Research into reductive extraction for bismuth and rare-earths (lanthanides) made ORNL chemical engineers particularly optimistic about the application of these processes to molten-salt breeder reactors. At the same time data was coming back to reactor designers on the dimensional stability of graphite at various temperatures in fast neutrons. The data on dimensional stability was not good, and was calling into question the way that graphite was intended to be used in two-fluid reactor designs.
These pieces of information were being compounded by the pressure to develop a concept for a follow-on breeding experiment to the MSRE, and this required that the eventual MSBR have at least a conceptual design into which the MSBE could trace.
This led ORNL MSRP leadership to undertake a serious shift in the design of the MSBR that we think may have been a mistake. They decided to abandon the two-fluid design, where thorium fertile material and uranium fuel material were kept separate, and to pursue a one-fluid design where thorium and uranium were combined in a single salt. At first blush, one might think that this could lead to a profound simplification of the chemical processing system, since there would no longer be a need to chemically remove U or Pa from the blanket, allow it to decay, and chemically add it to the fuel salt. But the reality of 233Pa’s decay time and its propensity to absorb neutrons, coupled with ORNL’s desire to compete with the LMFBR as a breeder with a short doubling time, led to a more complex chemical processing system than the two-fluid design.
The chemical processing that was proposed for the one-fluid reactor is depicted in Figure 1 and in greater detail in Figure 2. There are several key differences from the chemical processing flowsheet for the two-fluid reactor, and a number of similarities.
In the one-fluid reactor, thorium, uranium, protactinium, and fission products are all mixed together in a single salt. Separation of thorium from lanthanide fission products is rather challenging because of their chemical similarities. (This is the same reason why thorium is generally found in rare-earth deposits.) In each case, protactinium is extracted from the salt so that it can decay outside of the reactor. But in the case of the one-fluid reactor the need to extract protactinium is more pronounced. This is because of the strong desire to reduce the overall reactor inventory of fissile 233U (and shorten the doubling time) the fuel salt (now containing protactinium) is exposed to a greater time-averaged fluence than is the case in the two-fluid design. In the two-fluid design a simple way to reduce the time-averaged fluence is to increase the blanket inventory, but this is not a realistic option in the one-fluid design because of the aforementioned desire to reduce the fissile inventory and doubling time.
The proposed processing scheme was detailed in the report ORNL-TM-3579, Design and Cost Study of a Fluorination-Reductive-Extraction-Metal Transfer Processing Plant for the MSBR, released in May 1972. Fuel salt was first held up for cooling and decay of the shortest lived fission products, then routed to the primary fluorinator, where most of the uranium was removed by fluorination to UF6 using gaseous molecular fluorine (F2) as the fluorination agent. The salt, now stripped of most of its uranium, was routed to an extraction column where metallic bismuth containing lithium and thorium as reductants were contacted with the salt. The remaining uranium, protactinium, and zirconium in the salt were reductively extracted to the bismuth, leaving a salt now that only contained fission products (beyond its base composition of LiF-BeF2-ThF4). This salt entered another reductive extraction column where bismuth containing lithium contacted the salt to remove lanthanide fission products and some thorium. The salt then passed to a reduction column where UF6 was reduced to UF4 in the salt, refueling it and preparing it for return to the reactor. Makeup BeF2 and ThF4 were also added and any residual bismuth was removed from the salt. After a final cleanup step and valence adjustment the purified salt was returned to the reactor.
The bismuth containing some uranium, protactinium, and zirconium was directed to a hydrofluorination column where the metallic solutes in the bismuth were oxidized into their fluoride forms in the presence of a decay salt. The decay salt, containing UF4, PaF4, ThF4, and ZrF4 passed into a decay tank where 233Pa was allowed to decay to 233U. This uranium generated by protactinium decay was removed through fluorination to UF6 and routed to the reduction column to refuel the purified fuel salt.
The bismuth that had been used to carry the proactinium, having been scrubbed of its chemical passengers in the hydrofluorination stage, was routed to the “metal transfer” stage of the processing system where it was combined with bismuth containing lanthanide fission products that had been extracted from the fuel salt. These bismuth streams contacted a salt stream of lithium chloride. Lanthanides transfer to the LiCl but thorium is left behind, accomplishing a decontamination between these two steps. The LiCl is then was successively contacted with streams of bismuth containing metallic lithium reductant which removed the divalent and trivalent lanthanides in separate columns. The bismuth stream containing trivalent lanthanides was hydrofluorinated in the presence of a salt stream that had been designated for waste. The bismuth stream containing divalent lanthanides was combined with the one emerging from the protactinium extraction column and hydrofluorinated into the decay salt. Hence, both the decay salt and the waste salt were contaminated with fission products. Decay salt was the precursor for the waste salt as it was periodically discarded every 220 days. A final fluorination step captured any decayed uranium before discard.
The fluorinators would use F2 as the reagent; the hydrofluorinators would use HF, and the reduction column would use H2. Based on the production and consumption rates, a recycling system for these reagents was proposed. An electrolytic cell would split HF into F2 and H2, which would then be used in the fluorinators and reduction column, respectively. HF emerging from the reduction column would be used in the hydrofluorinators or routed to the electrolytic fluorine cell for production of F2 and H2. Mixed streams of HF and H2 would be separated in an HF distillation system. HF would be sent to the electrolytic cell while H2 would be cleaned up in a caustic scrubber using potassium hydroxide (KOH) in order to capture any residual fluorides. The H2 stream would be recycled to the system but a small amount (5%) would be directed to an alumina absorber, where any fission products like selenium hexafluoride or tellurium hexafluoride would be trapped. The hydrogen stream would also pass through a charcoal absorber to capture noble gases like krypton and xenon before being released up the stack.
Overall, the chemical processing system required for the one-fluid reactor was substantially more challenging than that required for the two-fluid reactor. The fundamental reason for this challenge is the chemical similarity between thorium and the lanthanide fission products, but it was also compounded by the need to extract protactinium rapidly and its connection to fissile inventory.
It should be noted that all of these challenges applied to the goal of a short-doubling-time molten-salt breeder reactor, which was the challenge ORNL faced in the 1960s as these design concepts were being evaluated. If the reactor wasn’t attempting to achieve a high breeding gain, or if it was perhaps not even a breeder at all, but an enriched-uranium-fueled reactor, then many of these challenges might not apply and there would be the potential for tremendous simplification of the chemical processing system. It is important to view these chemical processing systems in the context of the design objectives they were attempting to achieve, which were ambitious then and remain ambitious now.
The capital costs for the chemical processing system of the one-fluid MSBR were given in Table 10 of ORNL-TM-3579 as $35.6M in 1970 dollars. This was based on a 1000-MWe reactor, a reactor fuel volume of 1683 ft3, and a processing cycle time of 10 days. This cost correlates to a cost of $214M in 2013 dollars, or $0.214/watt installed.
Continuous Fluorination Experimental Development
During the time that the one-fluid breeder reactor was the reference design, progress continued to be made in the development of continuous fluorinators, which retained an important position in fuel salt processing. Experimental studies of fluorination of molten salt were carried out in a 1-in.-diam., 72-in.-long nickel fluorinator that allowed countercurrent contact of molten salt with fluorine. In these tests, molten salt (41-24-35 mole% NaF-LiF-ZrF4) containing UF4, was countercurrently contacted with a quantity of fluorine in excess of that required for the conversion of UF4 to UF6. Experiments were carried out with temperatures ranging from 525 to 600°C, UF4 concentrations in the feed salt ranging from 0.12 to 0.35 mole%, and a range of salt and fluorine feed rates. The fraction of the uranium removed from the salt ranged from 97.5% to 99.9%.
Axial dispersion in the salt phase was anticipated to be important in the design of continuous fluorinators, and gas holdup and axial dispersion were measured in columns having diameters ranging from 1 to 6 in. using air and aqueous solutions. Data were obtained for wide ranges of viscosity, surface tension, and superficial gas velocity. Correlations for gas holdup and axial dispersion were developed which were believed to be applicable to countercurrent contact of molten salt and fluorine in a continuous fluorinator. These correlations and the data on uranium removal in the 1-in.-diam continuous fluorinator were used for estimating the performance of larger diameter continuous fluorinators.
The combination of molten salt and fluorine results in a highly corrosive environment, and a future continuous fluorinator will need to protect against corrosion by maintaining a layer of frozen salt on surfaces that would otherwise contact both molten salt and fluorine, preventing molten salt from reaching the surface will allow passivation of the nickel to occur.
The feasibility of maintaining frozen salt layers in gas-salt contactors was demonstrated in tests in a 5-in.-diam, 8-ft-high simulated fluorinator in which molten salt (66-34 mole% LiF-ZrF4) and argon were countercurrently contacted. An internal heat source in the molten region was provided by Calrod heaters contained in a 3/4-in.-diam pipe along the center line of the vessel. A frozen salt layer was maintained in the system with equivalent volumetric heat generation rates of 10 to 55 kW/ft3. For comparison, the heat generation rates in fuel salt immediately after removal from the reactor and after passing through vessels having holdup times of 5 and 30 min are 57, 27, and 12 kW/ft3, respectively.
Operation of a continuous fluorinator with nonradioactive salt required a means for generating heat in the molten salt that was not subject to corrosion. Radio-frequency induction heating in fluorinator simulations was studied using nitric acid as was autoresistance heating using 60-Hz power with molten salt (65-35 mole% LiF-BeF2) in a 6-in.diam fluorinator simulator. Successful operation with auto-resistance heating rates as high as 14.5 kW/ft3 was carried out; the expected power density in processing plant fluorinators is 12 kW/ft3. Autoresistance heating was the preferred method, since it could be used over a wider range of operating conditions and since the electrical power supply is much simpler than that required for induction heating.
Reductive Extraction Experimental Development
Reductive extraction, which was considered as a protactinium removal technique while the two-fluid reactor was the reference design, assumed a much larger role in the one-fluid design. Consequently, ORNL researchers operated a salt-bismuth reductive extraction facility in which uranium and zirconium were extracted from salt by countercurrent contact with bismuth containing reductant. More than 95% of the uranium was extracted from the salt by a 0.82-in.-diam, 24-in.-long packed column. The inlet uranium concentration in the salt was about 25% of the uranium concentration in their one-fluid reference MSBR. These experiments represented the first demonstration of reductive extraction of uranium in a flowing system. Information on the rate of mass transfer of uranium and zirconium was also been obtained in the system using an isotopic dilution method, and HTU values of about 4.5 ft were obtained.
Correlations were developed for flooding and dispersed-phase holdup in packed columns during countercurrent flow of liquids having high densities and a large difference in density, such as salt and bismuth. These correlations, which were verified by studies with molten salt and bismuth, were developed by study of countercurrent flow of mercury and water or high-density organics and water in 1- and 2-in.-diam. columns packed with solid cylinders and Raschig rings varying in size from 1/8 to 1/2 in. Data was also obtained on axial dispersion in the continuous phase during the countercurrent flow of high-density liquids in packed columns, and a simple relation was developed for predicting the effects of axial dispersion on column performance.
The successful operation of salt-metal extraction columns was dependent upon the availability of a bismuth-salt interface detector. To this end, a successful demonstration was made of an eddy-current-type interface detector that consists of a ceramic form on which bifilar primary and secondary coils are wound. Contact of the coils with molten salt or bismuth was prevented by enclosing the element in a molybdenum tube. Passage of a high-frequency alternating current through the primary coil induced a current in the secondary coil whose magnitude was dependent on the conductivities of the adjacent materials; since the conductivities of bismuth and salt are quite different, the induced current reflected the presence or absence of bismuth. The detector appeared to be a practical and sensitive indicator of either salt-bismuth interface location or bismuth level.
Design and development work was initiated on a reductive extraction process facility that would allow operation of the important steps for the reductive extraction process for protactinium isolation. The facility would have allowed countercurrent contact of salt and bismuth streams in a 2-in.-diam., 6-ft-long packed column at flow rates as high as about 25% of those required for processing a 1000-MWe MSBR.
Metal Transfer Process Experimental Development
All aspects of the metal transfer process for the removal of rare earths were demonstrated in an engineering experiment. The equipment consisted of a 6-in.-diam compartmented vessel in which were present about 1 liter each of MSBR fuel carrier salt, bismuth saturated with thorium, and LiCl. The fluoride salt initially contained 147NdF3 at the tracer level and LaF3 at a concentration of 0.04 mole fraction. During the experiment, the rare earths were selectively extracted into the LiCl along with a negligible amount of thorium. Provision was made for circulating the LiCl through a chamber containing bismuth having a lithium concentration of 38 at.%, where the rare earths and thorium were removed. The distribution ratios for the rare earths remained constant during the experiment at about the expected values. About 50% of the neodymium and about 70% of the lanthanum were collected in the Li-Bi solution. The final thorium concentration in the Li-Bi solution was below 5 ppm, making the ratio of rare earths to thorium in the Li-Bi greater than 105 times the initial concentration ratio in the fuel salt and thus demonstrating the selective removal of rare earths from a fluoride salt containing thorium.
A larger metal transfer experiment was put into operation that used salt and bismuth flow rates that are about 1% of the values required for processing a 1000-MWe MSBR, and the preliminary design was carried out for an experiment that would have used a three-stage salt-metal contactor and flow rates that are 5 to 10% of those required for a 1000-MWe MSBR.
Fuel Reconstitution Experimental Development
To reconstitute the fuel salt, UF6 would be directly absorbed in MSBR fuel carrier salt containing UF4, resulting in the formation of soluble non-volatile UF5. Gaseous hydrogen reacts with dissolved UF5 reducing it to UF4
Since both UF6 and UF5 are strong oxidants, experiments were conducted primarily to find a material that was inert to these species. They showed that, at 600°C, nickel, copper, and graphite are not sufficiently inert but that gold is stable both to gaseous UF6 and to salt containing up to 6 wt % UF5. Consequently, subsequent studies were conducted in gold apparatus.
Results from several experiments showed that UF5 dissolved in molten salt slowly disproportionates to UF6 and UF4 and that the rate of disproportionation is second order with respect to the concentration of UF5. The studies also indicate that the solubility of UF6 in the salt is low.
Removal of Bismuth from Fuel Salt
In a processing plant, the fuel salt would be contacted with bismuth containing reductant in order to remove protactinium and the rare earths. It would be necessary that entrained or dissolved bismuth be removed from the salt before it is returned to the reactor, since nickel is quite soluble in bismuth (about 10 wt %) at the reactor operating temperature. Efforts to measure the solubility of bismuth in salt have indicated that the solubility is lower than about 1 ppm, and the expected solubility of bismuth in the salt under the highly reducing conditions that will be used is very low. For these reasons, bismuth can only be present at significant concentrations in the salt as entrained metallic bismuth.
In order to characterize the bismuth concentration likely to be present in the salt after it is contacted with bismuth, ORNL periodically sampled the salt in engineering experiments involving contact of salt and bismuth. The results indicated that the bismuth concentration in the salt in most cases ranged from 10 to 100 ppm after countercurrent contact of the salt and bismuth in a packed-column contactor; however, concentrations below 1 ppm were observed in salt leaving a stirred-interface salt-metal contactor in which the salt and metal phases are not dispersed. One of the difficulties was that of preventing contamination of the samples with small quantities of bismuth during cleaning of the samples and the ensuing chemical analyses.
It was expected that contact of the salt with nickel wool would be effective in removing entrained or dissolved bismuth, since a large nickel surface area can be produced in this manner.
A natural circulation loop constructed of Hastelloy N and filled with fuel salt was operated by the Metals and Ceramics Division for about two years; a molybdenum cup containing bismuth was placed near the bottom of the loop. Reported concentrations of bismuth in salt from the loop (<5 ppm) were essentially the same as those reported for salt from a loop containing no bismuth. No degradation of metallurgical properties for corrosion specimens removed from the loop containing bismuth was noted.