Atomic Energy for Military Purposes (The Smyth Report)
CHAPTER X. THE SEPARATION OF THE URANIUM ISOTOPES BY GASEOUS DIFFUSION
10.1. It was in February 1940 that small amounts of concentrated fractions of the three uranium isotopes of masses 234, 235, and 238 were obtained by A. O. Nier using his mass spectrometer and were turned over to E. T. Booth, A. von Grosse, and J. R. Dunning for investigation with the Columbia University cyclotron. These men soon demonstrated that U-235 was the isotope susceptible to fission by thermal neutrons. It was natural, therefore, that this group, under the leadership of Dunning, became more interested than ever in the large-scale separation of the uranium isotopes.
10.2. The diffusion method was apparently first seriously reviewed by Dumling in a memorandum to G. B. Pegram, which was sent to L. J. Briggs in the fall of 1940. This memorandum summarized preliminary investigations that had been carried on by E. T. Booth, A. von Grosse and J. R. Dunning. Work was accelerated in 1941 with financial help provided by a contract that H. C. Urey had received from the Navy for the study of isotope separation - principally by the centrifuge method. During this period F. G. Slack of Vanderbilt University and W. F. Libby of the University of California joined the group. An OSRD contract (OEMsr-106) calling specifically for diffusion studies went into effect on July 1 , 1941 , and ran for a year. The work continued on an expanding scale under a series of OSRD and Army contracts through the spring of 1945. Up until May 1943 Dunning was in immediate charge of this work; Urey was in charge of statistical methods in general. From that time until February 1945 Urey was in direct charge of the Columbia part of the diffusion work, with Dunning continuing as director of one of the principal divisions. On March 1, 1945, the laboratory was taken over from Columbia by Carbide and Carbon chemicals Corporation. Early in 1942, at the suggestion of E. V. Murphree the M. W. Kellogg Company was brought in to develop plans for large-scale production of diffusion-plant equipment and eventually to build a full-scale plant. To carry out this undertaking, a new subsidiary company was formed called the Kellex Corporation. In January 1943, Carbide and Carbon Chemicals Corporation was given the responsibility for operating the plant.
10.3. As stated in Chapter IV, by the end of 1941 the possibility of separating the uranium hexafluorides had been demonstrated in principle by means of a single-stage diffusion unit employing a porous barrier (for example, a barrier made by etching a thin sheet of silver-zinc alloy with hydrochloric acid). A considerable amount of work on barriers and pumps had also been done but no answer entirely satisfactory for large-scale operation had been found. Also, K. Cohen had begun a series of theoretical studies, to which reference has already been made, as to what might be the best way to use the diffusion process, i.e., as to how many stages would be required, what aggregate area of barrier would be needed, what volume of gas would have to be circulated, etc. Theoretical studies and process development by M. Benedict added much to knowledge in this field and served as the basis of design of the large plant.
10.4. Reports received from the British, and the visit by the British group in the winter of 1941-1942, clarified a number of points. At that time the British were planning a diffusion separation plant themselves so that the discussions with F. Simon, R. Peierls, and others were particularly valuable.
THE PRINCIPLES OF SEPARATION BY DIFFUSION A SINGLE DIFFUSION STAGE
10.5. As was explained in the last chapter, the rate of diffusion of a gas through an ideal porous barrier is inversely proportional to the square root of its molecular weight. Thus if a gas consisting of two isotopes starts to diffuse through a barrier into an evacuated vessels, the lighter isotope (of molecular weight M1) diffuses more rapidly than the heavier (of molecular weight M2). The result, for a short period of time, at least, is that the relative concentration of the lighter isotope is greater on the far side of the barrier than on the near side. But if the process is allowed to continue indefinitely, equilibrium will become established and the concentrations will become identical on both sides of the barrier. Even if the diffusate gas (the gas which has passed through the barrier) is drawn away by a pump, the relative amount of the heavy isotope passing through the barrier will increase since the light isotope on the near side of the barrier has been depleted by the earlier part of the diffusion.
10.6. For a single diffusion operation, the increase in the relative concentration of the light isotope in the diffused gas compared to the feed gas can be expressed in terms of the separation factor r or the enrichment factor, r-1, both defined in paragraph 9.8 of the last chapter. A rather simple equation can be derived which gives r-1 in terms of the molecular weights and the fraction of the original gas which has diffused. If this fraction is very small, the equation reduces to r = α, the "ideal separation factor" of paragraph 9.14. If the fraction diffused is appreciable, the equation shows the expected diminution in separation. For example, if half the gas diffuses, r-1 = .69(α-1), or for uranium hexafluoride r = 1.003 compared to the value of 1.0043 when a very small fraction of the original gas has diffused.
10.7. To separate the uranium isotopes, many successive diffusion stages (i.e., a cascade) must be used since a = 1.0043 for U235F6 and U238F6, a possible gas for uranium separation. Studies by Cohen and others have shown that the best flow arrangement for the successive stages is that in which half the gas pumped into each stage diffuses through the barrier, the other (impoverished) half being returned to the feed of the next lower stage. For such an arrangement, as we have seen, the ideal separating effect between the feed and output of a single stage is 0.69(α-1). This is often called ε, the "overall enrichment per stage." For the uranium hexafluorides, ε = 0.003, in theory but it is somewhat less in practice as a result of "back diffusion," of imperfect mixing on the high pressure side, and of imperfections in the barrier. The first experimental separation of the uranium hexafluorides (by E. T. Booth, H. C. Paxton, and C. B. Slade) gave results corresponding to ε = 0.0014. If one desires to produce 99 percent pure U235F6, and if one uses a cascade in which each stage has a reasonable overall enrichment factor then it turns out that roughly 4,000 stages are required.
GAS CIRCULATION IN THE CASCADE
10.8. Of the gas that passes through the barrier of any given stage, only half passes through the barrier of the next higher stage, the other half being returned to an earlier stage. Thus most of the material that eventually emerges from the cascade has been recycled many times. Calculation shows that for an actual uranium-separation plant it may be necessary to force through the barriers of the first stage 100,000 times the volume of gas that comes out the top of the cascade (i.e., as desired product U235F6). The corresponding figures for higher stages fall rapidly because of reduction in amount of unwanted material (U238F6) that is carried along.
THE PROBLEM OF LARGE-SCALE SEPARATION INTRODUCTION
10.9. By the time of the general reorganization of the atomic bomb project in December 1941, the theory of isotope separation by gaseous diffusion was well understood. Consequently it was possible to define the technical problems that would be encountered in building a large-scale separation plant. The decisions as to scale and location of such plant were not made until the winter of 1942-1943, that is, about the same time as the corresponding decisions were being made for the plutonium production plants.
10.10. The general objective of the large-scale gaseous diffusion plant was the production each day of a specified number of grams of uranium containing of the order of ten times as much U-235 as is present in the same quantity of natural uranium. However, it was apparent that the plant would be rather flexible in operation, and that considerable variations might be made in the degree of enrichment and yield of the final product.
THE PROCESS GAS
10.11. Uranium hexafluoride has been mentioned as a gas that might be suitable for use in the plant as "process gas"; not the least of its advantages is that fluorine has only one isotope so that the UF6 molecules of any given uranium isotope all have the same mass. This gas is highly reactive and is actually a solid at room temperature and atmospheric pressure. Therefore the study of other gaseous compounds of uranium was urgently undertaken. As insurance against failure in this search for alternative gases, it was necessary to continue work on uranium hexafluoride, as in devising methods for producing and circulating the gas.
THE NUMBER OF STAGES
10.12. The number of stages required in the main cascade of the plant depended only on the degree of enrichment desired and the value of overall enrichment per stage attainable with actual barriers. Estimates were made which called for several thousand stages. There was also to be a "stripping" cascade of several hundred stages, the exact number depending on how much unseparated U-235 could economically be allowed to go to waste.
10.13. We have seen that the total value of gas that must diffuse through the barriers is very large compared to the volume of the final product. The rate at which the gas diffuses through unit area of barrier depends on the pressure difference on the two sides of the barrier and on the porosity of the barrier. Even assuming full atmospheric pressure on one side and zero pressure on the other side, and using an optimistic figure for the porosity, calculations showed that many acres of barrier would be needed in the large-scale plant.
10.14. At atmospheric pressure the mean free path of a molecule is of the order of a ten-thousandth of a millimeter or one tenth of a micron. To insure true "diffusive" flow of the gas, the diameter of the myriad holes in the barrier must be less than one tenth the mean free path. Therefore the barrier material must have almost no holes which are appreciably larger than 0.01 micron (4 x 10-7 inch), but must have billions of holes of this size or smaller. These holes must not enlarge or plug up as the result of direct corrosion or dust coming from corrosion elsewhere in the system. The barrier must be able to withstand a pressure "head" of one atmosphere. It must be amenable to manufacture in large quantities and with uniform quality. By January 1942, a number of different barriers had been made on a small scale and tested for separation factor and porosity. Some were thought to be very promising, but none had been adequately tested for actual large-scale production and plant use.
PUMPING AND POWER REQUIREMENTS
10.15. In any given stage approximately half of the material entering the stage passes through the barrier and on to the next higher stage, while the other half passes back to the next lower stage. The diffused half is at low pressure and must be pumped to high pressure before feeding into the next stage. Even the undiffused portion emerges at somewhat lower pressure than it entered and cannot be fed back to the lower stage without pumping. Thus the total quantity of gas per stage (comprising twice the amount which flows through the barrier) has to be circulated by means of pumps.
10.16. Since the flow of gas through a stage varies greatly with the position of the stage in the cascade, the pumps also vary greatly in size or number from stage to stage. The type and capacity of the pump required for a given stage depends not only on the weight of gas to be moved but on the pressure rise required. Calculations made at this time assumed a fore pressure of one atmosphere and a back pressure (i.e., on the low pressure side of the barrier) of one tenth of an atmosphere. It was estimated that thousands of pumps would be needed and that thousands of kilowatts would be required for their operation. Since an unavoidable concomitant of pumping gas is heating it, it was evident that a large cooling system would have to be provided. By early 1942, a good deal of preliminary work had been done on pumps. Centrifugal pumps looked attractive in spite of the problem of sealing their shafts, but further experimental work was planned on completely sealed pumps of various types.
LEAKS AND CORROSION
10.17. It was clear that the whole circulating system comprising pumps, barriers, piping, and valves would have to be vacuum tight. If any lubricant or sealing medium is needed in the pumps, it should not react with the process gas. In fact none of the materials in the system should react with the process gas since such corrosion would lead not only to plugging of the barriers and various mechanical failures but also to absorption (i.e., virtual disappearance) of uranium which had already been partially enriched.
ACTUAL VS. IDEAL CASCADE
10.18. In an ideal cascade, the pumping requirements change from stage to stage. In practice it is not economical to provide a different type of pump for every stage. It is necessary to determine how great a departure from the ideal cascade (i.e., what minimum number of pump types) should be employed in the interest of economy of design, repair, etc. Similar compromises are used for other components of the cascade.
HOLD-UP AND START-UP TIME
10.19. When first started, the plant must be allowed to run undisturbed for some time, until enough separation has been effected so that each stage contains gas of appropriate enrichment. Only after such stabilization is attained is it desirable to draw off from the top stage any of the desired product. Both the amount of material involved (the hold-up) and the time required (the start-up time) are great enough to constitute major problems in their own right.
10.20. It was apparent that there would be only three types of material loss in the plant contemplated, namely: loss by leakage, loss by corrosion (i.e., chemical combination and deposition), and loss in plant waste. It was expected that leakage could be kept very small and that - after an initial period of operation - loss from corrosion would be small. The percentage of material lost in plant waste would depend on the number of stripping stages.
10.21. Questions as to how the barrier material was to be used (whether in tubes or sheets, in large units or small units), how mixing was to be effected, and what controls and instruments would be required were still to be decided. There was little reason to expect them to be unanswerable, but there was no doubt that they would require both theoretical and experimental study.
SUMMARY OF THE PROBLEM
10.22. By 1942 the theory of isotope separation by gaseous diffusion had been well worked out, and it became clear that a very large plant would be required. The major equipment items in this plant were diffusion barriers and pumps. Neither the barriers nor the pumps which were available at that time had been proved generally adequate. Therefore the further development of pumps and barriers was especially urgent. There were also other technical problems to be solved, these involving corrosion, vacuum seals, and instrumentation.
10.23. As we mentioned at the beginning of this chapter, the diffusion work was initiated by J. R Dunning. The work was carried on under OSRD auspices at Columbia University until May 1,1943, when it was taken over by the Manhattan District. In the summer of 1943 the difficulties encountered in solving certain phases of the project led to a considerable expansion, particularly of the chemical group. H. C. Urey, then director of the work, appointed H. S. Taylor of Princeton associate director and added E. Mack, Jr. of Ohio State, G. M. Murphy of Yale, and P. H. Emmett of Johns Hopkins to the senior staff. Most of the work was moved out of the Columbia laboratories to a large building situated near by. The chemists at Princeton who had been engaged in heavy water studies were assigned some of the barrier research problems. Early in 1944, L. M. Currie of the National Carbon Company became another associate director to help Urey in his liaison and administrative work.
10.24. As has been mentioned, the M. W. Kellogg Company was chosen early in 1942 to plan the large scale plant. For this purpose Kellogg created a special subsidiary called The Kellex Corporation, with P. C. Keith as executive in charge and technical head and, responsible to him, A. L. Baker as Project Manager, and J. H. Arnold as Director of Research and Development. The new subsidiary carried on research and development in its Jersey City laboratories and in the laboratory building referred to in the paragraph above; developed the process and engineering designs; and procured materials for the large-scale plant and supervised its construction. The plant was constructed by the J. A. Jones Construction Company, Incorporated, of Charlotte, North Carolina.
10.24-a. The Kellex Corporation, unlike conventional industrial firms, was a cooperative of scientists, engineers and administrators recruited from essentially all branches of industry and gathered for the express purpose of carrying forward this one job. Service was on a voluntary basis, individuals prominent in industry freely relinquishing their normal duties and responsibilities to devote full time to Kellex activities. As their respective tasks are being completed these men are returning to their former positions in industry.
10.25. In January 1943, Carbide and Carbon Chemicals Corporation were chosen to be the operators of the completed plant. Their engineers soon began to play a large role not only in the planning and construction but also in the research work.
RESEARCH, DEVELOPMENT, CONSTRUCTION, AND PRODUCTION, 1942 TO 1945
PRODUCTION OF BARRIERS
10.26. Even before 1942, barriers had been developed that were thought to be satisfactory. However, the barriers first developed by E. T. Booth, H. C. Paxton, and C. B. Sladewere never used on a large scale because of low mechanical strength and poor corrosion resistance. In 1942, under the general supervision of Booth and F. G. Slack and with the cooperation of various scientists including F. C. Nix of the Bell Telephone Laboratories, barriers of a different type were produced. At one time, a barrier developed by E. O. Norris and E. Adier was thought sufficiently satisfactory to be specified for plant use. Other barriers were developed by combining the ideas of several men at the Columbia laboratories (by now christened the SAM Laboratories), Kellex, Bell Telephone Laboratories, Bakelite Corporation, Houdaille-Hershey Corporation, and others. The type of barrier selected for use in the plant was perfected under the general supervision of H. S. Taylor. One modification of this barrier developed by the SAM Laboratories represented a marked improvement in quality and is being used in a large number of stages of the plant. By 1945 the problem was no longer one of barely meeting minimum specifications, but of making improvements resulting in greater rate of output or greater economy of operation.
10.27. Altogether the history of barrier development reminds the writer of the history of the "canning" problem of the plutonium project. In each case the methods were largely cut and dry, and satisfactory or nearly satisfactory solutions were repeatedly announced; but in each case a really satisfactory solution was not found until the last minute and then proved to be far better than had been hoped.
PUMPS AND SEALS
10.28. The early work on pumps was largely under the supervision of H. A. Boorse of Columbia University. When Kellex came into the picture in 1942, its engineers, notably G. W. Watts, J. S. Swearingen and O. C. Brewster, took leading positions in the development of pumps and seals. It must be remembered that these pumps are to be operated under reduced pressure, must not leak, must not corrode, and must have as small a volume as possible. Many different types of centrifugal blower pumps and reciprocating pumps were tried. In one of the pumps for the larger stages, the impeller is driven through a coupling containing a very novel and ingenious type of seal. Another type of pump is completely enclosed, its centrifugal impeller and rotor being run from outside, by induction.
10.29. As in the plutonium problem, so here also, there were many questions of corrosion, etc., to be investigated. New coolants and lubricants were developed by A. L. Henne and his associates, by G. H. Cady, by W. T. Miller and his co-workers, by E. T. McBee and his associates, and by scientists of various corporations including Hooker Electrochemical Co., the du Pont Co. and the Harshaw Chemical Co. The research and development and plant requirements for these materials and other special chemicals were coordinated by R. Rosen, first under OSRD and later for Kellex. Methods of pretreating surfaces against corrosion were worked out. Among the various instruments designed or adapted for project use, the mass spectrograph deserves special mention. The project was fortunate in having the assistance of A. O. Nier of the University of Minnesota and later of Kellex whose mass spectrograph methods of isotope analysis were sufficiently advanced to become of great value to the project, as in analyzing samples of enriched uranium. Mass spectrographs were also used in pretesting parts for vacuum leaks and for detecting impurities in the process gas in the plant.
10.30. Strictly speaking, there was no pilot plant. That is to say, there was no small-scale separation system set up using the identical types of blowers, barriers, barrier mountings, cooling, etc., that were put into the main plant. Such a system could not be set up because the various elements of the plant were not all available prior to the construction of the plant itself. To proceed with the construction of the full-scale plant under these circumstances required foresight and boldness.
10.31. There was, however, a whole series of so-called pilot plants which served to test various components or groups of components of the final plant. Pilot plant No. 1 was a 12-stage plant using a type of barrier rather like that used in the large scale plant, but the barrier material was not fabricated in the form specified for the plant and the pumps used were sylphon- sealed reciprocating pumps, not centrifugal pumps. Work on this plant in 1943 tested not only the barriers and general system of separation but gave information about control valves, pressure gauges, piping, etc. Pilot plant No. 2, a larger edition of No. 1 but with only six stages, was used in late 1943 and early 1944, particularly as a testing unit for instruments. Pilot plant No. 3a, using centrifugal blowers and dummy diffusers, was also intended chiefly for testing instruments. Pilot plant No. 3b was a real pilot plant for one particular section of the large-scale plant. Pilot plants using full-scale equipment at the plant site demonstrated the vacuum tightness, corrosion resistance and general operability of the equipment.
10.32. In December 1942, the Kellogg Company was authorized to proceed with preliminary plant design and in January 1943 the construction of a plant was authorized.
10.33. As stated in an earlier chapter, a site in the Tennessee Valley had originally been chosen for all the Manhattan District plants, but the plutonium plant was actually constructed elsewhere. There remained the plutonium pilot plant already described, the gaseous diffusion plant, the electromagnetic separation plant (see Chapter XI), and later the thermal diffusion plant which were all built in the Tennessee Valley at the Clinton site, known officially as the Clinton Engineer Works.
10.34. This site was examined by Colonel Marshall, Colonel Nichols, and representatives of Stone and Webster Engineering Corporation in July 1942, and its acquisition was recommended. This recommendation was endorsed by the OSRD S-1 Executive Committee at a meeting in July 1942. Final approval was given by Major General L. R. Groves after personal inspection of the 70-square-mile site. In September 1942, the first steps were taken to acquire the tract, which is on the Clinch River about thirty miles from Knoxville, Tennessee, and eventually considerably exceeded 70 square miles. The plutonium pilot plant is located in one valley, the electromagnetic separation plant in an adjoining one, and the diffusion separation plant in a third.
10.35. Although the plant and site development at Hanford is very impressive, it is all under one company dealing with bur one general operation so that it is in some respects less interesting than Clinton, which has a great multiplicity of activity. To describe the Clinton site, with its great array of new plants, its new residential districts, new theatres, new school system, seas of mud, clouds of dust, and general turmoil is outside the scope of this report.
DATES OF START OE CONSTRUCTION
10.36. Construction of the steam power plant for the diffusion plant began on June 1, 1943. It is one of the largest such power plants ever built. Construction of other major buildings and plants started between August 29, 1943 and September 10, 1943.
10.37. Unlike Hanford, the diffusion plant consists of so many more or less independent units that it was put into operation section by section, as permitted by progress in constructing and testing. Thus there was no dramatic start-up date nor any untoward incident to mark it. The plant was in successful operation before the summer of 1945.
10.38. For the men working on gaseous diffusion it was a long pull from 1940 to 1945, not lightened by such exciting half-way marks as the first chain-reacting pile at Chicago. Perhaps more than any other group in the project, those who have worked on gaseous diffusion deserve credit for courage and persistence as well as scientific and technical ability. For security reasons, we have not been able to tell how they solved their problems - even in many cases found several solutions, as insurance against failure in the plant. It has been a notable achievement. In these five years there have been periods of discouragement and pessimism. They are largely forgotten now that the plant is not only operating but operating consistently, reliably, and with a performance better than had been anticipated.
10.39. Work at Columbia University on the separation of isotopes by gaseous diffusion began in 1940, and by the end of 1942 the problems of large-scale separation of uranium by this method had been well defined. Since the amount of separation that could be effected by a single stage was very small, several thousand successive stages were required. It was found that the best method of connecting the many stages required extensive recycling so that thousands of times as much material would pass through the barriers of the lower stages as would ultimately appear as product from the highest stage.
10.40. The principal problems were the development of satisfactory barriers and pumps. Acres of barrier and thousands of pumps were required. The obvious process gas was uranium hexafluoride for which the production and handling difficulties were so great that a search for an alternative was undertaken. Since much of the separation was to be carried out at low pressure, problems of vacuum technique arose, and on a previously unheard-of scale. Many problems of instrumentation and control were solved; extensive use was made of various forms of mass spectrograph.
10.41. The research was carried out principally at Columbia under Dunning and Urey. In 1942, the M. W. Kellogg Company was chosen to develop the process and equipment and to design the plant and set up the Kellex Corporation for the purpose. The plant was built by the J. A. Jones Construction Company. The Carbide and Carbon Chemicals Corporation was selected as operating company.
10.42. A very satisfactory barrier was developed although the final choice of barrier type was not made until the construction of the plant was well under way at Clinton Engineer Works in Tennessee. Two types of centrifugal blower were developed to the point where they could take care of the pumping requirements. The plant was put into successful operation before the summer of 1945.