High-power density monolithic fuel cell stack

The transportation sector is currently undergoing a technology shift from internal combustion engines to electric motors powered by batteries. However, their limited range and long charging times limit widespread adoption. Electried transportation powered by solid oxide fuel cells (SOFCs) offer an attractive alternative especially for heavy freight and long-range transportation, as this technology can provide high-eciency and exible fuel choices. Thus far, the technology is mostly used for stationary applications owing to the high operating temperature, low volumetric and gravimetric power density, and poor robustness towards thermal cycling and mechanical vibrations of conventional ceramic-based cells. Here, we present a novel metal-based monolithic fuel cell design to overcome these issues. Highly cost-competitive and scalable manufacturing methods are employed for fabrication, and only a single heat treatment is required, as opposed to two or three for conventional SOFCs. The design is further optimised through three-dimensional multiphysics modelling, nanoparticle inltration, and corrosion-mitigating treatments. The monolithic fuel cell shows exceptionally high power density (5.6 kW/L) revealing the vast potential of SOFC technology for transport applications.

such as in long-haul trucks, trains, or ships. The advantages is that they can be operated using less pure hydrogen, ammonia, methane, and other liquid fuels.
Although the rst usage of SOFCs as range extenders in vehicles was demonstrated during the last decade 15,16 , the technology was considered less suitable for vehicles because of the high operating temperature, limited thermal cycling robustness of the ceramic-based cells 17 , and low power density (typical commercial SOFCs currently operate in the range of 0.1-1.0 kW/L) 15,18 .
This study presents a novel concept for fabricating a metal-based monolithic, high-temperature fuel cell stack with exceptionally high power density (5.6 kW/L) using cost-competitive and scalable manufacturing methods.
The concept of the metal-based monolithic stack is illustrated in Figure 1 and compared to the conventional stacking of ceramic anode-supported fuel cell stacks with metallic interconnects. In a conventional stack (both metal and ceramic supported), the cell support, gas channels, and interconnects constitute the thickest layers (few hundred micrometers), while the active part of the cell is in the range of few tens of micrometers. The monolith concept integrates the cell support, gas channels, and the interconnects into a single layer, thereby reducing the stack height by a factor of 2-4. This considerably increases the volumetric and gravimetric power density of the stack, which are important parameters considering the space and weight constraints in vehicles. More speci cally, the power per volume (and per weight) is signi cantly improved, and values of 6-8 kW/L are feasible.
In addition to the advantage of high power density, the metal-based monolith stack is signi cantly cheaper than a conventional stack. The volume reduction alone reduces the materials costs, and further reduction is achieved by using cost-effective metals 19 (Fe is much cheaper than Ni and zirconia (ZrO 2 ) used in the conventional stack) for the thickest component of the monolith. The high metal content also reduces temperature differences across the stack during operation and provides greater robustness to fast thermal cycles. This is important for hybrid automotive applications integrating SOFCs and electrical battery systems 20 .
Previous work on monolithic SOFCs designs were reported by Saint-Gobain, which developed fully ceramic monolithic SOFC stacks 21 , and by Argonne National Laboratory, which designed metal-supported SOFC stacks ("TuffCell") 22 . However, our concept of metal-based monolithic SOFCs comprises several signi cant differences when compared to the one developed by Argonne National Laboratory, such as the processing techniques to integrate gas ow channels, the electrode materials and thicknesses, and the concept of integrated seals. Thus, our single repeating unit (SRU) monoliths are signi cantly thinner (> 2 times thinner) and integrate cell components, interconnects, gas distribution channels and seals in a single heat treatment step while the "TuffCell" incorporates seal and cathode only after co-sintering.
The following sections present the underlying concepts for monolith design, manufacturing methods, achieved module, and electrochemical performance of the SRU monolith.

Concept For Monolith Stack Design
The schematic illustration of the monolith stack is shown in Figure 1. The electrolyte, composed of scandia-doped yttria-stabilised zirconia (ScYSZ, 10 µm), is sandwiched between two porous electrodes (ScYSZ-Fe22Cr composite, 15 µm) with dense sealing strips (ScYSZ-Fe22Cr, 15 µm) along the sides. These are placed between thicker layers corresponding to gas distribution channels (250 µm) and metallic interconnects (dense Fe22Cr, 150 µm). The SRU monolith is co-sintered in one ring step, which results in more robust interfaces than those in conventional stack designs and reduces the total number of ring steps in the manufacturing process. The integrated seals placed between the metallic interconnects and the ceramic electrolyte are laterally joined to the cermet electrodes to ensure gas tightness and improved thermocycling stability. To avoid thermal expansion coe cient mismatches with the surrounding components, while maintaining the electronic insulating property required for SOFC applications, the seals are made of a 5:1 volume ratio of ceramic to metal.
The main design advantage of the monolith is the integration of gas ow channels into a co-sintered block containing both a metallic substrate and ceramic cell interlayers. This results in extremely high power density and improved in-plane thermal distribution. The gas channels are placed between the interconnects and the electrodes to e ciently feed gases to the electrodes and minimise the volume of the cell. They are formed via honeycomb pore-former tapes containing exclusively organic materials that burn off during the ring step.
The gas channel dimensions (250 × 500 µm 2 ) are studied via a 3D multiphysics model built in COMSOL Multiphysics and optimised to obtain a tolerable pressure drop. In this model, the overall conservation laws are solved on a homogenised volume, and the microstructural details are introduced through effective material parameters. The model includes current, heat, and mass transport, including gas ows in channels, and nally, an assessment of the mechanical stresses in the structure.
An example of the model simulation is shown in Supplementary Figure 1. Here, the calculated relation between the gas overblow, pressure drop, and inlet gas temperature, which ensures that the stack temperature never exceeds 700 °C, is reproduced when the stack is operated at 0.67 A/cm 2 . The calculations con rm that cooling of the monolith stack is feasible with achievable gas channel dimensions (250 × 500 µm 2 ) and a maximum pressure drop of 300 mbar. Note that the catalyst in ltration was not accounted for in the simulation of the pressure drops as in ltration primarily in uences the diffusion in the dense microstructure, but not the convection in the channels. This slight overpressure of air ow could also be bene cial to handle the heat production arising from the speci c compactness of the monolith stack design, especially once integrated in a SOFC system. This challenge may also be solved by segmenting the monolith unit, with separating layers allowing for e cient heat transport, or designing operation strategies to optimize the heat production.

Manufacturing
The monolith is fabricated using tape casting, lamination, co-sintering, and catalyst in ltration, which are well-known methods for the manufacture of high-temperature fuel cells 23,24 .
The manufacturing of monolithic stacks is challenging as it requires the sintering of various ceramic and metallic particles while maintaining a hierarchical microstructure for the main gas ows and gas diffusion paths. One of the most complex steps is the binder burnout of these large and compact monolithic cells. This challenge may limit the number of repeating units in a monolith stack and require segmenting the assembly.
To identify the process parameters required to achieve an optimized monolith (avoiding disintegration and warpage during debinding), a 3D model simulating the debinding process was developed using COMSOL Multiphysics. The model describes the pressure inside the hierarchical microstructure emerging as the organics are burned. Kinetic models for the combustion of organics were developed and calibrated by tting the measured weight change in thermo gravimetric analysis (TGA) to the simulated ones. These are combined with a description of convection in the gradually changing microstructure as the organics burn. The mass transport occurs in the hierarchical microstructure with features ranging from micrometres to several centimetres. To effectively describe this, a multi-scale model was employed. Here, the diffusion of decomposed organics in the porous microstructures are described by Fick's diffusion, whereas the ow in the channels being formed is described by the Darcy-Weissbach equation, which can be simpli ed to Darcy's law, when the structure is homogenously distributed in the considered domain 25,26 . The latter approach allows for describing both the diffusion and the convection at different length scales in the same computational homogeneous media 25,26 . The heat transport is both through convection and conduction. The reaction kinetics are described through a set of Arrhenius expressions and add to the mass balances in the diffusion equation.
The model was used to identify the optimal heating ramps and the composition of the organics to minimise the overpressure inside the microstructure. For example, the calculations indicate that using a graphite/poly(methyl methacrylate) (PMMA) mixture in the sacri cial "pore-former" tapes is bene cial compared to using only PMMA, as the decomposition is spread over a larger temperature span. Figure 2 shows SRU monoliths manufactured with different pore-forming materials (a. 100 % graphite, b. 50-50 % graphite-PMMA, and c. 100 % PMMA) as sacri cial agents to form gas channels. The model describes how the burning of the different pore-forming agents increases the pressure generated in the monolith while using xed gas channel dimensions and heating pro le. Photographs of post-sintered SRU monoliths manufactured using the different pore-former agents to form gas channels are presented below the graphs. The monolith made from PMMA exhibited cracks. This is in good accordance with the model, which predicted that the use of 100 % PMMA would lead to the highest pressure (14 mbar around 200 °C) among the three pore-forming agents investigated.
Another di cult task is the co-sintering of the SRU monolith. The sintering shrinkage of the different layers must be matched to ensure that the cell structure withstands unavoidable mechanical stresses and prevents cracking of the monolith. Hence, the shrinkage of each single layer was investigated by dilatometry and adjusted by optimising the particle size distribution, composition, shape, and ratio of raw materials and organic additives. Complex casting techniques, such as co-casting and side-by-side casting, were used to improve the adhesion of the layers. The composition of the FeCr/ScYSZ electrodes and seals was tuned to match the macroscopic thermal expansion of these layers with that of the metal support and the electrolyte (see Supplementary Figure 2). All these process optimisations were required to obtain at and crack-free monoliths as shown in Figures 1 and 3.

Fuel Cell Functionalisation
The electrochemical activity was introduced by in ltrating various suitable nitrate solutions, that is Gddoped ceria without (CGO) and with nickel (Ni-CGO), and lanthanum strontium cobalt (LSC), on the fuel and oxygen sides, respectively. Stoichiometric amounts of the required nitrates were mixed with deionised water and surfactant (Pluronic P123) to obtain a concentration of 1-2 M. Because the electrodes in the monolithic design do not have exposed surfaces, the nitrates must be injected through the gas channels and spread to the porous electrode structure. This was accomplished using a clamping device (see Supplementary Figure 3) that can hold the individual monolith (one SRU or an entire stack) in place, while a slight pressure pushes nitrates through the open gas channels. Subsequently, the monolith was heated to 325 °C 27,28 at 2 °C/min to evaporate water, decompose nitrate salts, and form the desired phase (see Supplementary Figure 4 for thermogravimetric analysis). The procedure was repeated to achieve su cient coverage; this ensures electrochemical activity and corrosion protection. The EDS mapping image with elemental distribution analysis of a typical coverage after three in ltration cycles is shown in Supplementary Figure 5. The weight gain of the SRU monolith increased after each in ltration cycle and plateaued after approximately 10 cycles (see Supplementary Figure 6).

Corrosion Stability
In addition to providing electrochemical activity, the second important function of the CGO layer is to provide corrosion protection of the metallic phase in the electrode backbone, gas distribution layer, and the interconnect. On the fuel side, the most corrosive location is near the gas outlet (highest water vapour pressure). To increase the overall system e ciency, high fuel utilisation is required; outlet gas compositions containing less than 20 % H 2 and more than 80 % steam are relevant operating conditions, as reported previously 29 . The corrosion rate on the air side is lower than that in steam, considering the relatively low target operating temperature (650 °C) 30-32 . Nevertheless, a protective layer is needed to prevent evaporation of Cr from the metal as it can poison the air electrode. It remains to be tested whether the CGO layer can provide su cient protection against Cr evaporation. As far as we are aware, the only measurement of Cr evaporation from steel with a Ce-based coating was performed by Grolig et al. 33 . The study shows that a 10 nm thin Ce layer does not reduce Cr evaporation from AISI 441 steel . However, this result is not directly comparable to the coating applied in our work, which is thicker and contains Gd in addition to Ce. In case the CGO layer is found to be insu ciently protective, there are other well-proven coatings that can be applied on the air side instead, for example MnCo 2 O 4 34 . matching scale). This temperature was chosen as it corresponds to a critical temperature in terms of break-away oxidation in fuel side atmospheres with a high H 2 O content. At higher temperatures, Cr 2 O 3 is more easily formed, protecting the steel against Fe-oxide formation 35 . From the images of the non-coated monolith, it is evident that corrosion protection is most critical at the fuel side of the monolith. After oxidation in air, the metal is covered by a thin Cr-rich layer (most likely Cr 2 O 3 ), while after oxidation in the fuel side atmosphere, a large amount of non-protective iron oxide is formed. By in ltrating the monolith once with CGO and pre-oxidising the monolith at 850 °C in H 2 -N 2 to densify the layer, oxidation was strongly mitigated. However, some areas of the electrode backbone exposed to the fuel atmosphere showed signs of iron oxide formation. This is likely due to insu cient coverage by the CGO layer. The coverage improved after three rounds of in ltration, and no iron oxide formation was observed (see Supplementary Figure 8). Pre-oxidation alone had a minor role in reducing the detrimental oxidation (see Supplementary Figure 8). Due to the small particle size (5-10 μm) of the steel in the electrode layers, there is a limited reservoir of Cr available to maintain the Cr 2 O 3 scale 36 . Although the CGO layer provided su cient protection to prevent Cr depletion within 100 h, tests of longer duration (1000s of h) are needed to con rm that the coating su ciently slows down the oxidation rate to allow the use of such small metal particles.

Electrochemical Performance
A special test xture was developed to test the SRU monolith, as described in the Supplementary Figure 9. Under the given test conditions, this corresponds to ~1.75 % of the O 2 at the oxygen electrode leaking to the fuel electrode side, assuming that the lower OCV compared to the theoretical Nernst potential is solely due to leakage from the oxygen side of the cell. From a fundamental diffusivity point of view hydrogen crossover will be more likely. However, based on the observations made on the tested cells (i.e. the presence of pin holes with clear corrosion problems) and the fact that the OCV is lower than the theoretical Nernst potential, it is more likely an oxidizing gas that crosses over and not the hydrogen. Note that, prior to characterisation, the SRU monoliths were leaktested using a set-up developed in-house (see Supplementary Figure 10), thus ensuring that the interconnect layers are su ciently dense. The leak could possibly have originated from the test xture and not the monolith itself. Nonetheless, the measured OCV is considered acceptable for a cell in the early stage of development. However, electrolyte gas tightness and sealing need to be further improved.
The tested SRU monolith had an active cell area of ~18 cm 2 ; based on the i-V curve depicted in Figure   4.a, this can be translated to an area-speci c-resistance of 0.49 Ωcm 2 for data from OCV to 0.6 A/cm 2 .
Note that at a current density of ~0.67 A/cm 2 , the course of the i-V curve suggests that mass transport and/or fuel starvation negatively affect the performance of the SRU monolith. Assuming a cross-over leak of 1.75 % of oxygen to the fuel electrode, the H 2 utilisation was 24% at 0.67 A/cm 2 indicating scope for improvement with regard to the H 2 feed and distribution to the active sites.  Figure 3.b compare the height of the SRU monolith design developed in this work with those of common commercially available stacks, spanning from 1.4 to 4 mm. The SRU monolith achieved an impressive value of 5.6 kW/L, on a par with the best performing SoA Ni/YSZ anode supported ceramic SOFCs which have been optimised over decades.
To conclude, a novel metal-based monolithic SOFC was fabricated using cost-effective and scalable manufacturing techniques. Initial electrochemical tests revealed an exceptionally high power density (5.6 kW/L), which indicates their immense potential for mobile applications. We expect that this design, or similar, will enable large-scale production of electri ed transportation powered by fuel cells with considerably increased range, decreased charging times, and lower cost. Furthermore, it is believed that these ndings will create renewed interest for the development of other metal-based monolithic designs for SOFCs, solid oxide electrolysers, and other electrochemical devices. The seal/electrode/seal structure was fabricated rst via side-by-side casting using a doctor blade trough, separated in compartments where the suspensions could be poured without mixing. Then, a pore-former tape previously cast and laser-cut into a honeycomb shape was laminated onto the electrodes using heated rolls in a double-roll set up. The interconnects were co-cast on top of the laminated structure. The complete SRU monolith was obtained by laminating a ScYSZ electrolyte between two blocks of interconnect/pore former/seal-electrode-seal rotated by 90°. Germany) in air atmosphere using green tapes rolled into a cylinder (5 mm diameter, 10 mm length). To study the sintering step, the measurements were carried out in TOM_metal (Fraunhofer-Center for High Temperature Materials and Design, Germany) under a reducing atmosphere using debinded rolled tapes.
The fractured and polished cross-sectional microstructures of the as-prepared and the tested monoliths were investigated by scanning electron microscopy (SEM) using a Zeiss Merlin scanning electron microscope equipped with a eld emission gun and a Hitachi TM3000 equipped with a Bruker energy dispersive X-ray spectroscopy (EDS) system.

Electrochemical performance tests
The single-cell monoliths were tested in a commercially available FuelCon Evaluator test rig. A metallic test house was constructed for these cells. The metallic test house enables fast heating and cooling. The monolith was placed in the metallic test house using mica sheets for external sealing. Gas tightness was achieved by joining the two parts of the cell house with screws. The voltage and the current probes were placed equidistantly from the sides of the cell. The fuel inlet/outlet was located on the long side of the monolith, while the air inlet/outlet was placed perpendicularly on the short side to ensure cross ow, as shown in the design of the monolith (see Supplementary Figure 9).

Data availability
The data that support the ndings of this study are available on demand from the corresponding.

Code availability
The codes that support the ndings of this study are available on demand from the corresponding author. Metal-based monolithic stack design. a, Illustrations of conventional and monolith fuel cell stacks with ve repeat units and a SRU monolith stack. The numbers 1, 2, and 3 correspond to the interconnects, fuel cells (electrolyte and electrodes), and cell supports, respectively. b, Photograph of three modules of SRU monolith. c, Illustration of an exploded view of a SRU monolith. The numbers 4, 5, 6, and 7 correspond to interconnects and gas distribution channels (400 µm thick including 250 µm high gas channels), electrodes (10 µm), electrolyte (10 µm), and sealing (10 µm), respectively. d, Enlarged cross-sectional view of a SRU monolith (for clarity, the scale of the layers thickness is not respected).

Figure 2
Effect of pore-forming material on built-up pressure during the debinding step. A multiphysics model was used to simulate the debinding step of the SRU monoliths manufactured with different pore-forming materials (a. 100 % graphite, b. 50-50 % graphite-PMMA, and c. 100 % PMMA) as sacri cial agent to form gas channels. The model describes how the burning of the different pore-forming agents increases the pressure generated in the monolith while using xed gas channel dimensions and heating pro le.
Photographs of post-sintered SRU monoliths manufactured using the different pore-former agents to form gas channels are presented below the graphs.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. RevisedSupplementaryInformationunmarked.docx