Winter 2007/
 Spring 2008
W. M. Keck Observatory 

 In this Issue:
 A Beast at the Heart of the
 Two Beams of Light
 Time on the Keck Telescopes
 Small Things that Count

By Linda Copman, based on an informal interview with Dr. Jerry Nelson
Images and photos courtesy of Dr. Jerry Nelson

Image: Schematic of the Keck I Telescope facility.
“I am endlessly curious and at the same time very skeptical. I tend to only believe something if I actually understand it at an intuitive level that satisfies me. My father taught me to question everything and to find and challenge unstated assumptions. As a scientist, these traits tend to make me very careful in what I do. They also make me look for new solutions to problems.” — Dr. Jerry Nelson
Jerry Nelson was the first kid in his rural California hometown of Kagel Canyon to go to college. His father was a tool planner at Lockheed, whose job was to figure out how to manufacture parts and then record this process in the form of a blueprint. Nelson’s father worked a lot with metals, massaging them thermally in particular ways in order to retain their malleability. “I subliminally picked this up,” says Nelson, “and I have always been reasonably grounded in the real world of manufacturing processes and in the properties of materials.”

At Caltech, Nelson worked part-time in the machine shop, where he learned more about what could and could not be done with various machines. His skills in building apparatus were significantly strengthened when, while still an undergraduate, he helped to design and build a 1.5-meter telescope to conduct an all-sky survey to find the brightest infrared sources in the sky.

Image: The 200-inch telescope at Palomar Observatory in California, built in 1948.
In mid-1977, after closely studying the designs of most existing large telescopes, Nelson began to appreciate how difficult it would be to make a very large telescope in the same fashion. The Russians were building a 6-meter telescope and people at UC (University of California) were talking about a 7-meter telescope. Casting aside conventional technological approaches, Nelson suggested that a segmented mirror approach was the only sensible approach for increasing the size of the primary mirror. With this approach UC could make a truly significant step, by doubling the diameter of the 200” telescope at Palomar to 10 meters.

Early on, Nelson started working closely with Terry Mast of Lawrence Berkeley National Laboratory (LBNL). Mast lived next door to Nelson in the dorm when they were undergraduates at Caltech, and both were confirmed experimentalists, or physicists who prefer to devise experiments to measure physical properties - as opposed to theoreticians, who prefer to write equations to describe physical properties. Nelson and Mast had been good friends since the age of 18 and they were used to brainstorming solutions to fundamental problems together. “The two of us spent essentially all of our time thinking about the challenges of making a very large telescope,” says Nelson. Nelson and Mast’s philosophical decision to pursue a segmented design was a step that led to the next generation of large telescopes. The two realized that there were theoretical limits to making large monolithic mirrors, which they referred to as the “last of the dinosaurs,” and that segmented mirrors would become the basis for bigger, better, and more economical telescopes of the future.

Image: Nelson and Mast made a key philosophical decision to pursue a design path that would lead to an entirely new generation of telescopes, not just a single giant telescope.
Traditional telescopes utilize large primary mirrors which are bigger, heavier, more expensive, more fragile, harder to support properly against gravity, and harder to coat since they require a big vacuum chamber in which to do this job. Segmented mirrors allow the optical element size to be limited to the segment size, which makes most of these problems become no harder than the optics for a 1.8m diameter telescope. With segments, one can imagine building telescope optics with basically no size limit.

Segmenting surfaces is a very old idea, and radio telescopes are frequently constructed of panels or segments. Early on (1977-1979) there was a lot of skepticism about the segmented design approach, and Nelson had to convince skeptics both within the project and people in the astronomical community at large of the viability of his idea. “Please do your best to prove me wrong” was his motto during this time. An alternate design to build a 10-meter monolithic mirror was proposed, and Nelson had to find a way to convince the UC leadership that this was the wrong way to proceed.

To convince his colleagues and the UC leadership to pursue a segmented design, Nelson and his team developed and built working prototypes of the key components of their design. The team developed algorithms for the active control system. They also designed and built everything from edge sensors (glass pieces that sense where the segment is with respect to its neighboring segments), to actuators (the motorized screws on the back of each mirror segment that cause them to move), to the passive support apparatus for supporting segments against gravity. They even polished a subscale off-axis mirror utilizing an entirely new polishing technique. This polishing technique involved extensive mathematics derived from the theory of elasticity, as well as very thoughtful apparatus. The polishing technique proved successful, and, together, these efforts were sufficient to persuade UC administrators that the segmented concept worked.

Some unique issues posed by a segmented mirror design were the need to actively control the mirror positions in real time, as the telescope temperature and zenith angle changed. Telescope mirrors must be accurate to within a small fraction of the wavelength of light, and they must be rigid enough not to deform as a result of gravity, wind, or temperature variations from night to night. To give some perspective, human hair has a typical diameter of 50 microns or about 100 times the wavelength of visible light. Telescope optics must be no worse than about 10% of the wavelength of light, or accurate to within 1/1000 of the diameter of a human hair. There was also the extremely challenging problem of polishing the optical surfaces of the off-axis segments which would be needed to form a gigantic hyperboloid-shaped primary mirror. These challenges were recognized fairly early on in the design process, but Nelson’s team spent several years developing solutions.

Photo: Nelson poses with one of the Keck I Telescope mirror segments during construction of the telescope.
By 1979 they had selected the actual pattern of segmentation that appeared best, consisting of 36 hexagonal segments. They considered a petal design, but this resulted in a less efficient use of materials than hexagonal segments, which have three-fold symmetry. Hexagonal segments wasted the least amount of very expensive materials required to make the mirror. The design required 6 different kinds of segments to fit together to form a hyperboloid. When manufacturing the segments for the Keck I Telescope, they ordered 7 of each kind of segment, so that there is one spare segment of each kind. The optics in the Keck II Telescope are identical to those in Keck I, so that the mirror segments in the two telescopes are interchangeable. “This is a real boon to the functional economy of the two telescopes,” explains Nelson.

Nelson’s team then engaged in building and testing prototype components for the active control of the segment positions. Coming up with a method of controlling the mirror segments so they could work as though they were a single mirror was perhaps the biggest overall challenge they faced. The solution involved utilizing a mixture of geometry, suitable control approach, precision actuators, and edge sensors. All of this was groundbreaking work and required ideas, development, exploration of various options, iteration on designs, and prototypes and revised prototypes. The end products were an unprecedented success: the Keck Telescope primary mirrors act like a continuous surface, and the gaps between segments lose less than 1% of the total collecting area and image quality.
“A very straightforward mathematical algorithm is used to operate the mirror at Keck Observatory. There are 168 edge sensors, 168 linear equations to solve, and 108 actuators to control. Computers do the math and the system works just like it’s supposed to. We used principles from high school mathematics to figure this out.”  — Dr. Jerry Nelson

Image: Diagram of the whiffletree support system for the mirror segments, with 36 points of attachment and multiple pivot points.
The team also had to develop and test a unique method for polishing the segments, called stressed mirror polishing. This technique applies external forces to the mirror blank to deform it into a spherical shape. “We figured out how to push the blank in just the right places to make it into a sphere,” explains Nelson. It is much easier to polish spheres than hexagons, so the circular segments were first polished and then cut into hexagons. The final step in the process involved ion figuring, or bombarding the mirror surface with high-energy ions to remove any remaining inaccuracies — one atom at a time. They had to convince optical companies to use the stressed mirror polishing method and then help these companies to properly apply the method. This process was expensive, very exacting, and fraught with technical and interpersonal challenges.

Concurrently, they were faced with the task of developing and prototyping an economical support system to hold the segments stable against gravity. This system, called a whiffletree, was loosely based on a whipple-tree horse harness system, which distributes the weight evenly across a team of horses and increases maneuverability. Similarly, the Keck Telescope’s whiffletree support system is designed to distribute the weight of the mirrors across a series of points, so that each point pushes on the mirror segments properly to minimized the segment deformations. The whiffletrees work on the principle of a see-saw, rotating freely about a pivot point. This allows the ratios of the loads at the support points to be maintained, even when the total load varies. The distributed load of the whiffletrees is concentrated down to 3 points of support for each segment. This is the number required for stability. “Consider how much more stable a 3-legged table is than a 4-legged one,” suggests Nelson.
“Jerry Nelson had been thinking about the segmented mirror concept for a while when he started on this project. He always relied on what he called ‘first principles,’ or the basics of freshman-level physics, to test his design ideas. He recruited a couple of hundred very smart people to assist him. We needed to use computer modeling to simulate and refine our designs. The fact is that the Keck Telescopes really could not have been built before they were, simply because computer technology was not advanced enough to do the kind of modeling we needed until that time.”  — Barbara Schaefer, Observing Support Coordinator at Keck Observatory

Image: Installation of one segment of the Keck I Telescope’s primary mirror shows the complex underbelly required to actively control and support each segment.
To perfect the segmented mirror design, Nelson’s team drew upon their expertise in many different scientific fields, including structural engineering, mechanics, theory of elasticity, electrical engineering, control theory, lots of applied mathematics and matrix theory, physical optics, and physics at every turn. “Of course we had lots of problems and surprises,” recalls Nelson. “We were paving new ground so it was essential that we had a very deep and fundamental understanding of our design. This mastery of the underlying principles allowed us to efficiently develop the design and the hardware, and when there were surprises, to solve them,” he says.

Nelson and Mast worked with Gary Chanan of UC Irvine and with Jacob Lubliner of UC Berkeley. Lubliner helped to perfect the stressed mirror polishing technique which was used to create off-axis parabolic segments. Steve Medwadowski, an engineering professor at UC Berkeley, was responsible for the structural design of the telescope. The segment passive support system was aided by the contributions of Bob Weitzman of the UC space sciences lab. George Gabor and Bob Minor (both of LBNL) played key roles in the development of the segment position actuators, the edge sensors, and for many other mechanical issues.

Image: An alignment camera is used to adjust the segments properly. Gary Chanan was largely responsible for development of the Keck Observatory alignment camera.
“Ideas were central to the development of the Keck Telescopes. Our ideas were new and exciting. We did lots of prototyping to successfully show that our ideas worked — and that we actually knew how to build the unusual and hard parts to make the telescopes practical and affordable.”  — Dr. Jerry Nelson

The strong support of the leadership at LBNL and at UC was essential in providing early funding needed for development of Nelson’s ideas. “Thanks to these institutions, we actually had all the money we asked for, so work progressed limited only by our ability to recognize and solve technical problems,” says Nelson. They could not have succeeded without the support and enthusiasm of Caltech, which convinced Howard Keck, a Trustee, to support the project.

They also had the strong support of UC and Caltech astronomers, who were responsible for the construction of the superb science instruments at Keck Observatory that would consistently deliver outstanding science. Nelson also relied on an excellent team of engineers at Keck Observatory. “These engineers were dedicated and inspired by building the world’s largest telescope, and this was and still is a great motivating force to get the best out of everyone,” explains Nelson.

Photo: The finished primary mirror on the Keck I Telescope.

Today Nelson continues to lend his expertise to Keck Observatory. “The first thing I do when I come to work in the morning is read the night logs from Keck,” says Nelson. “If a problem arises during the night, I help to identify solutions which I share with the staff at the Observatory.” Nelson is the first to admit that he does this out of love for Keck Observatory, “his baby.”

Image of the Egg Nebula taken with Keck Observatory’s AO system.
Nelson is now working on a 30-meter telescope (TMT) which will become the world’s largest when it is finished in 2016 (he hopes). The Europeans, not wanting to be outdone as they were with Keck Observatory, are now planning to build a 42-meter telescope with twice the area of TMT, and they are only a couple of years behind the TMT team. “Building a telescope like this is loads of fun, and loads of staying up late at night,” chuckles Nelson.

Beyond this size, Nelson thinks it will be some time before even larger telescopes are built. Several reasons suggest this: cost is critical of course; wind shake appears to be a significant problem that will not go away with larger telescopes, particularly spatially variable wind forces on the primary mirror; finally, to achieve the full value of giant telescopes one wants effective adaptive optics and this is harder and harder with larger telescopes.

Keck Observatory is well positioned to continue to lead the field with its innovative instrumentation and adaptive optics (AO) systems. “Improvements in performance are manageable at Keck with today’s and tomorrow’s technology,” says Nelson. Such improvements will increase the Observatory’s scientific productivity for years to come. Nelson sees Keck as being able to lead in AO, progressing to shorter wavelengths — opening the door to a whole new universe of research.

Image of the galactic center taken with Keck Observatory’s AO system.
“The Keck Telescopes are the most powerful and productive astronomical instruments on Earth, and we are the envy of the world’s astronomical community. AO is clearly the arena where further development is likely to produce the greatest future discoveries. This is because the improvements in angular resolution allowed by AO will give us much greater sensitivity as well as clarity, and as Keck uses AO at shorter wavelengths, there will be many great astronomical discoveries. The most distant galaxies, formed at the earliest times in the universe, should be very interesting targets for Keck Observatory’s enhanced AO system.”
 — Dr. Jerry Nelson 

Click here to return to the main page.