(405e) Inhibiting Corrosion by Molten Fluoride Salts: Investigations on Flinak

Hydrogen is hailed as the non-polluting energy carrier of the future. The catalyzed decomposition of water by thermo-chemical cycles could be a carbon-free production means, provided that the primary energy source is so, as would be heat from a high temperature nuclear reactor.

In the United States, the 2005 Energy Bill authorized the construction of the Next Generation Nuclear Plant, to be built by 2017, that would generate electricity, and send heat to the hydrogen production plant via an intermediate heat exchanger loop. The NGNP is expected to be a high-temperature gas cooled reactor, with the later possibility of liquid salt cooling as well. Both types would require a heat exchanger loop for hydrogen production.

Two candidate heat transfer fluids are under consideration for this loop: liquid fluoride salts and pressurized helium. Liquid fluoride salts are chemically very stable across a wide range of temperatures, with boiling temperatures of around 1400°C. Their low vapor pressures make large pressurization unnecessary. Their fluid mechanics properties in the range of temperatures considered for hydrogen production (500°C to 1000°C) are similar to those of room-temperature water. They are transparent: in-service inspections will be easier. They have very good heat transport qualities, and only benign interactions with dry air and CO2. The primary design issues which must be addressed to use them involve (1) preventing freezing due to their freezing temperatures of ~450°C, (2) controlling corrosion of container materials by maintaining low fluorine and oxygen potentials, and (3) preventing or controlling the generation of toxic HF under contact with water and sulfuric acid. Issues (2) and (3) are closely related to the approach used to control the salt chemistry.

Salt components are chosen according to their compatibility with structural alloy components, which should be noble compared to the salts. Operation of the Aircraft Reactor Experiment in the 50s and the later Molten Salt Reactor Experiment (MSRE) demonstrated the compatibility of a fluoride fuel mixture with Ni-based alloys at maximum operating temperatures of 710°C. The studies performed by ORNL concluded that the corrosion, driven by the solubility of the container material's fluorides and the temperature gradient in the loop, could be reduced to acceptable levels by keeping the fluorine potential of the solution low, and keeping the salt clear of ingress impurities such as moisture or oxygen.

This earlier experience is sufficiently promising that fluoride salts could be preferred to pressurized helium if strongly reducing chemical conditions can be maintained in the loop. For the fuel salts used in the MSRE, highly reducing conditions were not possible, because uranium was to be kept in solution with the solvent salt (flibe, Li2BeF4). Yet, acceptable corrosion performance was obtained. However, for the NGNP intermediate loop application, substantially higher temperatures are needed, approaching 900°C. This requires the use of high-temperature metal alloys, not explicitly designed for low corrosion with liquid salts, and results in much higher solubility for container materials due to the exponential dependence of their fluorides' solubility with temperature. For high-temperature heat transport, extremely reducing conditions are desirable and even potentially mandatory.

Three methods exist to control the redox state of molten fluoride mixtures, which seek to minimize the ?fluorine potential', defined by DF2 = RTln p(F2). Low potentials define more reducing conditions and thus, less extensive corrosion. They are: · gas-phase control: fixing the F2 partial pressure by · major-metal control: example in Flibe · dissolved salt control by use of a divalent fluoride The eutectic flinak (LiF0.465KF0.42NaF0.115) was here studied. Two previously identified design problems were considered: 1. mitigation of the structures' corrosion during normal operations, 2. mitigation of the consequences of accidental ingress of corrosive of SO2 and concentrated H2SO4 used in the SI hydrogen production process.

Chemistry control methods for maintaining reducing conditions can be thought of as: 1. ubiquitous, or local 2. permanent, or punctual.

Ideally, chemistry control would be ubiquitous (control of the redox state is provided throughout the loop) and permanent (in comparison to a punctual ?emergency' system). It could be obtained by dissolving buffer species into the liquid salt to capture the oxidizing species that might contaminate it.

An ubiquitous and semi-permanent solution would allow for the buffer to be punctually regenerated. The MSRE illustrates such a case : the redox control was provided by the uranium in solution which would switch from U3+ to U4+ depending on the salt's redox potential. The container walls were protected because their alloys were more noble than uranium. The fluorine potential was fixed by the ratio of the activities of UF4 to UF3. To change its value (regenerate U3+ from U4+), a rod of beryllium was periodically dipped for a few hours in the salt, reducing UF4 via 2UF4 + Be ® 2UF3 + BeF2 . This chemistry control method and the relatively low operating temperature of 710°C can partly explain the very low levels of corrosion that were reported for the MSRE reactor metallic and graphite structures.

In the case of flinak, none of the salt's components has the inherent buffer capacity of the U3+/U4+ couple, which will only be obtained via the multiplicity of valence states. Dissolving such species as lanthanide fluorides in the salt, as has been proposed by the ORNL team of D. Williams et al, would work in a similar way.

Our approach has been to look for an oxygen getter that would prevent attack of the walls by capturing oxidizing ingress species and delivering them to a cleaning trap to be removed, while new oxygen getters are injected in the salt. In this investigation, a metal-control method has been studied. It uses a combination of an aluminum fluoride solid (cryolite) with local contact of the salt with a liquid sodium bath, and a particle filter (via a cold trap if needed) to remove solid aluminum oxide precipitates. The aluminum fluoride reacts strongly with oxygen, and creates very stable alumina, that is insoluble in flinak. This releases fluorine ions in the salt, which then react with the sodium bath to be converted to sodium fluoride.

Recent experiments at Idaho National Laboratory have shown that beryllium metal (Be0) may have a solubility exceeding one mole percent in the liquid salt flibe (Li2BeF4). When dissolved into flibe, it may thus be capable of maintaining low fluorine potentials ubiquitously. Yet, although beryllium is a relatively strong oxygen getter, its chemical toxicity makes it unacceptable for use in the intermediate heat transfer loop.

This observation raises the question of whether similar behavior may occur with flinak and its most unstable metal (e.g. sodium). Because sodium is not an effective oxygen getter, a combination of sodium and aluminum may then provide an effective method for ubiquitous control of both fluorine and oxygen potentials in flinak.

A thermodynamic evaluation was performed to assess the extent of these reactions. The results of these evaluations have substantial uncertainty because no data exists for the activity coefficients of the different species in solution (as diverse as Na°, Na3AlF6(l), NaF, Na2O ?). Thermodynamic equilibrium calculations still provide indications of how a flinak/sodium/aluminum-fluoride system may behave. Equilibrium calculations, assuming unity activity coefficients, were used to study the fate of ingress oxygen, showing that the formation of Al2O3 can protect chromium(least stable metal in the typical structural materials). Figure 1 shows that chromium remains stable when cryolite is present.

In the case shown in Fig. 2 (both cryolite and sodium are added), Cr is protected from oxidation and Al2O3 is formed (instead of Na2O). Significant alteration of the salt's composition occurs (formation of NaF and NaF(l)).

The previous equilibrium calculations are only indications of potential behavior of the flinak/sodium/aluminum-fluoride system. They are inexact because activity coefficients are assumed to be equal to one ? which is likely erroneous in so ionic a liquid. The calculations were also made assuming the presence of all species at the same time and place, and that one mole of salt going through the liquid metal bath would be in contact with one mole of liquid sodium. Moreover, the most practical way to implement this system would be to dissolve cryolite in the cold part of the loop, and saturate the salt with it. Its concentration would therefore not exceed its cold part value , which is around 570 ppm at 550°C (as reported by Brooker, von Barner, Bjerrum and al). The cryolite's ability to capture oxygen by forming alumina would thus be limited by this solubility.

Yet, this provides us with an interesting hint to solving corrosion issues in flinak systems: given that the concentration of dissolved cryolite would be constant in the salt, small leaks would be easily detected since the reaction with H2SO4 creates gaseous H2. Hydrogen gas can be detected in the gas space where the sodium bath would be located.

This work is supported by the US Nuclear Energy Research Initiative.