In August 2006, the Steward Observatory Mirror Lab cast a 3.75-m mirror under the stands of the University of Arizona football stadium. Twenty years earlier, this mirror could have become the primary mirror for the sixth largest optical telescope in the world. Today, it¡¯s a piece of the test optic system that is guiding the manufacture of the 25-m primary mirror for the Giant Magellan Telescope (GMT).

The GMT is part of a wave of new, ever-larger telescopes that first came on the astronomical scene in the early 1990s. Astronomers will use them to study distant planets, stars, galaxies and black holes. In some cases, their goal is to understand the structure and evolution of the objects themselves. In others, they will study objects to reveal the fundamental structure and evolution of the universe as a whole. A large telescope can capture supernovae (catastrophic explosions of stars that have run out of nuclear fuel) at such great distances that scientists can use the images to trace the expansion and acceleration of the universe. And slight distortions in the images of distant galaxies can map the distribution of the invisible dark matter that appears to make up over 80 percent of the universe¡¯s mass.

With their unprecedented sensitivity and angular resolution, these telescopes open new windows onto the universe. Sensitivity scales with area, and resolving power scales with diameter if a coherent wavefront can be maintained. One quest that demands all of the sensitivity and resolution that can be squeezed out of a telescope is the direct imaging of planets around other stars. Several hundred extrasolar planets have been detected by their star¡¯s tiny oscillation around the common center of mass, or the slight darkening that appears when they pass in front of the star.

Direct imaging is much more difficult because the planet is so close to the billion-times-brighter star. Just last year, astronomers reported the first images of several large planets orbiting other stars¡ªplanets several times larger than Jupiter with orbits similar to those of Uranus, Neptune and Pluto. The new generation of telescopes, including the GMT, will be able to image mature planets down to about Jupiter¡¯s size, with orbits as small as Earth¡¯s, as well as smaller Earth-like planets that are young enough to glow in the infrared as their gravitational energy leaks out. Direct imaging will one day lead to spectroscopy and the ability to detect oxygen in a planet¡¯s atmosphere, the signature of life.

A brief history of telescopes

A previous burst of telescope development occurred during the early 20^th century, when George Ellery Hale led efforts that culminated in the 60-in. and 100-in. telescopes on Mt. Wilson and the 200-in. Hale Telescope at Palomar. Following the construction of the 200-in. device in the 1930s and its commissioning in 1948 after the war, growth in telescopes was put on pause for almost half a century. Detectors improved and new wavebands opened up, but mass, flexure and thermal inertia proved serious obstacles to larger mirrors.

A mirror needs to be stiff enough to hold its shape against the wind and (at least in the mid-20^th-century paradigm) against its own weight. But that stiffness implies mass, which drives up the mass and cost of the whole telescope structure. It also makes the mirror a huge repository of thermal energy that can¡¯t be shed fast enough to follow changes in nighttime air temperature, and that causes the same kind of turbulence and image blurring we see when looking across a hot pavement.

The mirror challenge was finally overcome in the late 1970s and 1980s by three groups in three different ways. Jerry Nelson, then at the University of California at Berkeley, led the development of segmented mirrors, in which a large primary mirror is synthesized by dozens or even hundreds of small segments. The difficulties in making and controlling the large mirror are traded for the challenge of keeping the segments aligned to a fraction of a wavelength.

This concept has worked superbly in the twin 10-m Keck telescopes, and it is used for several other 10-m-class telescopes in use and under construction. Looking to the next generation of telescopes, two telescopes have been designed with one-meter-class segments¡ªthe Thirty Meter Telescope, which is being developed by the University of California, Caltech and partners, and the 42-m European Extremely Large Telescope.

Another solution to the problem of massive mirrors came from Ray Wilson and colleagues at the European Southern Observatory (¡°southern¡± because their telescopes are in Chile), who realized they could turn the flexibility of a thin mirror into an advantage by supporting it on an active system of computer-controlled actuators. This concept, in the form of mirrors 8 m in diameter and 175-200 mm thick, is the basis for ESO¡¯s Very Large Telescope (four telescopes), the two Gemini Telescopes and the Japanese Subaru Telescope.

Today, all large mirrors use Wilson¡¯s active optics concept. They¡¯re supported by 100-200 actuators and actively controlled with feedback from sensors that measure the shape of the reflected wavefront. This is a slow correction¡ªbecause of the mirror¡¯s huge inertia and the need to average atmospheric fluctuations out of the wavefront measurement¡ªbut it can have an amplitude of several microns.

The third solution, now the basis for the GMT, was Roger Angel¡¯s development of honeycomb sandwich mirrors at the University of Arizona. The mirrors¡¯ structure, formed by melting the glass in a complex mold, is the 2D version of an I-beam, so these mirrors are about eight times stiffer than solid mirrors with comparable mass. They bend less due to weight and wind, and their short thermal time constant lets them follow the changing air temperature at the telescope. In order to reduce telescope dimensions, the Arizona mirrors have shorter focal lengths than other large mirrors. The Mirror Lab uses a dramatic spin-casting process to produce the honeycomb structure and the deep curvature, then refines the surface by machining and polishing. The figure above shows the GMT mirror right after its casting.

Honeycomb sandwich mirrors

Beginning in 1983, the Mirror Lab cast and polished a number of mirrors of 1.8 m diameter¡ªthen 3.5 m, 6.5 m and finally 8.4 m. The 6.5-m mirrors are in the MMT telescope in Arizona and the twin Magellan telescopes at Las Campanas Observatory in Chile. The Large Binocular Telescope on Mt. Graham in Arizona is the world¡¯s largest with two 8.4-m primary mirrors on a common mount. The short focal lengths of these mirrors (the LBT mirrors are f/1.1) forced the lab to develop technology that allows the efficient manufacture of highly aspheric mirrors. After spin-casting, the next development was an active polishing disk that changes its shape continuously to match the local curvature as it moves across the surface.

Optical testing is also an interesting problem for very aspheric mirrors. The standard technique is phase-measuring interferometry, in which the full mirror surface is illuminated with a coherent beam, and the reflected wavefront interferes with an accurate reference wavefront, ultimately giving a contour map of the surface with a resolution of about ¦Ë/100. The challenge is that the illuminating wavefront must match the desired mirror surface; this wavefront is the template that the surface is compared with. Any error in the template causes an error in the mirror shape. The interferometer¡¯s illuminating wavefront is typically spherical, and a set of optics known as a null corrector transforms it into a template wavefront of the right shape.