The Mirror Makers

In 1968, a stark cement building slowly took shape on the grounds of an optical manufacturing company called Perkin-Elmer, which sits on a hill in Danbury, Connecticut. The cavernous building was made especially for secret military programs, including the construction of cameras for KH-series spy satellites. Such work capitalized on the company’s expertise—grinding and polishing disks of glass or metal into optical mirrors.

Veteran optical physicist Terence Facey leans back in his chair and declines to elaborate on what was done in the facility from 1968 to 1975, before the company turned its skills toward manufacturing the mirrors for the Hubble Space Telescope.

“It was classified work. Period,” he says politely.

Facey is a London-educated physicist, a thin, white-haired man whose manner and accent give him an air of timeless wisdom. He could easily be cast in a science fiction film as a sage.

Facey has worked at this facility for 34 years and has watched its name change from Perkin-Elmer to Hughes Danbury Optical Systems to Raytheon Optical Systems to, this year, Goodrich Optical and Space Systems Division. The name has changed so many times that those in the optics branch simply refer to the factory as “Danbury.”

Major government-funded optical programs do not come along very often, whether they are top-secret programs or purely scientific efforts. That is why Facey is sitting in a conference room now, talking about why Danbury should be the mirror maker for the        $1 billion Next Generation Space Telescope. When the NGST joins Hubble in space, it will have the largest mirror ever deployed, at least as far as non-classified vehicles are concerned.

The telescope will be stationed not in low Earth orbit like Hubble but a million miles from Earth, at a so-called Lagrange point, where the pull of the Earth and that of the sun are in equilibrium. Here, after a 104-day trip to Lagrange point 2, a mirror six to seven meters (about 20 to 23 feet) across will catch faint infrared light waves emitted by clouds of gases during the first billion years after the Big Bang.

“What did galaxies look like before they were born? This telescope will tell us that, and that’s really exciting,” says astronomer Alan Dressler of the Carnegie Observatories, an organization that operates ground-based telescopes in Chile and California. Unlike many of Hubble’s pictures of relatively nearby galaxies and stars, he says, “the images have the potential to be visually stunning and things we’ve never seen.”

John C. Mather, an astrophysicist at NASA’s Goddard Space Flight Center in Maryland, says that between measurements of cosmic background radiation and observations of young galaxies lies a cosmic “dark ages”—a period in the formation of the universe as yet unseen. Mather, who used the Cosmic Background Explorer to confirm the existence of background radiation left over from the Big Bang, looks to NGST to provide insight into the period when the universe was between about one million and a few billion years old—when galaxies and stars began forming.

Last June, NASA released the ground rules for the competition to build the NGST spacecraft and instruments. The company that wins the competition will then decide who will build the telescope’s mirror: Ball Aerospace in Boulder, Colorado, Goodrich in Danbury, Connecticut, or Eastman Kodak in Rochester, New York.

The three companies have been experimenting with different mirror materials through a $20 million project called the Advanced Mirror System Demonstrator (AMSD), which is funded jointly by NASA, the Air Force, and the National Reconnaissance Office, which operates spy satellites. The three companies each received about $3 million to build competing versions of lightweight mirrors and test them in vacuum chambers.

Each contractor will have to show that mechanical actuators on the back of the mirror can adjust the instrument’s shape, or “figure.” Phasing experiments, in which the contractors will prove that different segments can work together without losing focus, will come later.

For the contractors, the creation of the Hubble’s primary mirror provides a benchmark and, of course, a cautionary tale. It was here at Danbury, back in 1980, that Hubble’s primary mirror was ground and polished very precisely to the wrong specifications. The mirror’s curve was off by just one-fiftieth of the width of a human hair, but that was enough to spread Hubble’s light across multiple focal points instead of one. Here is the plot twist: The engineers and physicists at Ball, Danbury’s rival in Colorado, are the same happy few who stepped in to build the optics that corrected the images bouncing off Hubble’s flawed mirror.

Danbury officials deny that the real reason they are chasing the NGST contract is to redeem the company’s reputation. Facey insists the Hubble error was an “anomaly” in a long list of groundbreaking company triumphs, which includes the new Chandra X-ray observatory and the soon-to-be-launched Space Infrared Telescope Facility. Chandra is giving astronomers unprecedented information about black holes, and SIRTF will obtain images of not only the early universe but also planets in our solar system and the cosmic dust and gas surrounding nearby stars.

Ball, Danbury, and Kodak have teamed up with California-based aerospace giants that know how to integrate sensitive optical equipment and powerful satellite frames. Danbury is working with Lockheed Martin of Sunnyvale, and Ball and Kodak are joined up with TRW of Redondo Beach. But the prime NGST contractor must select the best mirror, regardless of any teaming arrangements—otherwise, NASA won’t approve the design.

Out in Boulder, in the same low-slung buildings at the foot of the Rocky Mountains where Hubble’s corrective optics were built, members of the Ball team make it clear that the competition is a fierce one.

Unlike Danbury, Ball—the same company that made your grandmother’s canning jars—is not known as a maker of large optical systems like Hubble’s primary mirror, let alone one that would be 10 times bigger. Its niche has been the design and manufacture of small scientific instruments. It made most of the instruments and cameras that sit behind Hubble’s main mirror, converting raw light from the mirror into pictures of the heavens. Because of schedule delays, Ball got the nod to build the SIRTF telescope and instruments through subcontractors after NASA halted Danbury’s work on that telescope’s 4.9-meter mirror.

NASA officials want the NGST mirror to be six to seven meters wide. A mirror that size could not possibly be ground from a single piece of glass, the way Hubble’s 2.4-meter mirror was, and must be made from segments. In order to stay within the telescope’s 6,600-pound payload limit, these need to be seven to 10 times lighter per square meter of surface area than Hubble’s mirror. NASA wants the new mirror to weigh no more than 44 pounds per square meter.

The size of the mirror—more than double that of Hubble’s, larger than any conceived of for a spacecraft mirror—started as a bold challenge from NASA Administrator Dan Goldin. In January 1996, Goldin stood up in front of the American Astronomical Society in San Antonio, Texas, and flabbergasted the audience by calling for the Next Generation Space Telescope to be built with a mirror eight meters across.

Carnegie’s Alan Dressler remembers sitting in the audience. He had just finished leading a meeting of astronomers who hoped to whet NASA’s appetite for a successor to Hubble. The Dressler Committee report, “The Hubble Space Telescope & Beyond,” called for a telescope with a mirror at least four meters across, or two-thirds wider than Hubble’s—half the size of Goldin’s proposed mirror. Dressler figured that was about the right size to magnify infrared waves and capture images of the very early universe. A bigger telescope would improve the resolution of the images somewhat, but the main goal was to get a mirror to the Lagrange point, far from Earth’s heat, where the infrared waves would stand out against the cold background of space, he says.

To reach space, the mirror will have to be sent atop an unmanned rocket, folded up like the leaves of a table or the petals of a flower. It must unfold in space without jamming. Then it has to hold its shape for 10 years in temperatures close to –457 degrees Fahrenheit, the point at which matter no longer has any thermal energy.

Engineers aren’t entirely sure what the extreme cold will do to the segments, and flaws invisible to the naked eye would blur the telescope’s images. The mirror will be supported by numerous tiny, electrically controlled mechanical arms, called actuators, which will, in a technique called adaptive optics, nudge each segment into place after deployment. Then, perhaps once a month, NGST will look out to a reference star. If the image is blurry, ground controllers will send commands up to the telescope to nudge the mirror segments back into place.

The project is so daunting that the potential contractors are relieved that NASA has reined in the program slightly. Earlier this year, agency officials signaled their willingness to settle for a telescope of six to seven meters instead of eight.

On the other hand, some astronomers are convinced that bigger is better, and they completely support Goldin’s vision of an eight-meter NGST mirror.

“In the early universe, there was a time when the first stars and star clusters lit up and illuminated the world, so to speak,” says astronomer Simon Lilly of the Herzberg Institute of Astrophysics in Victoria, British Columbia. He worries that a smaller mirror will produce degraded image quality and necessitate additional observation time. “We tried to design NGST so that observing that epoch is within its grasp,” he says. “For that, we need every little bit of sensitivity we can get. The bigger the better.”

Mather agrees there are significant advantages to a larger mirror. “Since the aim is to see the first light from the first objects that formed after the Big Bang, we don’t want to spend a year taking data with a four-meter telescope that would have taken less than a month with the eight-meter,” he says. “This is such a big difference that we don’t think the four-meter telescope will be able to see what we want to see…. We are pretty sure that a four-meter telescope would take 16 times as long to collect the light from a primordial galaxy as the original eight-meter concept would have done…. We can partially compensate for having a smaller telescope by getting more and better detectors.”

However, Dressler thinks a mirror with a minimum diameter of four meters would do just fine. He is nervous that NASA might be pushing the technology too far, which could leave astronomers with a flawed telescope or one that devours NASA’s astronomy budget as agency contractors try to reconcile the various elements required of the NGST mirror.

In Rochester, New York, another traditional American company with a homespun reputation is hoping to solve the puzzle. Kodak is the “dark horse” in the race, says Bernie Seery, NASA’s NGST program manager. “But the company is aggressively engaged in the selection process and is a formidable competitor.” Although Kodak is on the same team as Ball and will share a portion of the work if TRW is selected to build NGST, Kodak is still in competition with Ball to provide the mirror. Kodak engineers are working with a familiar concoction of melted silica used in the company’s cameras since the first Brownie. “Glass has been a traditional mirror material, so a lot is known about its manufacture,” says Jeffrey Wynn, general manager of space science in Kodak’s commercial and government systems division.

Kodak’s engineers hope their method of fashioning the mirror will result in one that is lighter and less expensive than a conventional glass mirror. “Ours is a little bit different,” says Wynn. “It’s a semi-rigid design constructed of a glass core section that looks like a honeycomb.” Glass is fused into a top and bottom plate while the mirror blank is still flat. “Then we slump it over a mandrel [a metal mold] and shape it.”

Such an approach eliminates much of the labor-intensive manufacturing processes usually required in building glass mirrors. “It’s much easier to work on mirrors in the flat than in steep curvatures,” says Wynn. “That’s what takes a lot of time—the steep curvatures in large parabolic mirrors have to be perfect, and the most laborious parts in generating those curves are the machining operations and the polishing.”

Kodak’s AMSD experience produced a mirror in the ultralight range of less than 33 pounds per square meter, far lighter than Hubble’s massive 330-pound-per-square-meter mirror. Because of the size needed for the NGST mirror, Wynn hopes Kodak’s manufacturing techniques will prove that glass is the right choice.

But besides the choice of materials for the mirror, nearly every aspect of the telescope will prove difficult to create. “You’ve got to push people to do the very best they can, and at some point you’ve got to say ‘Okay…now I’ve got to really build it,’ ” Danbury’s Facey says. He notes that the same thing happened with the Hubble, which was originally slated to have a three-meter mirror.

Inside Danbury’s mirror manufacturing facility, Facey points to a giant 4.3-meter mirror that is resting horizontally on a metal support. He says it weighs 7,000 pounds and must be moved on a cushion of air.

NGST will be nothing like that. Under the current design, Danbury would fabricate nine segments that would be less than 20 millimeters thick—about two-thirds of an inch. They will be, in a word, flimsy. “You breathe on this thing and it’s going to change,” Facey says.

Danbury’s design calls for a central mirror that would be shaped like an octagon. Keystone-shaped segments would extend out from each flat side of this central mirror to create one large surface. The keystones would be hinged so that one could be folded forward, the next aft, and so on, around the central mirror. That is how the engineers plan to squeeze it inside an unmanned rocket’s payload shroud. The tolerances will be tight—Danbury engineers have spent a considerable amount of time studying the lessons of the Hubble program, in which the error resulted from engineers relying on a single measuring tool, which turned out to be flawed.

And the complexities of the mirror itself are greater. In contrast to Hubble’s single-piece mirror, Ball’s initial blueprint calls for 36 segments. Each must stay perfectly matched to the surrounding segments or the mirror’s delicate focus will be lost. That job will be more complex with more segments, Facey says. “It’s not a non-trivial issue to poke those optics with a couple hundred actuators and still get something that looks good,” he observes.

It does not take long to realize that each company has a radically different opinion about how many segments will be needed to make up the NGST mirror. Danbury figures at most nine, including the central mirror; Ball figures up to 36; Kodak’s AMSD mirror, while not strictly a prototype of its NGST design, uses 19.

Ball engineers are eyeing beryllium, the same metal that Danbury had so much trouble using for SIRTF’s primary mirror. It is highly reflective and would need no special coating. Glass mirrors, on the other hand, are coated with gold to make them reflective.

Beryllium, unlike glass, is not available in large disks, so a large beryllium telescope would have to be assembled from lots of smaller pieces. Yet the Ball team remains tempted by beryllium because it is much stiffer than glass, says Doug Neam, a former college wrestler who is now Ball’s NGST program manager. “One of the fundamental building blocks of this program is how you will get a mirror of extremely lightweight design,” he says. “That’s what got us into looking at beryllium mirrors.” Neam gestures toward a model of Ball’s proposed NGST mirror. “If you look at the back of this mirror here you will see that the actual face sheet is just a couple of millimeters thick. And then the mirror itself is a couple of inches think. Call it 50 millimeters-ish,” he says. The model is not to scale, but his point is that a beryllium mirror would be much lighter than an equivalent glass mirror.

Despite all the potential advantages, Ball astronomer Douglas Ebbets cautions that Ball has not yet ruled out glass. “That final decision won’t be made for another year, pending the outcome of tests of mirror technology developments that are currently under way,” he says.

Engineers at Danbury say they know a thing or two about beryllium. In fact, they are openly resentful about the credit Ball engineers appear to be taking for SIRTF, which operates with a mirror fabricated in Danbury. “The bottom line is they’re flying the hardware that was made, in part, with these hands right here,” says Mark Stier, a blunt-speaking astronomer whose job, if Danbury wins the contract, would be to oversee the assembly of NGST’s mirror system. “There are only two places in the world where beryllium mirrors are made; this is one of them,” he says.

Stier is doubtful the metal is the first choice for the NGST mirror. The problem with beryllium is that its crystalline structure is not the same in all directions. “[The crystals] tend to expand more in one direction than another,” Stier says. “It’s not as homogenous as glass.” When beryllium hits the cold of space, it will shrink. “The mirrors would probably change their shape more when you cooled them than you would desire,” Stier says.

The Danbury team’s hard-won experience with SIRTF’s beryllium mirror has influenced the direction its NGST mirror is taking. “Did we have a hard time with [beryllium]? Yes. Are we concerned about beryllium? Yes. But that was not the determining factor in terms of our decision,” says Ira Schmidt, who managed the company’s Chandra program and is now in charge of the NGST work. “We know some of the idiosyncrasies of beryllium that maybe our competitors don’t know,” he adds.

But for both Danbury and Kodak, the more traditional material—which was used to fashion the Hubble’s mirror—will most likely be chosen. “We believe that glass works fairly well at cold temperatures,” says Kodak’s Jeffrey Wynn. “And programs like the space-based laser [prove] that glass stays in shape, depending on the application.”

And there is another huge advantage to glass. “Look at how many skyscrapers have large windows of glass,” Terence Facey says. “Glass is a well-understood material available in almost any size you want.” That is why the Danbury team believes that it can build the mirror from just nine segments, compared with the Ball team’s 36.

And Wynn thinks that a decision on the NGST mirror may not spell the end of the two options that aren’t selected as the final design. “Three different types of technology are still in development,” he says. “Danbury glass, Kodak glass, and Ball beryllium. The prime [contractor] and NASA may jointly make the selection, or they may carry two designs for a while as a backup. This is new technology no one has ever tried before.”

But in this field, today’s innovations quickly become tomorrow’s antiques. Even after the NGST is boosted to its lonely post at Lagrange point 2, both glass and beryllium may be passé. Seery and other NASA officials hope to start a program—the Large Telescope Systems Initiative—that would start looking at other lightweight materials for mirrors. But for now, either glass or beryllium, in a six- or seven-meter mirror, built either by old hands at mirror making or by a team new to the game, will offer scientists their farthest look yet back in time.