The null corrector itself can be difficult to make and measure. The poster child for this difficulty is the Hubble Space Telescope¡¯s primary mirror, which was polished to match the wrong template to exquisite accuracy. Since then, the astronomical optics community has been hyper-sensitive about the need to validate the null corrector. My colleague Jim Burge at the University of Arizona¡¯s College of Optical Sciences has developed a technique in which computer-generated holograms (CGHs) are used as inverse null correctors, effectively mimicking an ideal primary mirror. These guarantee accurate measurements of the Arizona mirrors, including the most aspheric large mirrors made to date, the LBT primary mirrors.

Segments for primary mirrors and adaptive secondary mirrors

While most of the current 8-m-class telescopes use monolithic primary mirrors, no one is thinking about making a monolithic mirror for the next generation of 25- to 40-m telescopes. The new telescopes are segmented, and the only question is what size the segments should be. The primary mirrors for the Thirty Meter Telescope and ESO¡¯s Extremely Large Telescope will use 500 to 1,000 segments of about 1.4 m diameter.

The GMT, on the other hand, will use the largest segments that can be made, which are 8.4-m honeycomb sandwich mirrors similar to the LBT primary mirrors. They guarantee a smooth wavefront over 8.4-m subapertures. The telescope¡¯s 3.2-m secondary mirror is segmented to match the primary mirror, and alignment is controlled with the seven small, agile secondary segments.

The GMT secondary segments are particularly agile because they¡¯re also the deformable mirrors of the adaptive-optics system that will correct wavefront distortions caused by the atmosphere. What appears as twinkling (intensity fluctuations) to the eye shows up as phase variations across a large telescope aperture. A large telescope in space could form images with diffraction-limited angular resolution of about ¦Ë/D, or 4 milli-arcseconds for a 25-m telescope at 0.5-µm wavelength.

For a ground-based telescope, however, the resolution-limiting diameter is that over which the wavefront is flat within about a wavelength¡ªtypically 15 cm at 0.5 µm at a good mountain site. This limits the resolution to 0.7 arc-second¡ªmore than 100 times worse than the potential resolution.

To get around this severe limitation, astronomers have developed remarkable adaptive optics systems that sense and correct for the atmosphere¡¯s effects on the wavefront. Their efforts got a boost when a good deal of military research in this area was declassified in the 1990s. The basic principle of correction is simple: Measure the distorted wavefront and bend the opposite error into a deformable mirror somewhere in the optical system.

The hard part is that the atmosphere is constantly changing, so correction requires a new measurement and a new mirror shape roughly every millisecond for visible wavelengths. The update requirement is relaxed to every few milliseconds for the more-forgiving infrared wavelengths. Most large telescopes have some form of adaptive optics, and most are working in the infrared today. As faster actuators, wavefront sensors and processors are developed, astronomers are gradually pushing the techniques to visible wavelengths.

The adaptive optics system of the GMT builds on the ones developed for the 6.5-m MMT and the LBT, with the secondary mirror segments serving as deformable mirrors. Each 1.1-m diameter segment will be 2 mm thick and supported by about 1,000 voice-coil actuators. This system adds no additional reflections beyond the two that occur in any telescope, an important advantage because every reflecting surface adds thermal noise to the signal, especially at longer infrared wavelengths. The adaptive secondary mirror will help enable the search for warm, young exoplanets in the infrared.

In fact, we expect to be able to see in the infrared even mature Earth-like planets around the nearest stars (if there are any). This detection would go right to the limit of capability. It requires both the resolving power and the infrared sensitivity of a 25-m aperture with an adaptive secondary mirror. For the GMT, the adaptive secondary has the additional advantage of keeping the seven intertwined telescopes aligned and in phase, even if the large primary segments have relative motions of many microns.

Building up to the Giant Magellan Telescope

Not only the adaptive optics system but the entire GMT design follows naturally from the progression of larger telescopes using honeycomb sandwich mirrors. The collecting area increased first by using the largest mirrors possible, and then by combining multiple mirrors for even more powerful systems. The basic mirror design hasn¡¯t changed and comes with mature active support and thermal control systems. These mirrors have produced some of the best images ever obtained without adaptive correction. The figure to the left shows a couple of examples from the 6.5-m Magellan telescopes. The gravitationally lensed quasar represents a different kind of ¡°astronomical optics.¡± The image of the colliding Antennae Galaxies is extraordinary for its 0.27-arcsecond resolution over a nearly 2-arcminute field of view.

The GMT achieves 3.5 times the collecting area of the LBT with only a small increase in overall telescope dimensions. The major novelty of the GMT is that the seven mirrors are segments of a single 25-m f/0.7 near-paraboloid instead of separate 8.4 m paraboloids like the LBT. The project team identified the fabrication and testing of the off-axis segments as important enabling technologies. Thus, they initiated the manufacturing process while detailed design of the telescope was going on in parallel.

The first GMT primary segment was cast in the Mirror Lab¡¯s spinning furnace in July 2005. Twenty tons of E6 borosilicate glass from Ohara melted over the ceramic fiber mold at a temperature of 1,200¡ã C, with a viscosity similar to room-temperature honey. (See a movie of the glass melting.) About 1,700 hexagonal boxes formed the cavities in the honeycomb sandwich, leaving a segment that is a single piece of glass but lightweighted by a factor of 5. After three months of slow cooling, the segment was lifted off the furnace hearth by gluing a large steel frame to its top surface. It was turned into a vertical plane to give technicians access to the partially open back plate of the honeycomb structure. This allowed them to wash out the ceramic fiber boxes trapped inside the segment.

The flat rear surface was then ground and polished so that load-spreaders could be bonded to it. These form the interface between the mirror segment and its 160-actuator active support system. The segment was then turned right-side-up and mounted on a polishing support that mimics the telescope support but with passive hydraulic cylinders. The spin-casting produced an axisymmetric parabolic surface, so the aspheric shape of the off-axis paraboloid was created by diamond machining excess glass from the surface, removing an additional 14 mm along the ¡°radial¡± diameter that lines up with the telescope¡¯s optical axis.