Electrode design for direct-methane micro-tubular solid oxide fuel cell (MT-SOFC)

A R T I C L E I N F O


Introduction
Solid oxide fuel cells (SOFCs) are types of fuel cells generally known for their high operating temperatures ranging from 600 to 1000°C [1][2][3]. In addition to the typical attractive features of fuel cell technology, such as high efficiency and environmentally friendly operation, SOFCs provide some unique features, such as full solid-state operation and flexibility in fuel sources [4,5]. Hydrogen (H 2 ) is the most commonly used fuel for SOFCs, with excellent electrochemical performances frequently reported [6,7]. However, the vast majority of H 2 is produced from hydrocarbons through the steam reforming process, during which 20-30% of fuel value is lost [8]. As a result of this, the concept of direct hydrocarbon utilisation in SOFC has attracted research interest worldwide [9][10][11][12]. By using hydrocarbons, alternative cell materials need to be utilised in order to suppress carbon formation, which is an inherent problem associated with conventional nickel (Ni)cermet anodes [13,14]. In a qualitative index of various SOFC anode materials by Ge et al., Ni-based anodes show superior electrocatalytic activity over Cu-based anodes, however the latter has the advantage of better fuel flexibility and a lower tendency towards coke formation [2]. Gorte  good cell performances with no or trivial coke formation have been demonstrated [8,[14][15][16].
Fabrication is a primary challenge associated with the copper-based anodes. Copper and its oxides have relatively low melting points, viz. 1085°C, 1326°C and 1232°C for Cu, CuO, and Cu 2 O, respectively [17,18]. This makes the preparation of a cell with Cu-based anode through standard high-temperature ceramic processing techniques (i.e. tape casting and sintering) impracticable. Instead, impregnation has been almost exclusively applied to incorporate Cu in an additional step [14,16,19]. Geometrically, a micro-tubular design offers additional benefits compared to flat and tubular designs, including improved thermal shock resistance, quick start-up and shut-down, and enhanced volumetric power density [20,21].
Anode design is a critical aspect of the direct utilisation of hydrocarbons in SOFC operation. In this study, a micro-structured yttriastabilised zirconia (YSZ) micro-tube has been fabricated via a phase inversion-assisted process to develop micro-tubular (MT)-SOFC with a Cu-based anode which is operated using CH 4 as a fuel source. One major benefit of the phase inversion-assisted process is the flexibility in control and tailoring of the micro-structure. The microstructured YSZ scaffold is composed of self-organised micro-channels and a skin layer. The thin skin layer functions as the electrolyte, whereas the microchannels perform as a substrate for the anode materials (Cu-CeO 2 ). These anodic materials are firmly positioned into the micro-channels, thus, deterring their thermal movement. Furthermore, this unique micro-tubular feature helps avoid delamination and defect formation, contributing towards improved structural integrity. After a complete cell was constructed, a performance test was conducted with direct CH 4 utilisation.

Preparation of micro-structured YSZ micro-tubes
The ceramic suspension was prepared by mixing the YSZ powder with the solvent (NMP) and dispersant. The resultant mixture was milled for 60 h (MTI Corporation model SFM-1 Desktop Planetary Ball Miller). PESf was subsequently added to the suspensions and milled for another 60 h to attain a homogeneous mixture. Before spinning, the suspension was degassed by stirring under vacuum for 2 h to eliminate air bubbles. The suspension was transferred to a stainless steel syringe and extruded through a custom-designed spinneret into an external coagulation bath of DI water. The spinneret was immersed in the coagulant bath to eliminate any air gap. The extrusion rates of the suspension and the bore fluid were precisely controlled using two Harvard PHD 22/2000 Hpsi syringe pumps. The micro-tube precursors were left in water overnight to complete the phase inversion before sintering. Table 1 presents the fabrication parameters for the preparation of YSZ micro-tube. Fig. 1 shows the schematic diagram of the phase-inversion based extrusion process. The inset (Fig. 1a) shows the picture of the dualorifice spinneret used in this work. Fig. 1b illustrates the procedure for preparing a complete single cell. After completetion of the YSZ microtubes ( Fig. 1b(i)), a dual-layer cathode consisting of an inner LSM-YSZ layer (LSM/YSZ = 50/50 by weight) and an outer LSM layer was brushpainted onto the YSZ micro-tube and sintered at 1000°C for 2 h (Fig. 1b (ii)). Subsequently, the anode materials (Cu-CeO 2 ), were incorporated into micro-channels of YSZ micro-tubes via vacuum-assisted co-impregnation process. A mixed aqueous solution of copper nitrate and cerium nitrate was prepared prior to impregnation, with a concentration of 4 M and 1 M, respectively. The set-up for the impregnation process is shown in Fig S1 with additional description of the impregnation process. The co-impregnated micro-tube underwent heat treatment at 450°C for an hour to decompose the nitrates. This impregnation was repeated until a target loading of 25 wt % was achieved.

Reactor assembly and sealing
The MT-SOFC was placed into two alumina supporting tubes (Almath Crucibles Ltd., UK) and sealed with a ceramic sealant (Ceramabond 552-VFG, Aremco, USA). The sealant became gas-tight following successive heat treatments at 95°C and 260°C for 2 h at each temperature with a heating rate of 5°C min −1 . This assembly was placed inside a 300 mm long quartz tube with 20 mm outer diameter (OD) (Almath Crucibles Ltd., UK) with stainless steel end-caps made inhouse, and sealed to the quartz tube by Viton O-rings (Polymax, UK). The whole set up is shown in Fig. S2.

Characterisation
The morphology of the micro-tube was examined using scanning electron microscopy (SEM) (JEOL JSM-5610 and LEO Gemini 1525 FEGSEM). Samples were gold-coated under vacuum at 20 mA for 60 s (EMITECH Model K550), and the SEM images with various magnifications were acquired. Energy dispersive spectrometry (EDS, JEOL JSM-6400 electron microscope) analysis was undertaken to evaluate the elemental distribution of anodic materials. The gas-tightness of the sintered micro-tubes was assessed using N 2 permeation method. The N 2 permeance was calculated from the pressure drop over 8 h using the following equation: where where P denotes the N 2 permeance of the tested membrane (mol m −2 s −2 Pa −1 ); V is the volume of the test vessel (m 3 ); R is the gas   constant (8.314 J mol −1 K −1 ); T is the measured temperature (K); and p 0 , p a , and p t indicate the initial, atmospheric and final pressure readings (Pa), respectively. A m is the membrane area (m 2 ); t is the measurement time (s); L is the length of the micro-tube; and R o and R in represent the outer and inner radiuses of the tube, respectively. The mechanical strength was examined through a three-point bending method using a tensile tester (Instron Model 5544) with a load cell of 1 kN. The specimen was placed onto two sample holders with a gap of 30 mm. The bending strength (σ F ) was calculated from the obtained fracture force using the following equation: where F represents the measured fracture force (N); L, D o and D i denote the length (m), the outer and inner diameters of the micro-tube (m), respectively. The average porosity of the micro-tube was studied using pycnometer (Micromeritics Accupyc II 1340). The porosity (ε V ) was calculated using the following equations: Where ρ pyc represents the skeleton density (g cm −3 ) measured by pycnometer, m, l, D o and D i denote the mass, length, outer and inner diameters of the sample (cm), respectively. In addition, an assumption has been made whereby each sample has a uniform structure with identical dimension all through the micro-tube. For a single cell, the current was collected from the cathode by wrapping silver wires on cathode surface with additional silver paste to improve the contact. For current collection at the anode, silver wires were passed through the lumen, with additional silver wool and silver paste to enhance contact between the anode surface and silver wires. Both wires from anode and cathode were connected to a potentiostat/ galvanostat (Iviumstat, Netherlands) for current-voltage (I-V) measurement and impedance analysis. The cell current-voltage testing was conducted at 650-750°C using 30 ml min −1 of fuel (H 2 or CH 4 ) and 50 ml min −1 of air as the oxidant. Electrochemical impedance spectroscopy (EIS) was taken at open circuit voltage (OCV) in the frequency range of 10 5 -0.01 Hz with a signal amplitude of 10 mV. Fig. 2 shows SEM images in which YSZ micro-tube exhibits a highly asymmetric structure consisting of a dense and thin skin layer at the outer section, and organised micro-channels growing through the entire cross-section. It has been suggested in previous studies that Rayleigh-Taylor instability theory can be well used to mimic the formation procedure of micro-channels [22]. When two fluids are in contact, the low-density fluid (external coagulant) tends to push into the highdensity fluid (ceramic suspension) due to interfacial instability/perturbation, during which solvent/non-solvent exchange occurs. Generally, a sponge-like region is formed when the phase inversion goes towards completion. However, a mixture containing 60/40 wt% of NMP/ethanol was applied as the bore fluid to delay the precipitation of polymer phase, which subsequently enables micro-channels originating from the external surface to penetrate through the whole cross section, forming open entrances to the inner surface [23]. However, one undesirable characteristic of micro-tube with large micro-channels is a limited sponge-like region, which is the main contributor to the triple phase boundary (TPB) for electrochemical reactions.

Micro-structures of micro-tubes
The sintered micro-tubes have the average outer and inner diameter of 1.44 ± 0.01 and 0.96 ± 0.01 mm, respectively. Micro-channels with open entrances have been described to be effective in facilitating the gas transport. If the pore radius is greater than 1 μm, the concentration polarisation resistance through the pores may be considered negligible by facilitating the diffusion of gaseous fuels, as well as the exhaust gases [24,25]. It can be estimated from the SEM image that the pore entrances are approximately 30-40 μm (Fig. 2e). This is advantageous for SOFC operation as easier fuel transport to the TPB area would lead to better electrochemical performance due to a lower concentration polarisation. The dense YSZ skin layer had a thickness of approximately 10 μm (Fig. 2d), which could well serve as the electrolyte, while the micro-channels and porous structure are appropriate for the deposition of anode materials (Cu and CeO 2 ).

Gas-tightness and mechanical property
Adequate gas-tightness of electrolyte is critical for an SOFC, since the electrolyte must prevent direct interaction between the fuel and the oxidant, as well as transport oxygen ions from the cathode to the anode. This property was characterised via nitrogen (N 2 ) permeation measurement. Membranes can be considered as gas-tight if the N 2 permeance is of the magnitude 10 −10 mol m −2 s −1 Pa −1 [26]. In this study, the average permeance of the prepared micro-tubes was 5.3 × 10 −10 mol m −2 s −1 Pa −1 , indicating that the sintered YSZ microtube is appropriate for development into a complete cell.
Additionally, sufficient mechanical strength is essential to cell lifespan. A three-point bending test was used to examine the mechanical strength of the micro-tubes. The average bending strength of the green YSZ micro-tubes was 264 MPa, which is significantly higher than microtubes with similar structure reported elsewhere [27]. Earlier studies proposed that anode-supported micro-tubes with a bending strength of about 178 MPa are appropriate for development into a complete single cell [28,29], indicating that our micro-tube is suitable for the fabrication of the MT-SOFC.

Vacuum-assisted co-impregnation process
The incorporation of the anode material was conducted through the vacuum-assisted co-impregnation process. A number of impregnation cycles are required to ensure sufficient anode materials to form a continuous phase. Fig. S4 shows the relationship between the anode loading and the impregnation cycles. The anode loading is 28 wt% of the full cell after 15 cycles of impregnation (65:35 wt% for CuO:CeO 2 ). The loading of anode materials increased almost linearly with the number of cycles, suggesting appropriate controllability of this impregnation process.
The porosity of the YSZ scaffold was measured to be 43%, and the porosity of the micro-tube after 15 cycles of the co-impregnation process was 36%. The porosity decreases after the impregnation as the anode materials occupied the void of the YSZ scaffold. Both values for the porosity are within the required optimal anode porosity range, between 30% and 40% [30].
It has been previously reported that CuO could agglomerate and thus form isolated, spherical clusters [31]. This phenomenon can be linked to the low thermal stability of Cu as Cu and its oxides have considerably low melting temperatures. Thus, the co-impregnation approach has been adopted and no metallic phase agglomeration or blockage of the channel entrance was observed, as depicted in Fig. 3b. In comparison to Cu and its oxides, CeO 2 has better thermal stability due to its higher melting temperature of 2602°C [2]. Therefore, the presence of CeO 2 may assist in scattering Cu throughout the entire area, as well as improving the thermal stability of the anode composite.
It is noteworthy that one of the main advantages of using a composite Cu-CeO 2 anode is that both materials do not form a solid solution, with each species potentially retaining their specific functionality. In this study, Cu is required for electronic conductivity whereas CeO 2 functions primarily as a catalyst for fuel oxidation. One critical property of an anode is a continuous electronic conducting phase to ensure smooth electron transport, which lowers ohmic losses. SEM image (Fig. 3b) shows the homogeneous distribution of both particles onto the inner surface of the micro-tube, which could also be shown by elemental mapping. Both materials are uniformly dispersed over the entire region of the inner surface and into the micro-channels ( Fig. 3a and b). The vacuum-assisted co-impregnation process applied in this work has demonstrated good Cu continuity, which is crucial for fuel cell performance. Fig. 4 shows the chemical mapping of the co-impregnated microtube. As can be seen, both materials are uniformly dispersed over the entire region of the inner surface and into the micro-channels. In addition, the higher resolution of the SEM images (Fig. 3d) of the impregnated micro-tubes depict the flake structure of the Cu-CeO 2 . The thickness of the anode catalyst layer is approximately 5 μm. This particular structure could lead to an increase in surface area, which contributes to a larger TPB at which the electrochemical reaction takes place.

Electrochemical performance test
Electrochemical performance tests were conducted at 650, 700, and 750°C using a ramping of 5°C min −1 with 30 ml min −1 of dry H 2 or CH 4 as fuel, and 50 ml min −1 of air as the oxidant. The co-impregnated cell (15 cycles) was used to conduct this study. Three cells were tested with a standard deviation less than 10%. The estimated active area was calculated to be approximately 0.45 cm 2 . The open circuit voltage (OCV) can reach up to 1.19 V for H 2 at 650 and 700°C and 0.95 V for CH 4 , respectively, indicating a proper gas-tightness of the YSZ electrolyte. Note that the OCV for a cell operating under CH 4 is slightly lower than the predicted value determined from Nernst potential equation (∼1.04 V) [32]. This discrepancy for the CH 4 system could be due to the presence non-electrochemical surface reactions [33]. Nevertheless, both OCV values for systems operated with either H 2 or CH 4 as reported here agreed well with other studies [8,18,[34][35][36]. Fig. 5a shows that a reasonably high power density of up to 0.55 W cm −2 was obtained from H 2 fuel at 750°C. This cell shows a substantial improvement over other SOFCs with similar Cu-CeO 2 -YSZ composite anodes [18,37], which may be attributed to the unique hierarchical structure of the micro-tubes. The presence of micro- channels with open entrances should help facilitate gas transport to the reaction sites. The measured peak power density is comparable to previously reported cells with Ni-based anodes [27], suggesting that the Cu-CeO 2 composite is a promising replacement of Ni-cermet [3,38,39].
In comparison, the CH 4 -fuelled system demonstrated lower power density than that of H 2 (Fig. 5b). This finding could be related to the catalytic limitation, where it is believed that electrochemical oxidation of CH 4 is slower than H 2 . Earlier work has shown that CH 4 is less reactive than H 2 in heterogeneous oxidation [40] and has lower reactivity at the anode [16]. Another reason for lower performance in the CH 4 system could be linked to the rate-limiting electrochemical steps at the anode. Direct oxidation of CH 4 in one step seems to be difficult [41], and it has been proposed that several parallel reactions occur when using hydrocarbons as the fuel [42]. This means that CH 4 can either be partially oxidised by the oxide ions to carbon monoxide (CO) and hydrogen (H 2 ) (Equation (6)) or fully oxidised to carbon dioxide (CO 2 ) and water (Equation (7)). In addition, there is a possibility of gradual internal steam and dry reforming (Equations (8) and (9), respectively) taking place within the system.
However, a system fuelled by methane is also susceptible to carbon formation, according to Equations (10) and (11).
Equations (6)- (9) show that there are several reacting species at the anode (CH 4 , H 2 and CO), contributing to the more complex anodic reactions as compared to the H 2 -fuelled system. Furthermore, issues surrounding the transport limitations could also be related to the larger molecular mass of CH 4 , which causes slower diffusion and may result in greater concentration polarisation. This could be minimise by having internal reforming. The addition of reforming agents would reform the CH 4 into H 2 and CO which are relatively easy to undergo electrochemical reaction. However, in this study, we aim to operate the system directly with dry CH 4 , thus no additional reforming agents were added.
The difference in the electrochemical performance between these fuels is reduced when the cell is operated at higher temperatures. The MT-SOFCs fuelled by CH 4 exhibited peak power densities of approximately 0.05, 0.09 and 0.16 W cm −2 at 650, 700, and 750°C,  respectively. Operation using dry CH 4 at 750°C demonstrated a power density of 0.16 W cm −2 which, to the best of our knowledge, is the highest compared to previously reported values for similar anode materials (Cu-CeO 2 ), as shown in Table 2. Fig. 6 shows the impedance analysis for the cell with a signal amplitude of 10 mV at the frequency range from 10 5 to 0.01 Hz. The high-frequency intercept on the x-axis indicates the ohmic resistance of the cell (R Ω ), comprising the ionic resistance in the electrolyte, both ionic and electronic resistance in the electrodes, and the contact resistance at the interface and current collectors. The first semi-circle (high-frequency arc) corresponds to the activation polarisation (η a ), which is related to TPB area and the number of reactive sites, whereas the second semi-circle (low-frequency arc) is associated with the concentration polarisation (η c ), comprising the mass transport resistance through electrodes and interfaces. When combined, these three parts give the overall cell resistance (η), as indicated by the intercept at lowfrequency.
It can be seen that total cell resistance (η) decreased when the operating temperature increases for both fuels, with this decrease being more apparent for the CH 4 -fuelled system. This is consistent with the observation that incremental changes in power density can be caused by temperature. There is only a small difference in ohmic polarisation for cells operated at different temperatures, measured at 0.3-0.35 and 0.3-0.5 Ω cm 2 under H 2 and CH 4 operation, respectively. In comparison, it can be observed that for CH 4 , concentration polarisation dominates total cell resistance, with this value being much higher than H 2 . This is in agreement with the trend of the power density, where the lower performance of CH 4 -fuelled systems could be due to more complex reactions taking place. A higher concentration polarisation for the CH 4 -fuelled system might be associated with the rate-determining step of the anode reaction [41]. Direct oxidation of CH 4 in a single step, involving an eight-electron transfer, appears to be challenging. Therefore, it is anticipated that the CH 4 oxidation reaction involves various paths including the transformation to H 2 and CO (in the case of partial oxidation) and the complete oxidation to form water and CO 2 . These reaction products must be transported away from the electrolyte/anode interface and through the porous anode to the fuel stream. Such diffusion of various reactants and products (i.e. CH 4 , H 2 , CO, CO 2 , and H 2 O) from the reaction site could contribute to the increase in concentration polarisation. Meanwhile, the greater activation polarisation of the CH 4 -fuelled system may be attributed to the catalytic reaction constraint.
The resistance against carbon formation of this cell was also investigated. A stability test has been conducted under constant operation at 750°C with dry CH 4 at 0.7 V. As can be seen from Fig. 7, the current density of the single cell remained relatively stable at 0.14 A cm −2 for 30 h with less than 1% of degradation.
Post-test analysis using SEM-EDS was performed to inspect the carbon deposition at the anode. Two different regions of the anode surface were analysed. From Fig. 8, the chemical mapping shows the distribution of carbon (C) element on the anode surface whereby only trivial amount could be detected. The spectrum also indicates a minor contribution corresponding to carbon element. This is in agreement with the stable electrochemical performance since rapid performance degradation will be observed if there is severe carbon formation.

Conclusions
In summary, a unique YSZ scaffold has been developed via a phaseinversion and sintering process, which is composed of dense outer skin layer acting as the electrolyte and a series of micro-channels for the infiltration of anode materials. This distinctive structure allows the use of Cu, which has a lower melting point than the required temperature for the YSZ micro-tube sintering process, to be incorporated as the anode, whereby this process is extremely difficult via conventional fabrication routes. Such micro-structural design also provides well distributed gas flow passages and helps to reduce transport resistance. Outstanding electrochemical performances have been obtained from both H 2 and CH 4 operation. Electrochemical testing showed that the system fuelled by dry CH 4 yields a power density of 0.16 W cm −2 , which is one of the highest ever reported for Cu-CeO 2 anode (nickel free). It has been found that only minor carbon formation could be detected and negligible degradation has been observed during longterm test. This indicates that our cell developed from the micro-structured micro-tube not only contributes to a better cell structure but is also suitable for direct CH 4 utilisation.