The Perfect Machine

by Ronald Florence

  GE’s publicity had ultimately backfired. A few newspapers picked up on their final releases about the sixty-inch disk, and asked in editorials whether big businesses weren’t “exhausting [themselves] in attempting to outbid one another for the privilege of building the ‘world’s greatest telescope’ at a fabulous price.” Three years before, GE had been proud to count itself among the builders of the telescope. Now the publicity department scurried to dissociate itself from the project.

  McCauley went to work the day after the meeting in New York.

  Trained as a physicist, he shared Hale’s attitude about publicity. Although he had never been involved with a highly publicized project, McCauley had read enough journalists’ reports to know that when reporters were eager for news, they often “insisted on forgetting the plain scientific facts and achievements given them while publishing a dramatical romance of their own invention which whatever scientific facts used garbled as to render them unscientific.”

  Fortunately Corning was a quiet company, secluded amid the rolling hills and narrow winding valleys of the southern tier of New York State, well off the beaten paths to the vacation areas around the Finger Lakes or Niagara Falls. Few tourists came to Corning in the early 1930s, and the town was content to go its quiet way. The company had been known for its research—Corning had supplied the glass envelope for Edison’s incandescent filament to create the electric lamp, and Corning research had produced the Pyrex brand of heat-resisting glass—but the company had never been overeager to publicize itself. The Corning memorandums summarizing the meeting didn’t even mention the publicity issue. An internal memorandum was quietly circulated to request that all reporting on the new project be oral only, with no written statements that might get out to snooping reporters.

  McCauley had successfully cast a disk in his first try, years before, but he was a cautious man. He recruited Ralph Newman, a mold maker from the Corning Laboratory, and Wallace Woods, an experienced glass worker from the blowing room, for a series of experiments to explore the problems of molding of a large telescope mirror.

  Compared to the ferociously complex process of spraying fused quartz, molding glass seems simple. A melting tank of refractory brick is filled with sand, soda, lime, and borax and heated with gas jets suspended from an arched roof until the mix melts into glass. Depending on the intended use, the mixture is fined—maintained at working temperature—for a period of days or even weeks to allow the bubbles that form within the mixture to rise to the surface. The glass in the tank is then ready to be molded or blown into shape. The process is straightforward and tried; for forms of glass used in undemanding applications, glassmaking hasn’t changed much in thousands of years.

  For optical glass, which would be sensitive to contaminants in the raw materials, from the bricks and cements used to build the tank, and from tools used to stir or transfer the glass, the process is a little more complicated. In glass for a critical application, like a telescope mirror, contaminants can weaken the internal structure of the glass or introduce strains that would ultimately distort the disk.

  Even with ordinary plate glass, maintaining the purity and consistency of a large mass of molten glass while transferring it from a tank to a mold is a challenge. Borosilicate glasses, like Pyrex, are even more demanding, because at the extremely high temperatures required to melt borosilicates, tank and tool materials can give off gaseous contaminants or even begin to melt into the glass mixture.

  Typically, a large production tank in the glass factory might produce twenty tons of glass over a twenty-four-hour period. The glass is dribbled out, depending on the articles produced, at anywhere from a few pounds at a time to a continuous stream of thirty to forty pounds per minute. At that rate of use the ingredients for new glass can be melted at one end of the tank as fast as the ready glass is removed from the working end at the other.

  A large telescope disk would require twenty tons of glass all at once. No one had ever fabricated a piece of glass that large or come up with a procedure to relieve the stresses in the glass so it could be used for a high-precision instrument. The requirements for ribs in the backs of the disks, to reduce the thickness of the glass and the weight of the disk without compromising its rigidity, meant that McCauley needed a method of molding glass with complex structures. Corning had vast experience in molding and blowing complex shapes into small units of glass. But no one had experience with molds built to withstand the heat of borosilicate glasses for as long as it would take to fill a mold with twenty tons of Pyrex.

  McCauley began with some basic assumptions: Given the schedule and budget of the project, they would not conduct costly experiments to develop new glasses and methods of melting or working glass. They would not build large machines for tasks that could be performed by manpower at lower cost. A surplus melting tank, idled by lack of demand, would be used to melt the glass, rather than a custom facility.

  For the first experiments McCauley had Ralph Newman build test molds to a standard pattern, with two refractory bricks for the bottom, five bricks for the sides, and wire and angle irons to hold up the corners of the mold. Wally Woods, the experienced glass man, would then pour molten 702 Pyrex, the same formula that had worked in the earlier pours, from a test tank into each mold so they could study the interactions of glass and brick.

  Newman started with the same refractory bricks that had worked successfully for a test disk years before, selecting good bricks, without cracks, for his molds. Each time Wally Woods poured hot glass into one of the new molds, a cloud of bubbles would erupt in the glass. Nothing obvious was wrong. McCauley dug up the records on the bricks and discovered that the leftovers from the shipment that had worked for molds in 1929 had been stored in a damp room under one of the glass factories. The bricks had wicked up moisture, which turned to steam when the molten glass was poured into the mold.

  He had Newman switch to a new grade of brick. That cured the bubbles: Even when the new Armstrong C-25 bricks were soaked in water before the glass was poured, there were no steam bubbles in the molded glass. But success begat new problems: After a few hours in contact with the molten glass the surface of the brick emitted gas bubbles that left a rough surface on the glass. McCauley’s solution was to brush the mold surfaces with a paste of silica flour and water before the glass was poured.

  A satisfactory mold material was only the first step. At the meetings in New York, Hale had described the problems with the one-hundred-inch plate-glass disk on the Hooker telescope, especially the bubbles that marked the layers of glass from each of the three pours. Corning’s production lines for Pyrex consumer and laboratory goods worked by filling molds directly from a big melting tank. The technique wouldn’t work for a telescope mirror, because the mirrors had to be made of exceptionally pure glass, and the purest glass in any tank comes from the center of the batch, which is not in contact with the walls of the melting tank.

  McCauley’s solution was to try an old technology. Ladling, which had been abandoned for production glass work almost everywhere, had advantages for the work on disks. Ladles can take the glass from the center of the tank. They allow the glass to be inspected before it is poured, and because the glass begins to cure against the walls of the ladle, leaving a ladle heel that is later broken out, the glass that remains viscous enough to pour into the mold—generally between half and two-thirds of the contents of the ladle—has effectively only been in contact with glass, not with materials that could introduce contaminations. By introducing the glass into the mold in ladle-size installments, the mold does not face the sudden heat shock of a huge mass of hot molten glass. The alternative to ladles, arrangement of spouts to pour glass from pots into the mold, would have required experimental heated spouts, a turntable for the mold to provide even filling, and a procedure to cut off the spout from the mold and heal the scar. McCauley’s instinct was exactly opposite to those of Ellis and Thomson and GE: He wanted the simplest, cheapest, and safest process.

  Corning
had done little ladling, and whatever experience they had, years before, was long before the development of Pyrex. To duplicate the process of filling a large mold, McCauley had Wally Woods try filling the molds with a small ladle, pouring half the contents of the ladle into the mold and the balance into a cullet can, where cooled glass was collected for later reuse. When Woods took two or more pours to fill the mold, strata would appear in the glass. McCauley suspected that the glass on the surface of the tank, exposed to the air, was devitrifying. When Woods rotated the ladle in the tank to fill it, the surface glass of the tank would end up in the ladle, and ultimately on the top of the molded glass, producing a layer cake with devitrified frosting. The devitrified Pyrex not only marred the appearance of the glass but introduced strains that would degrade the optical performance of a mirror. It wasn’t good enough for a telescope.

  The answer was to skim the surface glass off the ladle before pouring the balance of the glass into the mold. Woods explained that glass has to be skimmed with a paddle rather than a spoon. He fashioned the device he needed from a blow iron, and when he skimmed the surface off each ladle of glass before pouring it into the mold, the castings emerged with no strata. McCauley went home smiling that evening. He was ready to cast telescope disks.

  Molding a large mass of glass is the first step of a complex process that physicists like McCauley were just beginning to understand. Large castings of glass are subject to strains that develop in the internal structure of the material as the mass of molten glass cools. The strains can be relieved, or sometimes eliminated, by annealing the glass—heating the mass of molten glass until it is free of strains, then gradually cooling the mass according to a precise schedule determined by the chemistry of the glass and the size of the mass being annealed. The larger the glass casting and the more critical the use, the more precise the annealing schedule must be. The process is demanding: If the cooling schedule is wrong, or if a failure of equipment drops the temperature too quickly, the casting will emerge with strains that show up as distortions, weaknesses, or instabilities. The mirror would be the largest piece of glass ever cast, and the optical demands put on the primary mirror of the two-hundred-inch telescope would be the toughest criteria ever applied to a large piece of glass. The annealing was crucial.

  After the meetings in New York, McCauley asked two engineers, Howard Lillie at the Corning Laboratory and George Morey at the Geophysical Laboratory of the Carnegie Institution in Washington, to experimentally determine the annealing properties of the 702 Pyrex. The figures they produced for the annealing constant were inconsistent, but the required temperatures were high enough that McCauley knew he would need robust ovens, fine control of the temperature, and lots of power. It was theoretically possible to build a gas-fired annealing oven, but the fine temperature control he needed demanded an electrical oven.

  Thanks to the many inventions of Edison and Thomson, and to General Electric’s production of high-tension transformers and power distribution switches, electrical grids were bringing the miracle of electricity to vast rural areas of the United States in the 1930s. Factories switched over to electrical power even before homes. Electric motors allowed machinery to be portable, instead of requiring overhead belt drives that distributed the power from a central shaft driven by a steam engine or waterwheel. But even as electric power became ubiquitous, it was far less reliable than we have come to expect today. Generators and power lines failed regularly, lightning took out entire networks, and switching equipment and transformers fatigued or shorted out. For most usages, in an era before electronics, long-term reliability wasn’t a critical issue. But for the heating elements and controls in an annealing oven that might have to run for a year or more, reliability was the main concern. It was the one area of the project in which McCauley assumed that GE’s own work would prove useful.

  In mid-November 1931, McCauley and his boss, J. C. Hostetter, went up to the GE laboratories in Lynn, armed with a letter of introduction from Gerard Swope. Ellis was friendly and accommodating. GE, Ellis assured them, could have built the disks of fused silica, and would have, if the new experiments with the one-hundred-inch telescope hadn’t eliminated the need for quartz disks. He knew about the prices Corning had quoted on the series of disks and openly contrasted the numbers with GE’s figure for a two-hundred-inch disk, somewhere between $750,000 and $1 million. When Hostetter pointed out that Corning had held discussions with the Observatory Council but didn’t yet have a firm order for the disks, Ellis assured him that he had inside information on the project and that the order was as good as placed.

  Elihu Thomson showed up at the lab, greeted the two men from Corning, and wished them well. He left after a few minutes, and Ellis talked freely about the professor’s bitterness and disappointment that GE had been forced to abandon the project before producing a big disk. He proudly showed off the two sixty-inch disks the GE effort had produced—one good clear quartz, but split radially; the other filled with bubbles—and described his project to split the cracked disk and fuse the two halves together so it could be used for a solar telescope or other critical use.*

  Much of the equipment in the GE laboratory had been paid for by and was therefore the property of the Observatory Council, but after surveying the huge building and facilities GE had created, McCauley concluded that the only useful equipment for Corning was some large transformers, pyrometric devices for measuring high temperatures, and a large crane. Ellis said that he would be using some of the equipment to finish his work with the sixty-inch disks, so the only transfer arranged was for three large transformers that GE had never used. Those, and the information that GE used a special industrial grade of nichrome as heating elements for furnaces, were McCauley’s inheritance from the years of work and hundreds of thousands of dollars GE had expended on the project.

  McCauley wasn’t a man to criticize openly someone else’s work or plans, but he hadn’t been impressed with what he saw in Lynn. The fiendishly complex apparatus and the yard full of quartz chunks, waiting to be sorted underwater by one man, seemed like parts of a process doomed to endless experimentation and cost overruns.

  His own plan was conservative. Although they would be working with a glass mixture that had never been used for large castings, the procedure of filling a mold with ladles was tried and true. Glassmakers had practiced those skills, with glass formulations simpler than Pyrex, for centuries.

  McCauley had considered and rejected two alternate techniques. One idea was to “sag” the glass into the mold, by placing a large block of pure glass over the mold and heating it until it flowed into the shape of the mold. It was an appealing idea because it eliminated the process of ladling, but for the ribbed disks that the Observatory Council wanted, McCauley would need a mold with cores on the bottom strong enough to support the weight of the entire block of glass. No one had experience building that sort of mold out of refractory brick, the only material that could withstand the heat of molten glass. The alternative was to sag a disk of glass large enough to cover the entire mold. That meant they would first have to cast the unribbed disk, which would entail the costs and risks of two moldings and the risk of handling heavy materials twice.

  Still, the sagging idea was promising enough to merit at least an experiment. One of the Corning melting tanks was scheduled to be shut down for repairs. Normally they would shut off the burners, letting the melt inside cool rapidly. The resultant cracks and strains would then leave a mix of small chunks of glass. Instead McCauley had insulation installed on the sides of the tank before the shutdown, and he slowly reduced the heat in the tank. When the tank was cool, large blocks of cullet were left behind. None was big enough for a two-hundred-inch telescope disk, but several were large enough to suggest that sagging chunks of pure glass was a possible alternative to ladling.

  Another physicist at Corning, Dr. George Littleton, was famed for out-of-the-ordinary ideas. When he saw the plans for ribbed disks, he went back to his lab, muddled over a scratch
pad, then suggested an alternative that he was convinced would be cheaper to fabricate, lighter, and substantially more rigid than a cast mirror. Instead of pouring glass into a shaped mold, Littleton proposed, they could start with Pyrex-brand custard cups, which Corning produced in immense quantities. The cups would be stacked inverted in layers in an open mold, which would then be filled with molten glass. The glass would flow around the cups and seal them into a single structure of regularly arranged thin glass walls enclosing a geometric pattern of air spaces. Littleton’s calculations demonstrated that the resulting honeycomb structure would be light and rigid, and the fabrication procedure avoided the complexities of building molds in complex geometric designs.

  Making a telescope mirror from standard Pyrex custard cups was too audacious not to try. McCauley brought in a supply of the cups and put the mold makers and then Woods to work. When Woods ladled the molten glass into the mold, it flowed into the gaps between the inverted custard cups, sealing off air cavities here and there in the pile. As more glass was added, the pockets of trapped air, expanding with the increasing temperature in the mold, formed odd-shaped bubbles, turning the neat geometric pattern into chaos. McCauley and Littleton, each with a Ph.D. in physics, had temporarily forgotten that hot air expands.

  The custard-cup experiment settled the method of making disks. McCauley would stick with his original conservative plan. He had the Corning purchasing department order large ladles, capable of holding three hundred pounds of glass, and told the foremen in the Corning factories to be on the lookout for experienced ladlers among their glassmakers.

  McCauley had been impressed with the heavy-duty nichrome heaters he had seen at Lynn and the big transformers that GE had transferred to Corning. Willing salesmen had plied him with specification sheets for electrical control equipment. No order was too small to ignore in the depression. He was especially impressed with the large theater dimmers Westinghouse manufactured. The controls were simple, ruggedly built, and fitted with calibrated dials that could be rearranged as a vernier scale, which would allow precise adjustment of the temperature in the annealing oven. For safety he had the GE transformers rebuilt so they would take 2,200 volts from the main feeder lines coming into the Corning factory and put out a nonlethal 35 volts to the heating elements instead of the stock 110 volts.