Enhanced methane steam reforming activity and electrochemical performance of Ni0.9Fe0.1-supported solid oxide fuel cells with infiltrated Ni-TiO2 particles

Ni0.9Fe0.1 alloy-supported solid oxide fuel cells with NiTiO3 (NTO) infiltrated into the cell support from 0 to 4 wt.% are prepared and investigated for CH4 steam reforming activity and electrochemical performance. The infiltrated NiTiO3 is reduced to TiO2-supported Ni particles in H2 at 650 °C. The reforming activity of the Ni0.9Fe0.1-support is increased by the presence of the TiO2-supported Ni particles; 3 wt.% is the optimal value of the added NTO, corresponding to the highest reforming activity, resistance to carbon deposition and electrochemical performance of the cell. Fueled wet CH4 at 100 mL min−1, the cell with 3 wt.% of NTO demonstrates a peak power density of 1.20 W cm−2 and a high limiting current density of 2.83 A cm−2 at 650 °C. It performs steadily for 96 h at 0.4 A cm−2 without the presence of deposited carbon in the Ni0.9Fe0.1-support and functional anode. Five polarization processes are identified by deconvoluting and data-fitting the electrochemical impedance spectra of the cells under the testing conditions; and the addition of TiO2-supported Ni particles into the Ni0.9Fe0.1-support reduces the polarization resistance of the processes ascribed to CH4 steam reforming and gas diffusion in the Ni0.9Fe0.1-support and functional anode.

On-cell methane (CH 4 ) reforming in Ni-based anodes is an attractive option for directly using CH 4 -based fuels for solid oxide fuel cells (SOFCs) with high fuel efficiency and simplified system design 1,2 . CH 4 steam reforming is a catalytic process for commercial production of H 2 or syngas at a H 2 :CO molar ratio of 3:1 according to the endothermic reaction of Excessive addition of H 2 O will further converts CO to CO 2 by the slightly exothermic water gas shift (WGS) reaction [3][4][5] . If these reactions are taking place in the anode of an SOFC, H 2 is consumed via electrochemical oxidation to generate electrical power 6,7 , forming by-product of H 2 O. Such in-situ formed H 2 O is simultaneously used for CH 4 steam reforming, which reduces the amount of externally added H 2 O to improve the electrical efficiency of the SOFC system. However, for on-cell CH 4 reforming in Ni-based anodes, coking is frequently observed in the anode when steam/carbon (H 2 O/CH 4 ) ratio is low, since Ni catalyzes CH 4 decomposition that produces deposited carbon in the form of filament or particle via either CH 4 cracking or the Boudouard reactions as follow 2 The soot-like carbon particles are distributed on the surface of Ni particles, occupying the active sites for electrochemical reaction and the pores for fuel gas transport 8 ; and the carbon filaments formed by carbon diffusion into/precipitation out the Ni particles 9 disintegrate the Ni-cermet anode by lifting out the Ni particles from the anode (dusting).
It has been demonstrated that infiltration of oxides, such as rare-earth doped CeO 2 10-12 , BaO 13 and CaO-MgO 14 , into the Ni-based anode is an effective way to enhance its coking resistance by suppressing carbon formation and promoting steam-carbon reactions. Although TiO 2 has not been investigated in SOFCs, it was used as a support in catalysts for steam reforming of hydrocarbons (methanol 15 , ethanol 16 and glycerol 17 ), CO 2 reforming of CH 4 15,18 and CO oxidation 19 ; and high coking resistance was demonstrated in CH 4 20 and ethanol 16 reforming. Stimulated by these investigations, TiO 2 was evaluated in direct-CH 4 SOFCs for the enhancement of CH 4 on-cell reforming in the present study.
Compared with electrolyte-and electrode-supported SOFCs, metal-supported SOFCs have some advantages in the aspects of electrical/thermal conductivity and mechanical ductility; consequently, the temperature distribution in and tolerance to thermal cycle of the cell are improved 21,22 . In our previous study, Ni-Fe alloy-supported SOFCs were investigated with the purpose of using wet (3 vol.% H 2 O) CH 4 as the fuel, and high performance (0.6 V at 0.4 A cm −2 and 650 °C for 50 h 7 ) was achieved. However, the Ni 0.9 Fe 0.1 -support used was not fully resistant to carbon deposition, and carbon lumps were formed in its large pores. In order to develop metal-supported direct-hydrocarbon SOFCs, Ni 0.9 Fe 0.1 -supported SOFCs were prepared with NiTiO 3 infiltrated into the Ni 0.9 Fe 0.1 -support. It was expected that NiTiO 3 would be reduced into TiO 2 -supported Ni particles in H 2 to enhance CH 4 reforming activity and resistance to carbon deposition of the Ni 0.9 Fe 0.1 -supported cells. Figure 1a-c show the XRD patterns of the as-synthesized and reduced NTO and co-fired powder mixture of NiO, Fe 2 O 3 and NTO. The as-synthesized NTO demonstrated a perovskite structure of NiTiO 3 (JCPDF# 76-0334), and the reduced product was a mixture of TiO 2 (JCPDF# 21-1276) and Ni (JCPDF# 04-0850). Figure 1d shows EDS mappings of Ni, Ti and O for the mixture. It indicates that the bright granules in surface are identified by EDS as metallic Ni, and the dark areas rich in Ti and O. Based on this result, it is expected that the infiltrated NTO particles on the surface of the scaffold of the cell support be reduced into TiO 2 -supported Ni (0) particles. It was confirmed in our previous study 7 that the sintered NiO-Fe 2 O 3 cell support is consisted of two phases of NiO and NiFe 2 O 4 , and its reduced form is Ni 0.9 Fe 0.1 alloy. With NTO powder added, the co-fired NiO-Fe 2 O 3 -NTO mixture contained NiO, NiFe 2 O 4 and NTO (Fig. 1a), which indicates that NTO was chemically compatible with NiO and NiFe 2 O 4 at temperatures up to 1000 °C and would remain as an independent phase in the scaffold of the sintered NiO-NiFe 2 O 4 cell support.

Materials and cell characterization.
Shown in Fig. 2 is the SEM microstructure of the fractured cross-section of the reduced cell with Ni 0.9 Fe 0.1 -support. As observed previously 7 , the sintered NiO-NiFe 2 O 4 cell support was reduced into a porous scaffold (58%) with a bimodal pore distribution. The average size of the large pores was around 10 μ m, which is beneficial for fuel gas transport in the support to the functional anode; and the small pores within the stem of the scaffold give a high specific surface area that is beneficial for CH 4 reforming reaction. The Ni-GDC functional anode was approximately 1α m thick and intimately in contact with the fully dense GDC electrolyte (~10 μ m) Reforming activity of infiltrated Ni 0.9 Fe 0.1 -supports. CH 4 reforming in the Ni 0.9 Fe 0.1 -support is a chemical process that in situ produces H 2 , which is electrochemically oxidized on the functional Ni-GDC anode to generate electrical power with byproduct of steam via the reaction of Thus the reforming activity of the Ni 0.9 Fe 0.1 -support is of critical importance for the performance of the cell with on-cell CH 4 reforming. Figure 4 shows the CH 4 conversion rate and reforming product distribution at 650 °C in the Ni 0.9 Fe 0.1 -supports loaded with different amounts of TiO 2 -supported Ni particles. The initial values of CH 4 conversion rate were approximately 50%, 55%, 58%, 61% and 60% for the Ni 0.9 Fe 0.1 -supports loaded with 0%, 1%, 2%, 3% and 4 wt.% of NTO (designated as 0NTO, 1NTO, 2NTO, 3NTO and 4NTO), respectively. This indicates that the addition of TiO 2 -suported Ni particles in the Ni 0.9 Fe 0.1 -support promoted its reforming activity with a limit of 3 wt.% NTO, more than which the conversion rate decreased, possibly due to the over-cover of the reforming active sites on the surface of the Ni 0.9 Fe 0.1 scaffold by TiO 2 and increased surface area of the small Ni particles for carbon deposition. The CH 4 conversion rate of 0NTO, 1NTO, 2NTO and 4NTO decreased obviously with time after approximately 12 h, only which of 3NTO remained relatively stable during the testing period of 24 h. The main reforming products were H 2 , CO and CO 2 ( Fig. 4b-d), and their concentrations varied accordingly with the testing time.
Cell performance. The cells with NTO-infiltrated Ni 0.9 Fe 0.1 -supports were evaluated at 650 °C with wet CH 4 (3 vol.% H 2 O) as the fuel; Fig. 5 shows their initial I-V-P curves. The open circuit voltage (OCV) of all the cells was around 0.78 V, due to the partial electronic conduction of GDC electrolyte 23 . The maximum power densities increased from 0.99 to 1.20 W cm −2 as the NTO loading was increased from 0 to 3 wt.%. Further increasing NTO loading to 4 wt.%, it decreased to 1.17 W cm −2 . Figure 6 shows the initial impedance spectra of the cells under a current density of 0.4 A cm −2 (Fig. 6a), from which the ohmic (R O ) and polarization (R P ) resistances were determined, and the corresponding distributions of relaxation time (DRT, Fig. 6b) 24,25 . The value of R O of each cell was similar, around 0.063 Ω cm −2 , and that of R P varied in an opposite direction to the cell voltage and power density. This tendency of cell performance change with the amount of loaded NTO in the Ni 0.9 Fe 0.1 -support is consistent with that of the activity for CH 4 steam reforming shown above, which suggests that cell performance improvement is due to the increased reforming activity of the Ni 0.9 Fe 0.1 -support and the consequent increase in the amount of H 2 available for the anode reaction.
The DRT G(τ ) was associated with the impedance Z(w) by the following expression: Where G(τ ) is defined as the DRT of impedance Z, τ is relaxation time, Z′ (∞ ) is the limitation of the real part of Z as angular frequency w approaches infinity. Consequently, impedance could be represented as series connection of infinite number of parallel polarization resistor G(τ )dτ and a capacitor τ /G(τ )dτ . For a more detailed description of DRT method and application were referred 26 .
After the initial evaluation, all the cells were further tested at 650 °C and a constant current density of 0.4 A cm −2 for up to 96 h; the results are shown in Fig. 7. The improvement on cell performance durability is in consistence with that on CH 4 steam reforming activity. The cells with 0NTO, 1NTO, 2NTO and 4NTO Ni 0.9 Fe 0.1 -supports performed 67, 78, 90 and 96 h before the sudden drop of the cell voltage; and the cell with 3NTO Ni 0.9 Fe 0.1 -support outperformed the others, degrading linearly at a slow rate of 0.5 mV h −1 during the testing period. Post-test examination confirmed that the sudden voltage drop at the end of the test was caused by cell disintegration due to dusting of the Ni 0.9 Fe 0.1 -support. The linear voltage decrease, at nearly the same rate for all the cells, may represent the intrinsic cell degradation that needs further understanding for mechanism, whereas the non-linear voltage decrease is attributed to carbon deposition in the Ni 0.9 Fe 0.1 -support and functional anode. Since the deposited carbon remained in the cell, its amount can be quantified from the temperature-programmed oxidation (TPO) profile of the post-test cells, as shown in Fig. 8. The area of CO 2 peak, an indication of the amount of CO 2 formed from deposited carbon, were 7.89 × 10 −8 , 6.93 × 10 −8 , 2.61 × 10 −8 and 3.15 × 10 −8 for the cells with 1NTO, 2NTO, 3NTO and 4NTO Ni 0.9 Fe 0.1 -supports, respectively. These values support the explanation of the durability testing results and indicate that the cell with 3NTO anode-support is the most resistant to carbon deposition among the cells investigated.

Discussion
According to previous studies 19,27 , the effectiveness of TiO 2 on improving reforming activity can be attributed to its enhanced capability of H 2 O adsorption and consequently the coking resistance. It is the H 2 O adsorbed on the catalyst that increases the reforming activity 19 ; and the prevalent presence of subsurface defects of TiO 2 in reduced atmosphere, such as oxygen vacancies and Ti interstitials, enhances H 2 O adsorption due to surface relaxation and  charge localization. On-cell methane reforming, constant adsorption of H 2 O in anode will shift the equilibrium reaction of Eqs (1) and (2) in a forward direction. Therefore, H 2 and CO 2 concentration increases whereas CO concentration decrease with increase in the amount of H 2 O. The increase in H 2 concentration and the decrease in CO concentration subsequently prevent possible carbon formation by shifting Boudard reaction (Eq. 3) and decomposition of CH 4 (Eq. 4) in a backward direction. In addition, the excess H 2 reacts with oxygen ion from electrolyte to product electrical power and steam, which enhances the water-gas shift reaction and retards CH 4 decomposition. In additional to the contribution of H 2 O adsorption on TiO 2 , the TiO 2 -supported Ni particles on the surface of Ni 0.9 Fe 0.1 scaffold are also considered to increase the reforming activity, due to its known tendency to form a strong metal-support interaction (SMSI) between TiO 2 support and Ni metal and widely used catalyst of CH 4 and ethanol steam reforming 16,28 .
Based on the DRT shown in Fig. 6b and the results reported in a previous investigation 25 , five polarization processes were identified for individual cells, which are two high-frequency processes ascribed to the gas diffusion  and charge transfer/ionic transport within the functional anode (P 2A and P 3A ), one high-frequency process associated with oxygen surface exchange and bulk diffusion within the BSCF-LSM cathode (P 2C ), one low-frequency process related to mass transport in the Ni 0.9 Fe 0.1 -support (P 1A ) and one low-frequency process attributed to CH 4 reforming in the Ni 0.9 Fe 0.1 -support (P Ref ). The contribution of each process to the total polarization resistance was obtained by data fitting the impedance spectra (Fig. 6a) using the complex nonlinear least-squares method and an equivalent circuit (inset in Fig. 6a) consisting of an ohmic resistor R O , two RQ elements for P 2A and P 3A , a Gerischer element (G) for P 2C , a generalized finite length Warburg element (W) for P 1A and another RQ element for P R . The change of the polarization resistance for each process, R 1A , R 2A , R 3A , R 2C and R Ref , with the amount of loaded NTO is demonstrated in Fig. 6c. R 3A and R 2C remained almost unaffected by NTO infiltration, since the cathode was identical for all the cells, and the electrochemical reaction in the functional Ni-GDC anodes was the same reaction of H 2 oxidation 25 regardless of the amount of NTO loaded in the Ni 0.9 Fe 0.1 -support. The resistance of diffusion of reformate in the Ni 0.9 Fe 0.1 -support and Ni-GDC functional anode, R 1A and R 2A , decreased with increasing NTO amount till 3 wt.% and then increased at 4 wt.%, which reflects the amount change of H 2 in the reformate. It is expected that higher concentration of H 2 in the reformate lead to lower diffusion resistance in porous cell support and functional anode due to the high diffusivity of H 2 . R Ref is assigned to CH 4 steam reforming process; its change with the amount of loaded NTO in the Ni 0.9 Fe 0.1 -support is consistent with that of the reforming activity. According to the data-fitting results and discussions, it may be concluded that the cell performance improvement with NTO infiltration in the Ni 0.9 Fe 0.1 -support is attributed to the improved CH 4 reforming activity and the decreased potential of carbon deposition; consequently the polarization resistances related to CH 4 reforming and reformate transport processes are decreased.
NTO infiltration into Ni 0.9 Fe 0.1 -supports was investigated with the purpose of enhancing CH 4 steam reforming activity, carbon deposition resistance and cell performance. Based on the obtained results and discussion, the following conclusions are drawn.
(1) The activity of the Ni 0.9 Fe 0.1 -support for CH 4 steam reforming is enhanced by infiltrated NTO, which is reduced into TiO 2 -supported Ni (0) particles in H 2 . The TiO 2 improves the resistance to carbon deposition by adsorbing H 2 O, while the supported small Ni particles promote CH 4 decomposition. (2) 3 wt.% of the weight of the half cell (anode-support | functional anode | electrolyte) is the optimal value for the amount of NTO infiltrated into the Ni 0.9 Fe 0.1 -support. Increased CH 4 reforming activity lead to the improvement of cell performance, durability and resistance to carbon deposition.

Methods
Cell fabrication. Ni 0.9 Fe 0.1 -supported cells were fabricated by tape casting-screen printing-sintering process.  stirring, and then stoichiometric amount of Ni nitrate (Ni(NO 3 ) 2 ·6H 2 O, Sinopharm) was added prior to the addition of citric acid (CA) and ethylenediamine tetraacetic acid (EDTA) as the chelants. The molar ratio of metal ions:CA:EDTA in the solution was 1:1:1.5. Ammonia solution was used to adjust the pH value of the solution to approximately 7. Such prepared solution was infiltrated into the pores of the sintered NiO-Fe 2 O 3 scaffold and calcined in air at 1000 °C for 2 h to form crystallized NTO nano particles. This infiltration process was repeated to achieve the desired amounts of loaded NTO in the scaffold. The crystal structure of NTO and its chemical reactivity with NiO and Fe 2 O 3 were determined by X-ray diffraction (XRD, X'Pert) using a NiO-Fe 2 O 3 -NTO powder mixture co-fired in air at 1000 °C for 2 h. The NTO powder was obtained by calcining the dried solution in air at 1000 °C for 2 h, and its reduced form (650 °C in H 2 for 2 h) was characterized by XRD for phase identification and examined by using a scanning electron microscope (SEM, FEI sirion 200).
Steam reforming activity evaluation. To evaluate the catalytic activity of the infiltrated Ni 0.9 Fe 0.1 -support for CH 4 steam reforming, the NiO-Fe 2 O 3 support sintered at 1450 °C in air for 5 h was sealed in a ceramic housing using a Ceramabond TM sealant (Aremco Product, Inc.) and reduced at 650 °C in H 2 for 2 h. Then a mixture of 10% CH 4 , 10% H 2 O and 80% He was fed into the porous support at a constant rate of 100 ml min −1 . The steam content in the mixture was controlled by flowing dry CH 4 and He gases through a saturator containing distilled water at 50 °C according to the following equation 31  Compositional analysis of the effluent gas from the reactor was conducted with an on-line Pfeiffer Vacuum Mass Spectrometer. The steam reforming was performed at temperatures between 500 and 700 °C, and the CH 4 conversion rate (X (%)) was estimated using the following equation. Cell testing and characterization. The cell performance was evaluated at 650 °C with wet (3 mol.% H 2 O) CH 4 as the fuel and ambient air as the oxidant at a flow rate of 100 ml min −1 . Using a power supply of Solartron 1480A in 4-probe mode, the current density (i)-voltage (V)-power density (P) polarization curves were obtained at a scanning rate of 5 mVs −1 from 0 to 1 V, and electrochemical impedance spectra (EIS) were acquired within a frequency range from 100 KHz to 0.01 Hz and an AC signal amplitude of 10 mV. The microstructure of the cell was examined by using a SEM. The resistance to carbon deposition of (the amount of deposited carbon in) the Ni 0.9 Fe 0.1 -supported cell was characterized by temperature-programmed-oxidation (TPO) method at a flow rate of 20 ml min −1 of pure oxygen.