, or flibe, is the primary candidate coolant for the fluoride-salt-cooled high-temperature nuclear reactor (FHR). Kilogram quantities of pure flibe are required for repeatable corrosion tests of modern reactor materials. This paper details fluoride salt purification by the hydrofluorination–hydrogen process, which was used to regenerate 57.4 kg of flibe originating from the secondary loop of the molten salt reactor experiment (MSRE) at Oak Ridge National Laboratory (ORNL). Additionally, it expounds upon necessary handling precautions required to produce high-quality flibe and includes technological advancements which ease the purification and analysis process. Flibe batches produced at the University of Wisconsin are the largest since the MSRE program, enabling new corrosion, radiation, and thermal hydraulic testing around the United States.
Fifty years ago, the fluid-fueled molten salt reactor experiment (MSRE) went critical using a fluoride salt mixture of (61–29.1–5–0.9 mol%) . The mixture, referred to as fuel salt, has a melting point of 450°C and circulated through the reactor at temperatures up to 654°C at nearly atmospheric pressure . The combination of high-temperature operation and passive safety features owed to the fluid-fuel are still highly unique. The MSRE operated for over 15,000 h and was regarded as a resounding success . However, its technology was never expounded upon due to the lack of nationwide support for its successor, the molten salt breeder reactor .
In order for the next generation of nuclear reactors to be economically and politically competitive with hydrocarbon energy sources, they must be highly thermally efficient, passively safe, and produce minimal waste. Fluoride-salt-based reactors have the potential to satisfy all of these constraints [3,4]. One of these next-generation designs is the fluoride salt cooled high temperature nuclear reactor (FHR). It plans to employ solid TRISO fuel, rather than the dissolved uranium fluoride used by the MSRE, to achieve a core outlet temperature of up to 700°C. The desired high outlet temperatures require the use of materials that have not yet been fully evaluated for their compatibility with molten fluoride salts. To support the development of the FHR, these new materials must be tested. The salt used for these tests must be prototypical of what would be used in the reactor and therefore chemically purified, a necessity discovered during the MSRE.
The backbone of the MSRE and its predecessor, the aircraft reactor experiment, was a fluoride salt production facility . From 1956 to 1971, the facility produced and purified over 59,874 kg of fluoride salt mixtures between two processing units that had a capacity of around () each . Around 12,020 kg of salt was produced at this facility for the MSRE. Out of this total, 5172 kg was fuel salt for the primary loop and 6886 kg was coolant salt. The coolant salt, also known as , , (66–34 mol%), or flibe, served as both a “flush salt” to clean the primary loop before inserting fuel salt and as the coolant for the secondary loop . The facility stopped production after the end of the MSRE. Since the MSRE, at least one company capable of producing research-grade salt, run by former MSRE staff, has come and gone. With each MSRE anniversary, the pool of those with the first-hand experience needed to produce fluoride salts decreases. Fortunately, excellent documentation of the purification procedures, such as “Preparation and Handling of Salt Mixtures for the Molten Salt Reactor Experiment,” has been preserved . However, these have become slightly dated over time. With the new influx of interest in molten halide salts in both nuclear and solar thermal fields, it is worth revisiting the importance of salt quality as discovered by those at ORNL. This paper aims to include more details on the necessary requirements to safely produce high-quality salt, while introducing the batch purifier at the University of Wisconsin–Madison.
Another way of understanding Eqs. (1)–(3) is that as they proceed to the right, and they oxidize the redox potential of the salt. The redox potential is an intrinsic property of all electrically conductive solutions, such as ionic molten salts, which indicates the tendency for that solution and all dissolved constituents to undergo an oxidation or reduction reaction. The redox potential is solely determined by that solution’s chemical composition. Changing the redox potential will change the relative equilibrium of all reactions, including corrosion reactions, in the salt. In the case of Eqs. (1) and (2), the introduction of oxides and hydroxides—many of which are soluble in fluoride salts—creates a more oxidizing redox potential, and therefore results in a higher corrosion rate [5–8]. Hydrogen fluoride produced in Eq. (3) can either dissolve in the salt or react with the container material to create a soluble metal fluoride; both routes lead to oxidation of the redox potential.
Hydrolysis of fluoride salts can be reversed by reaction with hydrogen fluoride to produce water vapor at liquidus temperatures, as shown in Eqs. (1) and (2) . However, Eq. (3) indicates that hydrogen fluoride will oxidize the redox potential of the salt, fluorinating the purification vessel. To prevent this, hydrogen, a relatively weak reduction agent, can be added to the hydrogen fluoride to reverse Eq. (3) . By mixing a reduction agent with a fluorinating agent and sparging it through the salt, the redox potential can be set to a suitable level to allow removal of oxides, while minimizing the equilibrium concentration of dissolved metal fluorides in the salt. This is the only way to preserve the integrity of the purification vessel. Nickel is commonly used as a purification vessel material with ratios of hydrogen fluoride to hydrogen, allowing for quick oxide removal without excessive corrosion . More inert purification vessel materials, such as carbon, molybdenum, or tungsten, could be used with higher hydrofluorination ratios. A ratio was found to be acceptable with carbon crucibles . This dramatically shortens purification times.
Other fluorinating agents have been used to purify fluoride salts; however, their chemistry is too corrosive, or has process limitations. To remove from the initial fuel salt charge in the MSRE, fluorine gas was bubbled through the salt. Fluorine was capable of oxidizing the salt so dramatically that not only did it fluorinate oxygenic impurities but also fluorinated dissolved uranium tetrafluoride to volatile uranium hexafluoride, which bubbled out of the salt. The redox potential imposed by a pure fluorine sparge was found to cause extensive corrosion to the purification vessel [10–12].
Another fluorinating agent proposed by Scheele et al. is nitrogen trifluoride. At room temperature, nitrogen trifluoride is a relatively nontoxic gas, making it popular at facilities where a higher factor of safety is desired. However, nitrogen trifluoride thermally decomposes at 400°C into a gaseous mixture with behavior very similar to that of fluorine. In the work done at Pacific Northwest National Laboratory, it was found that the temperatures imposed by molten salt would convert nitrogen trifluoride into fluorine, allowing the mix to potentially remove oxide impurities from fluoride salts. Just as in the MSRE, the fluorine would be extremely corrosive to the purification vessel. Adding hydrogen to control fluorine’s oxidizing nature exothermically produces hydrogen fluoride, which brings the chemistry back to hydrofluorination [13,14]. The price of nitrogen trifluoride, nearly five times that of hydrogen fluoride, and its chemistry prevents its immediate implementation.
Afonichkin et al. have used ammonium bifluoride to purify fluoride salt mixtures [15,16]. At room temperature, ammonium bifluoride is relatively safe; upon heating, it decomposes into hydrogen fluoride and ammonium at around 230°C [15,17]. Ammonium is able to pass through the salt without reaction, whereas the hydrogen fluoride purifies the salt. However, if the effluent ammonium and hydrogen fluoride mixture is allowed to cool below its decomposition temperature, it recombines and deposits solid ammonium bifluoride, potentially leading to clogs. Of all fluorination techniques, the most suitable for a batch scale process is hydrofluorination because of its price, chance of success, and strong supporting literature.
Hydrofluorinated salt will inevitably contain dissolved hydrogen fluoride and metal fluorides, both of which lead to an undesirable redox potential [5,18]. Sparging with hydrogen removes hydrogen fluoride through substitution according to Henry’s law and reduces dissolved metal fluorides, mainly nickel fluoride and iron fluoride, to nickel and iron. The reduction of nickel and iron fluoride produces hydrogen fluoride, which is then removed through the effluent stream. By extracting dissolved hydrogen fluoride, nickel fluoride, and iron fluoride, the redox potential of the salt is set to an acceptable level. However, hydrogen is incapable of reducing any chromium fluoride, which will remain dissolved in the salt . To further purify chromium fluoride-containing melts, a metallic reducing agent must be added. Two of the most commonly used metallic reducing agents are beryllium, a strong reducing agent, and zirconium, a mild reducing agent . Beryllium and zirconium fluorides are more thermodynamically stable than chromium fluoride and therefore will reduce chromium fluoride, given enough time. Care must be taken with metallic reduction agents; over-additions to salt can cause undesirable behavior, such as wetting and graphite attack [9,19].
Although not a problem while reprocessing fluoride salts, commercially supplied fluoride salts procured during the MSRE contained undesirable sulfate concentrations. Upon melting, these sulfates partially converted into sulfur dioxide and trioxide. All three forms of sulfur are capable of causing corrosion in any salt system. It was found that the hydrofluorination process, along with hydrogen sparging, was sufficient to convert the small concentrations of sulfur present in commercially available salt into hydrogen sulfide. This hydrogen sulfide was then removed through the effluent stream . Sulfur is an exemplary contributor to impurity-driven corrosion–rapid corrosion, which occurs during the first few hundred hours of corrosion and ceases once all contaminants come to equilibrium. By removing sulfur, the rate of initial corrosion is rapidly reduced.
Wisconsin Batch Purifier
The need for high-purity flibe at the University of Wisconsin to support the development of the FHR exceeded the capabilities of a small purifier. After examining purification options, a batch purifier was designed to produce up to 40 kg of flibe from raw materials, and up to 100 kg of flibe in a liquidus state. Production of the batch purifier, and all supporting facilities, took around 1 year.
Pressure vessels for large-scale salt purification require careful tuning between cost, size, and durability. The fluoride salt production facility in the MSRE used a 6′ (1.83 m) tall by 12″ (0.30 m) NPS, sch 40 stainless steel pressure vessel with a 0.125″ (3.18 mm) nickel liner . Stainless steel, which has around seven times the strength of nickel at process temperatures, served as the load-bearing structure, whereas the nickel liner protected the stainless steel from fluoride salt corrosion. Portions of the MSRE stainless steel purification vessel that came in contact with hydrogen fluoride were plated with nickel. The nickel-plating seemed to have failed at least once, corroding the stainless steel and leading to the dislodging of fluoride corrosion-product scale into the salt . To avoid this problem, the University of Wisconsin purification vessel was constructed out of commercially pure nickel. The nickel purification vessel, shown in Fig. 1(a), was made of 0.25″ (6.35 mm) thick nickel sheet, rolled into a 11″ (0.28 m) diameter cylinder 36″ (0.91 m) tall, and welded up the seam. The 0.25″ (6.35 mm) thickness was chosen to account for corrosion during a potential pure fluorine or nitrogen trifluoride purification. The purification vessel was fitted with a nickel charging port and five tube stubs, which served as ports for a gas inlet and outlet, pressure sensing, salt transfer, and a thermocouple well.
The nickel purification vessel was wrapped with a custom, two-zone, 5 kW heating system and situated concentrically inside of a stainless steel pressure vessel with a flanged top. The stainless steel pressure vessel was kept around 0.5 psi (3.5 kPa) of argon relative to the nickel purification vessel by means of a differential pressure gauge and switch. If pressure in the nickel purification vessel increased, the pressure in the stainless steel vessel also increased by the same amount. This action minimized the pressure differential at high temperature, preventing the nickel purification vessel from deforming. As seen in Fig. 1(b), the 3.5″ (8.89 cm) gap between the two vessels was filled with pourable, high-temperature Microtherm insulation to allow the nickel purification vessel to reach purification temperatures. The stainless steel pressure vessel also served a secondary purpose as an extra barrier from hydrogen fluoride. In the event that thermal cycling caused a fitting to leak on the purifier, the overhead pressure in the stainless steel vessel would force argon into the purification vessel, preventing hydrogen fluoride from escaping. Gas and salt lines were fed out from the nickel purification vessel through compression fittings welded to the stainless steel pressure vessel. To prevent buckling under thermal expansion, all tubings inside of the stainless steel vessel were bent in the middle. To keep track of the salt as it was loaded or transferred to a storage vessel, the entire purification system was placed on an 800 kg scale with an accuracy of 50 g.
The MSRE used around 70 storage vessels to transport salt from the purification facility within the Y-12 plant to the MSRE . Storage vessels were also used at the University of Wisconsin to store up to 60 kg of purified salt after filtration. To preserve the purity of the salt after purification, nickel was chosen as the storage vessel material. Hastelloy® N, an alloy developed for high-temperature fluoride salt use during the MSRE, would have been desirable if available. Nickel storage vessels also presented the same high-temperature strength limitations as the purification vessel. Flat lids for the 12″ (0.30 m) outer-diameter nickel storage vessels were insufficient to contain the pressure of pushing salt at 550°C without using excessive quantities of material. To circumvent this issue, the nickel storage vessel tank heads were domed to ASME standards by Tank Component Industries. Both the body and the tank heads were made out of (4.76 mm) thick nickel sheet. By using domed vessel heads, the nickel storage vessels were able to withstand a maximum of 43 psi (297 kPa) for extended periods. This greatly increased the factor of safety at high temperatures, decreasing the chance of a flibe leak in an accident. Each pressure vessel was fitted with four 0.5″ (12.7 mm) tube stubs for gas introduction and venting and one 0.75″ (19.7 mm) tube stub for salt transferring. Figure 2 shows one completed pressure vessel was marked “7” on its exposed lifting lugs in high-temperature paint, standing for enriched , whereas the second was marked “N” for natural lithium. Vessels were baked in hydrogen at 550°C for 1 day before the salt was introduced. A detailed discussion of the storage vessel preparation is given in Sec. 5.4.
The acute and chronic hazards of beryllium and beryllium-containing materials were well understood during the MSRE and have been thoroughly covered in modern material as well [5,20–22]. Beryllium-containing material used for the MSRE was handled in a high-airflow, high-bay room by personnel in a plastic fresh-air suit . However, MSRE documentation does not provide numbers on beryllium contamination during typical operations. This prevented the optimization of engineering controls when designing the University of Wisconsin beryllium facilities. To mitigate the health risk associated with beryllium work, a negative-pressure walk-in fume hood, similar to that used in the MSRE, was constructed. The room is roughly with a 10′ tall ceiling () and is built into the laboratory wall on three sides. It is large enough to contain the purification system as well as a line of corrosion tests. Its structural support comes from () carbon steel tube frame anchored into the walls. Cross beams were positioned every 48″ (1.22 m), extending from the front of the frame to the back which served as attachment points for frames supporting lighting fixtures and HEPA filters.
Duct work was added to connect a large fume fan to the HEPA filters mounted on the ceiling of the room. The fan is capable of moving through the HEPA filters, which exhaust to the outside of the building. This airflow is capable of regenerating the room’s air supply once every 40 s, which is slightly slower than the MSRE purification facility’s regeneration period of 20 s. A grated air inlet was installed to the outside of the fume hood, acting as a source of air inflow. Lastly, Tyvek® coveralls, booties, gloves, and a full-face respirator with a P100 cartridge are always used to prevent beryllium exposure while in the walk-in room.
Hydrogen Fluoride and Gas Handling.
Industrial gas supply has improved since the production of salt for the MSRE. During the MSRE, hydrogen and helium used for the purification of the salt had to be purified on-site. Reduction of water and oxygen to 10 ppm or less in inert gases was achieved by reaction with high-temperature titanium sponge. Hydrogen, which would react with titanium, was purified over hot uranium metal. Now, ultrahigh purity hydrogen, argon, and helium can be purchased for reasonable prices with water and oxygen content less than 1 ppm. These ultrahigh purity gases were used for all salt applications and greatly simplified gas handling. Titanium sponge was occasionally used for purification of industrial inert gases to approximately MSRE levels, when ultrahigh purity cylinders were not available.
Anhydrous hydrogen fluoride (AHF) is one of the more daunting chemicals of the purification process. It is commonly handled as a clear gas, which turns into a white, cigarette-smoke-like vapor when exposed to water in an open atmosphere. It comes in a regular gas cylinder as a liquid with a boiling point of 19.5°C as shown in Fig. 3. In liquid form at room temperature, AHF has a concentration of 49.5 M and is well documented as toxic and highly dangerous. However, as a standard temperature and pressure gas, AHF has a concentration of 0.04 M, making small leaks in air more manageable with proper ventilation and personal protection equipment. AHF requires no regulator and can be attached directly to tubing with a CGA 670 fitting. To control the pressure, the gas cylinder is heated, which boils off some liquid. During the MSRE, 200 lb (90.72 kg) bottles of AHF were heated in a hot-air bath to roughly 100°F (38.8°C), corresponding to around 10–15 psi (69–103 kPa). The University of Wisconsin uses a low-wattage resistance heater to heat a 44 lb (19.96 kg) bottle of AHF to 27°C, corresponding to a pressure of around 4 psi (28 kPa).
The AHF at the University of Wisconsin was kept inside a toxic gas cabinet attached to a fume fan, ensuring considerable air flow over the bottle. In the event of a leak, a window to the cabinet could be opened and the gas bottle could be turned off. Hydrogen fluoride sensors were also installed in the gas cabinet and in the walk-in fume hood. In an emergency, these would automatically close a pneumatic valve in line directly after the AHF bottle. All tubing and instrumentation in the AHF delivery system were made out of 316 stainless steel, which is capable of forming a fluoride passivization layer as long as its environment is kept completely anhydrous. Lines that contained the wet hydrogen fluoride effluent stream used tubing, valves, and fittings made of nickel and Monel®. To prevent clogs via condensation, all hydrogen-fluoride-containing lines were heated to at least 60°C.
Mass flow controllers were used to regulate the flow of all gases into the purification vessel. This marks a large improvement from the MSRE production method, which required manual calibration of gases. To confirm the flow rate of the AHF during the MSRE salt purification, the inlet flow was diverted into a water column for a set period of time. AHF readily dissolves in water; by titrating the dissolved hydrogen fluoride, the total flow could be measured. MSRE personnel would then adjust the AHF flow rate with a throttle valve until the desired flow was obtained. To control the AHF flow at the University of Wisconsin, a stainless steel, maximum, Teledyne Hastings variable mass flow controller was used. The mass flow controller required a maximum of 7 psi (48 kPa) inlet and had a pressure drop of only 0.5 psi (3.5 kPa), allowing for safe, low-pressure utilization of AHF. Flow of hydrogen and argon was controlled through two stainless steel Aalborg variable mass flow controllers, each with a maximum flow rate. All gases were mixed directly after the mass flow controllers and fed into the purification vessel’s sparge tube.
A series of water baths was used to capture waste hydrogen fluoride at the University of Wisconsin. Four 20 L and two 9 L commercially supplied high-density polyethylene (HDPE) carboys were implemented for this task. Each carboy’s lid was outfitted with two Monel® bulkhead, compression fittings and sealed with Teflon® packing. The 9 L carboys were kept with a few centimeters of water or neutralizer solution, and served as a reservoir for any effluent-stream liquids. The 20-L carboys were filled with 10 L of water or neutralizer solution, and outfitted with a submerged dip tube, forcing the effluent stream to bubble in the carboy. Typically, one 9-L carboy followed by two 20-L carboys was attached to the effluent line in that order. A two-way Monel® valve was installed in the effluent line before the carboys to stop the effluent flow and purge the carboys with argon. Argon purges were performed before removing carboys to prevent the release of any undissolved hydrogen fluoride. Another Monel® venting valve was added between the two dip-tube carboys to release any hydrostatic pressure, preventing accidental backflow of hydrofluoric acid during carboy switches.
Originally, excess hydrogen fluoride was neutralized online during purification. To do this, sodium bicarbonate was dissolved until saturation with a slight amount of bromocresol purple indicator into the two 20 L carboys. This method clogged the 0.25″ (6.35 mm) diameter Monel® sparge tube after 2 h of hydrogen fluoride flow, terminating the first purification. Upon disassembly, it was found that sodium fluoride and corrosion products sealed the sparge tube. To prevent clogging, the 0.25″ (6.35 mm) Monel® sparge tubes were replaced with 0.5″ (12.7 mm) Teflon® tubes, and the sodium bicarbonate solution was replaced with saturated potassium bicarbonate solution in all carboys. Potassium carbonate reacts with hydrogen fluoride to produce potassium fluoride and potassium bifluoride, which is more soluble in water than sodium fluoride.
While the neutralizer modifications allowed the baths to run continuously, it was found that the bulk of the excess hydrogen fluoride was able to dissolve in the first 9-L carboy, which had no dip tube. The potassium carbonate in the 9-L carboy was able to neutralize the effluent; however, the lack of agitation in the carboy created a crystalline layer assumed to be potassium fluoride and potassium bifluoride. This crystalline layer marked a distinct line between acid and base in the carboy, as indicated by the phenol red in solution. As the neutralization reaction proceeded, the crystalline layer allowed the buildup of carbon dioxide pressure under it. At a certain pressure, the carbon dioxide would force its way through the crystalline layer, causing vigorous bubbling and a 1–2 psi (7–14 kPa) pressure spike in the purification vessel over the course of a minute. This occurred during the entire second purification.
Online neutralization was eliminated after the pressure spikes. Instead, waste hydrogen fluoride was dissolved in water in a 9-L carboy and a 20-L carboy. A third carboy was not used after it was previously found that the large portion of hydrogen fluoride dissolved in the first carboy, while the remainder was captured by the second carboy. When the hydrogen fluoride concentration in the first carboy had the potential to reach 10 wt.%, the flow was paused; all the carboys were removed to a fume hood, and fresh carboys were put in their place. Neutralization of the hydrofluoric acid was done by slowly adding sodium hydroxide solution until the phenol red changed from yellow to magenta, taking breaks when the solution became hot. During the third purification without online neutralization, the bulk of the hydrogen fluoride was once again found to dissolve without sparging in the first carboy. Offline neutralization proved satisfactory for the entire 76 h of the third purification and is recommended for future purifications.
The filter is arguably the most important part of molten salt purification. It must quickly and effectively remove insoluble contaminants as the salt is transferred from the purification vessel to the storage vessel. If a clog is encountered, the transfer tube and filter would likely have to be removed, exposing purified salt to atmosphere for extended periods, potentially negating the purification process. During the MSRE, two filter designs were evaluated: sintered and felt metal [24,25]. Sintered filters were typically of pure nickel construction, whereas felt metal filters were tested in 347 SS, Monel®, and Inconel® 600. Several pore sizes from 10 to 40 μm, were evaluated. A 40-μm pore-size filter is the most commonly referenced size.
During the construction of the University of Wisconsin batch purifier, sintered nickel filters could not be found for reasonable prices with the specifications required for large batches of salt. Therefore, the clear choice was to use woven filters made by Porous Metal Filters Inc (Longwood, FL). These filters were offered in Monel®, Hastelloy® C-276, and stainless steel with varying pore sizes. Ultimately, a 316 SS filter, 1.5″ (3.81 cm) diameter and 2.5″ (6.35 cm) tall with a 40-μm rating, was selected. This filter was able to remove contaminants from 52.5 kg of purified MSRE coolant salt at a rate around with a pressure difference of 3 psi (21 kPa) at a transfer temperature of 600°C. The sintered mesh filter, shown in Fig. 4, was welded into a knife-edge flange housing with VCR fittings on each side and added to the transfer line before the storage vessel. For smaller-size filtrations, 316 SS, compression-fitting T-filters were used with a rating of 40 μm. Compression-fitting sintered filters were much smaller in size and were found to clog after filtering 1 kg of properly reduced salt with an initial metals concentration of 144 ppm iron and 46 ppm of chromium.
One of the largest improvements to the purification process since the 1970s is the introduction of personal computer control. During the MSRE, three 8-h shifts were used to make salt 24 h a day for 6 months . Many portions of the production facility had to be checked and altered by hand. The University of Wisconsin employed a LabVIEW™ cRIO with a highly automatic control program (the home page shown in Fig. 5) during the purification. By inputting the total salt mass, max flow rate, and termination time, the purification was able to control itself and change modes from hydrofluorination to hydrogen reduction and to hot storage. Errors, such as mass flow disagreement, over-pressure, hydrogen fluoride leaks, or temperature anomalies, halted the purification automatically. If the error subsided or was manually fixed, the purification was able to resume where it left off. To heat the salt and gases, 13 heater zones were used, each controlled through a proportional-integral-derivative controller loop built in the program, and monitored by 32 thermocouples. The duty cycle and frequency factor of all heater zones were alterable, thereby controlling how aggressively the zone was heated. Slow heating, from the top down, was performed while bringing a large billet of salt through its melting point. Top-down heating is done to prevent the formation of confined, liquid salt puddles, which can expand and burst the purification vessel. While in the molten state, where phase transition bursting was not an issue, more assertive heating was used to raise salt temperatures.
The LabVIEW™ program was completely accessible through a remote panel, allowing personnel to monitor the reaction remotely on any desktop connected to the university network. By tunneling into the network, the purification could be monitored at home. Although the program was readily accessible from a computer, it was also designed to send periodic e-mail updates on critical values to allow monitoring without tunneling in. This combination of automation and accessibility allows purification with one person, rather than three shifts of full-time staff.
MSRE Coolant Salt
The MSRE coolant salt was made from 24 batches of hydrofluorinated, -enriched flibe over the course of 6 months in 1964. These 120-kg batches of flibe were individually loaded into the dump tank for the secondary loop, which was then used to charge the secondary loop. The secondary loop circulated for 23,566 h and was finally drained and stored in the dump tank after the MSRE shutdown in December 1969 [6,26]. Over the next 50 years, valuable components, such as remotely actuated valves, were removed from the coolant loop and repurposed, leaving the coolant loop and salt exposed to atmosphere. The coolant drain tank cell is located in a humid environment, which partially flooded at least once while open to atmosphere. Therefore, it is very likely that the coolant salt was able to absorb considerable moisture during its interim storage.
In 1999, the salt was remelted in the drain tank and transferred into five vessels. One of these five vessels, shown in Fig. 6(a), made of Hastelloy® C-276, containing roughly 350 kg was used to conduct “pool melt experiments” in preparation for the decommissioning of the MSRE. After completion of these experiments, the salt remained dormant until 2012, when it was transferred through stainless steel tubing into a series of five stainless steel vessels, as shown in Fig. 6(b). The last of these vessels, containing 57.4 kg of salt, was donated to the University of Wisconsin for salt experiments in nuclear environments, where tritium production by natural lithium would be unacceptable [27,28]. Hastelloy® C-276 and stainless steel are composed of 16% and 16–18% chromium by weight percent, over double that of Hastelloy® N. It is predicted that the hydrolyzed salt was able to absorb chromium from these containers.
Coolant Salt Regeneration
The donated 57.4 kg of MSRE coolant salt was regenerated at the university in three progressive steps: a 550 g batch, a 16.2 kg batch, and a 52.5 kg batch. Purification was performed in progressive steps as part of a shakedown. This proved to be valuable, as unforeseen issues occurred, mainly with the effluent stream handling as discussed in Sec. 3.5.
The goal of the first purification was to test the hydrogen fluoride delivery, neutralization system, and control over transferring small quantities of salt. To do this, a small purifier was constructed out of 6.75″ (17 cm) tall, nickel , 2.5″ (6.35 cm) NPS, sch 10 pipe with two 0.375″ (9.53 mm) thick nickel end caps. The top cap was machined with four holes: a 0.25″ (6.35 mm) outer diameter dip tube, a 0.25″ (6.35 mm) outer diameter thermowell, and two 0.5″ (12.7 mm) gas and salt inlets and outlets. Both 0.5″ (12.7 mm) tubes were flush with the bottom of the top cap; only the thermowell and dip tube extended to the bottom.
Before MSRE coolant salt was transferred into the small purification vessel, 0.1 g of B-26-D grade, 99.5% by metals basis, beryllium from Materion was loaded into purification vessel to serve as a redox agent during melting and purification. This was commonly done during the MSRE salt production process to prevent excessive corrosion by salts of unknown redox potential . The MSRE coolant salt was remelted from top to bottom, heated up to 550°C, and transferred through 0.25″ (6.35 mm) stainless steel tubing into the small purifier. All hydrogen fluoride lines were purged with dry argon multiple times to prevent corrosion as discussed in Sec. 3.4. Then, the hydrogen fluoride gas bottle was opened, allowing the gas to diffuse into the lines. After opening, the bottle was heated to 27°C to pressurize the hydrogen fluoride, and trace lines were heated to around 60°C to prevent condensation. After heating, an actuated valve was opened and the hydrogen fluoride was fed into the vessel at , and mixed with hydrogen flow set at .
Hydrogen fluoride flow lasted for around 30 min before behaving erratically, fluctuating from no flow to maximum flow several times per minute before ceasing. Suspecting condensation clogs, the hydrogen fluoride delivery lines were heated with a hot air gun, but this was not found to affect the flow. Hydrogen fluoride flow was turned off, while hydrogen was kept on overnight. The next morning, the hydrogen was found to be flowing according to the mass flow controller, but the first carboy in the effluent stream neutralizer had ballooned up and appeared to have pressurized. The purification was terminated. Pressure was released, and the dip tubes that neutralized the hydrogen fluoride were examined. It was found that the first dip tube had clogged. Upon clearing the clog, the hydrogen fluoride would flow again.
The second purification treated 16.2 kg of MSRE coolant salt. Salt was transferred from the as-received ORNL stainless steel vessel, through a stainless steel line, into the batch purification vessel described in Sec. 3.1 at 550°C. The coolant salt was immediately sparged with AHF and hydrogen at AHF and for 48 h at 550°C. Hydrofluorination was terminated when the pH in the first, online neutralizing, 9-L carboy dropped below 6. The hydrofluorinated salt was then hydrogen-sparged at at 550°C for 36 h. An empty, small nickel storage vessel designed to hold 3.1 kg of flibe was baked at 550°C in a hydrogen atmosphere for 1 day to remove surface oxides. After baking, the storage vessel was connected to the purification vessel’s stainless steel transfer tube under argon purge to preserve its oxide-free surface. Roughly 3190 g of purified flibe was pushed at 550°C into the storage vessel. Salt was allowed to solidify in the storage vessel under an ultrahigh purity argon atmosphere. After solidification, lines were strategically disconnected under flowing ultrahigh purity argon and capped. Once disconnected and sealed, the storage vessel was transferred into the glove box.
The salt was remelted in the glove box and pushed through a stainless steel transfer line containing an inline 40 μm filter and into a nickel crucible. The salt appeared crystalline but was green in color, indicating dissolved metal fluorides and therefore a less-than-desirable redox potential. To remove the remaining metal fluoride, the small nickel storage vessel was removed from the glove box, heated to 650°C, and sparged with ultrahigh purity hydrogen at a rate of for 3 days. Salt was once again moved to the glove box and poured through a stainless steel transfer line containing a 40 μm filter into a nickel crucible. The filter was found to clog. New 40 μm filters were ordered from Swagelok® and installed into the line. Following the filters’ installation, two nickel crucibles were filled. Crystalline grains were moderately sized; salt color appeared transparent in some locations with a light gray tint (Fig. 7).
As a final step to remove the gray tint, 1.77 g of 99.5%, B-26-D grade, alcohol-cleaned beryllium shavings from Materion were added to the 1840 g of remaining flibe in the small nickel storage vessel. The 0.25″ (6.35 mm) charging tube was found to clog with beryllium shavings during the addition. The small nickel storage vessel was cooled and pressurized to 50 psi (344 kPa), allowing a large portion of the beryllium shavings to blow back out of the tube. It is estimated that a slight portion of the 1.77 g of the initial beryllium addition dissolved in the salt; therefore, a smaller 1.1 g charge was selected for the second introduction to avoid over-reduction. The salt was heated to 650°C to facilitate the reaction of the beryllium with impurities. After 36 h, the treated salt was pushed through a filter, which was clogged and then replaced. After installing a new filter, three crucibles, each containing around 300 g of salt, were poured. The first two crucibles had small, sparse, but visible debris in them. These were thought to be unreacted beryllium metal that was too small to be caught by the filter. The final crucible of flibe, shown in Fig. 8, solidified to white and translucent appearance with no visible discoloration. The crystalline grains in these three crucibles appeared larger than the flibe poured without beryllium addition. The beryllium-reduced flibe did not wet the nickel crucible and was easily broken out of the containers after solidification.
Before the final purification, a 1.8 kg sample was withdrawn from the bottom MSRE coolant salt tank to serve as a reference of salt before purification. Immediately after the withdrawal, the final 39.5 kg was transferred into the purification vessel through a stainless steel transfer line at 550°C. With 13.2 kg of MSRE coolant salt already in the vessel from the second purification, the total amount of salt purified in this batch was 52.5 kg.
After increasing the temperature of the salt from 550 to 600°C, hydrofluorination commenced at a total flow rate of , with AHF and hydrogen. Effluent neutralization, which was done online previously, was avoided this time. Waste hydrogen fluoride was dissolved into a series of water carboys and neutralized with a sodium hydroxide solution in a fume hood, when the maximum hydrogen fluoride concentration had the potential to reach 10%. This proved to work much better than online neutralization. The hydrofluorination was performed for 76 h. The quantity of hydrogen fluoride sparged through the flibe per unit mass during the third purification was higher than that of purifications performed from raw materials during the MSRE. It is assumed that the third purification thoroughly regenerated the salt .
Hydrogen sparging was performed following the hydrofluorination at a flow rate of for 3 days. Salt temperature was brought up to 631°C for this step to expedite the reaction. After 3 days, the salt was cooled down to the transfer temperature of 600°C, and the sparge was reduced to . Salt was transferred to a storage vessel through a 40 μm filter after 5 days of hydrogen sparging. Flow rates during the transfer were with a 3 psi differential (21 kPa). After cooling, the salt was stored under pressurized argon atmosphere in preparation for further use.
Experience With Beryllium During Regeneration.
Two forms of beryllium monitoring were used during flibe operations: airborne monitoring and surface monitoring. Both of these measurements are performed before, during, and after each operation and are key to preserving employee health. Ultimately, airborne concentrations are the most important measurement, but surface contamination is measured as it can contribute to airborne measurements if disturbed. Two of the most common operations that could lead to exposure involve transferring salt and installing equipment in a potentially contaminated environment. Typical on-person air-sampling measurements, analyzed at the Wisconsin Occupational Health and Safety Laboratory (WOHL), are presented in Table 1 for these operations. Transfers involve opening up a molten salt vessel to atmosphere; inserting a dip tube; connecting the dip tube to the transfer line; transferring, removing and disposing of the dip and transfer tubes; and closing the vessel. A thin layer of flibe exists on all dip tubes when pulled from storage tanks, thus creating a chance for exposure. Furthermore, flibe was found to leak through a fitting in the transfer line during the transfer shown in Table 1, increasing the chance for exposure. Despite all of this, airborne beryllium concentrations were found to be under detectable limits. However, large amounts of dust were found, likely from the high-temperature insulation used. Similar measurements performed while working with and around potentially contaminated equipment showed undetectable levels of airborne beryllium. Dust measurements were quite high, indicative of the nature of the work. Although airborne beryllium was found to be below detectable limits, full-face respirators and Tyvek® coveralls were worn during all beryllium operations.
Surface measurements have also been implemented to validate the cleanliness of operations. Swipe measurements taken after the MSRE coolant salt transfer, measured by the WOHL, are shown compared to a control swipe in Table 2. While the airborne concentration of beryllium was always below detectable limits, surface measurements have found beryllium contamination. All beryllium contamination samples have been below legal and industrial limits for surface contamination without personal protection equipment. After samples were taken, floors and surfaces were cleaned with a standard, all-purpose kitchen cleaner and disposable, heavy-duty towels. A large improvement was seen after cleaning; many surfaces’ beryllium contamination was reduced below detectable limits with this method.
During the MSRE, considerable effort was placed on determining the salt composition. Led by the Reactor Chemistry Division, chemical techniques to determine Li, Be, Cr, Ni, Fe, and U in the salt were developed. Many of these methods require considerable chemistry background and equipment. The University of Wisconsin modernized these techniques for salt work with excellent results. Lithium and beryllium content was determined by inductively coupled plasma -optical emission spectrum (ICP-OES) analysis at the Wisconsin Occupational Health Laboratory (WOHL). To test this method, three melts were analyzed: the MSRE coolant salt as-received from ORNL and two melt batches of flibe produced at the University of Wisconsin from raw materials. All salts were unpurified; this was not a problem, as ICP-OES determines only metals and would not differentiate between oxide, hydroxide, and fluoride species. By assuming that all beryllium and lithium found was in the form of a fluoride salt, the mole percentage composition was determined. As shown in Table 3, it was found that beryllium content of the MSRE coolant as measured by the MSRE and the WOHL matched very well. Both were outside the theoretical composition of flibe by nearly 2 mol%. Lithium concentrations did not agree in both MSRE coolant samples. This is expounded upon by Thoma, who notes that the chemical lithium analysis was systematically biased and routinely disagreed with the weight of lithium fluoride used in the salt’s preparation and melting point . Analysis by the WOHL confirms Thoma’s suspicions that the lithium in the MSRE coolant salt was much closer to the nominal composition than indicated by ORNL measurement. Both small, Wisconsin-made batches of flibe were found to reside close to flibe’s nominal composition. It is predicted that a closer composition can be achieved in a larger batch where scale error is negligible.
Trace metal concentrations were determined to be within acceptable limits by neutron activation analysis (NAA) performed by the Massachusetts Institute of Technology (MIT). As shown in Table 4, five samples were analyzed: coolant salt as-received from ORNL, flibe made from raw materials, and three progressively purified MSRE coolant salt samples. As-received from ORNL, the coolant salt contained unacceptable levels of metal impurities, with 144 ppm of iron and 46 ppm of chromium. These unacceptable levels of contaminants were likely introduced from corrosion initiated by hydrolysis in the Hastelloy® C-276 and stainless steel containers used to store the flibe. In flibe manufactured from raw materials in an open air nickel crucible, this iron and chromium content was much smaller. However, poor chemical conditions caused by melting salt in the open air likely caused corrosion of the nickel crucible, dissolving 64 ppm of nickel into the salt.
The last three samples analyzed, “ORNL Poorly Reduced,” “ORNL Properly Reduced,” and “ORNL Beryllium Reduced,” show the same MSRE coolant salt as it progressed through the second purification discussed in Sec. 5. The poorly reduced salt was hydrogen-reduced for 36 h at 550°C, which was neither hot enough nor long enough to remove all the iron and nickel fluoride from the salt—both of which are completely reducible by hydrogen. As compared to the as-received flibe, iron was reduced by around 131 ppm, the nickel was increased by 10 ppm, and the chromium was reduced by around 14 ppm. The remaining impurities caused the salt to appear with a slight green tint, which is further discussed in Sec. 5.3. Sparging the poorly reduced salt with more hydrogen at 650°C removed essentially all of the nickel and nearly all of the iron, as indicated in the sample “ORNL Properly Reduced.” However, as predicted, chromium was unaffected by hydrogen. A final insertion of beryllium, followed by a filtration, lowered all metal components in the coolant salt to levels below the MSRE Metal Standards.
Salt Quality and Appearance.
Appearance is a useful indicator of salt quality. Clean flibe should pour uniformly translucent with no suspended particles. Upon solidifying, it should be white to translucent, with large crystalline grains on the surface—indicating the presence of locally ordered molecules. Usually, divots form on the surface as the salt cools and shrinks. Several impurities can potentially be inferred upon visual inspection alone. Salt with a speckled black or uniformly gray regions, such as shown in Fig. 9(b), may contain metallic contaminants. However, if the black precipitates float on the surface of the salt, they are likely carbon. Both metallic and carbonaceous impurities can be removed through filtration. Salt with a light-green tint, such as that in Fig. 9(a), can be ascertained to contain quantities of corrosion product metal ions, such as chromium, iron, or nickel fluoride. Nickel and iron fluoride, as mentioned earlier, can be removed through extended hydrogen sparging. In general, uniform, off-color salt may be remedied through exposure to a reduction agent, such as beryllium, followed by filtration.
Another definitive measurement of salt quality can be garnished from corrosion performance. A properly purified salt will have a redox potential which minimizes corrosion. If corrosion results match previous results recorded from the MSRE, the redox potential, and therefore the salt quality can be assumed to be acceptable. Thus far, salt made from the aforementioned purifications has produced repeatable, low-corrosion rates in out-of-pile tests. Measured by the metric of maximum chromium depletion depth via energy-dispersive X-ray spectroscopy performed with a scanning electron microscope, Hastelloy® N, the alloy used for the MSRE plumbing, experienced a chromium depletion depth of 1.4–3.5 μm over 1000 h at 700°C in a static environment. Stainless steel, a less-traditional salt alloy, experienced chromium depletion of 11.5 μm in a stainless steel crucible and 22.5 μm in a graphite crucible over 3000 h at 700°C; projected corrosion estimates are 17.1 and , respectively, . In comparison, corrosion tests performed at the University of Wisconsin with unpurified LiF–NaF–KF (46.5–11.5–42 mol%) salt in similar conditions yielded chromium depletion depths in 316 stainless steel of 40 μm in 1000 h, roughly quadruple the corrosion seen with purified flibe over 3000 h in a stainless steel crucible . The corrosion rates shown with purified flibe are greatly reduced compared to unpurified salts, and match closely with the MSRE, attesting to their quality.
Preserving Chemical Quality.
After the lengthy purification process, salt should be handled with extreme care to preserve its integrity as much as reasonably possible. While dormant, salt should be stored in helium leak-checked, sealed vessels, back pressurized with ultrahigh purity argon or helium to around 5 psi (35 kPa). This technique, used during the MSRE, insures that water vapor does not creep in over time through small cracks or poor seals that may have developed through cyclic heating.
Whenever possible, containers that contact molten flibe should be made of copper or nickel. Copper, nickel, and their alloys have oxides that are easily reduced by a high-temperature hydrogen bakeout. By flowing ultrahigh purity hydrogen over the insides of the vessels at temperatures greater than 450°C, nickel and copper oxide can be reduced to their metallic components and water, which is then carried away. It was found that overnight hydrogen bake-outs were suitable for removing of TIG weld-heat tint leftover from manufacturing, creating a uniform metallic surface. After baking, vessels must be kept away from atmospheric exposure, especially at high temperatures. Graphite, another common container material, requires elevated hydrogen baking at temperatures in excess of 1200°C to stop adverse behavior with the salt. Unbaked graphite was found to suspend in the salt. Baking temperatures of 850–950°C were performed at the University of Wisconsin with acceptable results .
At times, it becomes difficult to avoid the use of stainless steel, usually due to component availability. While the use of stainless steel with coolant salt seems promising, its use should be limited to short intervals until further study [29,31]. The main issue with stainless steel is its chromium content. On the surface, stainless steel has a layer of . This chromium sesquioxide cannot be reduced by hydrogen and therefore will be fluxed away by the salt on contact. By flowing sacrificial salt over surfaces, oxides and contaminants can be dissolved and washed away, allowing clean salt to be introduced without oxide contamination. This technique was used to clean the Hastelloy® N fuel loop of the MSRE, as well as thermal convection loops, and is suitable for removal of chromium oxides [6,31].
Many experiments require the ability to handle salt at room temperature. In these cases, molten salt can be transferred into a suitable-sized storage vessel, disconnected and sealed under a flowing inert gas, and handled freely in a glove box. Glove boxes are routinely kept at and without extensive measures. The storage vessel can be then heated to 100°C beyond the melting point of the salt, pressurized to fraction of a psi above hydrostatic pressure with an inert gas, and pushed through a heated line with an open end into a suitable crucible. This process is shown in Fig. 10. By venting the pressure in the storage vessel, the transfer can be promptly stopped. Upon cooling, flibe ingots poured in a glove box less than 5 cm thick were easily broken with a rock pick or pestle and removed from the crucible.
When handling room-temperature salt in the glove box, special care should be taken to avoid the introduction of dust, especially from high-temperature insulation that is composed of oxides. In small samples of salt, minuscule dust particles and trace water exposure can cause large contamination issues. While not in use, poured salt should be stored in a sealed container, such as a mason jar, to limit trace dust and water contact. Fresh gloves or tongs should be used to manipulate salt. By following these practices, repeatable results can be obtained.
By far, the largest issue encountered while purifying flibe was clogging. Clogs are extremely dangerous, as they can cause pressure buildups. At high temperatures, these pressures can easily deform and rupture the nickel vessels required for fluoride salts, or move salt up dip tubes. Due to the length of time it takes to melt and purify salt, these pressure buildups can happen unexpectedly and must be dealtwith immediately. Pressure buildups can occur from a wide variety of phenomena. In the effluent line, hydrogen fluoride, water, and trace salt vapor will create corrosion products from even the noblest alloys, which will threaten to clog the line. To mitigate this, all effluent lines should be of as large a diameter as possible and be replaced regularly. Another source of clogging comes from filters. Filters should be overengineered with large surface areas and inspected after the transfer for their performance. The last area of clogging occurs with salt freezing. A dip tube or sparge tube positioned at the bottom of a rapidly filled vessel can serve as an outlet, leading to the movement of salt into unheated lines, where it will eventually freeze and clog. These salt freezes can be remedied by blow torching; however, they can be difficult to locate. It is recommended that all dip tubes and sparge tubes be inserted into a vessel after salt is introduced.
The purification of fluoride salts is a complex project that is absolutely required for high-quality, repeatable results. With the literature available on the subject and proper fabrication support, it is possible to repeat the steps carried out at the MSRE, even in a university environment. Chemical exposure, while engineered for, has been minimal; airborne beryllium has been undetectable, while floor contamination is minimal and easily cleaned. Hydrogen fluoride is readily dissolved in water and neutralized with standard caustic solutions. The hydrofluorination process has been predictable, slow, and has produced over 50 kg of pure salt.
Many improvements to the purification process have been made solely due to the technology improvements. LabVIEW™ has been used to automate the purification process, allowing it to be run by a single person from the lab or a remote location. Gases are controlled through mass flow controllers, eliminating the need for periodic flow-rate calibration. Improvements in commercial gas availability make it possible to directly obtain process gases with higher purity than that of the gases used at the MSRE. Wet chemical analysis of salt has been eliminated and replaced by ICP-OES and NAA. These new technologies have made salt purification and quality control more accessible than ever.
Pure molten salt is the basis for all types of experimental research leading up to a molten salt reactor. It should be considered the first focus of a fledgling program. Purification eliminates the uncertainty in interpreting the results of all experiments. All facets of reactor technology development benefit from purification. Purification eliminates or reduces effects caused by undesirable chemical reactions, creates uniform, fluid salt for accurate thermal hydraulics experiments, and removes impurities, which may affect the behavior of salt in radiation fields. This has been recognized by the FHR program. Purified salt made at the University of Wisconsin has already produced high-quality data in static corrosion tests. Future work plans to use salt from this facility in thermal convection and forced loops, both in and out of reactor cores, to evaluate the performance of the materials critical for the development of the FHR.
The authors would like to thank Michael Ames for his NAA services, L. ‘Mac’ Toth for taking the time to communicate his valuable experience with flibe, and Fred Peretz, ORNL, for packaging the secondary loop MSRE salt used in this study as well as providing a complete history of the salt. This work was made possible by U.S. Department of Energy Nuclear Energy University Program, Contract No. DE-AC07-05ID14517.