Corrosion of 316H Stainless Steel in Flowing FLiNaK Salt

Type 316H stainless steel samples were exposed to flowing FLiNaK salt for 1000 h in a monometallic thermal convection loop (TCL) with a maximum temperature of 650°C and a minimum of 540°C. Samples in the hottest part of the TCL lost mass, with a maximum mass loss of 1.4 mg/cm 2 , while samples in the coldest parts of the TCL gained mass, with a maximum mass gain of 1.0 mg/cm 2 . Analysis of the samples that gained mass showed an Fe-rich layer on the sample surfaces, suggesting that Fe, not Cr, accounted for the majority of the mass transfer in the TCL. However, Cr loss was apparent to a depth of ~5 µm in the HL. Post-exposure analysis of the salt showed major increases in the Cr, Fe, and Mn contents. The TCL was modeled using the TRANSFORM code. Modeled values matched the experimental temperature measurements showing that TRANSFORM is capable of accurately simulating the TCL conditions.


Introduction
Molten salt reactors (MSRs) are a Generation IV technology that has gained increasing attention in recent years [1,2]. While the use of molten salts as coolant and/or fuel media offers many advantages, salt compatibility with structural materials must be ensured before deployment of an MSR. Materials testing for molten salt compatibility dates back to the Aircraft Nuclear Propulsion (ANP) program at Oak Ridge National Laboratory (ORNL) in the 1950s [3], and continued throughout the Molten Salt Reactor Experiment (MSRE) and Molten Salt Breeder Reactor (MSBR) programs in the 1960s and 1970s [4][5][6][7][8][9]. The primary concern is not necessarily the corrosion of materials in contact with the salt but that mass transfer due to interaction between the salt and structural alloys will deposit material in colder sections of the circuit and Because of the concern about mass transfer, ORNL scientists made extensive use of natural and forced convection flowing experiments to test alloy compatibility with molten salts [3,[9][10][11][12][13]. While a static test is easier and less expensive, the eventual saturation of the static salts with corrosion product limits the usefulness of static testing for studying compatibility and predicting corrosion rates. By circulating salt through a loop that imposes a change in temperature as salt traverses the circuit, corrosion associated with mass transfer can be driven indefinitely without saturating the salt with corrosion products. The driving force for corrosion over long time scales in a flowing system is the collection of temperature-dependent equilibria that result in the depletion of alloy components in the hottest parts of the loop, and the deposition of corrosion products in the cold parts of the loop [3,6,14,15]. Loops with a temperature differential mimic the conditions in an MSR in which salt would be hottest in the reactor core, and colder in the heat exchanger.
Thermal convection loops (TCLs) have been used since the ANP program to test material-salt combinations and evaluate salt-alloy combination before moving to forced convection loops [3,13]. Central to the TCL design are a pair of vertical tube lengths, one heated and one not heated. The difference in salt density between the two vertical sections creates a natural convective flow that circulates the salt in the TCL without the expense and complexity of a pump or valves. Since the MSRE era, TCLs have been used to study compatibility with salt and liquid metal for fision and fusion energy [16][17][18][19][20]. With the recent surge in interest in MSRs, TCLs have been operated at the Kurchatov Institute [21] and The University of Wisconsin [22] and ORNL [23,24] A great deal of early work on material compatibility with fluoride salts was focused on Ni-based alloys. Alloy N, developed at ORNL specifically for use as a fluoride salt-facing structural material [25,26], was used extensively in the MSRE [27]. Recent attention has turned to evaluating the compatibility of molten salts with stainless steels [28][29][30][31]. Stainless steels were the subject of some study during the MSRE era as well [5,8,32] with higher mass losses compared to alloy N [9]. Type 316H is the high-carbon version of type 316 stainless steel (SS), and is ASME Boiler and Pressure Vessel Code qualified for higher-temperature service. It is attractive because it is widely available in common forms, and is less expensive than Ni-based alloy N.
To move beyond the static testing available in the recent literature, the purpose of this work was to evaluate corrosion and mass transfer of type 316H SS in flowing fluoride salt to better understand the mechanisms involved. Prior studies only reported mass change data with no characterization of the reaction products [8,9]. For this first experiment, coupons were exposed to flowing FLiNaK salt with a maximum temperature of 650°C in a TCL. The samples were characterized to understand the extent and nature of mass transfer and compared to similar specimens exposed in static experiments. A TRANSFORM model was created to understand the heat flow in the TCL.

TCL Construction
Exposures were conducted in a redesigned 316H stainless steel TCL. A photograph, 2D drawing and a 3-D rendering of the TCL are shown in Figure 1. The loop section was constructed from ¾" schedule-40 type 316H stainless steel pipe, with an outer diameter of 26.7mm, an inner diameter of 20.8mm, and a wall thickness of 2.79mm. All dimensions are nominal, and based on ASTM A53 [33]. An overflow tank was positioned at the top of the hot leg (HL), and a SS fill tank was connected above that. Three 900 W heaters were positioned on the loop, two on the HL and one on the bottom leg. The whole loop was wrapped in heating tape to allow for greater control over temperature. Thermocouples were positioned in thermowells around the loop, as indicated in Figure 1b.

Salt preparation
The FLiNaK salt used for this work was prepared by Electrochemical Systems, a now defunct company that specialized in salt purification for research projects. It was recovered from oldstock, stored in airtight Ni alloy vessels. Salts were handled exclusively in Ar-filled gloveboxes (O2 and H2O <0.1 ppm), so moisture did not come in contact with the salts at any stage. Due to the age of the salt, details of its provenance are unknown, although hydrofluorination is believed to have been the primary method of impurity removal [34]. The as-received salt was analyzed in triplicate with inductively coupled plasma mass spectroscopy (ICP-MS) by the University of Wisconsin Hygiene Lab, and the results are shown in Table 1. Due to the lack of a reliable method for measuring moisture content, only metallic impurities are reported. Previous reports of similar FLiNaK salts from the same source suggested it was relatively pure with low corrosivity [35]. The 366 ppm of Zr in the salt is notable, and may have accounted for the low observed corrosion rate [36]. Cheng et al. observed Zr deposition on stainless steel samples exposed to FLiNaK salt with Zr addition, [37] although the Zr concentration in the salt used for that work (1.9%) was much higher than in the salt used for the present work

Sample preparation
Samples were cut from 316H plate. The composition of the heat is given in Table 2. Composition was measured using ICP-MS and optical emission spectroscopy, inert gas fusion, and combustion analysis. Two types of samples were prepared. 12 coupons measuring 25.4 x 19 x 2 mm and 28 SS-3 tensile specimens. The specimens were arranged in chains held together by 316 SS wire and are shown in Figure 2. Spacers were positioned at the top and middle of each chain to ensure that the chains remained centered within the vertical tubing of the TCL. . 316H sample chains that were hung in the vertical portions of the loop, one in the hot leg, and one in the cold leg.

Capsule testing
As a preliminary step, 316H samples measuring 12.7 x 6.4 mm were statically exposed to the same FLiNaK salt in 316H or low carbon arc cast molybdenum capsules. Each cylindrical capsule measured 7.6 cm in length by 2.5 cm in diameter. Samples were tethered to one end of each capsule, and capsules were filled with 30 g of solid FLiNaK salt. The capsules were welded shut (gas tungsten arc) in the same Ar-filled glovebox without exposure to air. The capsules were then sealed (electron beam welded) within outer SS capsules to provide secondary containment. The capsule testing procedure has been described in detail elsewhere [38]. A total of 12 capsules were heated in a box furnace for 1000h in 316H or molybdenum capsules at either 550 or 650°C.

TCL Testing Procedure
Sample chains were hung from the top of the vertical portions of the loop with type alloy 600 wire measuring ~16" long on the HL, and ~24" on the cold leg (CL). This wire represents the only dissimilar metal in the TCL, and its surface area was relatively small and unlikely to affect corrosion of the other specimens. The upper fill pot was loaded with ~5 kg of FLiNaK salt in an inert glovebox, sealed, and positioned on top of the TCL. With the TCL sealed, it was pumped down to a vacuum of ~7x10 -3 Pa, heated to ~200°C, and allowed to out-gas for ~2.5 days.
To begin the experiment, the entire loop including the fill pot was heated above the melting point of FLiNaK (454°C) using the furnaces and heating tape. The SS valve below the fill pot was opened, and the salt was allowed to drain into the body of the TCL, filling it to a depth approximately half way up the pot above the HL. The HL furnaces were used to achieve a maximum target temperature of 650°C at the top of the HL, and 550°C at the bottom of the CL, which was sufficient to drive convective flow around the loop.
The TCL was operated for 1000 h at steady state. At the conclusion of the 1000 h, the SS valve at the bottom of the CL was heated and then opened, allowing the salt to immediately flow into the drain tank, where it cooled to room temperature, along with the TCL. The drain tank was removed from the TCL and stored in a glovebox so the salt could be analyzed. The solid salt was briefly exposed to air (<10 minutes) during this transfer, but surface area was minimal due to the single solid form of the salt. The TCL was filled with deionized (DI) water and drained 3 times to remove residual salt from the samples, followed by a rinse with acetone before the sample chains were removed.

Characterization
Individual samples were cleaned by sonicating in ~50°C DI water for at least 1 h to remove any further traces of residual salt. Specimen mass change was measured with a Mettler Toledo model XP205 balance with an accuracy of ±0.04 mg. Samples were cross-sectioned, mounted in epoxy, polished, and imaged optically with a Leica model MEF4A microscope. Samples were also characterized with scanning electron microscopy (SEM) and energy dispersive spectroscopy (EDS) mapping on a Hitachi S-4800 equipped with EDAX hardware. Salts were analyzed by ICP-MS on a Thermo-Finnigan Element 2 instrument. Samples were digested using microwave digestion in 7.0 mL of 16 M nitric acid, 3.0 mL of 12 M hydrochloric acid, 2.0 mL of 28 M hydrofluoric acid and 3.0 mL of DI water. After digestion, samples were diluted using several different ratios to account for a wide range of concentrations of analytes.

Results
After cleaning, the samples exposed to static FLiNaK salt in capsules were weighed and the individual and average mass loss results are shown in Figure 3. The samples exposed in 316H capsules had less mass loss than samples exposed in Mo capsules. In both cases, the samples exposed at 650°C lost more mass than the samples exposed at 550°C. The coupon samples exposed in the TCL were also measured for mass change, and the results are shown in Figure 4 where the specimen mass change is plotted versus the estimated exposure temperature. The temperature of each sample was not measured directly, but modeled in TRANSFORM (see discussion section). To determine the temperature of each sample, a linear rate of temperature change was assumed between the thermocouples in the HL and CL, and a temperature for each sample was assigned based on its position within the leg. The hottest samples were at the top of the HL, and the coldest samples were at the bottom of the CL. All the samples in the HL lost mass, with a maximum mass loss of 1.8 mg/cm 2 . Samples in the CL gained or lost mass, with a maximum gain of 1.1 mg/cm 2 and a maximum loss of 0.6 mg/cm 2 . The mass change as a function of temperature appears roughly linear in Figure 4.    . EDS scans of samples exposed to FLiNaK salt for 1000 h at various locations in the TCL Figure 8. EDS map of 316H samples H1 and C11 exposed to flowing FLiNaK salt for 1000h at 647 and 552°C respectively To further investigate the composition of the deposition layer, transmission electron microscopy (TEM) micrographs of samples H6 and C7 are shown in Figure 9. On H6, the metal is mostly Fe and Ni with Cr depleted from the matrix, but Cr and Mo-rich precipitates are visible throughout the matrix. Mo-rich precipitates are also visible near the surface and along what appears to be a grain boundary. On sample C7, the deposition layer is visible above the metal surface. The layer was Fe-rich oxide, although this may be due to air oxidation of the focused ion beam (FIB) lamella, as the Cr-depleted deposition layer would be susceptible to air oxidation. Cr and Mo precipitates are also visible in the metal matrix, and Mo-rich precipitates are visible in the deposition layer. Figure 9. STEM-EDS map of 316H samples H6 and C7 exposed to flowing FLiNaK salt for 1000h at 582 and 604°C respectively The concentration of metallic species in the salt was measured with ICP-MS before and after exposure, and the results are shown in Figure 10. The content of Fe, Cr, and Mn in the salt increased dramatically during exposure, while the Ni content decreased slightly. Figure 10. Concentration of metallic species in FLiNaK salt before and after flowing for 1000h in the 316H TCL.

TCL Temperature Model
A thermal-hydraulic system model was created utilizing the Transient Simulation Framework of Reconfigurable Models (TRANSFORM) code developed at Oak Ridge National Laboratory [39]. TRANSFORM has been used to model natural and forced circulation molten salt loops [40]. However, the previous natural circulation models were larger, and thus did not have the low Reynolds number flow regimes in the present work. TRANSFORM was chosen as the modeling tool for this work because of its drag-and-drop setup, complex linear and non-linear solution routines, and emphasis on fundamental constitutive equations that allow for fast and accurate solutions with potential for easy reconfiguration. The generic default heat transfer and fluid flow models of TRANSFORM were used, i.e. no additional models specific to the problem were needed. The dimensions utilized in the model were taken from the drawing shown in Figure 1.
In the simulation, shown in Figure 11, each pipe was modeled as a single connected loop in TRANSFORM with multiple nodes within each pipe. The HL contained one node for each coupon, spacer and tensile bar, with a final empty node for the remaining empty section of pipe. A heat flux boundary condition of 1800 Watts (two 900 W heaters) was imposed on the pipe heat transfer surface. The sloped bottom leg, which was heated by another 900 W heater, was modeled with internal heat generation only, a simpler method. Generally, pipes that do not contain test samples such as the bottom leg can be modeled with less detail.
The CL pipes were modeled as pipes with a single heat transfer surface, as with the HL. The heat transfer surface is then given a convection resistance condition that is connected to an ambient temperature boundary condition of 21°C with the surface area determined by the geometry of the pipe. Typically, a boundary condition such as this will have a heat transfer coefficient α determined by a Nusselt number correlation that relates the Raleigh number to the Nusselt number. However, due to the presence of heat tape added to the piping components this correlation will not hold true and must be adjusted. To correct for the presence of a non-uniform heat tape, thermocouples were added to the model to match the positions at the inlet and outlet of each leg of the TCL and monitored by a PID controller that was instructed to adjust the constants of the heat transfer coefficient to maintain the ΔT between the upper pipe and CL to match the measured ΔT. Resistances were added to the model to represent the various bends and pipe losses throughout the natural circulation TCL and calibrated to give the proper ΔT across the HL.  Figure 12a, and the results of the TRANSFORM simulation in Figure 12b. The dotted black line at 24 h represents the steady state benchmark that was chosen for the simulation. Data are interspersed in their time scale. The first 10 datapoints were taken at approximately 5 to 10 minute intervals to ensure temperature was maintained at the beginning of the test, and later datapoints were taken 3-12 h apart once the loop was maintaining steady temperatures. The simulated thermocouples in the TRANSFORM model match those taken by the loop thermocouples. Temperatures once the simulation reached steady state are shown in Table 3. With these general results in hand, the detailed simulation temperature distributions for the fluid surrounding the coupons will be representative of the fluid temperatures and flows during the test. Steady state mass flow values through the loop were determined to be 9.61x10 -3 kg/s. The velocity profile around the coupons and spacers is shown for the HL and CL in Figure 13. Velocity was also measured during the experiment by locally heating an exposed section of the piping near the thermocouple at the bottom of the CL, and measuring the time for the temperature spike to reach the thermocouple at the bottom of the HL. This measurement yielded a velocity of 1.8 cm/s. Figure 14 shows the simulated temperature profiles in the HL and CL. Position is indicated by the vertical distance from the bottom of the top pot.

Discusson
When analyzing the corrosion results, it is important to consider the ratio of the volume of salt to the surface area of wetted 316H surfaces. Figure 15 shows mass change divided by the ratio of salt volume to wetted surface area for the static and TCL experiments. This figure accounts for all wetted 316H surfaces including the tubing and the capsule walls.
The volume to wetted surface area ratio in the TCL is 0.9 3 2 , and the ratio is 0.2 3 2 in the 316H capsules (the coupon area is 2.2 cm 2 and the wetted capsule area is 71.4 cm 2 ). The figure shows that the normalized mass loss of the HL sample was greater than the mass loss of the static experiments. This can be attributed to the mass transfer that occurred as salt flowed from the HL to the CL, and either leached or deposited corrosion product due to temperature-dependent reaction rate constants. The mass gain observed on the CL sample from the TCL confirms this mass transfer. In the static capsule experiment at 550°C, a mass loss was observed, Figure 15. Figure 15. Normalized mass change of 316H samples exposed to FLiNaK salt for 1000 h in capsules or a TCL. Actual temperatures of the loop specimens were 552°C and 647°C. Figure 8 and 9 show that the deposited layer on the CL samples was rich in Fe compared to the other alloy components. This is surprising, since most studies discuss Cr transport and Figure 10 shows that Cr is the most abundant alloy constituent in the salt. Unfortunately, solubility data are not available for Fe and Cr in FLiNaK salt at these temperatures. DeVan [14] reported the amount of alloying elements (including Fe and Cr) found in the salt after exposure of Ni~16%Mo model alloys to a U-containing FLiNaK salt. The Cr values were 20-80 x 10 -3 mole% with 3-11%Cr in the alloy compared to 15 x 10 -3 mole% with 4%Fe present but not easily comparable to the current conditions. Several factors may be considered. First, it is possible that the Cr concentration in the salt never reached a high enough value in the HL, so the rate of deposition in the CL remained low. Dissolution of Cr into the salt would slow as the 316H surfaces became Cr depleted as shown in Figure 7. Another possibility is the kinetics of Fe deposition could be faster than the deposition kinetics of Cr. It is also possible that the solubility of Fe in the salt changes more strongly with temperature than the solubility of Cr. If ~250 ppm Cr ( Figure 10) is soluble in FLiNaK at 550°C, then there would be no driving force for deposition in the CL. However, the ~70 ppm Fe in the salt may reflect the solubility limit in the salt at 550°C and if higher Fe levels were present in the HL, mass transfer of Fe would be favored [9]. Furthermore, since all of the surfaces are Fe-rich, the dissolution of Fe would continue throughout the experiment (unlike Cr which becomes depleted from all the specimens, Figure 7). Figure 5 shows that all the samples in the CL, and the lower temperature samples in the HL, had a deposited layer on their surfaces. However, Figure 4 shows that only samples C10, C11, and C12 gained mass, indicating mixed mode corrosion in the other samples with a deposit. One possibility is that the samples lost their mass (e.g. Cr depletion in Figure 7) in the initial stages of exposure, while the salt corrosivity was presumably at its highest, and before the deposited layer formed on the sample surfaces. Another possibility is that Cr was still attacked after the formation of the deposited layer, but this is less likely because of the slow kinetics of Cr diffusion.
From a single TCL observation for one time and one temperature gradient, it is not possible to resolve these issues and further work is needed to fully understand this mechanism. It should be noted that similar Fe-rich deposits have been observed for an alloy 600 (Ni-14Cr-7Fe) TCL with Cl salts [41]. In both cases, interpretation is confounded by the lack of Cr and Fe solubility data in either salt. However, the current results do suggest that 316H is reasonably compatible with FLiNaK at 650°C and a reasonable candidate for further MSR-related assessments. The amount of mass transfer was minimal compared to some systems [10]. With the proper inputs, modeling efforts [42] are in progress to explain these results.

Conclusions
Demonstrating classic mass transfer behavior, type 316H samples in the hot leg of the FLiNaK TCL lost mass and samples in the cold leg gained mass. Samples in the intermediate range exhibited mixed mode corrosion with observed deposits, but overall mass loss. The hottest TCL sample at 650°C lost more mass than a 316H sample exposed for the same time in the same FLiNaK salt but in a static capsule, exhibiting the accelerating effect of mass transfer. Examination of the TCL cold leg samples showed that Fe was the primary deposited alloy constituent in the cold leg. The hot leg samples showed typical corrosion patterns, with Cr depletion near the surfaces. A TRANSFORM model of the TCL matched the measured temperature profile, showing that it is a valid tool for modeling flowing molten salt in this experiment.