The world did not lack methods for separating isotopes when it discovered the possible utility of a kilogram of uranium-235 (U-235). Known techniques, pursued simultaneously in Germany and the United States, included ultra-centrifugation, diffusion across thermal or osmotic pressure barriers, and deflection in electric and magnetic fields. The last method appealed to Lawrence, who had made his reputation on the precise control of beams of charged particles. In principle the technique is simple. When passing between the poles of a magnet, a monoenergetic beam of ions of naturally occurring uranium splits into several streams according to their momentum, one per isotope, each characterized by a particular radius of curvature. Collecting cups at the ends of the semicircular trajectories catch the homogenous streams.

Schematic diagram of uranium isotope separation in the calutron.

Most physicists in 1941 doubted that electromagnetic separation would succeed in practice because they expected that the mutual repulsion of the like-charged ions would prevent the formation of narrow beams. But Lawrence, who had seen a line of positively-charged ions pour from his cyclotron, guessed that negative particles formed in the air kept the beam from dispersing under its own electrical influence. He had the 37-inch cyclotron modified to demonstrate the feasibility of electromagnetic separation of uranium isotopes using the principle of the mass spectrograph. "It will not be a calamity," he wrote Compton, if uranium turned out to have no military applications; but if "fantastically positive and we fail to get them first, the results for our country may well be a tragic disaster." By December 1941 the uranium ion beam was passing 5 microamperes to the collector; a small amount to be sure, but enough to assure Lawrence that space charge would not be a formidable problem.

The fact that beams of uranium ions could be defined well enough to yield small quantities of isotopes suitable for laboratory research by no means assured that electromagnetic separation could be worked on the industrial scale necessary to make a kilogram of U-235. The process has little to work on, only the very small difference in mass-1.25 percent-between uraniums 235 and 238. Because the lighter ions respond slightly more readily to the magnetic field than the heavier, their trajectories bend in a-tighter arc. At the end of their semicircular travel, the ions of U-235 are relatively more plentiful on the inside than on the outside of the beam. But the maximum separation even in the ideal case is small, only one tenth of an inch for an arc with a diameter of 37 inches. Actual beams are far from ideal.

Frank Oppenheimer (center right) and Robert Thornton (right) examine the 4-source emitter for the improved alpha calutron.

Many technical problems had to be solved before even a prototype could be tested in the field of the nearly-completed 184-inch magnet. The beams, though small, could melt the collectors during long hours of operation; the staff therefore installed water cooling for the collectors and tank liner. They contrived electric arcs to ionize the uranium chloride feed. They devised ways to extract the enriched uranium that collected at the receiver, and the still valuable feed material that condensed along with chloride "gunk" (to use their technical term) all over the inside of the tank. They made scrapers to clean the exit slits of the feed sources regularly lest the accumulated "crud" (another word of art) cut down beam strength. Lawrence's optimistic conclusion: by the fall of 1942 ten "calutrons" (as he called the electromagnetic separator), each with a 100 milliampere source and all operating within the 184-inch field, would produce four grams of enriched uranium a day. The "S-1" committee that oversaw the uranium project for OSRD recommended expending $12 million to create a plant with 25 times that capacity before the fall of 1943. Lawrence did not doubt that other means, particularly reactor production of fissile plutonium, might ultimately be the most efficient way to a bomb. But in mid-1942 no reactor worked, and the calutron did.

The calutron design settled on in 1942, called "alpha," provided for enrichment of natural uranium to about 15 percent U-235. Extravagant effort went into designing powerful ion sources and aptly shaped, eventually parabolic collecting slots. The many modifications and security codes proliferated whimsical names: sources Plato, Cyclops, Bicyclops, and Goofy mated with receivers Gloria, Irene, Mona, or Zulu. Ions from Plato and his friends traversed an arc 48 inches in radius to reach collector slits placed 0.6 inch apart. The guiding magnetic field was shimmed not by the old black art but in obedience to calculations. Accurately machined and installed, the shims greatly increased the usable beam that reached the collectors.

Installing magnet shims in an alpha calutron tank to increase output of uranium-235.

Among results obtained with the 184-inch magnet was a design superior to it for large-scale calutrons, the so-called "XA." The prototype of the magnets to be installed at Oak Ridge, XA was a rectangular, three-coil magnet giving a horizontal field in which the calutron tanks could stand side-by-side. It had room for four alpha tanks, each with a double source. By the spring of 1943, convinced that the Germans might be ahead, Groves decided to skip the scheduled pilot plant; from the XA and a scale model of the production magnet alone would come procedures for alpha operation at Oak Ridge. Tests of the first, full-scale system installed there, the XAX, were scheduled for July.

The spring and early summer of 1943 brought hundreds of trainees to Berkeley from Tennessee- Eastman Company, the operator for the Oak Ridge plant. The Laboratory labored to ensure that the test XA magnet system and alpha units were working by April in spite of delays in delivery of steel. Between April and July the training sessions ran continuously. In June a migration that by 1944 would reach 200 started for Oak Ridge. Laboratory expenditures exceeded half a million dollars a month.

The first wave of Berkeley workers at Oak Ridge had to see that the XAX magnet worked. Then runs could begin on the first production system, or "racetrack;" a 24-fold magnification of the XA that could hold 96 calutron alpha tanks. To minimize magnetic losses and steel consumption, the assembly was curved into an oval 122 feet long, 77 feet wide and 15 feet high. Want of copper for the large coils to produce the magnetic fields prompted a solution possible only in wartime: Groves drafted 14,700 tons of pure silver from a government vault for the purpose. Late in the summer of 1943 the XAX was ready for testing. After a week of difficulty, it cleared the hurdle for full-scale racetrack runs.

Control panels and operators for calutrons at Oak Ridge. The operators, mostly women, worked in shifts covering 24 hours a day.

The first two of five projected racetracks started up in November and failed from contaminated cooling oil; the second was limping in January, but produced 200 grams of uranium enriched to 12 percent U-235 by the end of February 1944, its fifth of the total goal of one kilogram of enriched uranium per month. By April four racetracks were functioning, including the repaired number 1. They required constant attention. Many people from the Laboratory helped to modify the units to reach production goals. Responsibility for operation passed entirely to Tennessee Eastman after the spring of 1944, and the Laboratory staff at Oak Ridge turned their attention to redesigning the calutron system for higher efficiency.

Many at the Laboratory, especially Lofgren and Kamen, thought that a second stage would be necessary to reach the required enrichment. Groves approved the idea. In the spring of 1943, during training at Berkeley for alpha operations, design began on the second or beta stage. Because beta would have only the enriched product of alpha as feed, it would process proportionately less material; its beam therefore did not need to be as broad, nor its dimensions as large, as alpha's. Beta design emphasized recovery, not only of the further enriched output but also of the already enriched feed. The first units were tried at Oak Ridge in late February 1944, but the sources had to be redesigned, and even by June difficulties persisted in recovering the precious beta feed strewn throughout the calutron. Process efficiencies stayed low: only 4 or 5 percent of the U-235 in the feed ended up in the output. A better source of enriched uranium feed would have to be found to create the 10 kilograms or so of 90 percent U-235 that Oppenheimer thought necessary for a bomb.

The "C" shaped alpha calutron tank, together with its emitters and collectors on the lower-edge door, was removed in a special "drydock" from the magnet for recovery of uranium-235.

The gaseous diffusion procedure for separation of uranium isotopes, which had consumed more money even than the calutron, had not met its design goals by late 1944. Groves decided that it could not be counted on to produce high enrichment, and that whatever it did produce would have to be supplemented with other slightly enriched uranium and processed through beta calutrons. To augment the calutron feed, the MED constructed still another plant at Oak Ridge, this one working by thermal diffusion, a method developed by Abelson.

In the critical production period in the first months of 1945, the calutrons, particularly the six betas of 36 tanks each, produced weapons-grade U-235 using feed from the modified alpha calutrons, the small output from the gaseous diffusion plant, and whatever the new thermal process had to offer. Virtually all the U-235 sent by courier on the train to Chicago and on to Los Alamos had passed through the beta calutrons. From these shipments Oppenheimer's physicists assembled the bomb that was to destroy Hiroshima.