Following the machining operation, the surface is ground with a sequence of finer abrasives and gradually brought to the desired shape. The atmosphere sets the accuracy requirements. The wavefront delivered to the telescope is extremely smooth on small spatial scales but has large-scale irregularities increasing to many microns on scales of 8 m or larger. The mirror surface needs to be several times better than the atmosphere on all scales. If it is, the surface errors are insignificant when adaptive optics are not used to correct for the atmosphere, and they are corrected for free when adaptive optics are used.

The use of active optics to control the shape of the primary mirror further relaxes the accuracy requirements on the largest scales, because we can bend in low-order aberrations like astigmatism with small changes in support forces. This relaxation is valuable for the difficult optical test in the lab. One can make wavefront measurements in the telescope by using the telescope¡¯s natural imaging property; these measures do not have to rely on a null corrector, so they¡¯re more accurate on large spatial scales than those made in the lab. Small misalignments in the null corrector would cause large-scale errors in the polished surface, and these can be fixed with active optics once a more accurate wavefront measurement is made in the telescope. Likewise, we can neglect slight mirror support errors and temperature variations in the glass that cause low-order aberrations during lab testing.

Measuring the off-axis segments

So why do we need a 3.75-m mirror to measure a GMT segment? Because the null corrector for the off-axis segments has to do so much more than any null corrector ever built. It has to make the template wavefront, with its 14 mm of aspheric departure, to an accuracy of about 1 µm on large scales and a smoothness of a few nanometers on small scales.

Among traditional, axisymmetric telescopes, the LBT has the most aspheric primary mirrors¡ªeach one is a symmetric paraboloid¡ªwith 1.4 mm of aspheric departure. Its null corrector consisted of a pair of lenses separated by 67 cm. The GMT null corrector, designed by Burge, is shown in the figure on the facing page. In addition to the 3.75-m mirror, it contains a second smaller mirror and a computer-generated hologram. Oblique reflections off of the two mirrors do most of the shaping of the wavefront, and the CGH cleans up the remaining aberrations.

The most challenging aspect of the GMT null corrector is alignment. The small package containing the interferometer, CGH and smaller mirror requires an alignment accuracy of about 10 µm, and the larger dimensions between that package, the larger mirror and the GMT segment must be controlled to about 100 µm. To get this level of accuracy in a non-axisymmetric system, we rely heavily on holograms and laser trackers. Holograms can be aligned optically to the wavefront (aligned to return a null wavefront to the interferometer), and they provide both optical and mechanical references so that other components can be aligned to the wavefront. We use laser trackers (distance-measuring interferometers coupled with sub-arcsecond angular encoders) to measure the positions and orientations of these reference holograms as well as the mirrors.

For previous mirrors, we could validate the interferometric measurement with a small CGH inverse null corrector. The GMT test wavefront is more than 3 m in diameter by the time it leaves the null corrector¡ªway too large to validate with a CGH. But we can achieve the same goal with an independent measurement of the segment figure, a test that¡¯s sensitive to the low-order aberrations that we would get wrong if there were a misalignment of the null corrector. For this purpose, Burge developed a scanning pentaprism system that scans the surface with a narrow collimated beam, which is focused on a detector in the mirror¡¯s focal plane. The displacement of the focused spot is proportional to the slope error on the surface.

These slope measurements are surprisingly accurate¡ªto about 0.1 arcsecond rms surface slope. That¡¯s because the scanning pentaprism, with two internal reflections, deflects the beam by a constant 90¡ã angle independent of small rotations of the prism. We use a second, fixed pentaprism to compensate for misalignments and instability in other components. A set of four scans at different diameters determines the first eight low-order aberrations to an accuracy of 50 to 100 nm rms, similar to the predicted accuracy of the principal optical test and well within the correction range of active optics.

The principal test and scanning pentaprism test work only on a polished surface. We measure the ground surface by scanning it with a laser tracker. (An infrared version of the principal test is possible but complicated by the use of holograms.) The tracker measures distance and angles, giving the coordinates of each sample point in 3D. It has sub-micron accuracy in its distance measurements, so it gives similar accuracy in the surface measurement if the tracker is located near the mirror¡¯s center of curvature.

The measurement is sensitive to drift in the position of the segment and the laser tracker during the scan, so we monitor fixed references at the edge of the mirror with a standard distance-measuring interferometer and compensate for any motion. There are similar references and an in situ calibration that compensate for errors in the angle measurements. Our goal is sub-micron accuracy in the GMT measurement.

The full set of test optics for the GMT segments has been installed in a new 28-m test tower at the Mirror Lab. The new tower replaced the original 24-m tower, which was used for all mirrors through the LBT primaries but wasn¡¯t quite large and stiff enough to accommodate the GMT tests. The GMT project and Steward Observatory have invested a tremendous effort in developing an accurate, redundant and convenient suite of tests for these segments. Combined with the casting, machining and polishing equipment at the Mirror Lab, these test systems provide a complete manufacturing plant for efficient serial production of the GMT segments.

The optics for the GMT build on successful experience with honeycomb sandwich mirrors in the MMT, Magellan telescopes and LBT. The GMT design combines the smoothness and stability of large honeycomb sandwich segments with low-noise adaptive secondary segments, to provide a dramatic advance in sensitivity and resolution.

This material is based in part on work supported by AURA through the National Science Foundation under Scientific Program Order No. 10, as issued for support of the Giant Segmented Mirror Telescope for the United States Astronomical Community, in accordance with Proposal No. AST-0443999 submitted by AURA.