A vitreous carbon crucible was placed in a cylindrical vessel made of refractory steel, closed to the top by a stainless-steel lid cooled by circulating water. The walls of the vessel were protected from chloride vapours by a graphite liner. The atmosphere of the cell was argon gas (99.995% purity, Linde). Heating was ensured by a programmable furnace and temperature was measured by a chromel-alumel thermocouple.

The electrolytic bath consisted of an eutectic mixture (58.2 mol.% - 41.8 mol.%) of LiCl (Sigma Aldrich, purity 99.98%) and KCl (Fox Chemicals, purity 99.99%). Li₂O (Fox chemicals, purity 99.9%) was used as a solute to target desired O²⁻ concentrations. All salts and solutes were stored in a glovebox under an argon atmosphere.

A classical three-electrode set-up was used for electrochemical measurements.

Working electrodes (WE) were made of metallic wires or cylindrical metallic pieces. Stainless steel 316L (69Fe-18Cr-10Ni-3Mo) and its main components were chosen for this study and supplied by Goodfellow. Table 1 gathers the shape and purity of each metal and alloy used in this study.

For electrochemical measurements a tungsten counter-electrode (CE) was used. It was ensured that the immersed surface area of the CE was greater than the WE one. The pseudo-reference electrode was a two millimeter diameter glassy carbon

rod. A 316L stainless steel plate was also used for the long-term immersion test.

Linear sweep voltammetry, square wave voltammetry and galvanostatic electrolysis were performed with an Autolab PGSTAT302N controlled by the software Nova 2.1.5.

The immersed samples were embedded in a conductive resin and polished to be further analysed by Scanning Electron Microscopy (SEM) coupled with Electron Dispersive Spectroscopy (EDS).

The oxide ions anodic behaviour in LiCl-KCl at 550˚C was first studied by Li₂O additions using square wave voltammetry. Figure 1 presents square wave voltammograms performed at 81 Hz on a tungsten electrode for various O²⁻ quantities.

The exponential signal (B), observed at 0 V is associated with chloride ions oxidation into chlorine gas and was used as an internal reference potential.

When Li₂O is added, a single anodic peak (A) is observed at −0.8 V/(Cl₂/Cl⁻) and its corresponding peak current density increases with the addition of Li₂O: it is thus associated with the oxidation of O²⁻. Despite the bubbling observed at higher anodic potentials (between −0.5 and 0 V/(Cl₂/Cl⁻)), the shape of the peak is not disturbed, allowing an accurate measurement.

The differential peak current densities measured at −0.8 V/(Cl₂/Cl⁻) are reported against O²⁻ concentration in Figure 2 .

A linear relationship is exhibited following the equation:

$\text{δip}=\left(0.85\pm 0.04\right)\left[{\text{O}}^{2-}\right]+0.0087$ (2)

where δip is the differential peak current density (A·cm⁻²) and [O²⁻] is the oxide ions molality in the salt (mol·kg⁻¹).

Using this linear relationship, the residual amount of oxide ions in the salt was estimated using the standard additions method to be around 0.01 mol·kg⁻¹. The slope of the calibration curve is 0.85 A·cm⁻² (mol·kg⁻¹)⁻¹ and was used to accurately quantify the oxide ions concentration in the salt before performing a corrosion experiment.

To investigate pure metals and SS316L behaviour in LiCl-KCl-Li₂O at 550˚C the same methodology was used. Linear sweep voltammetry were first performed in LiCl-KCl at 550˚C at different oxide ions concentrations, potential scan rates were determined and optimised to have the least effect on the reaction kinetics. To complete the linear sweep voltammetry results, pure metals were then immersed in a bath containing 0.150 mol O²⁻ kg⁻¹ for 8 hours and SS316L for 168 hours in a bath containing 0.08 mol O²⁻ kg⁻¹. The immersed samples were then studied by SEM-EDS analysis.

Figure 3 gathers the linear sweep voltammograms obtained on a 2 mm diameter chromium rod for different oxide ions additions and the SEM micrography taken after immersion in LiCl-KCl at 550˚C.

Figure 3(a) displays three linear voltammograms and their logarithmic representations obtained on chromium for different oxide ions contents. An exponential increase in the current density is observed close to the open-circuit potential (OCP), indicating a corrosion phenomenon: thus, this potential can then be attributed to the corrosion potential (E_corr). To better understand the Cr behaviour, logarithmic representations were used to accurately measure the corrosion potential and to estimate the corrosion rate.

For the lowest oxide ions content, chromium is directly dissolved into the bath. The oxide ions concentration increase leads to a decrease of both corrosion potential from −2.02 V/(Cl₂/Cl⁻) to −2.22 V/(Cl₂/Cl⁻) and corrosion rate from 3.4 10⁻⁴ A·cm⁻² to 6 × 10⁻⁵ A·cm⁻²: the oxide ions presence seems to stabilise the Cr metal. The formation of a protective layer on the electrode surface may be envisaged. To confirm it, an immersion test was performed and the sample analysed by SEM-EDS spot analysis ( Figure 3(b) ). Pure chromium is observed at the core, as well as a 50 µm-layer at the salt/metal interface. Elemental analyses showed an average molar composition of 38.5% Cr and 61.5% O, corresponding to Cr₂O₃.

Those results are in agreement with thermodynamic prediction regarding the spontaneous reaction between Cr and oxide ions. The Gibbs energy at 550˚C of Cr₂O₃ formation was calculated using the HSC 6.12 software:

$2\text{Cr}+1.5{\text{O}}_2={\text{Cr}}_{\text{2}}{\text{O}}_{\text{3}}\Delta G_{550˚\text{C}}^0=-457\text{kJ}\cdot {\text{mol}}^{-1}$ (3)

This highly negative value of the Gibbs energy indicates the Cr₂O₃ spontaneous formation in the presence of oxide ions when Cr and O²⁻ are simultaneously present in the salt.

Cr₂O₃ formation was already observed in different molten chloride salts [16] but was reported to be non-protective and easily dissolved [13] .

The same methodology was applied to iron and Figure 4 gathers the electrochemical signals and the SEM-EDS analysis.

The linear voltammograms ( Figure 4(a) ) exhibit three exponential signals towards the OCP showing a corrosion phenomenon whatever the oxide ions concentration. Moreover, no notable change is observed in the logarithmic representations in both corrosion potential (−1.86 V/(Cl₂/Cl⁻)) and corrosion rate (around 4 × 10⁻⁴ A·cm⁻²). Thus, oxide ions content has no influence on iron corrosion behaviour in LiCl-KCl at 550˚C.

The cross-sectional SEM micrography of the iron wire ( Figure 4(b) ) after the 8-hour immersion in LiCl-KCl ([O²⁻] = 0.150 mol·kg⁻¹) shows a highly degraded salt/metal interface: a complementary EDS analysis was performed where only pure iron was observed. Thus, no protective oxide layer is formed and iron is dissolved in the molten solvent.

However, Feng and Melendres observed the formation of an unstable oxide film progressively dissolved by the chloride medium in molten LiCl-KCl at 450˚C in the presence of oxide ions ( [17] . Thus, the following assumption can be made: the iron dissolution comes from either directly the metal oxidation or from the formation of an unstable Fe-O species.

Figure 5 presents the experimental results obtained on a nickel wire.

The linear voltammograms show three exponentially-shaped signals obtained on a nickel wire where oxide ions content seems to have a low impact on the Ni behaviour ( Figure 5(a) ). The logarithmic representation gives new information compared to the sole linear voltammetry. As oxide ions concentration increases, E_i=0 drops from −1.53 V/(Cl₂/Cl⁻) to −1.84 V/(Cl₂/Cl⁻) while the exponential potential remains identical remains identical. A very low current density plateau (≈1 × 10⁻⁴ A·cm⁻²) is observed in between, highlighting the corrosion resistance of nickel.

To confirm these results, an immersion test followed by SEM-EDS analysis was performed on nickel. The cross-sectional micrography taken after an 8-hour

immersion ( Figure 5(b) ) exhibits a nickel wire with an intact interface. EDS analysis was also performed and only pure Nickel was detected. Nickel is indeed resistant to corrosion and oxide ions have no influence on its behaviour, unlike Fe and Cr.

These observations are consistent with various studies showing a better corrosion resistance of nickel in molten chlorides compared to iron as an alloying compound [10] [18] [19] .

Figure 6 summarises the experiments carried out to study the influence of oxide ions on a molybdenum wire.

The linear voltammograms plotted on molybdenum ( Figure 6(a) ) highlight the influence of oxide ions on the metal. For the low oxide ion molality (4 × 10⁻³ mol·kg⁻¹), an exponential signal is observed on the linear voltammograms. However, on the logarithmic representation the corrosion potential is 0.6 V lower than the exponential and a low current density domain is present in between, thus Mo exhibits a corrosion resistance. With the oxide ions content increase, the corrosion potential decreases. Moreover, a new electrochemical signal is evidenced at -1.3 V/(Cl₂/Cl⁻). Its peak current density is proportional to [O²⁻] and the signal is characteristic of a diffusion limited gas evolution, suggesting the formation of a Mo-O gaseous species.

After immersion of a pure molybdenum wire in LiCl-KCl ([O²⁻] = 0.150 mol·kg⁻¹), EDS analysis showed an average molar composition of 72.3% oxygen and 27.7% molybdenum corresponding to MoO₃. According to the HSC 6.12 database, the formation of this oxide is spontaneous $\Delta G_{550˚\text{C}}^0=-535\text{kJ}\cdot {\text{mol}}^{-1}$.

MoO₃ was studied by Volkovicha et al. [14] in NaCl-KCl at 700˚C and observed that it reacted with NaCl to form gaseous MoO₂Cl₂. Thus, MoO₃ is spontaneously formed when Mo is in contact with O²⁻ and reacts with Cl⁻ to create a gaseous species, confirming the electrochemical observations. The oxide ions have then a negative effect on Mo corrosion behaviour.

Finally, the same experiments were carried out on 316L stainless steel, the linear voltammograms are displayed in Figure 7 .

The linear voltammograms and their logarithmic representations exhibit the complex behaviour of SS316L in the presence of oxide ions. When the O²⁻ content is low, 316L stainless steel shows a dissolution behaviour. Its corrosion potential is equal to −1.72 V/(Cl₂/Cl⁻) and SS316L is thus unstable in a bath with low oxide ions content. As the oxide ions content increases, E_corr shifts from −1.72 to −1.9 V/(Cl₂/Cl⁻). A new electrochemical signal emerged at E ≈ −1.45 V/(Cl₂/Cl⁻) followed by a low current density plateau (i ≈ 1 × 10⁻³ A·cm⁻²) between −1.4 and −1.15 V/(Cl₂/Cl⁻). A shift of the exponential potential towards higher potentials is observed. These observations indicate the formation of a passivating compound.

A 168 h immersion of a 316L stainless steel plate in LiCl-KCl at 550˚C containing 0.081 mol O²⁻ kg⁻¹ was then carried out. A micrograph was taken and a cross-sectional element mapping was performed, results are presented on Figure 8 .

The SEM micrograph ( Figure 8(a) ) highlights the degradation of the alloy where a 75 µm-depth attack is observed after 168 h. Two distinct zones of attack can be identified, the first starting at the edge of the plate and reaching a depth of 35 µm (I symbol in Figure 8(a) ), and the second one from 35 µm to 75 µm (II symbol on Figure 8(a) ). This figure also displays the element mapping where a low concentration is symbolized by a darker colour and a high concentration is symbolized by a lighter colour.

The behaviour of alloying compound can be individually analysed and compared to its pure form.

Chromium is reactive towards oxygen and diffuses towards the interface where its relative concentration increases ( Figure 8(b) ). Comparing Figure 8(b) and Figure 8(d) , Cr is located next to the O element, confirming a spontaneous reaction.

Iron, main compound of the alloy (69%), diffuses towards the surface and dissolves into the bath ( Figure 8(c) ). In the zone I, the dissolution of iron reaches 90% of the original content. In the zone II, iron concentration is halved compared to its core composition. Iron in the alloy exhibits the same behaviour as its pure form.

Nickel was observed to be resistant to corrosion whatever the oxide concentration. During this immersion test, no depletion of nickel is observed ( Figure 8(e) ).

A slight depletion of Molybdenum is observed in the zone where O element is found ( Figure 8(f) ). The experiments conducted on pure molybdenum showed its spontaneous reactivity with oxides, first forming MoO₃, and then a gaseous MoO₂Cl₂ species, explaining the Mo depletion.

To resume, each alloying compound has the same behaviour as its pure form.