Unlocking the Value of End-of-Life JÜLICH Solid Oxide Cell Stack Interconnect Assembly: A Combined Experimental and Thermodynamic Study on Metallic Resource Recyclability

Abstract

The present study provides fundamental information on the resource recyclability of the interconnect assembly, i.e., the steel interconnector and the nickel meshes, from an end-of-life JÜLICH Solid Oxide Cell Stack—F10 design. The interconnector is composed of iron, chromium, and less than 4 wt.% of other alloying elements, mainly cobalt and manganese. Calculated blended compositions with the nickel meshes revealed their potential as a raw material in the production of 4xx, 2xx, or 3xx stainless steels. The melting behavior of the interconnect assembly was investigated under different conditions, i.e., in inert and oxidizing atmospheres, with and without the addition of slag-forming fluxes. The results demonstrated preferential oxidation of chromium in a trivalent state within the stable cubic spinel phase. Finally, the experimental results were compared with the thermodynamic equilibrium calculations based on the available databases (FToxid, SGTE, and SGPS) in FactSage 8.1 software. The calculated tendency to oxidize is in the order of Cr > Mn > Fe > Co > Ni at P(O₂) greater than 10⁻¹⁰ bar, validating the experimental results.

1. Introduction

The race against time towards climate neutrality has driven several countries to adopt renewable energy alternatives, including the large-scale use of decarbonized “green” hydrogen [1,2]. The global hydrogen market is expected to grow rapidly in the coming decades to meet the climate targets set by the Paris agreement [2,3]. In Germany alone, deployment efforts are already in place to generate 90–110 TWh of energy from green hydrogen by 2030 [2,4]. In particular, the H₂Giga project was commenced in 2021 with the aim of reaching GW-scale electrolysis capacities [5]. This planned expansion will generate new waste streams in the form of end-of-life (EoL) fuel and electrolysis cells. Proactive approaches in the discussion and management of these resources must therefore be undertaken to support decarbonization efforts. In line with this, the ReNaRe subproject has been launched as a part of the H₂Giga technology platform to develop recycling and re-utilization solutions tailored to prospective electrolysis technologies [6].

Solid oxide cells (SOCs) are among the promising candidates for industrialization, owing to their superior electrical efficiency and lifetime performance [7]. Additionally, one salient feature of SOCs is their fuel feed flexibility, i.e., hydrogen, carbon monoxide, and hydrocarbons in fuel cell mode or steam and carbon dioxide in electrolysis mode, which enables the implementation of carbon capture and utilization strategies [8,9]. An illustration of a planar SOC stack developed by Forschungszentrum Jülich (JÜLICH) for stationary applications is illustrated in Figure 1. To increase electrical capacities, individual SOCs are stacked together, with interconnectors alternately inserted between them [8]. Interconnectors have two-fold purposes: to provide a serial electrical connection between cells and to separate the oxidizing and reducing gases [8]. An SOC stack is operated at temperatures of at least 700 °C under high oxygen partial pressure gradients for long durations. Therefore, the selection of interconnector materials requires stringent criteria, i.e., they must have a compatible coefficient of thermal expansion (CTE) with the cell, strong mechanical integrity, high thermal and electrical conductivities, and cost-effectiveness. Special alloys have been developed for this purpose, such as Crofer 22 H/APU alloys by JÜLICH [10,11,12] and Sunfire [13] and the CFY alloy in Hexis SOC stacks [14]. In recent JÜLICH designs, special coatings are applied on the air side, while Ni meshes applied on the fuel side of the alloy improve chromium retention and electrical contact [15,16].

Manufacturers of stationary SOC stacks aim for design runtimes up to 80,000 h [16]. These values are, however, not often achieved due to cell and interconnector degradation leading to premature failure. Interconnectors suffer considerable degradation due to, among other reasons, localized gas leakages and combustion, uneven temperature distribution, and element interdiffusion in areas of direct contact with adjacent cells, all of which may result in an overall decrease in electrical conductivity and mechanical properties [11,12,13,14,15,16,17]. Refurbishment of EoL interconnectors might be possible, e.g., via blasting off accumulated dirt, re-coating, and heat treatment; however, up to which extent is currently not well known. Ultimately, it is not arguable that interconnectors will inevitably reach the end of their useful life and will therefore require viable recycling solutions. While studies have been made to characterize and develop recycling strategies for the ceramic cells in EoL SOC stacks [18,19,20], similar efforts have yet to be conducted for the interconnectors. Steel-based interconnectors from JÜLICH SOC stacks are anticipated to be of high intrinsic value, and their recovery through remelting is potentially feasible [18]. However, the level of contaminants at the end of their lifecycle remains uncertain. Therefore, this paper provides fundamental information on the composition and metallic resource recyclability of an EoL JÜLICH SOC interconnect assembly. To achieve this, extensive sample characterization was performed, and the average compositions of the blended components, i.e., the interconnectors and the nickel meshes, were calculated. Finally, element distributions between metallic and oxide phases when melted under increasing oxidizing conditions were assessed through experimental work and thermodynamic equilibrium calculations.

2. Materials and Methods

2.1. Characterization

In the present study, an interconnect assembly consisting of an interconnector and two Ni meshes (Ni mesh 1: 2.0 × 0.6 mm²; Ni mesh 2: 0.2 × 0.125 mm²; mesh size x thickness) from one JÜLICH SOC stack of 10 × 10 cm² cells (F1004-60) was investigated. The interconnector was made up of Crofer 22 APU alloy and coated with MnCo_1.9Fe_0.1O₄ (MCF) and LaMn_0.45Co_0.35Cu_0.2O₃ (LCC10) coatings from the air side. The four-layer stack was operated for 3000 h at 710 °C in fuel cell mode using H₂-20%H₂O and air as fuel and oxidizing gases, respectively. The stack was then manually disassembled, and the top-most interconnect assembly was used for analysis. Specimens measuring 4 × 4 × thickness mm³ in dimension were removed from the interconnector’s longitudinal length using a laser cutting machine and subsequently divided into five zones, as shown in Figure 2. Zone 1 represents the edges of the interconnector, while zones 2 to 5 cover its inner region, as depicted in Figure 2c. Zones 4 and 5 are located on the rib-channel region in contact with the adjacent cell. Notably, Ni mesh 1 was glass-welded into the interconnector and could not be completely separated during sampling. The acronyms of techniques utilized to characterize the specimens are listed in

Table 1. Nomenclature of acronyms.

Acronym	Description
SEM	Scanning electron microscope
EDS	Energy-dispersive X-ray spectroscopy
ICP-OES	Inductively coupled plasma–optical emission spectroscopy
DTA	Differential thermal analysis
DSC	Differential scanning calorimetry
CAD	Computer-aided design
XRD	X-ray diffraction
BSE	Backscattered electrons
TGA	Thermogravimetric analysis
EBSD	Electron backscatter diffraction

.

Three sets of zonal specimens, each specimen weighing between 250 and 350 mg, were collected for characterization. The first set was coated with gold approximately 10 nm thick via sputtering to minimize electric charging during imaging. This was followed by nickel electroplating to deposit a nickel layer about 10 µm thick to preserve the oxides. Subsequently, the specimens were embedded in epoxy resins, ground, and polished. The polished surfaces were then analyzed with an SEM (MERLIN, Zeiss, Oberkochen, Germany) equipped with EDS (X-max 80, Oxford Instruments, High Wycombe, UK). Pure elements were used as standards for quantitative EDS analysis. The specimens from the second set were analyzed with an ICP-OES (iCAP 7600, Thermo Fisher Scientific, Waltham, MA, USA). On the last set, DTA (STA 449 C, NETZSCH equipped with rhodium furnace and Al₂O₃ protective tube) was performed to determine phase transition temperatures. A schematic of the device used for DTA is shown in Figure 3a. DTA signals were recorded using a DSC sample holder (Type S, Pt/Pt10%Rh thermocouple), as illustrated in Figure 3b. Zonal specimens were then placed in Pt/Rh crucibles with Al₂O₃ liner, heated to 1600 °C at 15 °C/min and then held for 5 min under 40 mL/min argon gas. A METROTEC oxygen sensor detected an O₂ content of around 620 ppm in the outflow gas stream. Transition temperatures were recorded from the second or third cycle when sample homogeneity was ensured. Temperature calibration was also conducted to ensure accurate phase transition temperatures. This involved heating up standard pure substances (In, Sn, Bi, Zn, Al, Ag, Au, and Ni) above their melting temperatures and adjusting the measured endothermic onset peaks to align with their known melting points. A calibration curve, derived from the differences between measured and reference melting points, was obtained and subsequently used to correct the experimental transition temperatures. To calculate the blended compositions, a three-dimensional CAD model was created using the open-source FreeCAD 0.19 software. The volumes of the various parts of the model were determined through geometric calculations by this software, and corresponding masses were calculated using reference densities. Finally, the blended compositions were determined from the component masses and the elemental composition of the specimens obtained from ICP-OES analyses.

2.2. Remelting and Smelting Experiments

To evaluate the element distribution between the metallic and oxide phases under increasing oxygen partial pressures, representative samples of approximately 1700 mg each were utilized in this study. The samples were comprised of zonal specimens and nickel mesh and possessed compositions close to that of the interconnect assembly. The samples underwent a series of remelting (without fluxes) and smelting (with fluxes) experiments. In the smelting experiments, lime and silica fluxes were added at a ratio of CaO/SiO₂ = 1.7 (w/w) and 10 wt.% of the sample. This ratio was decided based on the recommended basicity range of chromium-bearing steel slags to achieve optimum chromium recovery [21,22,23]. Readily available amorphous SiO₂ powder (Alfa Aesar, Kandel, Germany, 99.5%) was used, whereas CaO was obtained by decomposing CaCO₃ (VWR, Leuven, Belgium, 99.0%) at 1000 °C for 3 h. An XRD analysis with PANalytical EMPYREAN Diffractometer (Malvern, UK) using Cu Kα radiation in a Bragg–Brentano geometry confirmed the powder purity at around 95 wt.% CaO and 5 wt.% Ca(OH)₂. The fluxes were then combined and sintered at 800 °C for 3 h. Experiments were performed using the NETZSCH STA 449 F3 Jupiter ^® equipped with graphite furnace and Al₂O₃ protective tube. Samples weighing between 1600 and 1800 mg were placed within open 0.7 mL Al₂O₃ crucibles (KYOCERA, Mannheim, Germany). These crucibles were then positioned inside open 3.4 mL Al₂O₃ crucibles (NETZSCH, Selb, Germany) and fixed to the sample holder that also serves as a type-S thermocouple. The sample holder layout is shown in Figure 3c. The furnace was purged with argon gas and evacuated twice to remove any atmospheric oxygen. The furnace was then heated to 1600 °C and held for 30 min under increasing oxygen partial pressures at heating and cooling rates of 10 °C/min. Thermogravimetric analyses (TGAs) were simultaneously performed wherein sample weight changes were recorded during heating and cooling cycles. Experimental conditions used are listed in

Table 2. Experimental conditions.

Experiment	Sample Code	Ar/O2 Gas Mixture (mL/min)	Calculated P(O2) (Bar)
Remelting	R0	60/0	0
	R1	55/5	0.08
	R2	50/10	0.17
	R3	40/20	0.33
Smelting	S0	60/0	0
	S1	55/5	0.08
	S2	50/10	0.17
	S3	40/20	0.33

. Ultrapure argon and oxygen gases (99.9992 vol. %) provided by AIR PRODUCTS (Germany) were utilized. The total gas flow rates and pressures were constant at 60 mL/min and 1 bar, respectively. Subsequently, samples were cross-sectionally cut, embedded in resins, ground and polished, and analyzed with SEM and EDS. The results were compared with equilibrium thermodynamic calculations based on the available databases (FToxid, SGTE, SGPS) in FactSage 8.1 software. All oxide phases in FToxide and all metallic phases in SGTE and SGPS thermodynamic databases were employed in the calculations.

3. Results and Discussion

3.1. Sample Characterization

Figure 4a shows the cross-sectional microstructural layers on the zone 4 specimen. On the air side of the Crofer 22 APU alloy substrate, there were two main oxide layers—a perovskite top layer based on LCC10 and a second spinel layer based on the MCF coating. The perovskite and spinel layers were measured to be 96 and 132 µm thick, respectively, while the bulk of the thickness was attributed to the alloy substrate. The outermost Ni layers were incorporated via electroplating during the metallographic sample preparation to preserve the structural integrity of the oxide coatings. Additionally, thermally grown Cr₂O₃ and Cr₂MnO₄ were present along the alloy interfaces in both fuel and air sides. The oxide coating stoichiometric formulas were calculated based on the average atomic composition from three EDS point analyses per oxide (Figure 4b). The results revealed minor differences from the reference formulas of the coatings. BSE images and phase images derived from EDS mapping of the alloy–MCF coating interface (Figure 4c) showed the presence of an oxide layer, Mn_0.6Co_1.35Cr_1.05O₄, approximately 6 µm thick. This layer served as an interdiffusion zone between the MCF and Crofer 22 APU regions.

The perovskite and spinel coatings were similarly present in zones 2 to 5 (not shown), although they were absent in zone 1. Instead, a glass matrix consisting of Al, Si, and Ca oxides was found (Figure 4d) in a few zone 1 specimens. This can be attributed to the Type 87 ZYBF-2 glass sealant used to join the stack components. Finally, Ni mesh 1 was present on some welded areas in zones 4 and 5 due to the difficulty in completely detaching the mesh (Figure 4e).

The elemental compositions shown in

Table 3. Elemental composition and standard deviation (wt.%) values of zonal specimens.

Element	Zone 1	Zone 2	Zone 3	Zone 4	Zone 5	Mean (Zones 2–5)
Fe	76.3 ± 0.03	74.3 ± 0.60	73.76 ± 0.60	74.4 ± 0.60	73.7 ± 0.90	74.2 ± 0.63
Cr	22.9 ± 0.11	22.2 ± 0.10	22.0 ± 0.30	22.3 ± 0.30	22.2 ± 0.30	22.2 ± 0.24
Co	0.00 ± 0.002	1.83 ± 0.030	2.32 ± 0.040	1.21 ± 0.040	1.60 ± 0.040	1.74 ± 0.034
Mn	0.52 ± 0.005	1.45 ± 0.050	1.77 ± 0.050	1.15 ± 0.050	1.31 ± 0.050	1.42 ± 0.045
La	0.09 ± 0.003	0.09 ± 0.003	0.09 ± 0.003	0.81 ± 0.007	0.58 ± 0.019	0.39 ± 0.009
Ni	0.02 ± 0.001	0.02 ± 0.001	0.02 ± 0.001	0.06 ± 0.002	0.09 ± 0.001	0.05 ± 0.001
Traces	0.13 ± 0.005	0.08 ± 0.008	0.05 ± 0.003	0.05 ± 0.003	0.05 ± 0.002	0.06 ± 0.005

and based on ICP-OES analyses indicate that the zonal specimens were predominantly composed of Fe and Cr, while other alloying elements made up less than 4 wt.%.

Among the minor alloying elements, Co and Mn were the most abundant, with an average of 1.74 and 1.42 wt.%, respectively. This was expected due to the relatively thicker MCF layer compared to other oxide layers. The concentration of Co and Mn was lowest in zone 1, indicating the absence of MCF coating in this region. The concentration of La was highest in zones 4 and 5 at 0.81 and 0.58 wt.%, respectively, demonstrating the exclusive presence of perovskite coating on the active region in direct contact with the cell. The concentration of Ni was minimal at <0.1 wt.% for all zonal specimens; therefore, mesh sintering on non-welded areas was insignificant. The presence of Ca, Al, and Si from the glass sealant in zone 1 was not demonstrated here due to its absence on the analyzed specimen.

DTA measurements of the zonal specimens in Figure 5 show that the onset temperatures of the endothermic peaks, which correspond to melting points, range from 1514 °C to 1519 °C. These temperatures fell within the expected range for Crofer 22 APU (1510–1530 °C) [24], suggesting a negligible influence of the coatings on the alloy substrate’s melting point. Additionally, small endothermic peaks between 500 °C and 700 °C suggested the occurrence of solid–solid transformations. The experimental melting temperatures of zonal specimens were within ±3 °C from the melting temperature in the Fe-Cr binary phase diagram calculated with FactSage 8.1 using the SGTE database (Figure 6).

The melting point of Ni mesh 2 was found to be 2 °C lower than that of the pure Ni (1454 °C) obtained through temperature calibration as described in Section 2.1, indicating the high purity of the mesh. Therefore, it can be concluded that the Ni mesh is unaffected by operation and can be treated as pure for the sake of analysis. Conversely, exothermic peaks revealed severe undercooling (16–36 °C). Hence, the recorded crystallization temperatures were not reliable.

The interconnector volume was determined from the CAD model created using FreeCAD 0.19 software. Due to significant compositional variation between zone 1 and the region covering zones 2 to 5, their volumes were calculated separately. Corresponding mass values were determined based on these volumes and the reference densities for Crofer 22 APU [24]. The volume and mass of Ni mesh 1 were calculated similarly using the reference density for Ni [25]. In contrast, the mass of the detached Ni mesh 2 was directly measured.

Table 4. Component volumes and masses.

Component	Volume (mm3)	Mass (g)
Interconnector * (Zone 1)	21,564.60	166.0
Interconnector (Zones 2–5)	29,694.51	228.6
Ni mesh 1 **	2424.10	21.6
Ni mesh 2	-	13.6

Density: * Crofer = 7.7 g/cm³; ** Ni = 8.9 g/cm³.

presents the volume and mass values of the components. Using the composition of the interconnector regions, and assuming that the meshes were of pure Ni, blended compositions were computed and are shown in Figure 7. Accordingly, there are three steel families close to these compositions—AISI 4xx, 2xx, and 3xx series steels. Selected lean steel grades with the closest compositions are listed in

Table 5. Elemental composition (wt.%) of selected steel products from AISI 4xx, 2xx, and 3xx series.

AISI Series	Grade	EN No.	Fe	C	Cr	Ni	Mn	Si	Others	Others	Ref.
4xx	Crofer 22 APU	1.4760	Bal	≤0.030	20.0/24.0	-	0.50	≤0.50	≤0.050 P	≤0.020 S	[24]
2xx	2304	1.4362	Bal	≤0.030	22.0/24.0	3.5/5.5	≤2.00	≤1.00	≤0.035 P	≤0.015 S	[26]
3xx	304	1.4301	Bal	≤0.07	17.5/19.5	8.0/10.0	≤2.00	≤1.00	≤0.045 P	≤0.015 S	[26]

. Production of a 4xx-series steel product through the manufacture of Crofer 22 APU enables closed-loop recycling; however, this path highly depends on the successful separation of the interconnector from welded Ni meshes upon stack dissection. One viable approach to this is via automated stack dismantling with the use of artificial intelligence [18]. On the other hand, open-loop recycling is possible via the production of 2304 and 304 steels.

3.2. Thermogravimetric Analysis

In

Table 6. Mean and standard deviation (wt.%) values of predicted elemental composition of representative samples.

Element	Composition (wt.%)
Fe	68.4 ± 0.43
Cr	20.2 ± 0.61
Ni	8.5 ± 0.25
Co	1.3 ± 0.01
Mn	1.0 ± 0.35
La	0.3 ± 0.15
Traces	0.1 ± 0.02

, the mean predicted compositions and standard deviations of the representative samples are given. Remelted and smelted samples pertain to samples with and without the addition of 1.7CaO/SiO₂ (w/w) fluxes at 10 wt.% to the total feed. Both the remelted and smelted samples were observed to exhibit a relatively equal mass increase when exposed to low P(O₂) up to 0.17 bar (Figure 8). However, a substantial deviation occurred beyond this level, from which remelted samples experienced twice the weight increase compared to smelted samples. This demonstrates the beneficial effect of oxide fluxes in inhibiting oxidation, which was particularly evident under high oxidizing conditions. Notably, one remelted sample exposed to P(O₂) = 0.33 bar underwent complete oxidation while the corresponding smelted sample was only moderately affected.

The results from TGA as shown in Figure 9 demonstrate an overall weight increase as a function of P(O₂), assumed to be correlated to the quantity of oxygen absorbed by the sample to form oxide phases. Notably, oxidation reactions started at earlier timepoints in smelted samples compared to remelted samples. This was likely due to slag forming at lower temperatures in smelted samples, as supported by the oxide phase distributions (wt.%) calculated with FactSage 8.1 (SGTE, SGPS, and FTOxid databases) and shown in Figure 10. According to calculations, slag starts to form at a lower temperature (1243 °C) in a smelted sample in contrast to a remelted sample when both are subjected to atmospheric air condition. On the other hand, the completion of oxidation could be temperature-dependent, as evident from the observed slowing down of weight changes upon cooling.

Finally, a dip in weight change between 460 °C and 500 °C was observed only in smelted samples in Figure 9. This could be explained by the thermal dehydration of residual Ca(OH)₂ from the CaO-SiO₂ flux mixture. The corresponding reaction enthalpy and equilibrium temperature as calculated from FactSage 8.1 using the FactPS database are shown:

Ca(OH)₂(s)→CaO(s) + H₂O(g); ΔH = + 99.46 kJ/mol, T = 524 °C

3.3. Microstructural Analysis of Oxide and Metallic Phases

The oxide phases near the alloy–oxide interface were analyzed, where equilibrium phases that formed at 1600 °C were likely quenched and preserved. The metallic phases in the center of the alloy specimens were also investigated. The phase atomic compositions and stoichiometric formulas were determined by averaging 3 to 5 EDS point analyses per phase.

Table 7. Oxide phases observed on the alloy–oxide interfaces in remelted and smelted samples. Chemical compositions of slags exclude minor phases constituting less than 2 wt.%.

Sample Code	Phase	Stoichiometric Formula/ Chemical Composition	Short Notation
R0	Spinel	Cr2.4(Mn, Al)0.8O4	Spin
	Lanthanum chromium oxide	CrLaO3	CrLa
R1	Spinel	(Cr)2(Fe, Mn, Ni)O4	Spin
	Chromia	Cr2O3	Chro
R2	Spinel	Cr2FeO4	Spin
	Wüstite	FeO	Wus
R3	Spinel	(Cr, Fe)2(Ni, Co, Fe)O4	Spin
	Hematite	Fe1.5(Cr, Ba, La)0.5O3	Hem
S0	Hibonite	(Ca)(Al0.9Cr0.1)12O19	Hib
	Calcium aluminosilicate	Ca2Al2SiO7	CAS
	Anorthite	CaAl2Si2O8	Ano
S1	Spinel	(Cr0.9Al0.1)2(Fe0.7Ni0.10Mn0.20)O4	Spin
	Chromia	Cr1.8Al0.2O3	Chro
	Slag	26CaO-46SiO2-25Al2O3-3FeO (wt.%)	Slag
S2	Spinel	(Cr0.9Al0.1)2(Fe0.8Mn0.20)O4	Spin
	Slag	13CaO-66SiO2-12Al2O3-9FeO (wt.%)	Slag
S3	Spinel	Cr2FeO4	Spin
	Slag	11CaO-61SiO2-11Al2O3-17FeO (wt.%)	Slag

lists the oxide phases and their compositions, and Figure 11 depicts the corresponding BSE images on the alloy–oxide interface.

In the remelted samples, spinel was found to be the predominant oxide phase. Spinel and lanthanum chromium oxides were present at inert conditions (R0, Figure 11a) due to inherent oxides as coatings, which likely dissolved and transformed into equilibrium oxides. In spinel structures, Cr consistently occupied the octahedral sites, while the tetrahedral sites contained substitutional Fe, Mn, and Ni atoms. Breakaway oxidation was observed in R1 (Figure 11b), where fragmented chromia phases were located directly at the alloy–oxide interface, while the spinel phase was approximately 40 µm away from the interface. Conversely, the spinel in R2 was in direct contact with the alloy and embedded within a wüstite matrix (Figure 11c). In R3, where complete oxidation occurred (Figure 11d), the spinel constituted the matrix with precipitated hematite grains.

In the smelted sample under inert conditions (Figure 11e), the observed oxide phases were likely transformation products of the CaO-SiO₂-Al₂O₃ slag system. Under oxidizing conditions, on the other hand (Figure 11f–h), spinel consistently appeared as the main oxide phase embedded in the slag matrix. The slag showed a relatively high amount of aluminum, indicating significant crucible dissolution. The stoichiometric formula of the spinel was similar to that observed in the remelted samples, with the addition of minor aluminum atoms occupying the octahedral sites. Chromium was found to be concentrated in the spinel phase, while the slag contained a very low amount of Cr (less than 2 wt.%) in all studied oxidizing conditions. As the amount of oxidized Fe increased as a function of P(O₂), the amount of dissolved FeO in the slag also increased.

Overall, these observations strongly showed a high concentration of trivalent Cr oxidized in oxide spinel under the studied atmospheric conditions. Furthermore, the spinel phase exhibited high stability, embedded in either a wüstite or slag matrix or as the matrix itself.

Microstructural characteristics of alloy phases in remelted and smelted samples were also analyzed. BSE images of the microstructure of S0 (Figure 12a) demonstrated a two-phase microstructure consisting of a matrix and darker phases along the boundaries. Phase maps derived from EDS mapping of a region shown in Figure 12c are depicted in Figure 12d,e and highlight the distribution of delta-ferrite within the matrix. EBSD analysis revealed that this matrix was primarily composed of martensite with a minor amount of retained austenite, while the darker intergranular phase was confirmed to be delta-ferrite. These characteristics were also observed for sample R0. By contrast, the samples exposed to oxidizing conditions exhibited a predominantly single-phase matrix with a lath-martensitic microstructure, such as what is demonstrated in the BSE image of S1 in Figure 12b.

The ferrite matrix fractions for all samples are summarized in Figure 13. For the purpose of determining the overall alloy composition, the phase fractions between martensite and retained austenite comprising the matrix were not further distinguished since martensite possesses the same composition as the retained austenite. Elemental compositions were calculated from five EDS point analyses per phase, and the results are shown in Figure 14. The results suggested an overall decrease in Cr with increasing P(O₂). However, it showed a steeper decline in remelted samples, from which complete Cr removal occurred at P(O₂) = 0.17 bar. Furthermore, there was a gradual increase in concentration of Fe, likely with Cr depletion until P(O₂) = 0.17 bar was reached, after which it showed a slightly decreasing trend in the case of smelted samples. Finally, Ni concentration had a gradually increasing trend, while cobalt concentration remained relatively constant for both remelted and smelted samples. Manganese concentration is not shown as it is too close to the detection limit of EDS analysis (<0.5 wt.%) to observe any clear trend. This experimental work demonstrated the drastic decrease in Cr content in the alloy, indicating its high susceptibility to oxidation towards the spinel phase under studied P(O₂) conditions. Furthermore, the apparent ease of oxidation of elements was in the order of Cr > Fe > Co > Ni. To confirm this tendency, thermodynamic calculations were performed using FactSage 8.1 software.

3.4. Theoretical Element Distribution between Metallic and Oxide Phases

FactSage 8.1 was utilized to determine the element distribution in the equilibrium phases formed upon smelting of the blended composition described in Table 6 at 1600 °C under increasing oxygen partial pressures. The relative abundance of the most stable phases was calculated based on the Gibbs free energy minimization approach using available thermodynamic databases.

The distribution ratio, L^oxide/metal of an element M during the smelting process, was defined as the ratio of the equilibrium concentration of M between the oxide phases (M_oxide) and solvent metal phase (M_metal). This ratio is represented by the following equation [27]:

$$L^{oxide/metal}=\frac{M_{oxide}\left(\mathrm{wt}.\%\right)}{M_{metal}\left(\mathrm{wt}.\%\right)}$$

Several authors defined the ratio as the partitioning of the element between the metal and slag [27,28,29]. However, the term “oxide” was used in this equation to include all the predicted oxide phases at equilibrium. A positive value indicates preferential partitioning of M to the oxide phases and hence favorable removal from the solvent metal phase. Figure 15 shows the logarithmic distribution ratios of input elements greater than 1 wt.%. Computation predicted that complete oxidization of liquid alloy occurs at P(O₂) = 10⁻⁶ bar, where the graph ends. Notably, the ease of element oxidation is initially in the order Mn > Cr > Fe > Co > Ni from P(O₂) = 10⁻¹⁵ to P(O₂) = 10⁻¹⁰ bar. However, this changes to Cr > Mn > Fe > Co > Ni at higher oxygen partial pressures, which could be attributed to the formation of the cubic spinel from P(O₂) = 10⁻¹⁰ bar that favorably oxidizes and consumes Cr. This order of oxidation confirms the experimental results in Figure 14.

The calculated equilibrium phase fractions and the element distribution under different oxygen partial pressures are shown in Figure 16. Only the main phases with relative abundance exceeding 2% are included in the analysis. Phase formulas are provided in

Table 8. List of resulting equilibrium phases. Sublattice constituents with site fractions below 10⁻³ are excluded in the reference formula.

Phase	Reference Formula
Liquid metal	(Fe, Cr, Ni, Mn, Co)
Silicate	(Ca+2, Mn+2)2(Si+4)(O−2)4
Chromia	(Cr+3)2(O−2)3
Wüstite	(Fe+2)(O−2)
Spineltetragonal	(Cr+3, Mn+2)2(Cr+3)(O−2)4
Spinelcubic	(Cr+3, Fe+2, Fe+3, Co+2, Ni+2)2(Fe+2, Fe+3, Cr+3, Co+2, Ni+2)(O−2)4
Liquid slag	(Fe+2, Fe+3, Cr+2, Cr+3, Ni+2, Co+2, Mn+2, Mn+3, Si+4, Ca+2)x(O−2)y

. Caution was applied in comparing theoretical results with experimental data, as experimental conditions were also influenced by external factors such as mass transfer and reaction kinetics. As an example, the complete oxidation predicted at P(O₂) = 10⁻⁶ bar in equilibrium conditions contradicted the experimental data, wherein the alloy was still present at P(O₂) = 0.33 bar. It is also crucial to note that the P(O₂) values in equilibrium calculations do not directly correspond to those in Table 2. Instead, they are correlated to the oxygen potential experienced by the metallic and oxide phases in direct contact at thermodynamic equilibrium conditions. According to equilibrium calculations, transient phases such as corundum, silicate, wüstite, and tetragonal spinel occur at relatively low oxygen partial pressures ranging from P(O₂) = 10⁻¹¹ to 10⁻⁸ bar (Figure 16a). In contrast, cubic spinel and slag exhibit stability at high oxidizing conditions, with an increasing amount of cubic spinel as P(O₂) rises. X-ray diffraction analysis (Figure 17) confirmed that this spinel in smelted samples exhibited a cubic structure with Fd-3m symmetry. Chromium shows consistent partitioning into the cubic spinel over the slag (Figure 16b), which was also observed experimentally. Iron partially partitions to wüstite at P(O₂) = 10⁻⁸ bar before transitioning to the slag and cubic spinel phases. Manganese initially shows a preferential oxidation within the tetragonal spinel but tends to go to the slag at higher oxygen partial pressures. Finally, Ni and Co exhibit delayed complete oxidation, after which they partition between slag and cubic spinel phases.

Thermodynamic calculations and experimental findings (see Table 7 and Figure 11 and Figure 16) consistently demonstrate the preferential partitioning of Cr to the spinel over the slag at high P(O₂) values. Chromium is a critical component of stainless steels; thereby, strategies for its optimum recovery must be considered in process modeling. Furthermore, potential steel products, such as those specified in Table 5, do not have cobalt as an intentional alloying element. While nickel-containing stainless steels may unavoidably contain Co as a trace element up to 0.3 wt.% [30], this value is below the measured Co concentration in the samples and must be reduced if this recycling pathway is taken. Nevertheless, Co remains a critical element in the production of technology materials, particularly in lithium-ion batteries [31,32]. Therefore, it is crucial to look into Co recovery rather than allowing it to be diluted within the alloy. Cobalt is exclusively present on the MCF oxide coating; hence, pretreatment techniques prior to steelmaking, such as blasting and pickling of end-of-life interconnectors, should be explored.

4. Conclusions

Experimental work was conducted to assess the composition of an end-of-life JÜLICH SOC stack interconnect assembly. Based on extensive characterization studies performed on the sample, the following conclusions were made:

• There was only a minor difference between the stoichiometric formulas of the standard coatings (LCC10 and MCF) and the coatings on the end-of-life interconnector.
• In addition to the coatings, thermally grown oxides Cr₂O₃ and Cr₂MnO₄ were observed on the coating–alloy and alloy–fuel gas boundaries.
• The interconnector was predominantly composed of iron and chromium with less than 4 wt.% of other alloying elements, mainly cobalt and manganese.
• Melting temperatures of the zonal specimens fell within the expected range for Crofer 22 APU alloy substrate, demonstrating negligible influence of the oxide coatings to the alloy’s transition temperatures. The melting point of the nickel mesh was close to that of the pure nickel obtained from calibration, indicating the mesh’s high purity.
• Stainless steels in the 4xx, 2xx, and 3xx AISI series can be potentially produced based on the calculated blended compositions with and without the addition of nickel meshes.

Furthermore, the feasibility of recovering metallic resources from the end-of-life JÜLICH SOC stack interconnect assembly was evaluated. This was performed through small-scale remelting and smelting experiments on samples representative of the interconnect assembly and thermodynamic equilibrium calculations. Based on the results, the following conclusions were made:

• Smelted samples had an overall lower degree of oxidation than remelted samples within the studied P(O₂) conditions (0–0.33 bar), which confirms the practical and beneficial effects of the oxide fluxes in suppressing oxidation.
• There was an overall decrease in chromium as a function of P(O₂) relative to other elements in both remelted and smelted alloys. Iron concentration increased from P(O₂) = 0 to 0.17 bar, beyond which it showed a slightly decreasing trend. Nickel concentration was observed to have a consistently increasing trend, while the cobalt concentration remained constant.
• Chromium preferentially oxidized and partitioned to the cubic spinel over the slag at high P(O₂) values. Chromium is a critical component of stainless steels; thereby, strategies for its optimum recovery must be explored.
• Thermodynamic calculations using FactSage 8.1 predicted an ease of element oxidation, quantified by distribution ratios, in the order of Mn > Cr > Fe > Co > Ni from P(O₂) = 10⁻¹⁵ to P(O₂) = 10⁻¹⁰ bar, but this changes to Cr > Mn > Fe > Co > Ni at higher oxygen partial pressures. This switch could be attributed to the formation of the cubic spinel from P(O₂) = 10⁻¹⁰ bar that favorably consumes chromium.
• Potential steel products do not contain cobalt as an alloying element. Cobalt is nevertheless a key component in energy applications and therefore must be recovered rather than permitted to be lost through alloy dilution.

Future work will involve comparing such characterization results with interconnectors from other end-of-life SOC stack designs and manufacturers. Additionally, a recycling route based on scrap-based steelmaking operations with pretreatment steps for contamination control will be explored to transform interconnector assemblies into steel products that meet existing market demands.