"In a Mechanism Far, Far Away," Feature Article, April 2004

in a mechanism
far, far away

The Webb telescope will go almost a million miles out to see how stars were born, so there's only one chance to make it right.

This article was prepared by staff writers in collaboration with outside contributors.

NASA has made many successful attempts to expand the envelope of mankind's knowledge of the universe and its origins. In doing so, it has also pushed the envelope of manufacturing. The latest case in point: the James Webb Space Telescope that is to take the place of the Hubble Space Telescope in 2011.

The Webb telescope is an orbiting infrared observatory, and the project is managed by the NASA Goddard Space Flight Center in Greenbelt, Md. It will study the previously unobserved epoch of galaxy formation, peering through space dust to witness the birth of stars and planets.

The Webb telescope may one day help us to understand how planetary systems form and interact, and how the universe built up its present chemical/elemental composition. Information from the telescope may also determine the shape of the universe.

A major manufacturing challenge, as with any telescope, lies in the light-gathering mirrors. The larger the mirror, NASA points out, the fainter and more distant the objects it can collect light from. But weight is a huge factor in the cost of launching satellites, so the lighter the mirrors, the lower the launch costs.

The primary mirror will be made up of 18 hexagonal segments 1.3 meters wide measuring a total of 6.5 meters in diameter, or more than 21 feet. That's 2.5 times larger in diameter than Hubble's primary mirror and six times larger in area. Optical resolution is to be about 0.1 arc-second in light wavelengths from 0.6 to 28 micrometers. It will be the largest of eight mirrors that generate images for a near-infrared and visible spectrum camera, a near-infrared multi-object spectrograph, and a mid-infrared instrument with camera and spectrograph.

The Webb Space Telescope will use extremely large-aperture, low-mass mirrors. Made in segments so they can be folded to fit into a rocket nose cone for flight, they will open and array themselves when they reach their destination. These robust mirrors must be fabricated rapidly and cost effectively. But the risksÑand costsÑof even the smallest error are huge.

Unlike the Hubble, which is winding down its service 375 miles above Earth, the Webb telescope will be almost a million miles away at one of the Lagrangian points in space, where the gravitational pulls of Earth and the sun are about equal. That puts it far beyond the reach of technicians should something go wrong.

The primary mirror will be made of beryllium. Beryllium's advantage over glass is a near-zero coefficient of thermal expansion in the deep-space operating range of about -375°F, or about 40 kelvin.

Brush Wellman Corp. in Cleveland supplies the beryllium blanks; Axsys Technologies in Cullman, Ala.,
machines the mirror blanks, and SSG/Tinsley of Richmond, Calif., grinds and polishes the blanks. Ball Aerospace in Boulder, Colo., is responsible for the overall telescope design.

Getting it built is the job of a team of contractors led by Northrop Grumman Corp. Manufacturing is being overseen by a NASA team, including the Smithsonian Astrophysical Observatory, which is part of the Harvard-Smithsonian Center for Astrophysics in Cambridge, Mass.

The Smithsonian effort is being directed by Lester M.
Cohen, chief engineer for structural analysis and design. Trained as a civil engineer, he began his professional career in 1973 at Stone & Webster Engineering, a Boston firm that designs and builds power plants.

The structure of the primary mirror for the Webb Telescope permits small adjustments.

Since this is an all-or-nothing gamble, Cohen and his team are modeling and simulating every step in the primary mirror's manufacturing, assembly, and alignment processes. They are using finite element analysis software from Ansys Inc. in Canonsburg, Pa., to monitor technical developments in the mirrors and to help predict the performance of each manufacturing and inspection process.

"At the observatory, we do a lot of work in measuring what happens in manufacturing and predicting what happens in space," Cohen explained. "We monitor the processes for all kinds of deformations, ranging from tiny increments due to gravity to about 40 times gravity during the launch. Our work here has become more intense and exacting as NASA demands that mirrors be larger, lighter, and more precise."

The Smithsonian Astronomical Observatory fulfills its commitments to NASA and the contractors as a sort of backseat driver. Cohen and his team try to keep pace with the designers and engineers at the contractor companies and, whenever possible, work ahead of them. That way, they can anticipate problems and have solutions ready to offer when problems do indeed crop up. While this might be interpreted as a duplication of effort, it is essential in a program that's this visible, risky, and costly.

For example, the designers have to account for the huge difference between manufacturing temperature, approximately 70°F or 290 K, and the operating temperature in a space of 40 K. "The goal here is to make sure that what is beautiful on the ground is beautiful in space," Cohen added.

To see how these differences would affect the mirror's precision, prototypes were taken to the NASA Marshall Space Center in Huntsville, Ala., for measurement at 40 K with a laser interferometer. Then they were warmed up to 290 K for additional measurement. To avoid creating any thermal strain gradients, cooling and warming cycles were done slowly, a few days each. The mirror stayed within required specifications.

There are significant manufacturing challenges in the composite backplanes for the primary mirror. These are to be made from boron composites for their stiffness. "No composite structure has ever been modeled and tested to the degree of fidelity we need," Cohen said. "ATK in Magna, Utah, will design and fabricate subscale parts to show that we understand the mechanics of what is going on, and that the overall thermal and temporal stability requirements can be met."

The analysis and manufacturing challenges in the backplane are the adhesives used to combine all of the composite parts and the uniformity to which the composites themselves can be manufactured. "As glues dry, they shrink, and that introduces lots of orthotropic properties into the material, lots of stresses and strains in a variety of directions," Cohen said. "These can cause dimensional changes with temperature gradients as small as a fraction of 1°F. Even without the pull of gravity, this is enough to affect stability. The composites themselves can distort in an undesirable fashion as the temperatures change in orbit."

The mirrors are attached to the backplane via hexapod actuators, which can adjust the tip, tilt, and piston of each mirror segment. A radius of curvature actuator can also correct for the large (20mm) shrinkage of the mirror radius between room temperature and 40 K.

Cohen estimated that the computer models of the mirror have up to 12 million degrees of freedom. "We made these huge models," he pointed out, "because we wanted to have the feeling that we really knew the model."

This meant approximations, simplifications, and extrapolations were to be strictly avoided. Otherwise his solutions and predictions could be open to question, and conflicting interpretations might be argued.

"For each million DOFs you need between one and two gigabytes of RAM to handle the analyses," Cohen said. He and Smithsonian mechanical engineer Michael Eisenhower use Hewlett-Packard Itanium workstations. The 64-bit Itanium processors handle up to 24 GB of RAM so they accommodate big FEA scratch files very well. Scratch files for models this size need about 10 gigabytes of disk space, Eisenhower said.

Eisenhower and Cohen also rely heavily on Ansys
capabilities for substructuring. Essentially shorthand
for model builders, substructuring saves modeling and run time by reducing sets of thousands of identical elements to an equivalent single element, often called a superelement. "Without substructuring we may not be able to run these models in anything like real time," Cohen said.

According to Cohen, "Because every optic is a completely unique project, you have to always be looking at the learning curve, down which you will proceed only once for each piece. You always have to have a crosscheck for the engineers and the analyses. That's what we do at SAO."