Late in 1941 Lawrence had installed such a 180-degree mass spectrometer in place of the dees in the Berkeley 37-inch cyclotron. By running it continuously for a month his crews produced a partially separated 100-microgram sample of U235.1877 That was several hundred million times less than the 100 kilograms Robert Oppenheimer had originally estimated would be necessary to make a bomb. The demonstration proved the basic principle of electromagnetic separation even as it dramatized the method’s monumental prodigality: Lawrence was proposing to separate uranium atom by individual atom.
Magnetic field perpendicular to plane of drawing
Enlarging the equipment, increasing the accelerating voltage, multiplying the number of sources and collectors set side by side between the poles of the same magnet were obvious ways to improve output and efficiency. Lawrence had committed his time to winning the war; now he committed his beautiful new 184-inch cyclotron. Instead of cyclotron dees he had D-shaped mass-spectrometer tanks installed between the pole faces of its 4,500-ton magnet. Making the new instrument work, through the spring and summer of 1942, solved the most difficult design problems. It acquired a name along the way: calutron, another tron from the University of Cali fornia.
To separate 100 grams—about 4 ounces—of U235 per day, Lawrence estimated in the autumn of 1942, would require some 2,000 4-foot calutron tanks set among thousands of tons of magnets. If a bomb needed 30 kilograms—66 pounds—of U235 for reasonable efficiency, as the Berkeley summer study group had just worked out, 2,000 such calutrons could enrich material enough for one bomb core every 300 days. That assumed the system worked reliably, which so far its laboratory predecessors had hardly done. Yet in 1942 electromagnetic separation still looked so much more promising to James Bryant Conant than either the plutonium approach or gaseous barrier diffusion that he had offered up for debate the possibility of pursuing it exclusively. Lawrence was self-confident but not foolhardy; he insisted that the two dark horses should continue to run the race alongside the favorite.
Groves was less impressed. So was the first Lewis committee that had visited Chicago and Berkeley when Fermi was building CP-1 in the winter of 1942. The Lewis committee judged gaseous diffusion the best approach because it was most like existing technology—diffusion was a phenomenon familiar to petroleum engineers and a gaseous-diffusion plant would be essentially an enormous interconnected assemblage of pipes and pumps. Electromagnetic separation by contrast was a batch process untested at such monumental scale; Berkeley planned a system of 4-foot tanks set vertically between the pole faces of large square electromagnets, two tanks to a gap and a total of 96 tanks per unit. To reduce the amount of iron needed for the magnet cores the arrangement would be not rectangular but oval, like a racetrack:
And racetrack it was called, though its official designation was Alpha. Berkeley could promise only 5 grams of enriched uranium per day per racetrack, but Groves thought 2,000 tanks well beyond Stone & Webster’s capability and cut the number back to 500, reasoning, as Lawrence recalled later, “that the art and science of the process would go forward and that by the time the plant was built substantially higher production rates would be assured.”1878 Five grams per day per racetrack with only five racetracks would mean 1,200 days per 30-kilogram bomb even if the Alpha calutrons produced nearly pure U235, which they did not—their best production was around 15 percent. Groves counted on improvements and forged ahead.
He had to begin building before he knew precisely what to build. He worked from the general to the particular, from outline to detail. Fully six months before he decided how many calutrons to authorize, his predecessors, Colonel James Marshall and Lieutenant Colonel Kenneth Nichols, had moved to solve one serious problem of supply. The United States was critically short of copper, the best common metal for winding the coils of electromagnets. For recoverable use the Treasury offered to make silver bullion available in copper’s stead. The Manhattan District put the offer to the test, Nichols negotiating the loan with Treasury Undersecretary Daniel Bell. “At one point in the negotiations,” writes Groves, “Nichols . . . said that they would need between five and ten thousand tons of silver. This led to the icy reply: ‘Colonel, in the Treasury we do not speak of tons of silver; our unit is the Troy ounce.’ ”1879 Eventually 395 million troy ounces of silver—13,540 short tons—went off from the West Point Depository to be cast into cylindrical billets, rolled into 40-foot strips and wound onto iron cores at Allis-Chalmers in Milwaukee. Solid-silver bus bars a square foot in cross section crowned each racetrack’s long oval. The silver was worth more than $300 million. Groves accounted for it ounce by ounce, almost as carefully as he accounted for the fissionable isotope it helped separate.
Stone & Webster had only foundation drawings in hand when its contractors broke ground for the first Alpha racetrack building on February 18, 1943. Groves had initially approved three buildings to house five racetracks. In March he authorized a second, Beta stage of half-size calutrons, seventy-two tanks on two rectangular tracks, that would further enrich the eventual Alpha output to 90 percent U235. Alpha and Beta buildings alone eventually covered more area in the valley between Pine and Chestnut ridges than would twenty football fields. Racetracks were mounted on second floors; first floors held monumental pumps to exhaust the calutrons to high vacuum, more cubic feet of vacuum than the combined total volume pumped down everywhere else on earth at that time. Eventually the Y-12 complex counted 268 permanent buildings large and small—the calutron structures of steel and brick and tile, chemistry laboratories, a distilled water plant, sewage treatment plants, pump houses, a shop, a service station, warehouses, cafeterias, gatehouses, change houses and locker rooms, a paymaster’s office, a foundry, a generator building, eight electric substations, nineteen water-cooling towers—for an output measured in the best of times in grams per day. An inspection trip in May 1943 awed even Ernest Lawrence.
By August, twenty thousand construction workers swarmed over the area.1880, 1881 An experimental Alpha unit saw successful operation. Lawrence was urging Groves then to double the Alpha plant. With ten Alpha racetracks instead of five he estimated he could separate half a kilogram of U235 per day at 85 percent enrichment. An Army engineer’s less exuberant summary, written six days after Lawrence’s, predicted 900 grams per month with existing Alpha and Beta stages beginning in November 1943, for a total of 22 kilograms of bomb-grade U235 in the first year of operation. Faced with new estimates from Los Alamos that summer that an efficient uranium gun would probably require 40 kilograms—88 pounds—of the rarer uranium isotope, Groves bought Lawrence’s proposal.1882 The doubling would add four new 96-tank tracks of advanced design designated Alpha II and a proportionate number of Beta tracks, at a cost of $150 million more than the $100 million already authorized. If everything worked at Y-12, Groves justified his proposal to the Military Policy Committee, he would then have a 40-kilogram bomb core around the beginning of 1945.
The Army had contracted with Tennessee Eastman, a manufacturing subsidiary of Eastman Kodak, to operate the electromagnetic separation plant.1883 By late October 1943, when Stone & Webster finished installing the first Alpha racetrack, the company had assembled a work force of 4,800 men and women. They were trained to run and maintain the calutrons—without knowing why—twenty-four hours a day, seven days a week.
The big square racetrack magnets wrapped with silver windings were encased in boxes of welded steel. Oil that circulated through the boxes was supposed to insulate the windings and carry heat away. The first magnets tested at the end of October leaked electricity. If moisture in the circulating oil was shorting out the coils, the normal heat of operation would correct the problem by evaporating the water. Tennessee Eastman pushed on. Vacuum leaks in the calutron tanks were numerous and hard to find—one supervisor remembers spending most of a month looking for one leak.1884 Inexperienced operators had trouble striking and maintaining a steady ion beam. Groves recalls that the powerful magnets unexpectedly “moved the intervening tank
s, which weighed some fourteen tons each, out of position by as much as three inches. . . . The problem was solved by securely welding the tanks into place, using heavy steel tie straps. Once that was done, the tanks stayed where they belonged.”1885
The magnets dried out but continued to short. Something was seriously wrong. Early in December Tennessee Eastman shut the entire 96-tank racetrack down. The company’s engineers would have to break open one of the windings and examine it. That was major trauma; the unit must then be returned to Allis-Chalmers and rebuilt.
The inspectors found disaster: two major troubles. “The first lay in the design,” writes Groves, “which placed the heavy current-carrying silver bands too close together.1886 The other lay in the excessive amount of rust and other dirt particles in the circulating oil. These bridged the too narrow gap between the silver bands and resulted in shorting.” Groves arrived seething from Washington on December 15 to view the remains. The design’s inadequacy forced the general to order all forty-eight magnets hauled back to Milwaukee to be cleaned and rebuilt. The second Alpha track would not come on line until mid-January 1944. They would lose at least a month of production.
Tennessee Eastman’s 4,800 employees reported for work in the shambles of gloomy halls. Rather than lose them from boredom the company scheduled classes, conferences, lectures, motion pictures, games. Serious men in double-breasted suits scouted the state for chess and checker sets. At the end of 1943 Y-12 was dead in the water with hardly a gram of U235 to show for all its enormous expense.
* * *
Gaseous-diffusion research had progressed at Columbia University since John Dunning and Eugene Booth had first demonstrated measurable U235 separation in November 1941. By the spring of 1942 Harold Urey could note in a progress report that “three methods for the separation of the uranium isotopes have now reached the engineering stage. They are the English and the American diffusion methods, and the centrifuge method.” With the authorization of the full-scale plant Dunning’s staff, which had grown to include about ninety people, increased in early 1943 to 225.1887, 1888 Franz Simon’s diffusion method would have operated at low gas pressures and in incremental ten-unit stages but required extremely large pumps; Columbia designed a high-pressure system with more conventional pumps, a continuous, interconnected cascade of some four thousand stages. In a postwar memoir Groves reviews the design, which was both reliably simple and expensively tedious:
The method was completely novel. It was based on the theory that if uranium gas was pumped against a porous barrier, the lighter molecules of the gas, containing U-235, would pass through more rapidly than the heavier U-238 molecules. The heart of the process was, therefore, the barrier, a porous thin metal sheet or membrane with millions of submicroscopic openings per square inch. These sheets were formed into tubes which were enclosed in an airtight vessel, the diffuser. As the gas, uranium hexafluoride, was pumped through a long series, or cascade, of these tubes it tended to separate, the enriched gas moving up the cascade while the depleted moved down. However, there is so little difference in mass between the hexafluoride of U-238 and U-235 that it was impossible to gain much separation in a single diffusion step. That was why there had to be several thousand successive stages.1889
In schematic cross section the stages looked like this:
“Further development of barriers is needed,” Urey had concluded in his progress report, “but we now feel confident that the problem can be solved.”1890 It had not been solved when Groves committed the Manhattan Project to a $100 million gaseous-diffusion plant, however; no practical barrier was yet in hand. The American process required finer-pored material than the British; the material also had to be rugged enough to withstand the higher pressure of the heavy, corrosive gas.
Columbia had been experimenting with copper barriers but abandoned them late in 1942 in favor of nickel, the only common metal that resisted hexafluoride corrosion. Compressed nickel powder made a suitably rugged but insufficiently fine-pored barrier material; electro-deposited nickel mesh made a suitably fine-pored but insufficiently rugged alternative. A self-educated Anglo-American interior decorator, Edward Norris, had devised the electro-deposited mesh originally for a new kind of paint sprayer he invented; he joined the Columbia project in 1941 and worked with chemist Edward Adler, a young Urey protégé, to adapt his invention to gaseous diffusion. The resulting Norris-Adler barrier in its nickel incarnation seemed in January 1943 to be improvable eventually to production quality, whereupon Columbia began installing a pilot plant in the basement of Schermerhorn Laboratory and Groves authorized full-scale barrier production. The Houdaille-Hershey Corporation took on that assignment on April 1, the day the gates began operating at Oak Ridge, planning a new factory for the purpose in Decatur, Illinois.
Suitable barrier material was the worst but not the only problem Columbia studied and Groves engineered. Hex attacked organic materials ferociously: not a speck of grease could be allowed to ooze into the gas stream anywhere along the miles and miles of pipes and pumps and barriers. Pump seals therefore had to be devised that were both gastight and greaseless, a puzzle no one had ever solved before that required the development of new kinds of plastics. (The seal material that eventually served at Oak Ridge came into its own after the war under the brand name Teflon.) A single pinhole leak anywhere in the miles of pipes would confound the entire system; Alfred O. Nier developed portable mass spectrometers to serve as subtle leak detectors. Since pipes of solid nickel would exhaust the entire U.S. production of that valuable resource, Groves found a company willing to nickel-plate all the pipe interiors, a difficult new process accomplished by filling the pipes themselves with plating solution and rotating them as the plating current did its work.
The plant that would hold thousands of diffusion tanks, the largest of them of 1,000 gallon capacity, would be necessarily monumental: four stories high, almost half a mile long in the shape of a U, a fifth of a mile wide, 42.6 acres under roof, some 2 million square feet, more than twice the total ground area of Y-12’s Alpha and Beta buildings. K-25, as the gaseous-diffusion complex was designated, needed more than a narrow ridge valley. The building and operating contractors, Kellex and Union Carbide, found a relatively flat site along the Clinch River at the southwestern end of the reservation; the first surveying, for the coal-fired power plant needed to run the factory, began on May 31, 1943.
Rather than designing and setting thousands of different columns for footings the construction contractors leveled and compacted the entire K-25 foundation area, plowing, drying and moving in the process nearly 100,000 cubic yards of red clay. That took months; the first concrete—200,000 cubic yards—was not poured until October 21. By then the continuing failure to develop an adequate barrier material had led Groves to decide to lop off the unfinished plant’s upper stages and limit its enrichment potential to less than 50 percent U235—it would have been capable of taking natural uranium all the way to pure U235 with its full complement of diffusers—and to use this enriched material to feed the Beta calutrons at Y-12.
Kellex succeeded in devising a promising new barrier material in the autumn of 1943 that combined the best features of the Norris-Adler barrier and the compressed nickel-powder barrier. The problem then was what to do about the Houdaille-Hershey plant under construction in Decatur, which was designed to produce Norris-Adler. Should it be stripped and reequipped to manufacture the new barrier at the price of some delay in starting up K-25? Or should the several barrier-development teams make a final concerted effort to improve Norris-Adler to production quality? Over these significant questions Groves and Harold Urey violently clashed.
Kellex wanted to strip the Houdaille-Hershey plant and convert it, preferring delay to the risk of failure. Urey thought abandoning the Norris-Adler barrier would mean forgoing the production of U235 by gaseous diffusion in time to shorten the war. In which case he saw no reason to continue building K-25; its high priority, he argued, would even hinder the war effort by dis
placing more immediately useful production.
Groves decided to submit the dispute to an unusual review committee: the experts who had worked on gaseous diffusion in England. With the renewal of interchange between the British and American atomic bomb programs that autumn the British had arranged to send a delegation to work in America. Led by Wallace Akers of ICI, the group included Franz Simon and Rudolf Peierls. It met with both sides—Kellex and Columbia—on December 22 and then settled in to review American progress.
The participants reconvened early in January 1944. The new barrier, the British concluded, would probably be superior eventually to the Norris-Adler, but they thought the months of research on the Norris-Adler must count decisively in its favor if time was of the essence. The new barrier had been manufactured so far only by hand in small batches. Yet K-25 would require acres of it to fill the planned 2,892 stages of the diffusion plant’s cascade.1891
Then Kellex set a trap: it proposed to produce the new barrier by hand by piecework—thousands of workers each duplicating the simple laboratory process Kellex had initially devised—and claimed that by doing so it could match or beat the Norris-Adler production schedule. When the British had recovered from their surprise at the novelty of the proposal they signaled their preference for the new barrier by agreeing that if production was possible it ought to be pursued. That agreement sprang the trap; with the British implicitly committed, the American engineers revealed that they could only manufacture the new barrier by stripping the Houdaille-Hershey plant and forgoing Norris-Adler production entirely.
Groves in any case had already decided, the day before the January meeting, to switch over to the new barrier; the British review then simply ratified his decision. By changing barriers rather than abandoning gaseous diffusion he confirmed what many Manhattan Project scientists had not yet realized: that the commitment of the United States to nuclear weapons development had enlarged from the seemingly urgent but narrow goal of beating the Germans to the bomb. Building a gaseous-diffusion plant that would interfere with conventional war production, would eventually cost half a billion dollars but would almost certainly not contribute significantly to shortening the war meant that nuclear weapons were thenceforth to be counted a permanent addition to the U.S. arsenal. Urey saw the point and withdrew; “from that time forward,” write his colleague biographers, “his energies were directed to the control of atomic energy, not its applications.”1892