function can reduce by an order of magnitude the photon flux required for accurate phase retrieval of low-order errors. This facilitates wavefront correction on large telescopes with very faint reference stars.
