Corrosion Behavior of VM12-SHC Steel in Contact with Solar Salt and Ternary Molten Salt in Accelerated Fluid Conditions

: Ternary low melting point mixtures with the addition of LiNO 3 and Ca(NO 3 ) 2 have been presented as direct system candidates for CSP technologies due to having better physical and chemical properties than those of Solar Salt. In this study, thermal, physical and chemical properties are measured as is the corrosive behavior of stainless alloy VM12 (Cr 12%) when in contact with Solar Salt, 60% NaNO 3 -40% KNO 3 (wt.%) and ternary 46% NaNO 3 -19% Ca(NO 3 ) 2 -35% LiNO 3 (wt.%). Gravimetric weight change measurements were performed on the test specimens, which were tested under accelerated ﬂuid conditions (0.2 m s − 1 ) at 500 ◦ C for 2000 h. This research conﬁrms the potential of this novel formulation as a thermal storage medium and validates the suitability of ferritic VM12-SHC stainless steel as a structural material for CSP technology with Solar Salt. Meanwhile, the results obtained by scanning electron microscopy and X-ray diffraction indicate a reduction in the protective character of the oxide layer formed on this alloy when the medium contains calcium and lithium components.


Introduction
Wealth and growth in society are associated with the capacity and independence of electricity generation, whereas conventional fossil fuel technology is a heavily polluting industry, mainly because it is an intensive producer of greenhouse gases (GHG) [1]; the depletion of natural resources also constitutes a reason to strengthen alternative renewable energy sources, such as concentrated solar power (CSP) plants.
CSP plant technology has a competitive status amongst other renewable sources, such as photovoltaic, hydraulic and wind energy, because it may be coupled with a thermal energy storage system (TES) that yields heat to the power block overnight or at peak demand. Between the two principal commercial CSP technology layouts, the central tower (CT) and the parabolic trough (PT), the latter is the more mature, fostering up-to-date production with higher electricity [2]. In commercial PT plants, the thermal oil used as a heat transfer fluid (HTF) releases heat to the thermal energy storage system (TES) with molten salt TES by means of a heat exchanger. Nevertheless, there is an inherent freezing risk of the nitrate mixture, which eventually clogs pipes and pumps due to the high melting point (~227 • C) of the current state-of-the-art Solar Salt (60% NaNO 3 -40% KNO 3 ); therefore, researchers are concentrating on new fluids with better physical-chemical properties than those of Solar Salt and a higher working temperature range, lower melting point and higher decomposition temperature [3].
The composition of the nitrates is detailed in Table 1. The impurity levels of chlorides and sulphates were Na(SO 4 ) 2 of up to 500 ppm and 250 ppm in the formulations, respectively. These impurities are also involved in molten salt container degradation and the physical-chemical property change [27]. In this investigation, thermodynamic properties were measured, including the melting and degradation points, specific heat, density and viscosity of the molten salts.

•
Melting point and thermal decomposition Differential scanning calorimetry (DSC) is the most widely used technique for the melting point determination of molten salts. [8,28,29]. The melting point was measured using a TA Instruments DSC mod. Q20 coupled with a cooling system, which allows a working temperature range between −90 • C and 550 • C and a temperature precision of ±0.1 • C. Hermetic aluminum pans were employed on the sample holder in the cell. The analysis was conducted in an inert atmosphere (N 2 ), from room temperature up to 500 • C and a heating rate of 10 • C/min, which is the optimum testing velocity to identify phase transitions and melting points [30].
For the thermal degradation study, Q600 TA instrument thermogravimetric analysis (TGA) equipment was used. Its horizontal balance, with a 200 mg capacity and 0.1 µg sensitivity, was placed in a furnace with a maximum working temperature of 1500 • C and programmable heating rates of between 0.1 • C/min and 25 • C/min. Both the heating rate and the atmosphere parameters affect the degradation temperature evaluation [31]. Commonly, a 10 • C/min heating rate and N 2 or Ar inert atmospheres are used [32,33]. In these experiments, the TGA of the specimens was performed in a platinum crucible from room temperature up to 800 • C under a flowing N 2 atmosphere of 100 mL·m −1 . Prior to the analyses, both instruments were calibrated with indium.

•
Specific Heat capacity measurement The TA Instruments DSC mod. Q20 was used to measure the specific heat capacity (Cp) by its modulated differential scanning calorimetry (MDSC) feature. The standard MDSC was conducted according to the ASTM standard E1269; it was also implemented by other researchers [34]. To ensure the reproducibility of the measuring setup because of the facility sensitivity, several experiments under similar conditions were conducted for each salt: "Solar Salt" and 46% NaNO 3 -19% Ca(NO 3 ) 2 -35% LiNO 3 . The sample masses (around 3 mg) of molten salts were adequately deposited, covering the bottom of the aluminum pan in every experiment.  Density measurement Density (ρ) (g/cm 3 ), is an important parameter to evaluate the mass flow in pipes and the volumetric capacity of tanks [35]. It is certainly a challenging lab measurement procedure at a high temperature (i.e., 500 • C) due to the lack of commercial apparatuses limited up to~300 • C [36]. Therefore, experimentally, this property is less reported in the literature [36]. Most of the apparatuses are based on the Archimedes principle [37,38], while others have been determined by the pycnometer concept [39]. Using this insight, an empty graduated borosilicate glass tube was duly weighted in a calibrated Radwag precision balance, mod. PS-3500 R2. Around 12 mg of powder salt was placed in the tube. Once fused in a verified furnace and stabilized for five hours, the liquid content was weighted, and the occupied volume was recorded. Density was determined at 250 • C, 350 • C, 450 • C and 500 • C, by repeating the measurements five times at each temperature. Then, by means of the relation between the recorded mass (m) and its volume (V), the material density (ρ) was obtained.

•
Viscosity measurement Viscosity µ (cP) is an important factor to take into account in order to minimize the energy consumption of the impeller pumps in CSP plants [40,41]. In this investigation, the dynamic absolute viscosity (µ) was measured, which consisted of an evaluation of the particles' internal resistance against shear stress (σ, (Pa)). This relates viscosity and the share rate (γ, (1/s)) by the following equation [42]: The behavior of the molten salts studied was examined by a TA instrument, mod. Discovery Hybrid Rheometer (DHR1), with an adjustable parallel plate fitting. The whole system was placed in a high temperature furnace capable of reaching 700 • C. In this study, viscosities were evaluated from 200 • C to 500 • C with every increase in 50 • C temperature and a shear rate variation from 10 to 1000 (1/s). Before the measurements, the equipment was calibrated with a reference polymer specimen.

Corrosion Study of the Substrates
Corrosion studies were performed at an isothermal temperature of 500 • C and accelerated fluid conditions in a pilot plant. The velocity of the test was set to 0.2 m·s −1 . This velocity was selected on the basis of the flow velocity, which is given in areas near to pipe bends and valves in the current CSP plants (0.2-0.5 m·s −1 ), which are areas where high levels of corrosion are found. This patented (ES2534850-B2) pilot plant, built in AISI 321 steel, enables one to simulate a medium flowing under real operational conditions. A place area with an alumina sample holder allows one to place specimens, with 20 × 10 mm size, parallel to the molten salt flow for degradation tests of materials. This facility was explained in detail in a previous work [16]; it was used to compare static conditions versus accelerated flow conditions. It is feasible equipment for simulating corrosion effects when salt circulates through steel tubes in a CSP plant.
VM12-SHC steel was the substrate selected for corrosion evaluation. The chemical composition of the material was characterized as received by SEM-EDX in a weight percent of 0.38% Si, 0.25% V, 12% Cr, 0.42% Mn, 1.78% Co, 0.36% Ni, 0.39% Mo and 1.42% W and Fe in balance. These values are in line with those reported by the manufacturer [20]. Moreover, an XRD was performed on the material before the test (see Figure 1). Once the samples were removed from the pilot plant, they were cooled slowly in warm distilled water to eliminate the remaining solidified salt with which they had been in contact. They were then dried and weighed, and an average was taken from five values of their weights. Thus, the formula used to calculate the mass variation over time was where m i is the initial mass of the specimen, m f is the mass of the sample at the selected time, and S 0 is the initial area of the specimen. Table 2 presents all the results obtained for the main physical-chemical properties of both mixtures studied. The DSC revealed a melting point of the Solar Salt at 227 • C, along with its characteristic endothermic phase solid-solid transition (α, β) of the KNO 3 at around 132 • C [43]. This signal also appears in higher compositional formulations and can be shifted at a different heating rate [38]. The exothermic curve with a maximum peak at 214 • C was produced in the cooling segment slope and delayed with respect to the melting point because of atom nucleation [44] (see Figure 2a). The decomposition temperature was 597 • C. This corresponds to a mass loss of 3% with respect to the stability slope (see Figure 2b). Generally, authors assign a working temperature range of 240 • C-565 • C [45]. The average specific heat capacity at 500 • C was 1.55 J/g • C (see Figure 3), in line with other investigations [46]. Regarding the viscosity, the average value at 300 • C was 5.06 cP, which is higher than the 3.26 cP reported by González-Roubaud [47]. Most of the researchers stated Arrhenius or polynomic equation types with a temperature dependence of the viscosity. In this case, the relation one fits better with the polynomic mathematic expression, µ = 0.002·T 2 − 0. 140·T + 33.805 (see Figure 4a).

Characterization of the Salt Mixture
The average value obtained for density for the range between 250 • C and 500 • C was 1.804 g/cm 3 (see Figure 4b). Antoni Gil et al. [48] determined a similar average of 1.870 g/cm 3 between 265 • C and 565 • C.
In comparison with "Solar Salt", the studied ternary mixture, 19% Ca(NO 3 ) 2 -46% NaNO 3 -35% LiNO 3 , melted at 176.04 • C and was stable up to 575 • C (see Figure 5). There were two endothermic peaks at 86.63 • C and 217 • C due to water loss from Ca(NO 3 ) 2 .4H 2 O during the first thermal cycle; the second moisture release required a higher amount of energy (17.01 J/g) [49]. When calcium content is reduced, mixtures tend to have higher stability as observed by Judith C. et al. [44] for ternary Ca-Na-K nitrate combinations. Moreover, the thermal stability enhancement is a consequence of lithium oxides, Li 2 O and Li 2 O 2 , as reported by T. Wang et al. [50]. Thus, this investigated ternary mixture increases thermal range with respect to Solar Salt due to LiNO 3 presence.  The viscosity values at 500 • C were 2.93 cP. Meanwhile, Solar Salt resulted in 1.12 cP, with an equation tendency, µ = 8.64·10 6 ·T −2.437 , for the curve represented at different temperatures (see Figure 6a). LiNO 3 presence enables one to reduce viscosity and raise Cp as evaluated by Coscia, Kevin et al. [51]. In this work, the ternary mixture 19% Ca(NO 3 ) 2 -46% NaNO 3 -35% LiNO 3 presented a specific heat capacity of 1623 J/g· • C at 500 • C, higher than that of Solar Salt (see Figure 3). This is similar to the average value collected by Zhang Pheng et al. [32] for the formulation 30% LiNO 3 -18% NaNO 3 -52% KNO 3 up to 380 • C; when formulation varied to 20% LiNO 3 -28% NaNO 3 -52% KNO 3 , the heat capacity suffered a significant reduction value, 1.091 kJ/kg·K at 390 • C [52].  The density evaluations resulted in a linear decrease with temperature according to the expression ρ = 2.1968 − 0.0011·T, and a lower density than that of Solar Salt was presented in all temperature ranges caused by the large amount of LiNO 3 (see Figure 6b).
The energetic density ( (a) (b) The density evaluations resulted in a linear decrease with temperature according to the expression ρ = 2.1968 − 0.0011·T, and a lower density than that of Solar Salt was presented in all temperature ranges caused by the large amount of LiNO3 (see Figure 6b).
The energetic density (ƍɛ) calculations for the ternary mixture and the reference "Solar Salt" were taken from the experimental measurement, being expressed as follows: ƍɛ = ρ·Cp·ΔT [52], where ρ is the density, Cp is the specific heat capacity, and ΔT is the working temperature range. The binary mixture presented 721 MJ/m 3 at 500 °C, similarly to the data reported by Wang et al. [53] for the same temperature, 756 MJ/m 3  So, the studied ternary mixture presented higher storage capacity for a parabolic trough system, although its drawback is the resistance to flow. Besides, its corrosive impact must be evaluated in different materials. Figure 7 shows the gravimetric results of the VM12 in Solar Salt and the new ternary mixture formulated at 500 °C with a molten salt flow rate of 0.2 m/s. There was barely any weight loss when in contact with 60% NaNO3-40% KNO3. A sinusoidal shape was observed during the test when in contact with 46% NaNO3-19% Ca(NO3)2-35% LiNO3, with a gained weight mass of 0.5 mg/cm 2 after 2000 h. ε) calculations for the ternary mixture and the reference "Solar Salt" were taken from the experimental measurement, being expressed as follows: The density evaluations resulted in a linear decrease with temperature according to the expression ρ = 2.1968 − 0.0011·T, and a lower density than that of Solar Salt was presented in all temperature ranges caused by the large amount of LiNO3 (see Figure 6b). The energetic density (ƍɛ) calculations for the ternary mixture and the reference "Solar Salt" were taken from the experimental measurement, being expressed as follows: ƍɛ = ρ·Cp·ΔT [52], where ρ is the density, Cp is the specific heat capacity, and ΔT is the working temperature range. The binary mixture presented 721 MJ/m 3 at 500 °C, similarly to the data reported by Wang et al. [53] for the same temperature, 756 MJ/m 3 . The investigated ternary combination of nitrates of this research, Li-Ca-Na, with 35 wt.% LiNO3 resulted in 867 MJ/m 3 , which is an outstanding value when compared with that of other authors. For instance, Parrado et al. [9]  . So, the studied ternary mixture presented higher storage capacity for a parabolic trough system, although its drawback is the resistance to flow. Besides, its corrosive impact must be evaluated in different materials. Figure 7 shows the gravimetric results of the VM12 in Solar Salt and the new ternary mixture formulated at 500 °C with a molten salt flow rate of 0.2 m/s. There was barely any weight loss when in contact with 60% NaNO3-40% KNO3. A sinusoidal shape was observed during the test when in contact with 46% NaNO3-19% Ca(NO3)2-35% LiNO3, with a gained weight mass of 0.5 mg/cm 2 after 2000 h. ε = ρ·Cp·∆T [52], where ρ is the density, Cp is the specific heat capacity, and ∆T is the working temperature range. The binary mixture presented 721 MJ/m 3 at 500 • C, similarly to the data reported by Wang et al. [53] for the same temperature, 756 MJ/m 3  So, the studied ternary mixture presented higher storage capacity for a parabolic trough system, although its drawback is the resistance to flow. Besides, its corrosive impact must be evaluated in different materials. Figure 7 shows the gravimetric results of the VM12 in Solar Salt and the new ternary mixture formulated at 500 • C with a molten salt flow rate of 0.2 m/s. There was barely any weight loss when in contact with 60% NaNO 3 -40% KNO 3 . A sinusoidal shape was observed during the test when in contact with 46% NaNO 3 -19% Ca(NO 3 ) 2 -35% LiNO 3 , with a gained weight mass of 0.5 mg/cm 2 after 2000 h. The weight variations correlated with a noticeable spallation in different areas of the sample as may be seen in a visual inspection of the coupons (see Figure 8).  The identified magnesium atomic percentages, #1:2.42 and #2:2.27 (see Figure 9), come from salt impurities, Mg(NO 3 ) 2 , forming unprotected complex oxides from a reaction with the outer Fe 2 O 3 (see Equation (3)). Besides, some researchers credited magnesium reactivity with a direct responsibility for NOx emissions during the plant commissioning [8].

Corrosion Behavior in VM12
Regarding the ternary mixture developed, although Ca(NO 3 ) 2 is one of the main components of the designed ternary mixture and molten salt impurities, the SEM-EDX atomic spectrum analysis (see Figure 10b) revealed a low presence of Ca specimens. That also means a low carbonate existence due to the large LiNO 3 amount and its greater activity. Severe delamination occurred (see Figure 10a), and a loss of corrosion products could be observed in the images of the surfaces (light color areas); this was confirmed by the increasing amount of Cr and Fe with lower O (22.43%) incidence after the SEM-EDX #1 chemical element analysis spectrum (see Figure 10b). The fluid movement provoked a pulling force that encouraged the mechanical separation of the corrosion products from the substrate.
When the material transversal section was compared after its exposure to the different molten salts, it was observed that the SEM cross-section of VM12 showed a low corrosion product generation with a 970 nm layer after being in contact with Solar Salt for 1000 h. General corrosion persisted; at the end of the experiment, however, there were uneven corrosion layers with a maximum thickness of 6.7 µm (see Figure 11a,b). The line scan analysis indicated a majority presence of Fe, Cr and O, as well as Mg to a lesser extent. These findings are in line with the posterior structural phases identified in X-ray diffraction. The outer corrosion product scales (see Figure 11b) are promoted and formed by the impurities in molten salts at a high temperature [55].
When compared with the sample cross-section after 2000 h of exposure to 19% Ca(NO 3 ) 2 -46% NaNO 3 -35% LiNO 3 , VM12 showed a layer thickness of 2.23 µm ( see Figure 12a) with a compact zone close to the interface lengthwise). In the SEM-EDX linear spectrum of the oxide layer (see Figure 12b), it can be correlated with the intensity change of the Cr signal, of which Fe and O together had the higher counts. Moreover, weak Mg and Ca signals were observed, which are correlated to the SEM-EDX results on the surface of the sample at the end of the test (see Figure 10b).  Figure 13a shows the XRD analysis of VM12 after "Solar Salt" exposure, It reveled that the crystalline complex CaFe 16 Mg 2 O 27 was the option that best suited their intensity and position peaks. This crystalline phase evidenced the interaction between hematite and Mg +2 and Ca +2 (see Figure 11b). It is worth mentioning only crystalline species can be detected, which means that existing species with amorphous structures are not identified by XRD. On the other hand, the XRD analysis of the material in contact with the ternary mixture (see Figure 13b) confirmed a corrosion multi-layer mainly composed of LiFe 5 O 8 , hematite, magnetite and FeCr 2 O 4 . Although NaNO 3 accounted for 46% of the total ternary mixture weight, any crystalline structure could be identified with Na. Their activity was probably slowed down because of the greater Li + activity. This is a positive phenomenon against impurity concentration enhancement, such as NaCl or Na 2 SO 4 [56]. The XRD diffractogram also corroborated that lithium reacted with Fe 2 O 3 to generate LiFe 5 O 8 (see Equation (4)), as well as the unprotected MgFe 2 O 4 oxide, which justifies the spallation observed.
There were barely any remarkable phase intensity peaks related to crystalline hematite, but magnetite signals may be allocated more precisely. Thus, there was a rapid consumption of the generated Fe 2 O 3 that reacted with the lithium oxide. This phenomenon was also observed by Wei-Jen Cheng et al. [7] in a similar chromium content alloy X20 (12 wt.% Cr) in the presence of a ternary nitrate mixture of Li, Na and K in static conditions; at the same time, outer corrosion scales corresponded to lithium oxide layers and the inner one to protective chromium spinel.
Furthermore, the protective specimen LiFeO 2 might be formed by depletion of the LiFe 5 O 8 that grew on the substrate [7]. This also could occur by a direct reaction between magnetite and lithium oxide Li 2 O: Despite these reaction types, the two above-mentioned mechanisms were not preferential with VM12 and this ternary fluid in the accelerated conditions, and the protective oxide was not identified in XRD. Figure 14a,b shows the corrosion mechanism proposed for the ternary mixture and "Solar Salt" as a result of this study. In short, there have been studies on neither ternary molten salt mixtures nor accelerated conditions in VM12 to date; nevertheless, the first investigation of VM12 immersed in 60% NaNO 3 -40% KNO 3 resulted in a good corrosion resistance, with a weight gained of ≈1 mg/cm 2 after 1000 h of tests in static conditions [57], which is higher than the mass obtained in this research, 0.3 mg/cm 2 . This difference can be explained by the fact that the working temperature was 580 • C. D. Fähsing et al. [25] Energies 2021, 14, 5903 13 of 16 investigated this material under Solar Salt at 560 • C and with an addition of 300 ppm Cl − and SO 4 −2 impurities, the latter in the same range used in this research. The material had 0.07 mg/cm 2 after 2000 h, and a protective oxide chromium scale in the inner part of the multi-layer and corrosion products that is consistent with those obtained in this investigation was detected. Nevertheless, the higher mass gained could be attributed to the effect of fluid circulation in this research. VM12 tended to form protective layers very quickly during the first test hours in contact with the accelerated ternary mixture. As a consequence of their insufficient compactness and stability, it suffered scale spallation from 500 h of running. In contrast, the substrate in contact with Solar Salt generated a more stable multilayer oxide scale.

Conclusions
This work assessed the physical-chemical properties of a ternary mixture containing 46% NaNO 3 -19% Ca(NO 3 ) 2 -35% LiNO 3 , the corrosiveness effect in the ferritic alloy VM12 and its comparison with the current state-of-the-art fluid. Corrosion tests were evaluated in a patented pilot plant with a fluid movement of 0.2 m/s during 2000 h of exposure at 500 • C.
The ternary mixture resulted in a melting point of 176 • C and 575 • C degradation, which would make it possible to increase the thermal working range in a direct thermal energy storage system, and a Cp of 1.6 J/g • C gave it higher energy density (MJ/m 3 ) than that of Solar Salt. In contrast, viscosity values were above those obtained for Solar Salt at up to 400 • C because of Ca(NO 3 ) 2 . This would imply more energy consumption from pump impulsion that, together with lithium price, aggravates the levelized cost of energy.
The compatibility studies between the material and fluids were performed with the same chloride and sulphate impurity levels. The corrosion results of the newly formulated fluid in VM12 (12% Cr) demonstrated a particular oxidation corrosion mechanism due to the presence of LiNO 3 , with a superior mass gained of the coupons to that of Solar Salt at the end of the test due to lithium basicity. The formation of compact layers of lithium oxide was not promoted, thus explaining the less protective crystalline structure identification of LiFe 5 O 8 in XRD rather than the upper protective scale LiFeO 2 . The lack of lithium protective oxides along with the complex oxides formed with impurities (MgFe 2 O 4 ) on the surface of the ferritic material yielded severe scale spallation in contact with the ternary mixture. Thus, according to the gravimetric, microstructural characterization and phase analysis of the corrosion products, it can be stated that what prevailed was the preference for the protective chromium oxide formation in the oxidation mechanism.
VM12 shows better compatibility with Solar Salt than with the ternary mixture in an accelerated fluid regime. The mass gained value in accelerated flow conditions harmonized