On October 5, 1942, General Groves paid his first visit to the Metallurgical Laboratory at the University of Chicago, as he explains in Now It Can Be Told: The Story of the Manhattan Project, where he met with Arthur Compton and “about fifteen of his senior men”:
Among them were two other Nobel Prize winners, Enrico Fermi and James Franck, together with the brilliant Hungarian physicists Eugene Wigner and Leo Szilard, and Dr. Norman Hilberry, Compton’s assistant.
[…]
With respect to the amount of fissionable material needed for each bomb, how accurate did they think their estimate was? I expected a reply of “within twenty-five or fifty per cent,” and would not have been greatly surprised at an even greater percentage, but I was horrified when they quite blandly replied that they thought it was correct within a factor of ten.
[…]
My position could well be compared with that of a caterer who is told he must be prepared to serve anywhere between ten and a thousand guests. But after extensive discussion of this point, I concluded that it simply was not possible then to arrive at a more precise answer.
[…]
This uncertainty surrounding the amount of material needed for a bomb plagued us continuously until shortly before the explosion of the Alamogordo test bomb on July 16, 1945. Even after that we could not be sure that Uranium-235 (used in the Hiroshima bomb) would have the same characteristics as plutonium (used in the test and later against Nagasaki), although we knew of no reason why it should be greatly different.
[…]
After the meeting, Compton and I resumed a discussion we had begun earlier with Szilard on how to reduce the number of approaches which were being explored for cooling the pile. Four methods—using helium, air, water and heavy water—were under active study. It was essential that we concentrate on the most promising and more or less abandon work on the others. By the end of the afternoon we settled on helium cooling. But within three months this decision was changed. The design problems early encountered in the comparatively small air-cooled reactor at Clinton indicated that the handling of any gaseous coolant in the much larger Hanford reactors would be very difficult. And as the operation of the Fermi test pite in December had proved that in a properly designed uranium pile water could be used as a coolant, it was adopted for the plutonium reactors we built at Hanford.
I left Chicago feeling that the plutonium process seemed to offer us the greatest chances for success in producing bomb material. Every other process then under consideration depended upon the physical separation of materials having almost infinitesimal differences in their physical properties. Under such circumstances, the design and operation of any industrial processes to accomplish this separation would involve unprecedented difficulties. It was true that the transmutation of uranium by spontaneous chain reaction into usable quantities of plutonium fell entirely outside of existing technical knowledge; yet the rest of the process—the chemical separation of the plutonium from the rest of the material—while extremely difficult and completely unprecedented, did not seem to be impossible.
Up until this time, only infinitesimal quantities of plutonium had been produced, and these by means of the cyclotron, a laboratory method not suitable for production in quantity. And by quantity production of plutonium, I do not mean tons per hour, but rather a few thimblefuls per day. Even by December, 1943, only two milligrams had been produced.
[…]
This was in accord with the general philosophy I had followed throughout the military construction program and to which we adhered consistently in this project; namely, that nothing would be more fatal to success than to try to arrive at a perfect plan before taking any important step.