The 10 meter, 36 segment Keck mirror
The Keck Telescope in Hawaii, a joint project of Caltech and the University of California, is - at ten meters in diameter - the largest optical and infrared telescope in the world. Its great aperture - four times the area of the 200 inch Hale Telescope on Mount Palomar - is made possible by a unique segmented design, developed by Jerry Nelson (now at the University of California, Santa Cruz): the ten meter mirror actually consists of 36 hexagonal segments, each 0.9 meters on a side, which fit together to form a virtually continuous piece of glass.The alignment of the segments "to form a virtually continuous piece of glass" is the responsibility of the wavefront-sensing group in the UCI Department of Physics and Astronomy, and is carried out with a device called an "alignment camera," built at UCI and delivered to the Keck Observatory in 1992. Alignment of the segments in the two "tip and tilt" degrees of freedom (rotation about axes in the plane of the segment) is a relatively routine matter - even at the required accuracy of a few hundredths of a second of arc, and was accomplished literally on the first night the alignment camera was turned on. Alignment in the piston or phase degree of freedom (motion perpendicular to the surface of the segment) is much more difficult because standard geometrical optics techniques are generally insensitive to phase, and has taken a great deal longer to perfect.
How well does the telescope need to be phased; that is, how small do the steps between mirror segments have to be in order for the telescope to achieve its maximum performance? At optical wave- lengths it actually makes little difference, because the angular resolution is limited to about 0.5 arcseconds by the blurring effects of the earth's atmosphere. But at near infrared wave- lengths - 2 microns or so - the optical consequences of atmospheric turbulence are much reduced and the familiar elementary considera- tions apply. That is, the angular resolution of a telescope is about lambda/D, where lambda is the wavelength and D is the diameter of the mirror. But for a segmented telescope, is D the 1.8 m diameter of a segment, or the diameter of the full 10 meter primary mirror? The answer is that if the steps between segments are large compared to a wavelength, then we should use the segment diameter, but if the steps are small compared to a wavelength (in practice about lambda/20), then the relevant D is the full 10 meters. Thus the Keck Telescope needs to be phased to about 1/20 of an infrared wavelength, or about 0.1 microns. Clearly it makes a big difference - more than a factor of 5 in angular resolution - whether or not the telescope is phased.
Since geometrical optical techniques are insensitive to phase, our phasing technique must exploit wave optical techniques such as interference or diffraction. The exact approach we use is a variation of Young's familiar two slit experiment. In this case the two "slits" are the two halves of a small circular aperture which straddles two adjacent segments. If the segments are in phase, we obtain a circular diffraction pattern, but if there is a phase difference between the two segments, the pattern is more complicated. The Figure shows actual diffraction images obtained at Keck using starlight and a charge-coupled device (CCD) detector. In the upper left panel the segments are in phase, and in each successive panel the step between segments has been increased by an additional 40 nm, or lambda/22 where the wavelength lambda is 891 nm. Note that in general the simple in-phase pattern splits and forms two separate peaks whose relative intensity varies with the phase difference. The seventh picture (not shown) is the mirror image of the fifth; the eighth is a mirror image of the fourth, and so on. After 11 steps, the segments are half a wave apart, so the optical path difference, which doubles on reflection, is a full wave, and the pattern repeats. The practiced eye can easily estimate the phase to the nearest picture, well below the required 100 nm tolerances. (Of course the computer can do this correlation much faster and better than the practiced eye.)
The alignment camera contains a "fly's eye" lens - actually an array of tiny (2 mm by 3 mm) prisms, which enables us to observe the diffraction patterns corresponding to all 84 intersegment edges simultaneously, so that all the segments can be phased in parallel. In practice the phasing measurements must be repeated in a second filter corresponding to a second, incom- mensurate wavelength, so that we can tell the difference between being in phase and being out a phase by an integer number of waves.
The phasing procedure takes about an hour from start to finish, and the telescope is stable enough that the procedure only needs to be repeated every few weeks. This fall we are scheduled to deliver a second camera to be used to align and phase the Keck II telescope - an identical twin of the first, now under construction 70 meters away from Keck I at the opposite end of the Keck Observatory building. Eventually the two Keck Telescopes will be linked together interferometrically to form an instrument of unprecedented power and resolution.