Lucky Imaging at Other Wavelengths

At longer wavelengths the optical path errors due to atmospheric turbulence are much less severe because they represent a smaller number of wavelengths. The spatial correlation scales r₀ and temporal correlation scales t₀ both vary in proportion to wavelength to the power of 1.2. This means for example that at K-band (2.2 microns wavelength) the length and timescales will be increased by a factor of approximately three when compared with the scales in I-band (850 nanometres wavelength). Similarly, the scales will be shorter and faster by a factor of two into B-band (450 nanometres wavelength).

Telescope sizes comparable to a 2.5m telescope at 850nm:

Pass band	Equivalent telescope diameter scaled as λ1.2
B, 450nm	1.17m
V, 550nm	1.48m
R, 650nm	1.81m
I, 850nm	2.5 m
J, 1.2 micron	3.78m
H, 1.6 microns	5.34m
K, 2.2 microns	7.83m

What this does mean is that the many experiments that we have carried out in I-band (850 nanometres) on the NOT 2.5 metre telescope on la Palma directly comparable with equivalent experiments carried out in K-band (2.2 microns) on 8 metre class telescopes such as the European Southern Observatory VLT.

A necessary prerequisite, for successful Lucky Imaging, is to have a detector system that can be run at high speed and yet have essentially negligible readout noise so that the technique does not lead to a very high intrinsic noise background. At present near infrared detectors such as those made by Rockwell have readout noises of perhaps eight electrons RMS at pixel rates of a few hundred kilohertz. Rockwell are working with Caltech and others to develop a much higher speed readout architecture running at megahertz pixel rates with the readout noise close to one electrons RMS. This sort of technology will be essential if Lucky imaging is to be used at longer wavelengths on larger diameter telescopes.

For more information on infrared detectors manufactured by Teledyne click here: http://www.teledyne-si.com/infrared_visible_fpas/

Potential for Wide field Surveys

The Lucky Image Selection method can be applied to most kinds of imaging. For example, we can use a CIRSI type approach, with an array of EMCCDs each selecting their own Lucky images which are then mosaiced after the exposure. We use 6 pointings to cover entire field of view. In one night can survey to very faint limits and with excellent resolution using the methods outlined elsewhere.  We have already made a mosaic of 9 fields in the globular cluster M15, and compared it with an image taken with a conventional CCD camera on the same telescope at the same time.

In each of the nine fields around the core of the globular cluster we used an 80 second run with 10% selection to produce each of the nine fields. These were then merged in the computer to produce a composite Lucky image that covers just under one arc minute square. The first image was taken on the same telescope with the standard CCD camera that is use there for general-purpose imaging. It is a thinned (back illuminated) 1024 by 1024 CCD and it was used image the centre of the globular cluster with an image scale of six pixels per arc second. The seeing was approximately 0.6 arc seconds as measured on that camera, and a sequence of Lucky images were taken immediately after the standard camera image was recorded.

In the image below, you can see a comparison between the original standard camera image and the high-resolution Lucky mosaic image. The Lucky mosaic is the sharper, higher contrast image.

The above picture was taken with a single detector that was moved from position to position to cover the whole field. It will be very straightforward in principle to build an instrument that had a mosaic of detectors in the way that was used for the CIRSI instrument. This has the sensitive areas of the detectors separated from one another by slightly less than the width of the detectors in one direction and slightly less than twice the height of the detectors in the other direction. By using six different pointings of the telescope images may be obtained over the whole sky. Each detector will have its own moments of excellent seeing and its own moments of very poor seeing. At the end of the sequence of six pointings we will have observed the entire region of the sky with the full resolution of the Lucky technique and will be able to produce images over very wide fields. For example with an array of 96 CCDs each of 1024 by 1024 pixels, in an array of 12 by 8 it would be possible to survey one quarter of the square degree in one night to limiting magnitude in the region of I ~ 24. The sort of mosaic would look rather like that shown in the following image:

This image shows the way that the frame-transfer L3CCDs might be laid out in order to allow a wide field survey to be carried out with only six telescope pointings. The area shown coloured indicates the light sensitive part of the CCD. The rest of the CCD is required for charge transfer and storage since these are all frame-transfer CCDs.

High-Resolution Lucky Imaging

The Problem: Lucky Imaging on Larger Telescopes

The inevitable demise of the Hubble Space Telescope (HST) will leave a huge gap in our astronomical capabilities. There is no planned space project able to equal its resolution or sensitivity in the visible. Astronomers will look to the instrumentation community to fill this gap. Adaptive optics techniques have so far barely managed to match Hubble resolution and then only in the near infrared (K-band: 2.2 µ). Laser guide stars have also been very slow to deliver on their promise. It is disappointing that so little science has come out of such a substantial investment worldwide in adaptive optics and laser guide stars despite over 20 years of development. Realistically there is little immediate prospect of this changing significantly. Techniques which actually do deliver Hubble resolution or better, either stand-alone or in conjunction with an AO system, must be developed as a matter of urgency. Ideally they should use faint natural guide stars (to assure good sky coverage), and provide a usable isoplanatic patch size (at least 20 arc seconds diameter). A successful technique must work with an acceptable photon gathering efficiency as well as being simple to build, operate and maintain and therefore of relatively low cost. Without such an imaging capability, Extremely Large Telescope (ELT) programs become much harder to make work well. Fortunately, we now have an exciting new way forward. New detector developments of noiseless, electron-multiplying CCDs, have allowed us to demonstrate all the technologies required to create a system that satisfies these requirements at relatively modest cost.

A programme in Cambridge to develop Lucky Imaging techniques has been extremely successful and led to a number of scientific papers. They have shown that atmospheric turbulence is more complicated than usually assumed by adaptive optic systems designers. Often the errors it introduces are so substantial that frequently no amount of correction will compensate for it (see figure 1).

Figure 1: A typical example of the variation in full width of a bright star image as tracked over about 80 seconds, with a frame time of 85 ms. Dramatic changes in the size at of images obtained are seen on every timescale right down to the shortest that the system could detect. These rapid, discontinuous changes are very difficult for adaptive optic systems to follow.

This behaviour is not predicted by conventional Kolmogorov turbulence theory but we have found these conditions are almost always present in practice. Our approach is to take images fast enough to freeze the atmospheric fluctuations. Each frame is then checked for quality and only those good enough are combined to create the final image. Under reasonably good conditions on a 2.5 m telescope in I-band we are able to use about 10% of the images to give 0.1 arc second resolution, 20% gives 0.13 arc second resolution and 30% gives 0.15 arc second resolution. Resolution much better than Hubble requires 7-10m telescopes (in I-band), but with telescopes of this size there is essentially no chance of a diffraction limited frame being obtained. This is because there are so many more turbulent cells across the area of a much larger telescope. It is the ratio of the diameter of the telescope to the typical cell size that determines whether a lucky imaging will work on any particular telescope/wavelength/seeing combination. Most of the power in the atmospheric turbulence is on the largest scales. These are relatively slowly varying and so can be measured relatively easily using rather faint reference stars. By using a low adaptive optics system in front of our lucky camera we effectively increase the cell size of the turbulence so that lucky imaging again becomes practical.

There is the additional problem that we always need to have a reference star so that we know which images are sharper. If the system is to be usable open much of the sky we need to be able to work with reference stars as faint as 18.5 magnitude. Traditional wavefront sensors used with adaptive optic systems such as the shack-Hartmann sensor simply cannot work at these magnitudes since the breakup the light into a large number of different cells. Our designs however use a rather different technique which is particularly sensitive when managing low order wavefront sensing. These curvature sensors should allow us to get to this sensitivity limit that we need. We are currently funded to build such a system for the European Southern Observatory's Very Large Telescope (VLT). these 8.2 m diameter telescopes should let us achieve an angular resolution which is 4-5 times that of Hubble. The telescope diameter is 3 times that of Hubble and by working not only in the far red but in the visible we should be able to get the additional resolution. Such a system will be very exciting and will open entirely new fields of research.

A paper about all this was recently published in Optics and Photonics News, andcan be found here.