The Perfect Machine (61 page)

The routine was deadening. At least the rough grinding had showed some minuscule, visible progress. Workers who had been on the project during the grinding quit in frustration at the monotony of the polishing. Marcus Brown was fair, but he wasn’t an easy boss. Shaping the mirror was a compulsion for Brownie. He had no time for arguments. “I know what I want done,” he would say. “Do it!” By 1938 the depression was easing. There were other jobs.

In September 1938 the disk was raised to a vertical position for the first optical test. Even for the largest mirror in the world, the testing procedure was straightforward. J. B. L. Foucault, who devised a pendulum experiment to demonstrate the rotation of the earth, had also developed a simple and reliable test for measuring whether a mirror had a true spherical shape. Foucault discovered that if a point source of light is shined at the center of curvature of a
perfect
spherical mirror, all the light is reflected back to a point. A knife edge moved so that it cuts off the light at the focus will cause the mirror to darken evenly all over. A slight movement of the knife edge across the point of focused light will cause no moving shadow in the mirror.

For testing the two-hundred-inch mirror, a Foucault test station, with a micrometer-adjustable eyepiece, knife edge, and point source of light was set up at the far end of the room, 120 feet away. The point source focused the light through a pinhole 1/1000 inch in diameter onto the mirror. Brownie or Anderson would then move the knife edge until it just cut off the light. If the mirror was perfect, the entire diameter of the disk would suddenly go black. But no mirror is perfect. The secret to the testing procedure was to read the significance of the patterns of light and shadow created by the knife edge, to identify areas that required additional polishing. Good opticians read the shadows the way a sailor reads dark spots on the water for wind, seeing significance in patterns that escape the eye of the amateur.

Brown and Anderson tested the mirror on Saturdays. The optical lab had been designed as much for testing the mirror as for polishing: The big five-hundred- and one-thousand-watt lights in the ceiling were set behind panels of heat-absorbing glass, and the vents in the walls and attic were oversize. The accuracy of the tests depended on the optical homogeneity of the air in the room. The air-conditioning and
ventilation equipment would all be shut down one half hour before testing began, to allow the air in the room to steady itself. The heavy cork insulation on the walls and ceiling were sufficient to maintain the interior temperature, in summer or winter, while tests were in progress.

On the first test of the two-hundred-inch mirror, the Foucault patterns weren’t hard to discern. Brownie and his crew had achieved a fair spherical surface in their preliminary polishing. Anderson reported that there was still a trace left of the worst of the fractures in the glass, which showed up as a “very fine dark line.” They could live with only one fracture. “Brown and I are jubilant over the results.”

Some zones of the disk were too high or too low, which wasn’t surprising for a relatively preliminary stage of the shaping of the mirror. The mirror also seemed to suffer from astigmatism. The curvature in the vertical plane was approximately one millimeter shorter than the curvature in the horizontal plane. The linear astigmatism was approximately 0.05 inches, which didn’t seem that bad on a disk almost seventeen feet in diameter that didn’t yet have what Brownie or Anderson would consider an
optically
smooth spherical surface.

Anderson and Brownie rotated the mirror ninety degrees and tested it again. The astigmatism was still there, and still in the vertical plane. That was troubling. If the astigmatism was in the shape of the mirror, it should have rotated when the mirror was turned. Anderson repeated the tests, checking that the error wasn’t introduced by the test procedure itself. The astigmatism remained.

There were enough surface errors—high and low spots revealed by the knife-edge tests—to keep Brownie and his crew busy with the polishing tools and rouge while Anderson worried about the astigmatism. More months of fine polishing improved the accuracy of the overall shape of the mirror from approximately 0.01 inch to 0.001 inch. But the vertical astigmatism was still there. As the accuracy of the surface permitted finer measurements, Anderson discovered that when the disk was tested with its axis horizontal, then rotated by one hundred eighty degrees, so that what had been the top was now the bottom, the astigmatism was not exactly the same, but differed by as much as 0.01 inch in the value of a focal length in the two directions—an amount too large to be accidental.

The results were enough to send Anderson back to his slide rule.

Physicists who choose optics as their field crave precision and predictability. An astrophysicist sometimes has to live with vague answers, theories based on the paltry fragments of evidence he or she can squeeze from observations at the limits of a telescope’s reach. In optics the materials are predictable. The qualities of glass, even new glasses like Pyrex, can be measured. The response of the glass to tension, compression, and heat can be measured and extrapolated. The optician
relies on that predictability to bring an optical surface to the incredible level of precision—on the order of one millionth of an inch—that a large telescope mirror requires.

John Anderson was a precise man. He wore a neatly tied bow tie and wire-rimmed glasses. His hair was carefully parted in the middle. He was a worrier. He had worried about an earthquake when the disk arrived at the optics shop. When Brownie began the surface grinding, Anderson worried that they would not get through the contaminations to good, clear glass. Now the behavior of the mirror disk troubled him even more, because he wasn’t sure he would explain it. The surface grinding seemed to have removed the checks and fractures. The glass that remained was consistent. Tests of samples from the disk, and George McCauley’s tests of glass from the batch of Pyrex that had been used to cast it, didn’t differ from the parameters they had used in designing the disk. They had gotten the glass they expected. But something was wrong.

Anderson suspected the support mechanisms in the back of the disk. The supports were supposed to compensate for any tendency of the disk to change shape as it moved from horizontal to vertical. The designs of the complex devices had been gone over and over; the parts had been machined to close tolerances in the astrophysics machine shop; and in tests the supports had performed exactly as the calculations predicted. In the disk they didn’t seem to be doing the job.

He had the supports removed and retested. The thirty-six supports were precision machines, an assembly of levers, counterweights, gears, and ball bearings like a fine wrist watch, but large enough that it took two men to carry each of them. They were machined to the precision of a watch. But in a watch, where parts move continuously, and the largest force is the turning of the hands, simple jeweled bearings are sufficient to keep the parts moving freely. The lever arms of the support mechanisms moved only short distances and infrequently; the loads they had to push were on the order of 850 pounds.

When the supports were retested, the mechanics found that some of the support mechanisms required three or four times as much force to overcome the stiction as the freer-moving supports. The mechanics went to work on the supports, stripping them down, testing batches of bearings, repolishing shafts. When the mechanics finished, some of the components were brought back to the optics shop and polished with rouge by the opticians, honing the surfaces to an optical smoothness. When the rebuilding was finished, every support mechanism tested within the same narrow parameters.

Even as he had the mechanics rebuild the support mechanisms, Anderson realized that the repairs wouldn’t fully fix the problem. It took him a long time to figure out what was really wrong.

The design of the support mechanisms was clever. The levers of the mechanisms were designed to push up, against the tops of the
pockets in the back of the disk, to counteract the force of gravity pulling down on the disk. The supports were four inches behind the actual surface of the disk. The distance from the support to the front of the disk acted as a lever, pushing the upper part of the surface in front of each support forward, and pulling the lower part back to create a vertical S. In the language of the optician, “The deformed condition will consist in the addition of a very weak convex cylinder to the upper half and a similar concave cylinder to the lower half.” With thirty-six supports, the thirty-six areas of the mirror should each have been equally deformed: thirty-six barely detectable Ss on the surface of the disk.

But there were no local Ss, or if there were, they were too small to measure. Instead the tests of the mirror showed a vertical concavity of the entire surface. It was as if each support were producing only the “concave cylinder” on the lower half of each area in front of the pockets. Together, those minuscule concave sections added up to a deepening of the shape of the disk in that plane—astigmatism.

Anderson stared at sketches on the blackboard. He ran numbers through his slide rule. He studied Russell Porter’s drawings of the support mechanisms. He combed engineering and optics texts, trying to find an analogy to the behavior of the disk. There were no data on the behavior of masses of glass as large as the disk. No one had ever analyzed the interaction of support mechanisms and a thin, ribbed disk. And no one had ever tried to figure and test an optical surface as large and demanding as the two-hundred-inch disk.

The problem, Anderson finally realized, was that despite the interlocked webbing around the pockets, the front of the disk was stiffer than the honeycombed back. As a result the effect of the support levers pushing upward in the pockets was asymmetrical: A smaller area would become convex from the upward and outward pressure of the lever arm, and a larger area below would become concave. The Ss had a tiny upper half and a giant lower half. Instead of the concave and convex distortions canceling one another, the net local effect was concave.

Anderson’s solution to the astigmatism was to design twelve “squeeze” correctors that would press on the rear portion of the edge of the disk. The edge correctors also worked with lever arms and counterweights. When the mirror was horizontal (the telescope was pointing to the zenith), they would have no effect; when the mirror was vertical (pointing to the horizon), they would exert the maximum squeeze on the edges of the disk. The compensating squeeze was calculated to compensate exactly the tendency to vertical astigmatism. The machine shop fabricated the supports, and they were installed in the mirror cell with the rebuilt pocket supports. The troublesome vertical astigmatism all but disappeared. Brownie and his crew of twenty-one men went back to work.

By 1938 the trajectory of the project had leveled off. The years when design work was in partial suspension, when a crew of opticians waited in the optics shop, busying themselves with experimental disks while they waited for news first from GE, then Corning, when work on the mountain was tentative, lest they build an observatory without a telescope—were long gone. Everywhere work was in full swing. On the mountain work crews worked on the dome and mounting for the two-hundred-inch and the forty-eight-inch Schmidt telescopes, auxiliary buildings, the residence for astronomers, cottages for staff, and the powerhouse and utility building. Sandy McDowell was taking daily reports from Palomar and firing off his memos in eight copies.

Even after three years, McDowell still didn’t get on with the scientists in the project. The objections didn’t arise because he was an outsider. Max Mason had stepped in from the Rockefeller Foundation without ruffling feathers. As chairman of the Observatory Council, he stayed out of the day-to-day work on the telescope, but he understood the technical problems and the working style of men like John Anderson and Sinclair Smith. Men like George McCauley at Corning, Rein Kroon from Westinghouse, or John Strong from Johns Hopkins, who had come to work on experiments on the coating of the mirror, were accepted almost immediately. They spoke the language of scientists and played by the rules of science—documenting their work, commanding the facts and figures to make their points, and producing solutions that could stand the test of the physicists.

McDowell had never learned the rules and style of science. Used to navy command and his own old-boy network of the Bureau of Ships and the companies that contracted for them, he had never gotten over the feeling that what the group really needed was a “kick in the pants.” He never fully grasped how different the telescope was from the machines he knew, like a battleship turret gun, or why the criterion of “good enough” that produced the best balance of cost and function in expendable military hardware wasn’t an appropriate criterion for a unique scientific instrument that was expected to function effectively for a century.

McDowell wrote reports on the progress of the project for journals and gave talks wherever he could. Even for an audience like the readers of
Scientific American
he would write that the control system had to be able to hold a rifle on a quarter at a distance of three miles, the sort of analogy that might be appropriate to naval gunnery but that missed the real challenge of a telescope control system that had to compensate for minuscule atmospheric refraction and periodic errors that were not a factor in holding a gun on a target. Early in the project Sinclair Smith investigated the possibility of using photosensitive devices as detectors for the telescope. He worked with Joel Stebbins, an astronomer who had pioneered work in this area, and Vladimir Zworykin, director of research at RCA, to find out the state of the art
in detection devices, including those used for early television cameras. McDowell’s conclusions from this research were: “Because of the possible public demand for seeing results of the 200-inch telescope, and particularly because of the desire not to allow the public inside the dome at night, it seemed that a television projection of planets, moon, etc., as seen through the 200-inch telescope, would be of interest.” It was the kind of suggestion that was guaranteed to make astronomers cringe.