dirty my hands like a painter's assistant.’” When Szilard announced that he had hired a stand-in, a young man whom Anderson remembers as “very competent,” Fermi acceded to the arrangement without comment. But he never again pursued an experiment jointly with Szilard.
The arrangement as finally consummated looked like this:
Szilard's Ra + Be source stands in the center of the tank, which holds 143 gallons of manganese solution; the fifty-two cans of UO2 gather around.
It worked. The three physicists found neutron activity “about ten percent higher with uranium oxide than without it. This result shows that in our arrangement more neutrons are emitted by uranium than are absorbed by uranium.” But the experiment raised puzzling questions. Resonance absorption, for example, was clearly a problem, capturing neutrons that might otherwise serve the chain reaction. The report estimates “an average emission [of secondary neutrons] of about 1.2 neutrons per thermal neutron” but notes that “this number should be increased, to perhaps 1.5,” because some of the neutrons had obviously been captured without fissioning — demonstrating the big capture resonance around 25 eV that Bohr had attributed on his graphs to U238.
Another problem was the use of water as a moderator. As Fermi's team had discovered in Rome in 1934, hydrogen was more efficient than any other element at slowing down neutrons, and slow neutrons avoided the parasitic capture resonance of U238. But hydrogen itself also absorbed some slow neutrons, reducing further the number available for fission. And it was already clear that every possible secondary neutron would have to be husbanded carefully if a chain reaction was to be initiated in natural uranium. George Placzek came down from Cornell, where he had found a new home, for a visit, looked over the arrangement and insightfully foreclosed its future. As Szilard tells it:
We were inclined to conclude that… the water-uranium system would sustain a chain reaction… Placzek said that our conclusion was wrong because in order to make a chain reaction go, we would have to eliminate the absorption of [neutrons by the] water; that is, we would have to reduce the amount of water in the system, and if we reduced the water in the system, we would increase the parasitic absorption of [neutrons by] uranium [because with less water fewer neutrons would be slowed]. He recommended that we abandon the water-uranium system and use helium for slowing down the neutrons. To Fermi this sounded funny, and Fermi referred to helium thereafter invariably as Placzek's helium.
In June the Columbia team wrote up its experiment and sent the resulting paper, “Neutron production and absorption in uranium,” to the
Szilard remained in the steamy city: “I was left alone in New York. I still had no position at Columbia; my three months [of laboratory privileges] were up, but there were no experiments going on anyway and all I had to do was to think.”
Szilard thought first about an alternative to water. The next common material up the periodic table that might work — that had a capture cross section considerably smaller than hydrogen's, that was cheap, that would be thermally and chemically stable — was carbon. The mineral form of carbon, chemically identical to diamond but the product of a different structure of crystallization, is graphite, a black, greasy, opaque, lustrous material that is the essential component of pencil lead. Although carbon slows neutrons much less rapidly than hydrogen, even that difference might be put to advantage by careful design.
Lewis Strauss was leaving for Europe the week of July 2. Hoping that the financier might coax support for uranium research from Belgium's Union Miniere, Szilard sent Strauss a last-minute letter arguing that a chain reaction in uranium “is an immediate possibility” but chose not to mention his new uranium-graphite conception. Apparently he wanted to discuss it first with Fermi; the same day, July 3, he wrote the Italian laureate at length. “It seems to me now,” he reported, “that there is a good chance that carbon might be an excellent element to use in place of hydrogen, and there is a strong temptation to gamble on this chance.” He wanted to try “a large-scale experiment with a carbon-uranium-oxide mixture” as soon as they could acquire enough material. In the meantime he thought he would set up a small experiment to measure more accurately carbon's capture cross section, only the upper limit of which was then known. If carbon should prove unsuitable their “next best guess might be heavy water,” rich in deuterium, though they would need “a few tons” of that scarce and expensive liquid. (Deuterium, H2, has a much smaller cross section for neutron capture than ordinary hydrogen.)
Across the one hundred sixty-third anniversary of the Declaration of Independence Szilard's ideas evolved rapidly. On July 5 he visited the National Carbon Company of New York to look into purchasing graphite blocks of high purity (because impurities such as boron with large capture cross sections would soak up too many neutrons). He wrote Fermi his finding the same day: “It seems that it will be possible to get sufficiently pure carbon at a reasonable price.” He also mentioned arranging the uranium and carbon in layers.
Fermi sat down in Ann Arbor at the end of the week to respond to Szilard's first report. Independently he had arrived at a similar plan:
Thank you for your letter. I was also considering the possibility of using carbon for slowing down the neutrons… According to my estimates a possible recipe might be about 39,000 kg of carbon mixed with 600 kg of uranium. If it were really so the amounts of materials would certainly not be too large.
Since however the amount of uranium that can be used, especially in a homogeneous mixture, is exceedingly small..
The idea of layering or in some other way separating the uranium from the graphite originated in calculations Fermi made in June for the manganese water-tank experiment. Fermi's calculations led both men to consider partitioning the oxide from the graphite in the new design they were independently evolving. Partitioning would give the fast secondary neutrons room to slow down, bouncing around in the moderator, before they encountered any U238 nuclei. Szilard's next letter, on July 8, mentions that “the carbon and the uranium oxide would not be mixed but built up in layers, or in any case used in some canned form.” Both the July 5 and July 8 letters apparently crossed with Fermi's letter in the mail.
By the time he heard from Fermi, Szilard had seen still farther and realized that small spheres of uranium arranged within blocks of graphite would be “even more favorable from the point of view of a chain reaction than the system of plane uranium layers which was initially considered.” The arrangement Szilard had in mind he called a “lattice.” (A geodesic dome would represent such a lattice arrangement schematically if it were a complete sphere and if all its interior volume were filled like its surface with evenly spaced points.) His calculations indicated somewhat larger volumes of material than had Fermi's: “perhaps 50 tons of carbon and 5 tons of uranium.” The entire experiment, he thought, would cost about $35,000.
If a chain reaction would work in graphite and uranium, Szilard assumed, then a bomb was probable. And if he had managed these conclusions, he further assumed, then so had his counterparts in Nazi Germany. He sought out Pegram in those early July days and tried to convince him of the urgent need for a large-scale experiment to settle the question. The dean resisted the assault: “He took the position that even though the matter appeared to be rather urgent, this being summer and Fermi being away there was really nothing that usefully could be done until the fall.”
For several weeks Szilard had been trying on his own to raise funds from the U.S. military. In late May he had asked Wigner to contact the Army's Aberdeen Proving Ground, its weapons-development facility in Maryland. While he was thinking through the possibilities of a uranium-graphite system he had talked to Ross Gunn about Navy support. Now Fermi's letter of July 9 and a July 10 letter from Gunn arrived to discourage him. Fermi wrote of layering the carbon and uranium but calculated in terms of a homogeneous system — of graphite and uranium oxide crushed and mixed together. Szilard concluded he was being mocked: “I knew very well that Fermi… computed the homogeneous mixture only because it was the easiest to compute. This showed me that Fermi did not take this
