Deeper Understanding of Ternary Eutectic Carbonates/Ceria-Based Oxide Composite Electrolyte through Thermal Cycling

: Due to a high conductivity of about 0.1 S · cm − 1 , Li-Na-K carbonate eutectic and Sm-doped ceria composite material is a good electrolyte candidate for hybrid fuel cells operating between 500 ◦ C and 600 ◦ C. The present paper aims at a deeper understanding of the species and mechanisms involved in the ionic transport through impedance spectroscopy and thermal analyses, in oxidizing and reducing atmospheres, wet and dry, and during two heating/cooling cycles. Complementary structural analyses of post-mortem phases allowed us to evidence the irreversible partial transformation of molten carbonates into hydrogenated species, when water and/or hydrogen are added in the surrounding atmospheres. Furthermore, this modiﬁcation was avoided by adding CO 2 in anodic and/or cathodic compartments. Finally, a mechanistic model of such composite electrical behavior is suggested, according to the surrounding atmospheres used. It leads to the conclusions that cells based on this kind of electrolyte would preferably operate in molten carbonate fuel cell conditions, than in solid oxide fuel cell conditions, and conﬁrms the name of “Hybrid Fuel Cells” instead of Intermediate Temperature (or even Low Temperature) Solid Oxide Fuel Cells. adminis-tration, A.R.; M.C. A.G.,

It is worth mentioning that the composite material suggested by Zhu et al. [15] is close to the usual electrolyte structure of Molten Carbonate Fuel Cells (MCFC) electrolyte, composed of a lithium aluminate (LiAlO 2 ) porous matrix embedding molten carbonates, and operating close to 600 • C. For this reason, usual binary eutectic compositions,

CO 3 Composite Electrolyte
The composite electrolyte was prepared by the solid-solid method according to the previously described procedure [32,44]. First, a ternary eutectic mixture of Li 2 CO 3 -Na 2 CO 3 -K 2 CO 3 (43.5-31-5-25.0) %mol., (noted hereinafter as LiNaK), was prepared by mixing commercial salt powders (purity ≥ 99.8%, Sigma Aldrich, Saint Quentin Fallavier 38297, France) in stoichiometric amount. The mixture was well ground, heated up to 650 • C (3 • C·min −1 ) and maintained at this temperature for 1 h under dry 30 %vol. CO 2 in synthetic air (Air Liquide, 75 Quai d'Orsay, Paris 75007, France). After cooling (3 • C·min −1 ), the eutectic mixture was carefully ground to attain the state of a fine powder. The second step was the mixing of as-prepared eutectic to a commercially available samaria-doped ceria (SDC) powder Ce 0.8 Sm 0.2 O 1.9 (purity ≥ 99.9%, Fuel Cell Materials, 404 Enterprise Dr, Lewis Center, OH 43035, USA) in proportion of 30-70 %wt. for SDC-LiNaK. Ground composite mixture was thermally treated under the same conditions as the eutectic, with one difference: the cooling was replaced by a quenching to room temperature; then the mixture was ground thoroughly again. All grinding operations during the composite synthesis were performed in a porcelain mortar.
Electrolyte pellets were made from the oxide-carbonate composite powder, by a dry press technique. Placed in a 13-mm-diameter mold, 0.9 g of powder was pressed by uniaxial pressure equivalent to 4 t applied for 5 min. Then, the pellets were sintered in air at 600 • C for 1 h, with a rate of 5 • C·min −1 for both heating and cooling steps. All as-prepared pellets were kept in an oven at 120 • C before and after electrochemical tests.

Electrochemical Analysis
Electrical behavior of SDC-LiNaK electrolyte was carried out by electrochemical impedance spectroscopy (EIS) using a two-electrode symmetric configuration [44]. Electrodes were formed as follows: M-0034 gold paste (Metalor, 74 Warstone Lane, Birmingham, UK) was deposited on both parallel faces of the pellet, heat-treated next for 1 h at 600 • C in air (heating and cooling rates were 1 • C·min −1 ), ensuring as-made electrodes' electrical resistance was lower than 1 Ohm.
To perform the conductivity measurements, the sample was introduced in Open-Flanges™ fuel cell test bench (Fiaxell, Lausanne, Switzerland), allowing to work in single or dual controlled atmospheres. Gold wires connected to gold meshes were used as current collectors on both sides of the pellet. The electrical behavior of such symmetrical half-cells was studied under synthetic air, nitrogen, hydrogen/nitrogen mixtures, with or without carbon dioxide, dry or wet atmospheres. Test conditions in terms of atmospheric compositions are displayed in Table 1. All high purity grade gases (≥99.999%) were supplied by Air Liquide (75 Quai d'Orsay, Paris 75007, France). Dry hydrogen of the same purity was produced from deionized water (≥18.2 MOhm·cm) by the laboratory HK-250 generator (LNI Schmidlin, Versoix, Switzerland). The resulting gas flow rate through the cell was 150 cm 3 ·min −1 , whatever the atmosphere used. When required, a constant vapor flow was generated by an integrated steamer (Fiaxell, Lausanne, Switzerland) from the same quality water inside the test bench.
Impedance spectroscopy measurements were performed using a Modulab Solartron Analytical potentiostat (Princeton Applied Research, 1100 Cassatt Road Berwyn, PA 19312, USA) with a FRA 1MHz frequency analyzer module, in a 10 MHz-0.1 Hz frequency range (11 points per decade), applying a 100 mV-amplitude alternative voltage signal, with respect to the system linear behavior. Between each measurement, once the desired temperature had been reached, a minimum hold of 15 min had been observed for the thermal stabilization of the sample. Data were collected by XM-Studio ESC software (Princeton Applied Research, 1100 Cassatt Road Berwyn, PA 19312, USA).

Phase and Microstructural Analysis
Phase composition of SDC-LiNaK composite electrolyte was determined before and after electrochemical tests by X-Ray diffraction using Bragg-Brentano geometry. All samples were analyzed as pellets. The irradiated surface was prepared by removing metallic electrode and then manually polishing on 2000-grit SiC sand disc to ensure a good flatness. The as-synthesized pellet was only polished on one side after synthesis. Measurements were performed on a D8 Endeavor diffractometer (Bruker, Billerica, MA, USA) equipped with a Co anticathode (λ K α 1 = 1.78897 Å, λ K α 1 = 1.79285 Å, λ K α 1 / λ K α 2 = 0.5) operated at 35 kV and 28 mA and an energy dispersive Lynxeye (Bruker Billerica, MA, USA) detector with integrated K β filtering. Data was collected from 20 to 100 • 2θ, with a step of 0.0257 • 2θ within 15 s. The sample holder rotation rate was kept at 10 rpm. The phase components were identified using the HighScore (PanAlytical, Almelo 7602 EA, The Netherlands) software and the PDF-4+ RDB database purchased by ICDD (International Center for Diffraction Data).

Thermal Analysis
Simultaneous Thermogravimetric and Differential Scanning Calorimetry analysis (TGA-DSC) were performed on Setsys Evolution ATG 16/18 easy fit analyzer (Setaram, Caluire et Cuire 69300, France). The powder electrolyte samples were placed in the alumina 100 µL crucibles so as to fill the crucible to 3 4 full. The same empty crucible was used as reference during acquisition of DSC signal by an S-type sensor (Pt/PtRh 10%, Setaram, Caluire et Cuire 69300, France). The temperature scanning rate was 2 • C·min −1 during heating and cooling stages, all performed in the 40-600 • C range under atmospheric pressure and 50 cm 3 ·min −1 gas flow rate. High purity argon (≥99.999%, Air Liquide, 75 Quai d'Orsay, Paris 75007, France) was used as an inert gas for analysis, and also as a base to form different percentage hydrogen mixtures. A dry hydrogen (≥99.999% of purity) was produced by the HK-250 generator (LNI Schmidlin, Versoix, Switzerland). Data were collected and processed using Calisto software (Setaram, Caluire et Cuire 69300, France).

Results and Discussion
Whatever the gaseous atmospheric composition, the composite impedance was measured at different temperatures, during two heating and cooling cycles. As reported in Figure 1, the conductivity does not follow a single Arrhenius law in all the temperature range explored, i.e., from 200 and 650 • C typically. A peculiar sigmoid shape described the composite electrical behavior, as already mentioned by several groups from the first studies dealing with these families of composites as fuel cell electrolytes [12,32,[44][45][46][47][48]. Indeed, three regions can be distinguished during the first heating, with two Arrhenius behaviors at low temperature (noted LT, below 360 • C) and high temperature (noted HT, higher than 400 • C), with activation energies equal to 2.0 eV and 0.27 eV, respectively, in nitrogen ( Figure 1a). It is worth noting that the activation energy calculated at low temperature when 20% of oxygen is added (Figure 1b) is half of what is determined in pure nitrogen, indicating the probable impact of oxide ions contribution to the total conductivity, limited in nitrogen atmosphere. However, no modification is observed at the highest temperature range explored and is ascribed to the molten carbonate contribution embedded in the solid oxide phase. The usual activation energy of ternary carbonate eutectic when the conductivity is measured from a molten bath is, however, slightly higher, between 0.26 eV and 0.42 eV, referring to Lair et al. [49] and Ward and Janz [50].
These two regions on Arrhenius plots are separated by a narrow region (called transition region) in terms of temperature (30 to 40 degrees), ending close to the melting temperature of the ternary carbonate eutectic mixture (397 • C) [51] as will be seen later by DSC measurements in the same atmosphere, yielding to a huge gain of a factor 10 3 to 10 4 in the conductivity from low to high levels. This transition zone corresponds, whatever the molten salt considered, to the salt mixture progressive melting process. The interface between the molten carbonate phase and the solid oxide phase SDC is considered to be important in relation to this transition region. Indeed, a large number of authors adopted one model, consisting of H + or HCO 3 − migration at the interface solid oxide/carbonate, thus promoting the conduction mechanism in the vicinity of the interface [15,17,34,43]. However, as the present results were obtained in dry air, no hydrogen (whatever its form) was present in the surrounding atmosphere, and thus, this model makes no real sense. According to Maier et al. [35], a space charge zone is formed at the interface between both conductive phases, with an increase in oxide ions vacancies concentrations compared to the volume content. This phenomenon, which is more realistic, is more pronounced as far as the temperature increases up to the ternary eutectic melting temperature, yielding to the molten carbonate percolation as the carbonate phase would be continuous in our case, with more than 30 %wt [52,53].
As in our previous studies, SDC-LiNaK composite electrical behavior was revealed in a nitrogen atmosphere in order to guarantee a reference-kind behavior. The conductivity variation as a function of the reverse temperature was as expected, with the previous described evolution, following three temperature regions. More interesting was the stable electrical behavior during two heating/cooling cycles. No change was observed as reported in Figure 1a,b, in N 2 and synthetic dry air, respectively.
Considering the electrical behavior in hydrogen-containing atmospheres, it is really quite different to that shown in Figures 2-4. One can note that there are still three regions during the first heating corresponding to the similar temperature ranges as described before in nitrogen and dry air. First, we observe that the end of the transition zone is slightly higher, 406 • C instead of 397 • C. Moreover, at higher temperatures, the activation energy decreased down to 0.21 eV. This could be justified by the appearance of a part of electronic conduction following doped ceria reduction in hydrogen, according to the simplified Reaction (1). This reduction reaction can be rewritten as Reaction (2), taking into account the defect chemistry, according to the Kröger-Vink notation: where O × O is an oxygen in its regular position in ceria lattice and V •• O is an oxygen vacancy, e is an electronic defect negatively charged, as an electron. Considering the electrical behavior in hydrogen-containing atmospheres, it is really quite different to that shown in Figures 2-4. One can note that there are still three regions during the first heating corresponding to the similar temperature ranges as described before in nitrogen and dry air. First, we observe that the end of the transition zone is slightly higher, 406 °C instead of 397 °C. Moreover, at higher temperatures, the activation energy decreased down to 0.21 eV. This could be justified by the appearance of a part of electronic conduction following doped ceria reduction in hydrogen, according to the simplified reaction (1). This reduction reaction can be rewritten as reaction (2), taking into account the defect chemistry, according to the Kröger-Vink notation: where O × is an oxygen in its regular position in ceria lattice and V  is an oxygen vacancy, is an electronic defect negatively charged, as an electron.   Considering the electrical behavior in hydrogen-containing atmospheres, it is really quite different to that shown in Figures 2-4. One can note that there are still three regions during the first heating corresponding to the similar temperature ranges as described before in nitrogen and dry air. First, we observe that the end of the transition zone is slightly higher, 406 °C instead of 397 °C. Moreover, at higher temperatures, the activation energy decreased down to 0.21 eV. This could be justified by the appearance of a part of electronic conduction following doped ceria reduction in hydrogen, according to the simplified reaction (1). This reduction reaction can be rewritten as reaction (2), taking into account the defect chemistry, according to the Kröger-Vink notation: where O × is an oxygen in its regular position in ceria lattice and V  is an oxygen vacancy, is an electronic defect negatively charged, as an electron.  The second point is that the composite electrical behavior drastically changed during the first cooling. Not only one but two drops in conductivity are now observed, as well as some changes in the slope values. This kind of evolution has already been pointed out in a previous work dealing with binary carbonate eutectic combined with Gd-doped ceria [2]. A partial transformation of molten carbonates into molten hydroxide-containing mixture was suggested, due to the associated recrystallization of LiOH-or NaOH-carbonatebased mixtures, close to 360 and 310 • C, respectively. Nevertheless, the most important result is that the electrical behavior of the composite seems to be stabilized during the second heating/cooling cycle as the transition temperatures remain the same as during the first cooling.
Moreover, by increasing the amount of hydrogen in the surrounding atmosphere, the level of conductivity is increased, which is clearly highlighted at intermediate and low temperature ranges. It is worth mentioning that the activation energy values, corresponding to the three new linear regions observed during cooling, are modified as shown in Figure 3: • At low temperature (<307 • C), before the first transition, the activation energy decreased from 1.58 eV to 1.02 eV, related to the first heating and cooling steps, respectively; • At high temperature (>410 • C) the activation energy decreased from 0.21 eV to 0.12 eV. This variation clearly indicates that the transport mechanisms and/or the species involved are modified. Moreover, the next change of slope is observed at around 406 • C down to 361 • C, and the activation value is 0.49 eV (Figure 3b). This value is close to the isolated molten ternary eutectic as reported by Ward et al. [50]. However, this should have been the same value of up to 490 • C, as in his work, which was not the case in our study; as this temperature is included in the temperature range, for which the activation is lower, being 0.12 eV. The other assumption is that this activation energy value of 0.49 eV quite often corresponds to protonic conduction through hydroxide ions [54][55][56]. The following slope, around 360-315 • C temperature range, is close to 1.2 eV, and is not fully identified. Nevertheless, assuming the partial transformation of carbonates into hydroxides, this could be ascribed to hydroxide transport mixed with carbonates into the porous solid matrix. The activation energy values corresponding to the different regions of the Arrhenius plots present in this paper are calculated in the same way, and can be found in the Supplementary Information.
in Figure 3: • At low temperature (<307 °C), before the first transition, the activation energy decreased from 1.58 eV to 1.02 eV, related to the first heating and cooling steps, respectively; • At high temperature (>410 °C) the activation energy decreased from 0.21 eV to 0.12 eV. This variation clearly indicates that the transport mechanisms and/or the species involved are modified. Moreover, the next change of slope is observed at around 406 °C down to 361 °C, and the activation value is 0.49 eV (Figure 3b). This value is close to the isolated molten ternary eutectic as reported by Ward et al. [50]. However, this should have been the same value of up to 490 °C, as in his work, which was not the case in our study; as this temperature is included in the temperature range, for which the activation is lower, being 0.12 eV. The other assumption is that this activation energy value of 0.49 eV quite often corresponds to protonic conduction through hydroxide ions [54][55][56]. The following slope, around 360-315 °C temperature range, is close to 1.2 eV, and is not fully identified. Nevertheless, assuming the partial transformation of carbonates into hydroxides, this could be ascribed to hydroxide transport mixed with carbonates into the porous solid matrix. The activation energy values corresponding to the different regions of the Arrhenius plots present in this paper are calculated in the same way, and can be found in the Supplementary Information.  Thus, one may assume that the hydrogen oxidation in carbonate media can yield to the transformation of carbonate ions into hydroxides, according to Reaction (3). That also means that CO 2 is locally formed and, consequently, it avoids a more or less deep decarbonation of the carbonate media and transformation of carbonates into alkali oxides. Furthermore, this phenomenon can be described by Reaction (4), considering the defect chemistry, fully involving defects originated at the interface, regarding oxygen vacancies from ceria phase side and hydrocarbonate and hydroxide ions formation on the carbonate phase side: where (OH) • O is hydroxyl ion occupying an oxygen regular site in cerium oxide phase and (HCO 3 ) • CO 3 represents the formation of hydrocarbonate occupying a carbonate position in the melt.
This last mechanism (Reaction (4)) has been suggested by Xiong et al. [57] studying composite based on yttrium-doped barium zirconate (BZY) associated with binary (Li,Na) 2 CO 3 carbonate eutectic. In the case of SDC-LiNaK, as cation defect concentration is elevated at the interface, one can assume the simultaneous formation of hydrocarbonate and hydroxide ions. Indeed, during cooling of the composite, the initial mixture can recrystallize into several phases combining alkaline metal cations from the eutectic and formed hydroxide ions. Thus as hydroxides are mixed with ternary LiNaK eutectic, it is well-known that the recrystallization temperature can be decreased, which could correspond to the second drop in conductivity around 360 • C [58,59], as shown in Figure 3b. The next drop could be ascribed to NaOH formation as its recrystallization temperature is lower, around 310 • C [60].
Furthermore, by increasing the amount of hydrogen in the atmosphere, even if no change is observed during the first heating, it is worth noting that both transition temperatures during the first cooling are slightly shifted towards a lower temperature, as illustrated in Figure 4a,b. This is confirmed during the second heating/cooling cycle as reported in Figure 4c,d. This corroborated the change of molten phase composition, from molten carbonate to molten carbonate-hydroxide mixture, amplified by the increase in the amount of hydrogen in the surrounding atmosphere; thus displacing the equilibrium (Reactions (3) and (4)) to the right side and, consequently, favoring the transformation of carbonate into hydroxides ions. Slight decreases in activation energies can be noticed as a result of easier transports, thanks to higher amounts of hydrogen in the atmosphere, according to Reaction (2).
The previous description finally involves extrinsic defects, thus not initially included in the material. This notion of extrinsic ionic conduction was first suggested by Marques et al. [48], whereas Zhu et al. [61,62] developed a conduction by protons, as well as Huang et al. [23]. The role of the carbonate ion has been elucidated next by Wang et al. [34] and more deeply in numerous studies [63][64][65]. Following this approach, proton interacts with both oxide and carbonate ions through metastable hydrogen bonds, thus the rotation and the mobility of carbonate ion favors proton mobility, breaking these metastable bonds. Above the melting temperature, this mobility is important enough to ensure high conductivity of composite material, with a low activation energy. Moreover, the recent works of Xiong et al. [57] showed experimentally for BZY-(Li,Na) 2 CO 3 , in parallel with DFT calculations, that hydrocarbonates were the preferential species involving protons in carbonates, more specifically at the interface with the oxide phase, according to Reaction (4).
Dealing with the transition temperatures (t t ), their corresponding values were also determined by DSC analyses, in order to corroborate the assumptions based on the impedance spectroscopy results. In Figure 5, DSC signals are reported during two thermal cycles, in pure argon and in hydrogen-containing dry atmospheres (23 %vol. and 50 %vol. in Ar), for the same sample. It could be noted, compared to pure Ar, a progressive decrease in the recrystallization temperature of the ternary eutectic was observed by increasing the hydrogen content, with ∆t = 20 and 27 • C, for 23% and 50% of H 2 in Ar, respectively, whereas the melting temperature is not affected by the atmosphere composition. Furthermore, some additional peaks appear during the first cooling around 368-365 • C, independent of the hydrogen content. It could be linked to KOH recrystallization, with a melting temperature being 400 • C [60]. Another peak after the main one, according to the screening direction, appears at 309-307 • C. It is also non-dependent of the hydrogen content and could be ascribed to NaOH recrystallization, as the melting temperature is close to 320 • C [60]. It is worth noting that the presence of hydroxide is well-known to induce a lowering of melting temperature of the LiNaK carbonate mixture [58,59], as observed in the Figure 5b,d during both cooling steps for the recrystallization temperature, as well as in Figure 5c, during the second heating after modification of the molten salts composition.
During the second heating, three peaks can be distinguished. The first one, exothermic, is not explained. The second one, endothermic, corresponds to the melting of one phase, more probably ascribed to sodium hydroxides as already assumed (transition temperature around 313-315 • C), not dependent on the hydrogen content even during melting. However, the peak which could have been related to the melting of potassium carbonate as this recrystallized phase has been identified during the first cooling, is not evidenced due to the probable overlapping of this peak with the enlarged eutectic melting DSC signal. Moreover, during the second heating, the higher the hydrogen content, the lower the eutectic melting temperature: 362 and 354 • C for 23% and 50% of H 2 in Ar, respectively, compared to initial 389 • C (first heating step). As during these DSC signal acquisitions, the temperature rate is quite low (2 • C·min −1 ), the hydrogen oxidation obviously can occur and forms some hydroxide and hydrocarbonate ions as previously described (Reaction (4)). Table 2 Energies 2022, 15 Dealing with the transition temperatures (tt), their corresponding values were also determined by DSC analyses, in order to corroborate the assumptions based on the impedance spectroscopy results. In Figure 5, DSC signals are reported during two thermal cycles, in pure argon and in hydrogen-containing dry atmospheres (23%vol. and 50%vol. in Ar), for the same sample. It could be noted, compared to pure Ar, a progressive decrease in the recrystallization temperature of the ternary eutectic was observed by increasing the hydrogen content, with Δt = 20 and 27 °C, for 23% and 50% of H2 in Ar, respectively, whereas the melting temperature is not affected by the atmosphere composition. Furthermore, some additional peaks appear during the first cooling around 368-365 °C, independent of the hydrogen content. It could be linked to KOH recrystallization, with a melting temperature being 400 °C [60]. Another peak after the main one, according to the screening direction, appears at 309-307 °C. It is also non-dependent of the hydrogen content and could be ascribed to NaOH recrystallization, as the melting temperature is close to 320 °C [60]. It is worth noting that the presence of hydroxide is well-known to induce a lowering of melting temperature of the LiNaK carbonate mixture [58,59], as observed in the Figure 5b,d during both cooling steps for the recrystallization temperature, as well as in Figure 5c, during the second heating after modification of the molten salts composition.
During the second heating, three peaks can be distinguished. The first one, exothermic, is not explained. The second one, endothermic, corresponds to the melting of one phase, more probably ascribed to sodium hydroxides as already assumed (transition temperature around 313-315 °C), not dependent on the hydrogen content even during melting. However, the peak which could have been related to the melting of potassium carbonate as this recrystallized phase has been identified during the first cooling, is not evidenced due to the probable overlapping of this peak with the enlarged eutectic melting DSC signal. Moreover, during the second heating, the higher the hydrogen content, the lower the eutectic melting temperature: 362 and 354°C for 23% and 50% of H2 in Ar,  respectively, compared to initial 389 °C (first heating step). As during these DSC signal acquisitions, the temperature rate is quite low (2 °C·min −1 ), the hydrogen oxidation obviously can occur and forms some hydroxide and hydrocarbonate ions as previously described (Reaction 4). Table 2 summarizes all the phase transition temperatures revealed by DSC measurements as a function of the atmosphere composition and the step in the cycling. Figure 5. DSC signals observed for SDC-LiNaK electrolyte powder as function of sample temperature in pure Ar, 23%-and 50%-H2-in-Ar gas mixtures at atmospheric pressure. The first heating/cooling cycle (a,b) was followed by second one (c,d) for the same sample. Scanning rate was two degrees per minute and its direction is indicated by arrows.  Figure 5. DSC signals observed for SDC-LiNaK electrolyte powder as function of sample temperature in pure Ar, 23%-and 50%-H 2 -in-Ar gas mixtures at atmospheric pressure. The first heating/cooling cycle (a,b) was followed by second one (c,d) for the same sample. Scanning rate was two degrees per minute and its direction is indicated by arrows.
Thanks to the previous observations, if carbonates are partially transformed into hydroxides, in hydrogen-based surrounding atmospheres, the corresponding reactions can be written as Reactions (5) and (6). This will first be induced by the partial reduction in hydrogen of Ce 4+ into Ce 3+ , thus to electron creation and, as a direct reaction, the formation of water molecules. As a result of the presence of H 2 O, some carbonate ions will react and lead to the formation of CO 2 , and hydroxide ions. As CO 2 is produced this could explain that decarbonation, i.e., the transformation of carbonates into oxide ions is avoided (Reactions (6)-(8)), displacing equilibrium and stabilizing the system. It is worth noting that this phenomenon was one of the controversies when discussing the apparent stability of these materials, operating without adding CO 2 in the surrounding atmospheres. In the usual MCFC, there is no possible reduction, probably as the lithium aluminate porous matrix is inert whatever the operating atmosphere, thus no electrons are produced and next to no CO 2 is formed. Therefore CO 2 is required in both atmospheres in the hybrid system. Moreover, in the current composite materials, it is worth mentioning that these electrons are not mainly introducing electronic conductivity into the composite electrolyte; they induce partial transformation of molten carbonates into hydroxide ions, leading to lower transition temperature during the first cooling step, stabilizing it as no change in the electrical behavior is observed. More interesting is the impact of water addition in oxygen-based and hydrogenbased atmospheres, as reported in Figure 6a,b, respectively. The electrical behavior in such moist atmospheres confirms the assumption of hydroxide formation, as added water favors the transformation of carbonates into hydroxides (Reaction (6)), even when no H 2 is initially present in the water-containing atmosphere (humid air). A hysteresis can be described in humid air, (Figure 6a) between the first heating and the other cycle steps even if its amplitude is not as important as in dry or humid hydrogen. The hysteresis is amplified when comparing the electrical behavior in dry hydrogen and humid hydrogen. The transition temperatures are shifted towards a lower temperature, whatever the moist atmosphere compositions. However, as previously concluded when dry atmospheres were surrounding the system, the composite material is transformed by the addition of water vapor in the atmospheres, yet almost stabilized after the first thermal cycle. Coming back to the role of the interface, the phenomenon can be described once again by defect chemistry according to the Kröger-Vink notation by Reactions (9) and (10), for humidified air and completed with Reaction (11) in humidified hydrogen. In the later reaction, one can notice that hydrogen is also a resulting product allowing the mechanism repetition [66].
Whatever the surrounding humidified atmosphere, hydroxyl and hydrocarbonate ions are formed, thus leading to specific conduction and the activation energies in the intermediate temperature range are still around 0.5 eV and 1.1 eV as described for the electrical behavior in dry hydrogen (Figure 3b). It is less evidenced in humidified air due to a shorter temperature range, although still present (Figure 6a).
Considering the addition of CO2 in both reducing and oxidizing atmospheres, as in MCFC working atmosphere conditions, no more hysteresis is observed during heating and cooling cycles (Figure 7). This means that CO2 is the required component in the atmosphere to fully stabilize the electrical behavior of molten carbonate/cerium-based solid oxide phase composite materials; as potential electrolytes, as shown by reaction 8, avoiding eutectic decarbonation. Nevertheless, it should be noted that above the carbonate ternary eutectic melting temperature, the electrical conductivity is slightly higher in an oxidizing atmosphere compared to the associated value in a reducing atmosphere, i.e., at 550 °C 0.17 S·cm −1 compared to 0.07 S·cm −1 , and at 600 °C 0.20 S·cm −1 compared to Coming back to the role of the interface, the phenomenon can be described once again by defect chemistry according to the Kröger-Vink notation by Reactions (9) and (10), for humidified air and completed with Reaction (11) in humidified hydrogen. In the later reaction, one can notice that hydrogen is also a resulting product allowing the mechanism repetition [66].
Whatever the surrounding humidified atmosphere, hydroxyl and hydrocarbonate ions are formed, thus leading to specific conduction and the activation energies in the intermediate temperature range are still around 0.5 eV and 1.1 eV as described for the electrical behavior in dry hydrogen (Figure 3b). It is less evidenced in humidified air due to a shorter temperature range, although still present (Figure 6a).
Considering the addition of CO 2 in both reducing and oxidizing atmospheres, as in MCFC working atmosphere conditions, no more hysteresis is observed during heating and cooling cycles (Figure 7). This means that CO 2 is the required component in the atmosphere to fully stabilize the electrical behavior of molten carbonate/cerium-based solid oxide phase composite materials; as potential electrolytes, as shown by Reaction (8), avoiding eutectic decarbonation. Nevertheless, it should be noted that above the carbonate ternary eutectic melting temperature, the electrical conductivity is slightly higher in an oxidizing atmosphere compared to the associated value in a reducing atmosphere, i.e., at 550 • C 0.17 S·cm −1 compared to 0.07 S·cm −1 , and at 600 • C 0.20 S·cm −1 compared to 0.10 S·cm −1 , respectively, combined also with some modifications of the activation energies. Referring to Rondão et al. [64] who studied the electrical behavior of CeO 2 -(Li,Na) 2 CO 3 composite in different atmospheres, they found that the conductivities vary in the following order σ(H 2 ) > σ(O 2 ) > σ(CO 2 ). Taking into account that conductivity values in oxygen and carbon dioxide are much closer, they assumed that this conductivity decrease in CO 2 atmosphere could be the result of dicarbonate ion C 2 O 2− 5 formation, according to Reaction (12): This possible transformation and the existence of dicarbonate ions as reported in [67][68][69], leads to a lower carbonate ions concentration, as well as a decrease in mobility of charge transport spaces due to the larger size of dicarbonate ions, resulting in an increase in activation energy and a decrease in conductivity in such operating conditions including CO 2 .
Moreover, dealing with the stability of SDC-LiNaK in these surrounding atmospheres, in particular in a mixture of water, hydrogen and carbon dioxide (MCFC standard anode atmosphere), one can assume that it is due to the chemical stability of carbonate phase in the composite. Even if cerium can be reduced by H 2 , different reactions can occur as already discussed, mainly the hydrogen oxidation in the presence of carbonate ions, as well as carbonates hydrolysis in the presence of water according to Reactions (6) and (7), respectively. The chemical stability is maintained thanks to the high partial pressure of CO 2 on the MCFC anodic atmosphere: 16 %vol. of CO 2 seems to permit the displacement of both 6 and 7 reaction equilibria, avoiding the formation of hydroxides and thus the recrystallization into new phases during the cooling.
This possible transformation and the existence of dicarbonate ions as reported in [67][68][69], leads to a lower carbonate ions concentration, as well as a decrease in mobility of charge transport spaces due to the larger size of dicarbonate ions, resulting in an increase in activation energy and a decrease in conductivity in such operating conditions including CO2.
Moreover, dealing with the stability of SDC-LiNaK in these surrounding atmospheres, in particular in a mixture of water, hydrogen and carbon dioxide (MCFC standard anode atmosphere), one can assume that it is due to the chemical stability of carbonate phase in the composite. Even if cerium can be reduced by H2, different reactions can occur as already discussed, mainly the hydrogen oxidation in the presence of carbonate ions, as well as carbonates hydrolysis in the presence of water according to Reactions (7) and (6), respectively. The chemical stability is maintained thanks to the high partial pressure of CO2 on the MCFC anodic atmosphere: 16 %vol. of CO2 seems to permit the displacement of both 7 and 6 reaction equilibria, avoiding the formation of hydroxides and thus the recrystallization into new phases during the cooling. X-ray diffraction analysis was performed on the as-synthesized pellet, and after impedance spectroscopy measurements in the different atmospheres as reported and discussed before. Sm-doped ceria is indexed as cubic structure (Fm3m) with the well-defined more intense and narrow peaks (Figure 8a,b). Nevertheless, the complete indexation is highly complex due to numerous small peaks corresponding to the carbonates and new phases appearing after exposition during several thermal cycles. This also explains the specific representation of the square root of the intensity recorded. Moreover, some assumed new phases based on carbonates, hydrogen carbonates, hydroxides and other potential and especially non-stoichiometric phases are not systematically referenced in XRD data bases, or with low accuracy.
Nevertheless, from obtained diffractograms we can clearly distinguish the change in phase composition of initial SDC-LiNaK after thermal cycling and recrystallization under reducing atmospheres (Figure 8a). Atmosphere containing 23% vol. of hydrogen gener- X-ray diffraction analysis was performed on the as-synthesized pellet, and after impedance spectroscopy measurements in the different atmospheres as reported and discussed before. Sm-doped ceria is indexed as cubic structure (Fm3m) with the well-defined more intense and narrow peaks (Figure 8a,b). Nevertheless, the complete indexation is highly complex due to numerous small peaks corresponding to the carbonates and new phases appearing after exposition during several thermal cycles. This also explains the specific representation of the square root of the intensity recorded. Moreover, some assumed new phases based on carbonates, hydrogen carbonates, hydroxides and other potential and especially non-stoichiometric phases are not systematically referenced in XRD data bases, or with low accuracy.
Nevertheless, from obtained diffractograms we can clearly distinguish the change in phase composition of initial SDC-LiNaK after thermal cycling and recrystallization under reducing atmospheres (Figure 8a). Atmosphere containing 23 %vol. of hydrogen generally induces the recrystallization of the same phases forming the initial eutectic mixture. At the same time, some new diffraction maxima appear, although their indexation remains problematic due to low intensity. This new phase can be more precisely indexed in the case of 100 %vol. hydrogen atmosphere as NaK 2 [H(CO 3 ) 2 ]·2H 2 O (PDF n • 04-010-8201). It is important to highlight that such a phase modification process goes with the SDC grain's coalescence, which is much stronger at higher H 2 content. The micrographs obtained by Scanning Electron Microscopy (Figure 9c,e,f) show this influence, which results in the growing up of the oxide phase into larger grains, thus reducing the interfacial oxide-carbonate surface, and making the phase distribution homogeneity much less favorable. This phenomenon is reflected by a decrease in the SDC phase full width at half-maximum (FWHM) on corresponding XRD patterns (Figure 8a). The same NaK 2 [H(CO 3 ) 2 ]·2H 2 O phase is formed in larger amounts in the sample exposed to humidified 23 %vol.-H 2 , which leads us to the same conclusion that the presence of water promotes the formation of hydrogenated carbonate phases in the composite electrolyte mixture. These results corroborate the conclusions drawn from Arrhenius diagrams obtained during thermal cycling under the same atmospheres. In addition, in the case of humidified air (Figure 8b) water itself can induce recrystallization of the same hydrogenated phase, as well as others, such as Na 3 [H(CO 3 ) 2 ]·2H 2 O (PDF n • 00-029-1447) and KHCO 3 (PDF n • 04-013-5503) compared with dry air. Interesting to note that in the two cases of these humidified atmospheres (containing hydrogen or not), the microstructure does not undergo any significant changes (Figure 9b,d) except a slight increase in SDC phase grain size. However, this oxide phase remains well-dispersed throughout the volume of the composite material. This indirectly suggests an important role of ceria reduction process by H 2 and its re-oxidation in the presence of water according to Reaction (11).
With regard to the standard operating MCFC atmospheres, the cathode-one (30 %vol. CO 2 in air) is not a source of a significant change in recrystallized phases compared to the initial electrolyte composition. The same phases are found in the sample resulting from the synthesis, as well as in that which has undergone thermal cycling in dry air. However, certain maxima present on the XRD pattern could indicate the presence of a hydrogenated carbonate phase, in particular NaK 2 [H(CO 3 ) 2 ]·2H 2 O. However, it is difficult to pronounce this as there are not as many maxima that can be attributed to this suspected phase, and compared to the sample studied under humidified air (where this phase is undoubtedly present), some are even absent. So, it could be another composition that is not listed in the database.
Moreover, the standard MCFC anode atmosphere (64 %vol. H 2 /16 %vol. CO 2 /20 %vol. H 2 O), induces a low recrystallization of NaK 2 [H(CO 3 ) 2 ]·2H 2 O, which is certainly due to the presence of hydrogen. Nevertheless, a diffraction maximum at 90.5 • 2θ remains unexplained. In addition, the formation of a sodium oxalate phase can be suspected after cycling under this atmosphere. Usually, oxalate ions can only be formed in the presence of carbon monoxide (Reaction (13)) in the surrounding atmosphere, which is not the case. Furthermore, the chromatographic analysis based on a protocol fixed by Meskine et al. [70] indicates the presence of CO in the outlet gas mix (cf. Supplementary Information). It can be assumed here that the formation of CO can be catalytically favored in the presence of samaria-doped ceria in the composite according to the reverse water gas shift reaction (RWGS) (Reaction (14)) [71].
The composite electrolyte morphology of sample cycled in standard MCFC cathodic atmosphere does not present any notable differences compared to the as-synthetized sample. It could be suggested that there are no chemical interactions between the phases making up the composite material under such conditions. The absence of this type of reactivity is systematically mentioned in the literature [17,72]. Referring to the last micrograph obtained for a sample being cycled under anodic MCFC atmosphere (Figure 9h), we can conclude that such conditions prove to be very favorable regarding the stability of the initial microstructure of the SDC-LiNaK composite, or even its slight improvement in terms of homogeneity of the phase distribution compared to the as-synthesized sample. It is clear that this structural stability of the electrolyte is partly responsible for the stable behavior during thermal cycling under such an atmosphere, and this positively influences its electrical behavior. Here, one may assume that a sufficient partial pressure CO 2 level plays a protective role, thus preventing the equilibrium displacement in the decarbonation direction of the composite material (Reaction (8)). bility of the initial microstructure of the SDC-LiNaK composite, or even its slight improvement in terms of homogeneity of the phase distribution compared to the as-synthesized sample. It is clear that this structural stability of the electrolyte is partly responsible for the stable behavior during thermal cycling under such an atmosphere, and this positively influences its electrical behavior. Here, one may assume that a sufficient partial pressure CO2 level plays a protective role, thus preventing the equilibrium displacement in the decarbonation direction of the composite material (reaction (8)).  EIS experiments are corroborated by DSC measurements in the same atmospheres when possible. These results indicate the formation of alkali hydroxides (NaOH and KOH) in amounts small, yet sufficient when mixed to carbonate eutectic, to reduce their melting temperature, initiating then the appearance of transition temperatures in the low temperature region (250-350 °C). Thus, a combination of these two experimental methods allowed us to review and establish, in the form of chemical equations, different possible interactions between the phases, constituting the composite electrolyte and surrounding atmospheres, while considering the predominant role of the oxide/carbonate interface in the formation of various ionic species responsible for the total conductivity of the electrolyte material. These equations are summarized in Figure 10. It is important to highlight that MCFC standard operating atmospheres do not strongly affect the conductivity of composite SDC-LiNaK electrolyte whatever the temperature range, except a slight decrease in the conductivity values in 400-550 °C range during thermal cycling in standard anodic atmosphere. This fact is likely to be assigned EIS experiments are corroborated by DSC measurements in the same atmospheres when possible. These results indicate the formation of alkali hydroxides (NaOH and KOH) in amounts small, yet sufficient when mixed to carbonate eutectic, to reduce their melting temperature, initiating then the appearance of transition temperatures in the low temperature region (250-350 • C). Thus, a combination of these two experimental methods allowed us to review and establish, in the form of chemical equations, different possible interactions between the phases, constituting the composite electrolyte and surrounding atmospheres, while considering the predominant role of the oxide/carbonate interface in the formation of various ionic species responsible for the total conductivity of the electrolyte material. These equations are summarized in Figure 10. EIS experiments are corroborated by DSC measurements in the same atmospheres when possible. These results indicate the formation of alkali hydroxides (NaOH and KOH) in amounts small, yet sufficient when mixed to carbonate eutectic, to reduce their melting temperature, initiating then the appearance of transition temperatures in the low temperature region (250-350 °C). Thus, a combination of these two experimental methods allowed us to review and establish, in the form of chemical equations, different possible interactions between the phases, constituting the composite electrolyte and surrounding atmospheres, while considering the predominant role of the oxide/carbonate interface in the formation of various ionic species responsible for the total conductivity of the electrolyte material. These equations are summarized in Figure 10. It is important to highlight that MCFC standard operating atmospheres do not strongly affect the conductivity of composite SDC-LiNaK electrolyte whatever the temperature range, except a slight decrease in the conductivity values in 400-550 °C range during thermal cycling in standard anodic atmosphere. This fact is likely to be assigned It is important to highlight that MCFC standard operating atmospheres do not strongly affect the conductivity of composite SDC-LiNaK electrolyte whatever the temperature range, except a slight decrease in the conductivity values in 400-550 • C range during thermal cycling in standard anodic atmosphere. This fact is likely to be assigned to dicarbonate ion formation at high CO 2 partial pressures. Therefore, it probably can be recommended during fuel cell operation, employing any composite oxide/carbonate electrolyte, to keep the CO 2 partial pressure at an optimal level on both cathode and anode compartments.

Conclusions
This study achieved a deeper understanding of the electrical behavior and thermal stability issues of SDC-LiNaK composite electrolyte under various reducing, oxidizing, wet and dry atmospheres. Impedance Electrochemical Spectroscopy coupled to thermal cycling, directly provided information about electrolyte conductivity values, activation energy values and their changes (variations) throughout at least two complete heating/cooling cycles in an intermediate temperature range (250-600 • C). SDC-LiNaK composite underwent different hydrogen contents; dry reducing atmospheres highlight a different electrical behavior during a cooling stage at the end of the first thermal cycle. It consists of the occurrence of different transition temperatures separating Arrhenius plots in different regions characterized by singular activation energy values. The more the surrounding atmosphere is rich in hydrogen, the greater the shift in transition temperatures toward lower temperatures. This transformation becomes remarkable when the electrolyte is exposed to humidified 23%-in N 2 -hydrogen mixture. Moreover, it was shown that water, present in air, forming a humidified oxidizing atmosphere, induces the same changes in SDC-LiNaK electrical behavior. Thus, in both dry and wet hydrogenated atmospheres, an Arrhenius plot high temperature region (350-600 • C) appears with an activation energy value around 0.45 eV, which indicates a conductivity mechanism involving hydrogenated ions.
Post mortem phase composition results reinforce our conclusions on conduction mechanisms revealed by EIS measurements. Indeed, during recrystallization at the final cooling stage, new hydrogen carbonate phases are formed in hydrogen-and/or water-content atmospheres: KHCO 3 , Na 3 [H(CO 3 ) 2 ]·2H 2 O and NaK 2 [H(CO 3 ) 2 ]·2H 2 O. Concerning samples undergoing MCFC standard operation atmospheres, the phase composition remains generally stable, except with regard to sodium oxalate formation in anodic atmosphere, due to a reverse water gas shift reaction (RWGS) manifestation and the resulting CO formation and its subsequent interaction with carbonate ions.
Morphological studies indicate a tangible change in the microstructure of electrolyte samples thermally cycled in hydrogen-content dry atmospheres: the oxide phase grains coalesce, and the process is more amplified with higher hydrogen content. Water prevents transformations of this oxide phase whatever the atmosphere used. In contrast, carbon dioxide-content and both MCFC standard operating atmospheres do not affect the SDC-LiNaK composite microstructure.
Finally, the authors expect that the present results on the electrical behavior and thermal stability of SDC-LiNaK can be extrapolated over a large family of composite oxide/carbonate electrolytes.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/en15072688/s1, Figure S1: Outlet composition of gas mixture during EIS thermal cycling analysis of SDC-LiNaK under MCFC standard reducing conditions at 441 • C and 546 • C. Table S1: Retention times of gases forming an outlet mixture during EIS thermal cycling analysis of SDC-LiNaK under MCFC standard reducing conditions. Figure S2. Nyquist EIS plots recorded in nitrogen atmosphere at different temperatures within the symmetric Au|SDC-LiNaK|Au electrochemical configuration. The amplitude of the applied sinusoidal voltage was V 0 = 100 mV. The numbers on the diagrams indicate the logarithm of the frequency of the applied signal. Table S2. Activation energy (E a ) and transition temperature (t t ) values calculated from Arrhenius plot during thermal cycling of SDC-LiNaK under dry reducing atmospheres containing different amounts of hydrogen. H-heating; C-cooling; LT and HT-low and high temperature regions respectively. Table S3. Activation energy (E a ) and transition temperature (t t ) values calculated from Arrhenius plot during thermal cycling of SDC-LiNaK under oxidizing and reducing atmospheres both containing 20% of moisture. H-heating; C-cooling; LT and HT-low and high temperature regions respectively. Table S4. Activation energy (E a ) and transition temperature (t t ) values calculated from Arrhenius plot during thermal cycling of SDC-LiNaK under MCFC standard oxidizing and reducing atmospheres. H-heating; C-cooling; LT and HT-low and high temperature regions respectively.