AISI441 ferritic stainless steel is classically used as interconnect raw material in CEA stacks for its interesting properties for high temperature SOEC application and its relatively low cost among all commercial ferritic stainless steel grade candidates. Its chemical composition is given in Table 1. To study the long-term oxidation resistance and Cr evaporation of this steel, bare 0.2 mm thick AISI441 steel samples of 20 x 20 mm² in the “as-received” state were coated on both sides with a combination of several layers: CeO2/Co (CeCo), Co/CeO2 (CoCe) and CeO2/Co/CeO2 (CeCoCe) elaborated by PVD-HiPIMS at CEA. AISI441 samples were also coated by CeO2/Co after a thermal treatment at temperature higher than 800°C used to stress-release AISI441 interconnects inducing a slight oxidation. For reference, bare and stress-released (SR) AISI441 were also considered. Before deposition or thermal treatment, AISI441 samples were cleaned in ultrasonic acetone then ethanol bath for 10 min each. Average thicknesses of ceria and metallic cobalt layers were measured at about 40 and 580 nm respectively after deposition, using SEM and TEM images of samples cross-sections prepared by focused ion beam (FIB) as shown in Fig. 1.
Table 1
Chemical composition (in wt%) of the AISI441 ferritic stainless steel used in this study and analysed by Glow Discharge Mass Spectrometry. * Given by the manufacturer
| Fe | Cr | Nb | Si | Mn | C | Ti | Ni | Mo | V | Co | Cu |
| Bal. | 17.80* | 0.44 | 0.45 | 0.25 | 0.015* | 0.12 | 0.15 | 0.02 | 0.01 | 0.02 | 0.08 |
All samples were heat-treated in static ambient air for 5 000 h into two AET Technologies tubular furnaces at 700 and 800°C respectively, with 1°C/min rate of heating-up and cooling-down (note: this rate is not respected down to room temperature due to furnace inertia). No air stream was flowing across the quartz tubes of the furnaces that are closed by circular metallic end caps with polymeric gasket in the cooled zone of each tube tip. The tube volume is big enough to avoid any air starvation on the samples surface during the entire oxidation treatment. Samples were positioned horizontally in specific quartz holders (Fig. 2).
Specific weight gains, expressed in mg/cm², were measured at 2 000 and 5 000 h thanks to a Mettler XS204 balance with a 0.1 mg resolution. As Co coated samples oxidize very fast into Co3O4 spinel oxide within the first hours of the oxidation test, 0.185 mg/cm² due to the total oxidation of a 580 nm Co layer (mass of the incorporated oxygen) was subtracted for the coated samples, to compare the oxidation rate only linked to the thermally formed oxides of the steel. This value was calculated from the following Eq. (1) with the Co thickness \(\:{th}_{Co}\):
$$\:\left(\frac{\varDelta\:m}{S}\right)=\:\frac{4\:*\:16\:*\:{th}_{Co\:}*\:{\rho\:}_{Co}}{3\:*\:{M}_{Co}}$$
1
where \(\:\frac{\varDelta\:m}{S}\) is the weight gain normalized by the sample surface, \(\:{\rho\:}_{Co}\) the density of the metallic Co, \(\:{M}_{Co}\) its molar mass. Each value of specific weight gain at 2 000 h and 5 000 h is an average calculated on 6 and 4 samples respectively.
To allow material screening regarding Cr volatilization over long durations of a few thousands of hours, a method was designed at CEA about 10 years ago with the aim to test several materials in the same experiment [15]. It differs from the transpiration method broadly used for such a study that allows measuring the absolute Cr volatilization rate for one sample but does not allow quickly comparing the Cr retention power of many samples in one single test [19]. This method is based on the Cr poisoning property of La0.8Sr0.2MnO3−δ (LSM): the volatile Cr species emitted by bare or coated interconnect samples are trapped by this getter material. The sample-holders are designed to maintain a fixed distance of 0.5 mm between the LSM getter and the sample. Further, in the present work, two quartz towers were positioned in the furnace in order to have two samples per AISI441/coating combination in the same testing conditions, evaluate reproducibility and get more reliable data. Once oxidation test finished, the LSM getters were dissolved in acidic solution for chemical analysis of Cr traces by ICP-OES (Agilent model 725 ICPOES), in addition to La, Sr, Mn, Ce, Fe, Co and Ti elements. La, Sr and Mn were analyzed by ICP-OES as reference elements present in the getter material to check their stability between samples and the reproducibility of the technique. It was also checked that none of Fe or Ti elements from the AISI441 steel was found in a higher amount than present in the initial pristine LSM, indicating a limited evaporation of those elements in the testing conditions and setup. Each element content is expressed as the mass of the element related the total mass of the getter. As mentioned previously, this method does not allow the determination of the absolute volatilization rate of Cr species from the interconnect material because all volatile Cr species are very likely not trapped by the LSM getter material. Nevertheless, in this specific study carried out in similar conditions for each experiment, it permits to evaluate the different coatings efficiency against Cr volatilization compared to bare AISI441 (material screening). For post-test microstructural characterization, cross-sections of each sample were cold vacuum embedded in epoxy resin and polished. Alumina and not silica particles were used in solutions of the polishing process. A carbon coating was sputtered on the surface to avoid charging effect on the samples when examined by a high-energy electron beam and improve signal to noise. Images with secondary and back-scattered electrons detectors were recorded with a Philips XL30 scanning electron microscope (SEM) equipped with an energy dispersive X-ray 6650 INCA Energy spectrometer for EDX chemical analysis. The average thickness of the oxide layers and of the substrates was measured on SEM images and EDX profiles as well.
CeO2/Co coating was also integrated on the air side in 4 Repeat Units (RU) of a 9-cell short stack operated at 750°C in electrolysis mode for a total duration of 5 200 h. The electrochemical performances of uncoated and coated RUs are only slightly in favor of the coated ones with degradation rates slightly lower in average. Post-test analysis of selected samples of stack cells and interconnects were carried out by SEM/EDX as well as time-of-flight secondary ion mass spectrometry (ToF-SIMS). Images were acquired at 10 kV in Zeiss Leo and Merlin SEM, and EDX with Bruker detectors at 10 or 20 kV. In ToF-SIMS 5 instrument from IONTOF GmbH, high-resolution imaging analyses were performed in positive ion mode using Bi+ (30 keV) as primary ion species.
Oxidation behavior in air up to 5 000 h
Average weight gains of each tested AISI441/coating combination are presented at 700 and 800°C in Fig. 3.
After 2 000 and 5 000 h at 700°C (Fig. 3(a)), there is no significant difference among bare, SR and coated AISI441 steel, in regard to scattering values, except for CoCe-coated AISI441 which presents the highest average weight gain. Visually, overall surfaces look uniform for all samples (not shown here).
The same conclusions about oxidation kinetics are given for the samples exposed for 2 000 h at 800°C (Fig. 3(b)). Nevertheless, at this higher temperature, CoCe-coated AISI441 does not show higher weight gain than the other coated samples. After 2 000 h, a few oxide delamination sites are visible on bare and SR samples (see light grey zones on images in Fig. 3(b)). After 5 000 h, the oxide spallation of those samples expanded to the major part of the samples surface, partially explaining their lower weight gain with high scattering compared to coated samples for which no spallation was observed (Fig. 3(b)). It is worth noticing here that all samples were curved after the oxidation test at 800°C, whatever their position in the sample holder. This geometry reveals plastic deformation of the samples during the exposure test. It is very likely induced by the isothermal oxide growth on the surface, which created compressive stresses in the oxide due to the subsequent volume expansion leading to strong deformation in the substrate. It is particularly true when the oxidation rate is high, which could explain why curvature is mainly observed at 800°C compared to 700°C and why a higher curvature is observed with bare AISI441 (faster oxide growth) compared to coated AISI441 (reduced oxide growth). Compressive stresses could also be due to the difference of thermal expansion coefficient (TEC) between the oxide and the alloy (slightly lower TEC for the oxide in this case), involving deformation during the cooling step down to room temperature at the end of the oxidation treatment. Those phenomena are often observed in literature on very thin samples after oxidation at high temperatures [20].
Thus, at that stage, it can be concluded that the slight surface pre-oxidation induced by the stress-release thermal treatment of interconnects does not improve significantly the oxidation resistance of AISI441 steel neither without nor with an additional coating (CeCo for instance). Moreover, the coatings are efficient to avoid oxides spallation after 5 000 h in air at 700°C and 800°C as well. Based on the present results, all coating combinations have a quite similar long-term effect on the oxidation behavior of AISI441 steel. Nevertheless, a question remains about the high weight gain measured for CoCe-coated AISI441 at 700°C. In any case, precautions must be taken when giving conclusion on the oxidation behavior of samples only based on weight gains. Indeed, a lower weight gain can reflect the formation of thinner oxide scale (better oxidation resistance) but higher Cr evaporation and oxide delamination as well (lower oxidation resistance). Consequently, those results need to be complemented by Cr volatilization measurements as follows and complete post-test microstructural characterizations of the oxide scale (nature and thickness of oxides) for a clear conclusion.
Cr volatilization in air up to 5 000 h
The Co content measured in the LSM getter material after 5 000 h of long exposure test is given in Fig. 4(a). It is slightly higher in LSM getters facing coated AISI441 than bare AISI441 samples in particular at 800°C, indicating that Co may slightly evaporate from the coating during the test. In regard to the Cr content (see Fig. 4(b)), the coatings show a clear retention of Cr as the Cr amount detected in the LSM getter facing the coated samples is lowered by a factor 3 to 4 compared to the bare and SR AISI441. The Cr concentration is quite the same with all CeCo, CoCe and CeCoCe coating combinations.
The improvement of Cr retention brought by the coatings in comparison to bare AISI441 reference is presented in Fig. 5. The Cr retention value is calculated by Eq. (2), leading to a Cr retention equal to 0 for bare AISI441 and equal to 1 when no Cr is detected:
$$\:Cr\:retention=1-\left(\frac{Cr\:content\:of\:a\:sample}{Cr\:content\:of\:bare\:AISI441}\right)$$
2
Based on this representation, it is shown that the coatings decrease by 70% ± 14% and 75% ± 15% the Cr evaporation compared to bare AISI441 at 700 and 800°C respectively. This value is lower than the Cr retention power value of 99% measured for Ce-Co coated AISI 441 in humid air at 800°C in [21–23]. This discrepancy is very likely linked to the experimental procedure used in that study which is less accurate to measure the absolute Cr volatilization rate that the denuder tube used in [21–23].
Post-test characterization
SEM/EDX microstructural characterization of samples revealed that after 5 000 h of oxidation in static ambient air, bare and SR AISI441 show a similar oxidation behavior. The classical duplex oxide made of an inner Cr oxide (most likely chromia) and an outer (Cr,Mn) oxide (most probably the spinel form) is observed by SEM (see Fig. 7(a)). The thickness of this weakly conductive duplex oxide appears lower when AISI441 is stress-released but it remains very high at 800°C and is associated with a strong oxide spallation as well (Fig. 6). A similar comment can be done for the couple formed by AISI441 + CeCo and SR AISI441 + CeCo but with a small difference of this oxide layer thickness between both samples and a better adherence of this oxide due to CeCo coating. Those results confirm the negligible impact of this thermal treatment on oxidation behavior in the tested conditions (700 and 800°C). It is worth noticing here that when a coating is applied, the duplex character of the Cr and (Cr,Mn) oxides layer was more or less attenuated in particular after 5000 h at 800°C as Mn diffuses into the Co oxide protective layer (see Fig. 7(b)). This results in a Cr oxide layer possibly containing small amounts of Mn. AISI441 + CeCo and AISI441 + CeCoCe present a lower thickness of Cr oxide than the duplex layer of bare and SR AISI441, confirming the good efficiency of those coating combinations on the oxidation kinetics (Fig. 6). This is particularly true at 800°C where oxide spallation strongly affects the measured oxide thickness for samples made of bare and SR AISI441.
The difference between the initial and final thicknesses of the AISI441 substrate measured on SEM images and EDX profiles showed that it lost up to 15 µm on bare and SR AISI441 due to the oxidation process at 800°C compared to only 5 µm for coated samples. The positive effect of CeCo-based coating is noticeable reducing the thickness loss by a factor 3 after 5 000 h at 800°C. No clear depletion of Cr is revealed below the oxide layers whatever the sample in agreement with the fast Cr diffusion in ferritic stainless steel at these high temperatures (Fig. 7 for bare and CeCo-coated AISI441 as example). From EDX analysis of the coated samples, the formation of a protective (Co,Mn) oxide with Mn coming from the steel is clearly observed on top. This (Co,Mn) oxide with a thickness of about 1 µm is quickly formed from the deposited metallic Co in the first instants of oxidation. The thin ceria layer is also clearly visible above the Cr oxide scale for the CeCo coating (Fig. 7(b)) and above the (Co,Mn) oxide for the CoCe coating (Fig. 8(a)). In some places, this layer appeared as a thin line of discrete ceria particles after 5000 h at 800°C while it remained continuous after oxidation at 700°C. No or low Cr content is detected in the Ce/Co-based coatings whatever the configuration and the temperature (Fig. 7(b) for CeCo-coated AISI441 at 800°C for instance). In parallel, no or low Fe content is detected in the CeCo and CeCoCe coatings (Fig. 7(b) for CeCo-coated AISI441). However, high levels of Fe are detected in the oxides and in particular in the CoCe coating after exposure at 700°C (Fig. 8(a) and (b) where Fe replaces almost all Mn in the coating at this location). For this sample, higher thicknesses of the protective (Co,Mn,Fe) oxide were measured at 700°C (about 2 µm in this case), in agreement with the higher weight gain previously obtained. This result looks consistent with literature data showing fast diffusion of Fe in a Co spinel coating on a Sanergy HT steel sample after 168 h exposure at 850°C in humidified air [11]. It also confirms the ability of Ce reactive element to promote chromia nucleation in the first instants of oxidation and limit cations diffusion, including Fe, from the steel when first applied on the steel surface [17].
Finally, based on all the SEM images and EDX profiles recorded in this work, it seems that a Si-rich oxide is present at the steel surface below the other oxide layers whatever the sample and temperature. Thanks to the present polishing procedure using alumina particles, it can be concluded that the Si oxide layer is not a pollution and is formed during the oxidation treatment. Discontinuous Si oxide layer has already been described in literature for this steel grade [24]. It is probably because Laves phase Fe2Nb formed in the steel thanks to small Nb addition is not able to trap the entire Si amount present in the steel. Nevertheless, this Si-rich oxide is not considered detrimental regarding the electrical conductivity at the steel/electrode interface, as far as discontinuous or mixed with other elements increasing its electrical conductivity [25].
Post-test observations of the operated short stack with 4 CeCo-coated interconnects out of 9 on the air side showed that the thickness of the weakly conductive duplex oxide is slightly higher for an uncoated RU (2-2.5 µm) compared to the Cr oxide scale observed for a coated one (1.5-2 µm). If this thickness value at 750°C is consistent with the values measured at 700 and 800°C at sample scale, the difference between uncoated and coated RUs is lower. It could be due to the presence of a La0.8Sr0.2MnO3−δ (LSM) contact layer limiting steel oxidation and/or the influence of electrical current and its direction in operation through electrochemical reactions [26]. As observed at sample scale, a Si oxide layer, thinner than 1 µm and fortunately discontinuous at this stage, is clearly seen on the steel surface (Fig. 9). Then, despite the possibly high resistivity of this compound, it is not considered as detrimental, as already mentioned at sample scale.
On the H2O/H2 side, the oxide scale formed on the interconnect surface looks very similar to the air side and is composed of a duplex Cr-rich and Cr/Mn-rich oxide, however with a thickness slightly higher (2.5–3.5 µm in average) as mentioned in some works [27]. The formation of Si oxide at the oxide/metal interface was also observed on the H2O/H2 side. Fortunately, no sign of dual effect, i.e. catastrophic oxidation on the air side when the interconnect is exposed to H2 atmosphere on the other side is observed for this tested stack whatever the coated or uncoated RU.
The presence of Cr species in the contact layer and the oxygen electrode, which is known to be detrimental for the stack performance [5–7], was investigated by ToF-SIMS analyses, as low Cr levels are not easily distinguished with SEM/EDX technique. Figure 10 below illustrates the slight but actual gradient of Cr amount detected in the LSM contact layer from the interconnect to the electrochemical cell in a standard RU while no Cr is detected when the CeCo coating was present. The presence of Cr in the oxygen electrode is also visible only in absence of protective coating.