Characterization of a Thermodynamic Reference Electrode for Molten LiF-BeF₂ (FLiBe): Part II. Materials Analysis

Electrochemical tests are conducted in Ref. 6 with two separate TREs, RE-01 and RE-02. Table I summarizes the details of RE-01 and RE-02. The electrode body is a 35 mm tall BN closed-end tube. A LaF₃ single crystal membrane provides ionic conductivity with the main salt bath; it is in contact with the main salt bath through a 0.3 mm diameter hole on the side of BN tube. As reported by the manufacturer, low-concentrations Eu (nominally 0.3% molar fraction) is present as a dopant to the LaF₃ crystal to increase its ionic conductivity. RE-01 operated for around 100 hours and showed a potential drift of more than 70mV vs. a glassy carbon quasi-reference electrode. RE-02 operated for roughly 15 hours. For additional details on the electrode design, electrode materials, salt preparation, and salt characterization the reader is directed to the previous publication by the authors.⁶

Table II lists the salt samples and BN samples collected for analysis and provides details about collection method and visual appearance. All sample preparation and sample storage are performed inside a glove box (LC-Technology) that maintains an argon atmosphere with oxygen and moisture below 1ppm. The BN tubes of RE-01 and RE-02 are cut open using a metal razor blade to extract the salt from the TREs. Salt samples are prepared by breaking up the salt with an agate mortar and pestle (Sigma-Aldrich, part Z112526). The samples are sealed in borosilicate vials in the glove box and sent for ICP-MS analysis. Even though the sample collection method aims to limit sample contamination, the contamination of salt samples by BN cannot be excluded due to the dusty nature of BN. BN samples are collected using a razor blade to separate two pieces (BN-01 and BN-02) from RE-01 and stored in borosilicate vials in the glove box.

The ICP-MS analysis is performed by the University of Wisconsin - State Laboratory of Hygiene and follows the procedure previously described in Refs. 8, 11. Samples are digested in acidic aqueous solution using two dissolution matrices, one with HF, one without HF. Both digestions proved to be equally effective at dissolving most of the elements and the analytical results from multiple digestions are averaged. Lanthanum, europium, and boron are exceptions. La and Eu ICP-MS values are lower and have more scatter when digested with-HF compared to digestion without-HF. This effect could be caused by volatilization or precipitation of species when HF is added to the acidic aqueous solution. For La and Eu, only the results from digestion without-HF are considered. Due to the limited size of the collected sample, RE-01-M is not digested without-HF; the La and Eu values in RE-01-M are inferred, with large uncertainty, from the values measured for the with-HF digest; details of this calculation are provided in Supplemental Material. Boron values are lower when digesting without-HF; for B, only results from digestion with-HF are considered.

The ICP-MS elemental content is reported in wppm, defined as the ratio of the weight of the element to the weight of salt sample analyzed. The mole fractions are calculated on a basis of a multi-component system comprised of LiF, BeF₂, and the solute species considered; for the purpose of these calculations, the LiF-BeF₂ molar ratio is assumed to be 2.00:1. This approach is equivalent to using a FLiBe molar weight of 33.0 g/mol; a similar convention is used in Ref. 9.

For X-ray diffraction (XRD) analysis, powdered salt samples are prepared on a glass microscope slide in the glove box. Double-sided tape is used to fix the sample to the surface of the slide. Kapton film (0.3 mm thick, SPEX Sample Prep) is wrapped around the slide to contain the sample and to limit air exposure of the sample. XRD is performed with a Bruker D8 Discover diffractometer equipped with a CuKα monochromated radiation source (λ = 1.541 Å) operated at 50 kV and 1000μA, using the software Diffrac. The area detector rotates in 10^o steps, 20^o to 80° and each step has a duration of 5 minutes. A 0.5 mm collimator is used and the sample stage is rotated at a rate of 0.3 rotations per minute to provide a random sampling of crystal orientations in the sample. The instrument is calibrated with corundum (Al₂O₃) as the standard material (NIST SRM 676a). XRD raw data undergoes Kα2 component stripping and background subtraction using the Diffrac.Eva v3.1 software. PDF-4+ 2016 (International Center for Diffraction Data) powder diffraction file (PDF) library is used for pattern identification and peak assignment. Raw XRD data is reported in Supplemental Material.

Scanning Electron Microscopy (SEM) is performed with a LEO 1530 instrument, which uses a Schottky-type field-emission electron source. Energy Dispersive Analysis System (EDS) (5 kV accelerating voltage, 27.5 take-off angle), which is integrated into SEM, is used to determine the elemental composition of sample surfaces. The SEM/EDS samples are prepared in the glove box. During transfer from the glove box to the SEM/EDS instrument air exposure is likely, for a duration of less than two hours.

BeF₂, like all beryllium-containing compounds, is a hazardous chemical. The glove box, fume-hood, and personal protective equipment provide protection from respiratory and dermal exposure to beryllium-containing chemicals. Beryllium contamination in the laboratory is monitored by surface swipes. The work in the present study is performed during the period of June 2015 through June 2016, during which a total of 202 surface swipes are analyzed. Any detection of beryllium above the detection limit of 0.025 μg/100 cm² (a total of 18 swipe samples with detectable Be) is followed by cleaning and decontamination actions. The housekeeping goal for the lab is 0.2 μg/100 cm²;¹² during the duration of the present study this goal is exceeded on two occasions, followed by updated standard operating procedures and extensive lab cleaning and decontamination.

Salt from RE-01 and RE-02 is removed as relatively intact cylinders of frozen salt. RE-02 frozen salt presents a green tint, which is typically observed when NiF₂ is added to FLiBe solution. Such a green tint was absent from the RE-01 salt. A large quantity of black powders is visible at the top and on all the outside surfaces of the salt cylinder in RE-01 and to a lesser extent in RE-02. To avoid the presence of floating or deposited insoluble components, RE-01-M and RE-02-M are salt samples collected from the center of the RE-01 and RE-02 salt cylinders; these two samples are assumed to be representative of the bulk salt in RE-01 and RE-02. The internal cavity of the BN tubes, originally of white color, is found partially covered with black residue. This phenomenon is most evident in RE-01 (Figure 1) and it is also present in RE-02 to a lesser extent.

Zoom In Zoom Out Reset image size

ICP-MS results demonstrate a change in the mass fraction of Ni, La, Eu and B after operation of the TREs, as reported in Figure 2. The results for these elements are discussed here, and the full ICP-MS dataset of 47 analyzed elements is reported in Figure SM-03 of Supplemental Material.

Zoom In Zoom Out Reset image size

Figure 2a shows that the Ni mass fraction in all the three RE-01 samples is lower than in RE-02. This demonstrates higher depletion of NiF₂ after 100 hours of RE-01 operation than after 15 hours of RE-02 operation. It is to be noted that the initial loading of NiF₂ in RE-01 was twice as high as in RE-02 (see Figure 2a). Nevertheless, after operation, the concentration of NiF₂ in RE-01 is smaller than in RE-02, indicating that the NiF₂ depletion occurs over a timescale of ten hours to hundreds of hours.

The measured NiF₂ in both RE-01 and RE-02 is lower than the NiF₂ that is originally added to them. It is possible that RE-02 depleted over the course of 15 hours; it is also possible that the observed decrease was due to the compositional fractionation during solidification. The lowest temperature datum available for NiF₂ solubility in FLiBe is 6200 wppm at 475°C, which is equal to the RE-02-M value (t-test, 95% confidence).¹⁰ The linearly extrapolated solubility at the melting point of FLiBe (459°C) is 5200 wppm, which is two standard deviations below RE-02-M. While there is uncertainty in the absolute values of NiF₂ loading above 5200 wppm, there is less uncertainty in the comparison between RE-01 and RE-02, because they experience the same thermal transient during solidification. Furthermore, this uncertainty does not affect the RE-01 values, all of which are below 5200 wppm.

Figure 2a shows no increase in the Ni mass fraction in Salt-01 compared to Salt-02 (t-test, 95% significance), indicating that BN is not penetrated by Ni ions. Figure 2c shows that the quantity of B found in Salt-01 is not statistically different than the B found in Salt-02 (t-test, 95% significance). Thus, there is no indication of corrosion products of BN in FLiBe. Salt-01 and Salt-02 are not in contact with BN surfaces at the time of sampling; and contamination by BN powders is not expected as a source of error for these samples. No other indications of BN degradation in FLiBe are observed.

Boron is found in the salts of RE-01 and RE-02 (Figure 2c). Given the dusty nature of BN, boron content in TRE salt samples could be due to contamination of the samples by BN powders; however, because the sampling process was similar for RE-01 and RE-02, a comparison between the two remains reliable. BN redox reaction with NiF₂ could produce soluble boron species in the salt. Higher boron content is observed in RE-01 than in RE-02. This indicates that the BN reaction occurs over a timescale of ten hours to hundreds of hours.

RE-01-M and RE-02-M show a statistically equivalent LaF₃ content, which averages to 5400 +/− 1000 ppm across RE-02-M, RE-01-M, and RE-01-T. Figure 2b shows that this amount is within one standard deviation of LaF₃ solubility of 6300 wppm at 480°C in LiF-BeF₂-UF₄ (62.8-36.4-0.8 mol%). Temperature-extrapolation to 459°C yields LaF₃ solubility of 4500 wppm,¹³ also within one standard deviation of the observed value. This indicates that LaF₃ reaches a solubility limit in FLiBe within less than 15 hours of equilibration time. Europium, nominally present at 0.3 mol% in the LaF₃ crystal procured from Crystran Ltd, UK, is also detected. Eu in RE-01-M is higher than in RE-02-M, indicating that its solubility has not yet been reached within 15 hours.

Salt-01 presents a statistically higher La content than Salt-02, and more than two orders of magnitude smaller than RE-01-M and RE-02-M. Similarly, Eu in Salt-01 is statistically higher than in Salt-02. Since the LaF₃ crystal has some solubility in FLiBe, a small flux of lanthanum and europium occurs through the small hole in the BN tube that allows for ionic contact between the TRE and the main salt bath.

Composition variability is observed among the three solid salt samples from RE-01. The bottom salt volume can include precipitated species. Ni metal (9000 kg/m³) and NiF₂ (4700 kg/m³) are of higher density than FLiBe (2000 kg/m³). The top salt volume could include powders that remain suspended due to surface tension effects. RE-01-T and RE-01-B have equal nickel content, higher than RE-01-M, and lower than the solubility limit. RE-01-M has less than 30% of the NiF₂ solubility limit and less than 5% of the initial NiF₂ added to RE-01, confirming Ni(II) depletion. RE-02-M is at the solubility limit of NiF₂ in FLiBe at its melting point.

Similarly, spatial distribution of La concentration is observed in RE-01. The higher La concentration in RE-01-B is attributed to two possible effects: i) solid fragments of the LaF₃ crystal that are physically trapped in RE-01-B, ii) saturated LaF₃ precipitated out of solution during cooling transients. RE-01-T and RE-01-M have equal LaF₃ content, pointing in the direction of imperfect sampling being the cause for the high LaF₃ content in RE-01-B. RE-01-B had 0.65% Eu/La, compared to 1.2% in RE-01-M and 1.0% in RE-01-T, and also compared to 0.85% in RE-02 and 1.1% in Salt-01. The lower Eu content in RE-01-B is closer to the 0.3% expected in the LaF₃ crystal. This observation indicates that small fragments of the LaF₃ crystal were collected with the RE-01-B sample. We conclude that samples collected from the middle of the electrode salt are most likely to be representative of the bulk composition of the melt.

XRD results performed on BN-01, BN-02 and BN-03 are reported in Figure 3. BN-characteristic peaks are labeled with their respective Miller indexes, from the Powder Diffraction Files (PDF) in Table III. Additional peaks of smaller intensity found in BN-01 between 2θ of 30° and 43° are attributed to constituents of the BN and are considered as part of the BN background when interpreting the BN-02 and BN-03 spectra. The diffraction patterns of BN-02 and BN-03 show two additional peaks, which correspond to the two main peaks of the metallic Ni. These peaks do not match any major peaks of NiF₂ (see Table III). These results demonstrate the presence of metallic Ni on the inner surfaces of the TRE.

Zoom In Zoom Out Reset image size

Figure 4 presents the results of the SEM / EDS analysis performed on BN-02. Three textures are observed (Figure 4a). EDS point scans are used to identify the elemental composition of each of the textures (Figure 4b). EDS cannot identify the contribution of Li, Be and B signals as they are grouped together below the lowest energy peak; EDS does differentiate among N, Ni, and F. Texture 1 is a layered structure that is present across all of the BN-02 sample and appears to be the underlying substrate. EDS on Texture 1 shows high N content, and no Ni or F. Texture 1 is attributed to BN.¹⁴ Texture 2 consists of micrometer-sized porous particles on top of Texture 1 and is observed across all the BN-02 sample. EDS on Texture 2 shows high Ni content, small amount of N, and no F. Texture 2 is attributed to metallic Ni. Texture 3 consists of 1–10 um droplet-shaped structures deposited on top of Texture 1. Texture 3 is only present in some regions on the BN-02 sample surface. EDS on Texture 3 shows high F content, small amount of N, and no Ni. Texture 3 is attributed to droplets of frozen salt on the BN surface. Additional SEM pictures are reported in Figure SM-02 of Supplemental Material.

Zoom In Zoom Out Reset image size

BN is expected to reduce Ni(II) to metallic Ni and form boron fluoride species. Figure 5 presents the Ellingham diagram for the chemical species relevant to possible reactions in the TRE. The reaction between NiF₂ and BN (Eq. 2) is thermodynamically favorable above 150°C.¹⁵

The decrease of Ni(II) concentration in the RE-01 salt being due to its reduction to metallic Ni by reaction with BN is confirmed by post-characterization of RE-01. After 100 hours operation at 600°C, visual inspection shows no green coloration of the RE-01 salt, indicating depletion of NiF₂. ICP-MS confirms NiF₂ concentration decreases to 0.090 ± 0.011 mol% after 100 hours of operation. Based on Equation 1, a TRE potential drift of 85 ± 5 mV would be observed when the concentration of Ni(II) in the TRE solution drops from its solubility value of 0.859 ± 0.016 mol% at 600°C to 0.090 ± 0.011 mol%. Electrochemical measurements report a drift of 70 ± 5 mV, based on time-dependent OCP measurements vs. a glassy-carbon quasi-reference electrode immersed in the main salt (shown in Figure 6 of the authors previous publication⁶). ICP-MS concentration measurements are within two standard deviations of electrochemical measurements for RE-01.

Zoom In Zoom Out Reset image size

The black powders found at the bottom of RE-01 and on the surface of the RE-01 BN tube are identified by XRD and SEM-EDS to be metallic Ni. Boron content is higher in RE-01 than in RE-02. These results support the hypothesized reaction in Eq. 2. Some amount of black powders is visually observed in RE-02, indicating that also in RE-02 some amount of NiF₂ is reduced to Ni. Boron content in RE-02-M is higher than in Salt-01 by more than one order of magnitude, indicating that at 15-hours the NiF₂ reduction reaction has already proceeded.

Reaction 2 postulates production of volatile BF₃(g). However, observed boron content is higher in RE-01 (0.0212 ± 0.0002  mol%) than in RE-02 (0.01764 ± 0.00009 mol%), and both are higher than in Salt-01 (0.0052 ± 0.00012 mol%). Bates and Quist probe the structure of BF⁻₄ anions in alkali-metal fluoroborate melts, including LiBF₄, by IR emission spectrum and Raman spectroscopy; they report that the samples decomposed too rapidly for reliable emission data to be collected.¹⁶ Polyakova et al. studied the electrochemical behavior of soluble boron in molten KBF₄+LiF-NaF-KF in the range of 2.6 to 15.9 mol% at 550°C to 700°C.¹⁷ Therefore, the observation of some amount of FLiBe-soluble boron is not surprising.

Formation of LiBF₄ (Eq. 3) is thermodynamically favorable and can be considered as a possible solubilization mechanism for BF₃(g) in FLiBe.

If the Reaction 3 is responsible for the consumption of NiF₂ in RE-01 from 1.2 mol% to 0.09 mol% and all of produced B remains soluble in the salt, then 0.74 mol% B would be expected in RE-01 salt. The measured value is lower by a factor of 40. If Reaction 3 is at equilibrium with 0.09 mol% NiF₂ and 1 atm partial pressure of N₂(g) bubbles, then 5 mol% B would be expected in RE-01 salt, 250 higher than the observed value. Another reaction to consider is the thermal degradation of LiBF₄:

If Reaction 4 is at equilibrium with 1 atm partial pressure of BF₃(g) bubbles, then 0.6 mol% LiBF₄ would be expected in RE-01 salt. The measured value is lower by a factor of 28. Challenging the bounding assumption of 1 atm BF₃(g), the partial pressure of BF₃(g) in equilibrium with the observed 0.02 mol% LiBF₄ is calculated to be 3600 Pa.

In summary, soluble boron species are observed in small concentrations in the TRE salt, assumed to be solvated by FLiBe as LiBF₄ (i.e. BF4⁻ anions). The majority of the boron produced by the reduction of NiF₂ is not retained in the salt.

An LaF₃ solubility limit of 0.140 ± 0.003 mol% is observed, based on samples collected in solid form from RE-01 and RE-02. This value is similar to the LaF₃ solubility of 0.14 mol% at 480°C and 0.11 mol% at 459°C, reported in LiF-BeF₂-UF₄ (62.8-36.4-0.8 mol).¹³ This indicates that with a surface area of < 0.5 cm², and a salt inventory of < 0.8 g, the equilibration time between the LaF₃ crystal and FLiBe is less than 15 hours. The impact of 0.14 mol% of LaF₃ on the activity coefficient of NiF₂ in FLiBe solution is not known. However, the reasonably short equilibration time provides for a way to ensure a stable and repeatable TRE potential.

The presence of Eu in solution indicates that Eu has a non-zero solubility in FLiBe. RE-02 salt (15 hours at temperature) has 0.0011 mol% Eu; RE-01 salt (>100 hours at temperature) has 0.0015 mol% Eu. It is uncertain if the dissolution of Eu is limited by Eu solubility in FLiBe, or by the dissolution rate of the LaF₃ crystal, or by the diffusivity of Eu in the LaF3 crystal. It is not expected for Eu at concentrations as low as 0.0015 mol% to impact the activity coefficient of NiF₂ and the electrode potential is expected to be stable even if Eu dissolution has not yet reached equilibrium.

Only a small fraction of the boron reaction product is retained in the RE-01 salt after 100 hours of operation: 3% retention of the boron products in RE-01 after 100 hours of operation with an initial NiF₂ loading of 1.2 mol%, and ≥16% retention of the boron reaction products in RE-02 after 15 hours of operation with an initial NiF₂ loading of 0.53 mol%. It is advantageous to retain in the salt only a small fraction of retained boron products because soluble BF₄⁻ could impact the complexation of Ni²⁺ in FLiBe solution, and hence change the NiF₂ activity coefficient. The impact of x_BF4- on the activity coefficient of Ni²⁺ in FLiBe merits further study.

The BN body with a 0.3 mm diameter hole and the LaF₃ crystal provide an efficient separation between the TRE salt and the main salt. This barrier is not penetrated by nickel as demonstrated by no increase in nickel content between Salt-02 and Salt-01 (t-test, 95% significance). In a FLiBe solution where only LiF and BeF₂ are present in the melt, such as the main salt bath, BN should maintain its stability (see Ellingham diagram in Figure 5). BN is chemically stable in FLiBe, as demonstrated by no increase in boron between Salt-02 and Salt-01.

Lanthanum is detected in Salt-01 at a value higher than in Salt-02. The LaF₃ dissolution rate into the main salt is limited by the small LaF₃ surface exposed, pointing to the importance of the BN TRE body in protecting the LaF₃ crystal from dissolution in FLiBe. The LaF₃ crystal maintains its structural integrity because the dissolution of the LaF₃ single crystal membrane is limited by the 0.14 mol% LaF₃ solubility limit in the TRE compartment and the flux to the main salt bath is limited by the small surface area exposed (<0.25 mm²). The LaF₃ membrane stability could be negatively impacted by temperature transients and freezing and melting cycles of the TRE, which could drive further LaF₃ removal from the crystal membrane instead of re-dissolving deposited LaF₃. The integrity of the LaF₃ membrane upon several thermal cycles merits further study. Based on the Ellingham diagram in Figure 5, the LaF₃ membrane should not be subject to reduction reactions in neither the NiF₂-loaded TRE salt nor the main salt bath.

To avoid interaction of NiF₂ with BN, alternative redox couples could be used for a FLiBe TRE. Some of the redox couples that are below BF₃ on the Ellingham diagram include: AlF₃, MnF₂, TiF₂, and TiF₃.¹⁵ The amount of information available for Ni ions in FLiBe (solubility values, formal potentials, the chemical and electrochemical characterization of the Ni/Ni(II) redox couple, Nernstian behavior and reversibility) and the relatively low cost and easy manufacturability of metallic Ni electrodes will need to be weighed against the inconvenience of managing the NiF₂-BN reaction.

To avoid NiF₂ depletion from the electrode cup caused by the identified reaction with BN, and to extend the duration of stability of the TRE, excess NiF₂ could be added to the electrode compartment. A NiF₂ content higher than three or four times its solubility value would most likely extend the electrode lifetime considerably. This solution will be tested in future experiments with this type of TRE. This engineering solution could be susceptible to buildup of soluble boron compounds in the TRE salt, which could cause electrode drift by impacting the activity coefficient of Ni²⁺. Hence this engineering solution must be accompanied by testing of the boron content in the TRE salt as a function of operation time.

Alternatively, the geometry of the TRE can be optimized to eliminate contact surface area between the TRE salt and the BN tube by containing the electrode salt in a cup machined from a LaF₃ crystal. The BN body on the exterior of the LaF₃ crystal is still needed to limit the contact surface area between LaF₃ and the main salt. This is important because with a sufficiently large amount of main salt the 0.14 mol% solubility of LaF₃ in FLiBe can lead to loss of the integrity of the LaF₃ membrane: 370 mg of LaF₃ would dissolve in 43 g of FLiBe, which is 45% of the LaF₃ crystal. Dissolution of LaF₃ in the TRE salt is not a concern because the TRE salt inventory is small relative to the size of the LaF₃ membrane (830 mg): 6 mg LaF₃ would dissolve in 700 mg of FLiBe, which is 0.7% of the LaF₃ crystal.