Design and characterization of novel glass-ceramic sealants for solid oxide electrolysis cell (SOEC) applications

In this work, three new glass-ceramic compositions are designed and characterised as sealant materials for solid oxide electrolysis cells (SOEC), having operating temperature of 850°C. The crystallization and the sintering behavior of the glasses are investigated by using differential thermal analysis (DTA) and heating stage microscopy (HSM), respectively. The glasses show glass transition temperatures of 715-740 o C, while the coefficients of thermal expansion (CTE) of 9.3-10.3 ×10 -6 K -1 (200°C-500°C) are measured for the glass-ceramics, matching with the CTEs of the other cell components. The compatibility between the glass-ceramic sealants, the 3YSZ electrolyte and the Crofer22APU interconnect is examined by means of SEM and EDS, in the as-joined condition and after 1000 hours at 850°C in air. Compositional changes in the glass-ceramic sealants are reviewed and discussed with respect to the formed crystalline phases before and after the ageing treatment at 850°C. three scans were made for each glass composition. The crystalline phase analysis of the glass-ceramics before and after thermal ageing was carried out by using PanAlytical X'Pert Pro PW 3040/60 Philips (The Netherlands), with Cu K α and the X’Pert software. The XRD analysis were carried out in the range of 2 theta 10° - 70°, with step size of 0.02626° and time per step 10.20 sec.


Accepted Article
This article is protected by copyright. All rights reserved.
interconnect surface was found to be a viable method to prevent adverse chromates formation, as studied up to 300 hours at 850 °C. Reddy et al.(44) studied new series of developed lanthanide containing diopside glass-ceramic sealants with SrO content in the 7-12 mol% range. The proposed glass-ceramics had excellent properties after joining, nevertheless all the reported glass-ceramics showed reduction in their CTEs during thermal ageing from 500 hours to 1000 hours. Moreover, the compatibility of these glassceramics with a Sanergy HT metallic interconnect and an 8YSZ solid electrolyte were discussed up to 500 hours.
This study provides a new insight into Ba free glass sealants for SOECs applications, for the working temperature of 850 °C. SrO was used as the main glass modifier in addition to CaO and MgO, and the concentration of B 2 O 3 was limited to a maximum of 6 ml%. SrO was preferred over BaO due to the fact that Ba tends to react with Cr more readily than Sr to form a chromate compound, though both chromate formation reactions are thermodynamically favorable with negative Gibbs free energies(45). The durability of these new sealants was studied at 850°C for 1000 hours in static air.
The main objective of this paper is to compare 3 new glass-based compositions with different amounts of SrO as a modifier, by reviewing its effect on crystal phases formation as well as on thermal and thermomechanical properties of the glass-ceramic sealant in contact with a Crofer22APU interconnect @ 850°C.

Accepted Article
This article is protected by copyright. All rights reserved.
The compatibility of glass-ceramics sealants with the Crofer22APU interconnect and 3YSZ electrolyte was investigated by SEM (Merlin ZEISS). For SOEC applications, 8YSZ is commonly used as electrolyte due to its high ionic conductivity(8). However, in this study 3YSZ is used, because it is typically used in the electrolyte supported cells thanks to its superior toughness as compared with 8YSZ(46, 47). To investigate the compatibility, the Crofer22APU/Glass/3YSZ joined samples were processed in a Carbolite furnace (CWF 13/5) in static air. Prior to the joining, each substrate, with dimensions of 1.5 cm x 1.5 cm, was cleaned with acetone. The glass was then deposited by spatula in the form of slurry containing the glass powder and ethanol in 70:30 wt%. During the joining procedure a load of 15 g/cm 2 was placed on the samples. The joining of the HJ1 and HJ3 sealants was done at 950 o C for 1 hour at a heating rate of 5 o C/min. However, for the HJ4 sealant the joining was carried out at 950°C for the dwelling time of 5 hours, at a heating rate of 2 o C/min. Further details about the selection of the different joining cycles are given in section 3.1.
The cross sections of Crofer22APU/glass-ceramics/3YSZ joined samples were metallographically polished up to 1 µm by diamond paste and investigated by SEM after coated with gold.

Thermal Analysis
The DTA curves corresponding to the three glass systems and their shrinkage behavior vs temperature, obtained from HSM, are shown in Figure 1. In Figure 1(a), the T g , T x and T p labels corresponds to glass transition temperature, onset crystallization temperature and peak crystallization temperature, respectively. The average characteristic temperatures (of three measurements) along with their standard deviations are summarized in Table 2.
Form the data in Figure 1(a), it is apparent that the DTA thermogram of HJ1 glass system showed sharp exothermic peaks of crystallization, while the intensity of the crystallization peak reduced

Accepted Article
This article is protected by copyright. All rights reserved. significantly in HJ3 glass. However, no crystallization peak was observed during the DTA analysis of the HJ4 glass system. Also, the HSM curves of the glasses (Figure 1(b)) indicate a clear difference in the sintering behavior of the different sealants. The T FS (temperature of first shrinkage), reported in Figure 1(b), corresponds to the temperature at which the sintering process was initiated by viscous flow, whereas T MS corresponds to the maximum shrinkage temperature.
The comparative study of HJ1 and HJ3 systems showed that by increasing the mass concentration of SrO in HJ3, a higher CTE of as-cast glass was measured as compared with the HJ1 system. Whereas, the as-cast HJ4 glass showed lowest value of CTE. On the other hand, the HJ3 and HJ4 glasses showed the glass transition (T g ) shifted to a higher temperature in spite of having a higher SrO concentration.
The glass-ceramics derived from the different parent glass compositions were obtained with the heat treatments mentioned in section 2. From the data obtained from DTA and HSM (Table 2), the heat treatment of 950 o C, 1h at a heating rate of 5 °C/min, was chosen to ensure maximum devitrification for all glasses. The CTEs of the obtained glass-ceramics for HJ1 and HJ3 systems, increased significantly as compared with their as-cast glasses. However, this heat treatment caused negligible increase in CTE of HJ4 glass-ceramic as compared with as-cast glass, probably due to a slight devitrification. Therefore, to ensure sufficient devitrification, the slow heating rate of 2 o C/min and long dwelling time of 5 hours was chosen as a heat treatment to prepare HJ4 glass-ceramic (as it will be shown in Figure 4). After suitable heat treatments, the CTEs of all the glass-ceramics, given in Table 2, were within the desired range (9-12 x 10 -6 K -1 ), taking into consideration the CTEs of the other cell components (CTEs for Crofer22APU and 3YSZ are 12 x 10 -6 K -1 and 10.5 x 10 -6 K -1 respectively) (48).

Accepted Article
This article is protected by copyright. All rights reserved.
The CTEs measured after ageing the glass-ceramics for 1000 hours at 850°C are also reported in Table 2. The CTEs of the HJ1 glass-ceramics were slightly reduced after ageing. For the HJ3 glassceramic no change in the CTE was observed. On the other hand, thermal ageing slightly increased the CTE of the HJ4 glass-ceramic.

XRD and microstructural analysis
The XRD of as-cast glasses for three different glass systems is shown in figure S1. The XRD patterns of the different glass-ceramics before and after ageing are shown in Figure 2. As-joined HJ1 glassceramic, in Figure 2 shows the XRD patterns of the HJ4 system treated at different temperatures and for different dwelling times ( Table 2). The HJ4 glass-ceramic treated at 950°C for 1 hour, contained only SrSiO 3 as the crystalline phase in addition to the residual glassy phase. However, an increase in the dwelling time to five hours at 950°C, resulted in the formation of cristobalite (SiO 2 ) as secondary phase in addition to SrSiO 3 as main phase. The XRD patterns of HJ1, HJ3 and HJ4 glass-ceramics after ageing at 850 °C, 1000 hours are also shown in figure 2(a), 2(b) and 2(c) respectively. The XRD patterns of pure phases shown in figure 2, corresponds to the simulated patters obtained from the X'Pert software data base. The different crystalline phases present in as-joined and aged glass-ceramics are summarized in Table 3.
The SEM cross-section images of the interfaces of the HJ1, HJ3 and HJ4 glass-ceramics with Crofer22APU and 3YSZ substrates are shown in Figure 3. The evolution of microstructure of the HJ4 system after processing at different heat treatments along with corresponding EDS analysis, is

Accepted Article
This article is protected by copyright. All rights reserved.
shown in Figure 4. The heat treatment of 950°C, 1h (Figure 4(a)) shows that a significant amount of residual glassy phase was still present (dark area). Only one type of crystalline phase was observed.
On the other hand, the heat treatment at 950°C, 5h at 2°C/min resulted in the formation of a new phase.

Discussion
In HJ1 composition CaO and MgO were added as the main modifiers. The SrO addition was minimal (9% mol) in HJ1 with the main purpose to act as a network modifier. SrO concentration was increased in HJ3 and further in HJ4 to have Sr containing crystalline phases in addition to having minimal SrO in the residual glass phase, thus maintaining a viscous glass behavior and to reduce the potential formation of Sr chromate. A proper balance of SiO 2 /SrO (equal to 1) is required to obtain a desired high CTE SrSiO 3 phase (10.9 x 10 -6 K -1 ) (49), however as the increasing of SrO contents also increases the possibility of formation of undesirable SrCrO 4 phase, improving one property could potentially come at the expenses of other functionalities, and the right balance is often difficult to achieve. Therefore, the SiO 2 /SrO in HJ3 and HJ4 was kept 2.3 and 1.99 respectively, slightly higher than some glasses reported in literature (28, 43) where formation of SrCrO 4 resulted in poor adhesion of glass-ceramic with interconnects. To this purpose, the proposed glass compositions require high silica contents in order to obtain high viscosities and desired crystalline phases, considering the high operating temperatures of the SOECs at 850°C and the potential reactivity of the Crofer22APU with SrO.

Accepted Article
This article is protected by copyright. All rights reserved.

Thermal Analysis
The difference in the crystallization behavior of different glasses, obtained from DTA (Figure 1(a)), indicate that the intensity of crystallization reduced from HJ1 to HJ4. In HJ4 glass, probably the crystallization was not enough to be detected during the DTA analysis. The HSM analysis of the HJ1 and HJ3 systems (Figure 1(b)), which have more devitrification than HJ4, showed a constant shrinkage for a certain temperature range after the completion of sintering (T MS ). On the contrary, HJ4, showed a continuous viscous flow at temperatures higher than the T FS, due to the fact of having low devitrification.
In order to obtain dense and consequently, leakage free sealants, it is necessary to complete the sintering before the crystallization starts (T MS < T x ), thus avoiding the formation of porosity in the sealant due to increased viscosity caused by crystal growth(7). As soon as the crystallization occurs, the glass viscosity will drastically increase, hindering the viscous flow of the glass and the adhesion to the metallic or ceramic substrates. Therefore, the crystallization mechanism of the glass-ceramic should be controlled and taken into account in the heat treatment schedule. In the HJ1 and HJ3 glass systems the sintering was completed prior to the beginning of the crystallization ( Table 2).
Addition of modifiers reduce the characteristic temperatures and improves the CTE of glasses due to increase in number of non-bridging oxygen atoms. This effect becomes more prominent with increasing atomic radii of modifiers used (23). The HJ3 and HJ4 glass system resulted in high Tg

Accepted Article
This article is protected by copyright. All rights reserved.
showed the formation of Anorthite (CaAl 2 Si 2 O 8 ) as main phase while solid solution of Akermanite-Gehlenite Ca 2 (Mg 0.5 Al 0.5 )(Si 1.5 Al 0.5 O 7 ) as secondary phase. The CaAl 2 SiO 8 phase has CTE of 4.9 x 10 -6 K -1 (52), and Ca 2 (Mg 0.5 Al 0.5 )(Si 1.5 Al 0.5 O 7 ) has CTE around 7.7-8.0 x 10 -6 K -1 (54). No Sr containing phase was detected in HJ1 due to its minimal concentration (9 mol%), thus indicating that the SrO was only present in the residual glassy phase. The HJ3 and HJ4 systems with higher SrO concentration formed Sr containing crystalline phases. In HJ4 glass-ceramic, the higher SrO content and a suitable SiO 2 /SrO resulted in the formation of desired SrSiO 3 . The long dwelling time (5h) resulted in the devitrification of secondary SiO 2 phase (cristobalite) in addition to SrSiO 3 . The presence of cristobalite could be an issue in particular if the joined samples would have been submitted to thermal cycles, since a phase transformation around 270°C (with a change in the specific volume) can lead thermomechanical stresses in the microstructure with possible crack formation, thus affecting the joined structure integrity.
The XRD of the aged glass-ceramics of HJ1, HJ3 and HJ4 systems as shown in Figure 2(a), 2(b) and 2(c) respectively, indicate that these systems were stable after ageing and no new phase was formed.
The SEM analysis of Crofer22APU/ as-joined glass-ceramic interfaces (Figure 3) show continuous interfaces, thus confirming a strong adhesion of the three glass-ceramics with both the Crofer22APU and 3YSZ substrates. There was no evidence of any cracks, gaps or delamination at both interfaces.
The microstructure of the glass-ceramics was dense (low fractions of pores) and uniform throughout the sample.
The detail examination of the Crofer22APU/HJ1 interface (Figures 3(a)) clearly showed a continuous thin layer of crystals as indicated by the white arrow. The growth of these crystals (1-2 µm) was most likely due to the heterogeneous nucleation where the Crofer22APU substrate served as the nucleation site. No strontium chromate was detected at the Crofer22APU/HJ1 interface (at least after the joining process in air atmosphere).

Accepted Article
This article is protected by copyright. All rights reserved.
The microstructure of the HJ3 glass-ceramic (Figure 3(c) and 3(d)) showed the presence of different types of crystalline phase in addition to the residual glassy phase. Figures 3(e) and 3(f) also show a very dense HJ4-based joint with little porosity, as well as its strong adhesion with both of the joined materials. The "viscous character" of this system led to a negligible level of porosity during the joining treatment.
The microstructure of the HJ4 system after processing at different heat treatments is shown in Figure 4. The microstructure of HJ4 glass-ceramic obtained after the heat treatment of 950°C, 1h (Figure 4(a)) shows the presence of only one crystalline phase in addition to the residual glass phase.
According to XRD (Figure 2(b)) that phase corresponds to SrSiO 3 . On the other hand, the heat treatment at 950°C, 5h at 2°C/min resulted in the significant evolution of microstructure. Apparently this heat treatment not only increased the volume fraction of the initially formed SrSiO 3 phase but also resulted in the devitrification of secondary SiO 2 phase (Figure 4(b)). The corresponding EDS spectrum of the dark zones in the SEM images of the HJ4 glass-ceramic microstructure indicated a SiO 2 rich phase (spot 1), whereas spot 2 showed a Si and Sr rich phase, referring to SrSiO 3 as investigated by XRD. EDS analysis of spot 3 was similar to spot 2, thus indicates SrSiO 3 phase. The different contrast between spot 2 and spot 3 was probably due to the presence of very thin glass layer on the crystals at spot 3. These EDS results validate the XRD analysis of HJ4 at different heat treatments, as discussed in section earlier.
The table 4 shows the EDS point analysis carried out at the residual glassy phases on HJ1, HJ3 and HJ4 as-joined glass-ceramics. In all the glass systems, the concentration of Sr was less than 10 at %, which was beneficial in order to maintain the viscous behavior of glassy phase. These results also rationalize the purpose behind increasing the SrO contents from HJ1 to HJ4 i-e to form SrO containing crystalline phases and to have minimal SrO in the glassy phase thus avoiding the potential formation of SrCrO 4 . The EDS line scans across Crofer22APU/ glass-ceramics interface ( Figure 5) also

Accepted Article
This article is protected by copyright. All rights reserved.
confirmed that there was no formation of chromates at interface, nor diffusion of elements from either side of interface.
The Crofer22APU/glass-ceramic interface for the HJ1, HJ3 and HJ4 systems ( Figure 6) investigated after the thermal ageing for 1000 hours at 850 o C showed a uniform and crack free microstructure.
The microstructures of these glass systems were dense and similar to their as-joined microstructures ( Figure 3). A good sinter-crystallization behavior of these systems, studied previously, is further confirmed, since no pores are detected in the microstructure.
The EDS line scans across the Crofer22APU/HJ1, Crofer22APU/HJ3 and the Crofer22APU/HJ4 interfaces after thermal ageing has shown in Figure 7. From the EDS line scan across Crofer22APU/HJ1 interface, the diffusion of Cr into the HJ1 glass ceramic was detected. Although, the crystalline phases in HJ1 reduced the CTE of the obtained glass-ceramic; therefore, HJ1 is the least promising sealing candidate among three studied glass systems. Nevertheless, the HJ1 glass system showed good adhesion with Crofer22APU even after aging for 1000 hours at 850°C, however, the diffusion of Cr can potentially form SrCrO 4 and can be critical in long terms. Moreover, presence of Cr can also alter the CTE of the glass. On the other hand, no Cr diffusion was detected in HJ3 and HJ4 glass-ceramics. These SEM results excluded the formation of SrCrO 4 and the consequent delamination at the Crofer22APU/glass-ceramic interface, thus making these systems as promising candidates for solid oxide cells seals at the working T of 850°C. The presence of low CTE Sr 2 Al 2 SiO 7 phase in HJ3 and cristobalite phase (with different polymorphs) (7) in HJ4 did not determine cracks within the glass-ceramics or at the interface with Crofer22APU even after 1000 hours at 850 °C.

Accepted Article
This article is protected by copyright. All rights reserved.

Accepted Article
This article is protected by copyright. All rights reserved.