Prevention of Sulfur Poisoning and Performance Recovery of Sulfur-Poisoned-Anode Electrode by Shifting Anode Electrode Potential

The possibility of performance recovery of the sulfurated Ni-yttria stabilized zirconia anode electrode by shifting the anode electrode potential to the stable region, in which nickel exists as metal, is investigated. The effect of controlling the potential of the anode electrode to the stable region to suppress the generation of nickel sulfide is also revealed. The surface of Ni particle reacts with sulfur to form Ni3S2 in a low temperature region; however, the nickel is reduced to metallic by shifting the electrode potential to −1.9 V vs reference electrode exposed to O2. When the potential of the Ni-based anode electrode is maintained at the value of ≤ −1.9 V vs the reference electrode potential, the sulfidation of nickel is inhibited. © The Author(s) 2015. Published by ECS. This is an open access article distributed under the terms of the Creative Commons Attribution 4.0 License (CC BY, http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse of the work in any medium, provided the original work is properly cited. [DOI: 10.1149/2.0061511jes] All rights reserved.

A small and high-performance solid oxide fuel cell (SOFC) system without any external equipment for fuel reforming has a potential as a power source for mobile machine like automobiles and robots. To realize the small SOFC system, the direct use of liquid hydrocarbon fuel, which has high energy density and is easy to storage and supply, is desirable. In order to use the existing fuel supply infrastructure, the SOFC should be operated by gasoline. For instance, gasoline sold in Japan contains H 2 S up to 10 ppm. Thus, the SOFC must be able to be stably operated at 10 ppm H 2 S-containing gasoline. A porous cermet of Ni and yttria-stabilized zirconia (YSZ), which has widely been studied as an anode electrode for the internal reforming operation SOFC, is one of promising anode electrode materials; however, performance degradation of anode electrode due to sulfur poisoning occurs by trace amounts of the impurity sulfur. [1][2][3][4][5][6][7][8][9] To date, extensive efforts have been taken to gain a profound understanding of the sulfur poisoning by both experimental and theoretical approaches. Based on the knowledge, promising ways to suppress the sulfur poisoning and recover the degraded performance have discovered as was summarized and reviewed in some papers. [10][11][12] Previous studies revealed that the sulfur poisoning of the Ni-YSZ anode is a two-step process. The first process is physical-and/or chemical-adsorption of sulfur onto the surface of Ni particles 2,3,5,8,13 to block the active sites for electrochemical oxidation of fuel, and causes a rapid increase in anode overpotential. The subsequent second process is arise from the microstructure change and the formation of sulfur compounds such as Ni 3 S 2 and NiS, and generates a gradual degradation that lasts for hundred hours or even longer. [2][3][4][5][6]8,[13][14][15] The sulfur poisoning rate of both processes is dependent on the partial pressure of the sulfur species in the fuel and temperature, 1,3,5,7,8,16,17 and the poisoning is stopped and recovered by introducing a fuel that does not contain sulfur. 1,3,[5][6][7]9,15,16 However, the recovery by the introduction of pure gases (e.g., H 2 , N 2 ) without sulfur species is unrealistic because it is difficult to equip the pure gases in a limited space of the mobile machines. To overcome the problems, advanced anode materials have been proposed. SOFC was successfully operated by using Cu-ceria based anodes. 18,19 Mixed ion electron conducting ceramics anode is another candidate such as BaZr 0.1 Ce 0.7 Y 0.2−x Yb x O 3−δ 20 and Sr 2 MgMoO 2−δ . 21,22 However, the poor catalytic activity for fuel oxidation of these anode electrodes limits the cell powers. Therefore, a suppressing method of sulfurpoisoning for the Ni-YSZ anode electrode is much-expected.
In this study, the effect of controlling the potential of the anode electrode in the stable region to suppress the generation of nickel sulfide is revealed. The possibility of performance recovery of the sulfurated anode electrode by shifting the anode electrode potential to the stable region, in which nickel exists as metal, is also investigated.

Equilibrium Electrode Potential by Thermodynamic Calculation
Dong et al. 4 and Chen et al. 8 revealed that Ni 3 S 2 is a sulfur composite, which is formed under the conditions of relatively lower partial pressure of the sulfur species and the cell temperature. Phase equilibrium of the Ni-O-S systems 3,9,23,24 also pledicts that Ni 3 S 2 is likely formed. Peterson et al. 25 reported that H 2 S is thermally-decomposed into elemental sulfur and hydrogen. Therefore, for simplification, only three kinds of equilibrium reactions indicated by equations 1, 2 and 3 are assumed to be occurred on the anode electrode during the power generation in a H 2 fuel containing H 2 S impurity.
The standard electromotive forces of each reaction for the various cell temperatures were theoretically estimated by using the Gibbs energy of the reactions calculated by the thermodynamic database MALT2. Then, the equilibrium electrode potentials were indicated against a reference electrode exposed to O 2 of 1 atm. For the potential calculation, the atmospheres of cathode and anode were supposed to O 2 and 10 ppm-H 2 S/3%-H 2 O/H 2 , respectively. Figure 1 shows the estimated equilibrium electrode potential for the reaction 3 and the stable-electrode-potential regions of metallicnickel and Ni 3 S 2 against the cell temperature. In this roughly estimate process, for simplify, the generation reactions of the sulfur oxide and other nickel sulfides, NiS, Ni 3 S 4 , and NiS 2 , were ignored. Thus the estimated value includes a certain degree of inaccuracy. However, the estimation indicates that the required anode potential to stabilize nickel particles as metal is a feasible range in the actual SOFC.

Experimental
Test cell preparation.-Surface-polished yttria stabilized zirconia (YSZ) pellets of 17 mm in diameter and 2 mm in thickness were used for the electrolyte. Pt cathode electrodes and Pt reference electrodes were prepared on the surface and side of the pellets, respectively, by firing a pure-Pt paste at 1300 • C for 1 h. An NiO paste was prepared by dispersing the NiO powder particles into an organic vehicle consisting with ethyl-cellulose, surfactant and organic solvent. Ni porous electrodes (10 mm diameter) or Ni model electrodes of grid-like pattern (widths of line and space : 1 mm and 1 mm) ( Figure 2) were formed on the other side of the pellets through a screen-printing, a firing at 1300 • C for 1 h and a reduction at 950 • C in H 2 . Powders of NiO and YSZ (NiO/YSZ = 75/25 (wt)) were thoroughly mixed in an appropriate amount of water using a planetary ball mill (200 rpm, 12 h), dried in an oven at 80 • C, and made into the NiO-YSZ paste using the organic vehicle. The NiO-YSZ paste was screen-printed in diameter of 10 mm and fired at 1300 • C for 1 h. The resulted NiO-YSZ electrode was reduced to the Ni-YSZ cermet electrode at 950 • C in H 2 .
Characterization.-Experimental studies were performed by using the Ni porous electrode, the Ni model electrode and the Ni-YSZ cermet electrode and a mixed gas of 9.31 ppm-H 2 S/H 2 including 3% steam as fuel. Microstructure change of the electrode was analyzed by scanning electron microscopy (SEM). Sulfur on the surface of the anode was detected by energy-dispersive X-ray spectroscopy (EDX). Crystal phase change of the anode electrode was analyzed by grazing incidence X-ray diffraction analysis (GIXRD) with a grazing incidence angle of 1 • using Cu K α radiation with D8 DISCOVER, Bruker AXS, Japan (40 kV, 40 mA) at 20 • C. Electrode performances were in- vestigated by electrochemical analysis using an electrochemical work station (SP150, Biologic SAS). The anode electrode potential was shifted against the reference electrode exposed to O 2 gas of 1 atm. The anode overpotential change by a change in the generation current was measured by a three-probe linear sweep voltammetry. The limiting current, which is the maximum current estimated by ignoring the cathode overpotential and electrolyte resistance, was compared to evaluate the performance change of the anode electrode. Anode electrode resistance was measured by a three-probe alternating current impedance method over the frequency range from 1 MHz to 100 mHz with an oscillation voltage of 10 mV on open-circuit potential. Overview of the anode electrode and test specifications are summarized in Table I.

Results and Discussion
Characterization of the Ni-YSZ anode electrode in the H 2 Scontaining fuel.-High temperature poisoning.-The Ni porous  Figure 5 Ni-model (grid) 9.31ppm-H 2 S/3%-H 2 O/H 2 900 • C, quench Figure 6 Ni-model (grid) Ni-porous (circle), poisoned at 400 • C   Figure 4 shows a time-dependency of the limiting current of the Ni model anode electrode after the introduction of H 2 S into the fuel. The performance degradation was happened quickly within 1−2 h after the introduction of H 2 S into fuel, just like the previous studies. 16,[26][27][28][29] The similar performance degradation was also confirmed by using the thermally stable Ni-YSZ cermet anode ( Figure 4). Hence, the degradation is the impact caused not by the morphology change but by the impurity H 2 S. In order to identify the existence form of sulfur on the Ni electrode at high temperature, the Ni model electrode was quenched in N 2 purge gas after the power generation in the 9.31 ppm-H 2 S/H 2 fuel at 900 • C for 40 h. The surface analysis data by EDX is shown in Figure 5. Sulfur has not existed on the surface  of the Ni electrode. If the nickel sulfide had be generated, it had to be detected by the EDX analysis also in the present work, just like several previous studies 8,30,31 detected the nickel sulfide by the ex-situ EDX analysis, because the nickel sulfide cannot be reduced to metallic nickel during the quench. On the other hand, Wang et al. 23 and Choi et al. 32 pointed out by their theoretical studies that sulfur poisoning in low H 2 S-concentration at elevated temperature is originated from the dissociation of H 2 S and the adsorption of atomic sulfur on the nickel surface. Several previous studies 3,5,8,33 also reported that the initial rapid performance degradation at elevated temperature in the atmosphere with low H 2 S concentration is caused by the adsorbed sulfur atoms on the nickel particles. If the sulfur existed as the adsorbed atom on the nickel particle, it cannot be detected by the ex-situ EDX analysis in the present work just like that Rasmussen et al. 34 could not detect the adsorbed sulfur by the ex-situ EDX analysis. Piecing together the knowledge from the previous studies and the result of Figure 5, the performance degradation by the power generation in the 9.31 ppm-H 2 S-containing fuel at high temperature of around 900 • C is not caused by the nickel-sulfide formation, but is a result of physicaland/or chemical-adsorption of sulfur onto the surface of Ni grain.
By changing the fuel from pure H 2 to 9.31 ppm-H 2 S/H 2 , the anode overpotential was increased in the small current region at first (Figure 3). Figure 6 shows the Nyquist plots and Bode plots of the electrochemical impedance spectra of the anode electrode of open circuit potential before and after the introduction of H 2 S into the fuel. The Nyquist plots were consists of one circles in both cases at 950 • C. The Bode plots apprise that the circles in the Nyquist plots were observed in the similar frequency region. In this measurement, the gas diffusion resistance is negligibly small because the model anode electrode  used here is dense and simple shape. Hence, the impedance in the small current region including the OCV state is assigned to the charge transfer reaction. When the current is sufficiently small, the anode polarization resistance is related to the anode exchange current density. In the case of small exchange current density, the overpotential in the sufficiently small current region became larger by the introduction of H 2 S into the fuel gas. It is well known that the number of the reaction site near the triple-phase boundary (TPB) of Ni, YSZ and gas phase governs the exchange current density, which means that the observed larger overpotential of the anode in the H 2 S-containing fuel in the small current region including the OCV state is attributed to the diminished number of the reaction site by the adsorption of sulfur. Consequently, the adsorbed sulfur reduced the limiting current in the high temperature region. Low temperature poisoning.-The Ni model electrode was slowly cooled from 950 • C to the room temperature in the 9.31 ppm-H 2 S/H 2 fuel by a rate of 1 • C/min, and then EDX mapping analysis of its cross section was performed (Figure 7). Sulfur existed on the surface of nickel particle and the interface of Ni particle and YSZ electrolyte. Since nickel atoms also existed at the same position, sulfur on the electrode ought to be present as nickel sulfide formed at the low temperature region during the slow cooling. In order to reveal the effect of the poisoning and identify the sulfide generated at the low temperature, the Ni electrode was poisoned at 400 • C. The fresh Ni model electrode was cooled slowly in pure H 2 gas from 950 • C to 400 • C by a rate of 1 • C/min, and the fuel gas was changed to 9.31 ppm-H 2 S/H 2 at 400 • C. The anode overpotentials for various current at 400 • C were measured before and after the fuel gas change (Figure 8). The overpotential was increased and the resultant limiting current was decreased in a short period of time for the H 2 S-containing fuel introduction. Then, although the power generation with constant voltage of −0.5 V against the reference electrode potential at 400 • C for 10 h by the H 2 Scontaining fuel was performed, significant performance degradation was not confirmed. This time dependence infers that the sulfur poisoning is happen quickly even at the low temperature around 400 • C and the subsequent progress is slow. After the power generation at 400 • C for 5 h in the H 2 S-containing fuel, the anode electrode was quenched in N 2 purge gas and its surface was analyzed by SEM and EDX mapping (Figure 9). Many of the nickel particles became sulfide and the shape of the particle was rounded. Figure 10 is a profile of  CIXRD analysis of the electrode surface. The rounded particle was Ni 3 S 2 . That is, the formation of Ni 3 S 2 is a rapid reaction which has happened in the short period of time for the temperature decreasing process. Interestingly, some of the nickel particles in the Figure 9 were metallic nickel even after 5 h of sulfur poisoning. The authors, at the moment, cannot clearly explain this phenomenon.
Performance recovery in the H 2 S-containing fuel of the sulfurpoisoned Ni anode electrode by sulfide generation.-The fresh Ni anode was cooled in the humidified pure H 2 from 950 • C to 400 • C, and the Ni anode potential was held at −0.3 V against the reference electrode potential in the 9.31 ppm-H 2 S/H 2 fuel including 3% steam at 400 • C for 5 h to be sulfur-poisoned. The anode electrode was heated again to 700 • C, and the various negative voltages from −0.5 V to −2.6 V against the reference electrode potential were applied to the anode electrode for 10000 s at each voltage in order to lower the anode electrode potential. Then, for each the applied voltage, the anode overpotential for various current was measured (Figure 11a). In Figure 11a, the initial limiting current at 700 • C in the humidified H 2 , which was measured on the way of the cooling from 950 • C to 400 • C, was also shown. Figure 11b depicts a relation between the limiting current and the applied negative voltage (anode potential). When the applied voltage was lower than −1.5 V (vs reference electrode potential in O 2 ), the limiting current was in-  creased. When the voltage was lowered to −1.8 V, the limiting current increased sharply with the applied voltage. The recovered performance reached approximately 90% or more of the initial performance. Considering that the performance degraded approximately 10% by the adsorption of sulfur shown in Figures 3 and 4, the anode-electrode shifting recovers the performance of the nickel anode electrode, which is poisoned by the formation of sulfide at lower temperature for the start-stop process of SOFC. Some of the adsorbed sulfur might have been removed by oxidation with the oxygen, which was formed by the electrolysis of steam under the negative potential of the anode electrode, at least temporarily.
The recovery-starting potential of the limiting current was lower than the value (approx. −1.3 V vs reference electrode potential in O 2 ) shown in Figure 1. The deviation of the experimental value of the reduction potential from the estimated value obtained by the thermodynamic calculation is mainly due to simplification of the reaction. In actual experiment, the reaction of sulfur and the oxygen generated by electrolysis of steam (e.g., S + O 2 = SO 2 ) must have occurred also at least. However, in the thermodynamic calculation for Figure 1, the oxidative reaction of sulfur was ignored to simplify. The overvoltage, which is caused by the affected by position of the reference-electrode, might have also increased the deviation. Because the reference electrode was attached on the side of the pellet (not on the same plane to the anode electrode), some of the applied voltage might have been divided to the overvoltage by the electrolyte resistance.  Figure 12 shows the limiting current change at 700 • C in the H 2 Scontaining fuel as a function of the application time at the various negative anode potentials. The data were obtained on the way to reach the steady value by the potential application for 18000 s (shown in Figures 11a and 11b). While an anode potential was maintained in each value, temporarily, potential application was stopped and the measurement was executed. The limiting current recovered to the almost same value after 75 min independently of the applied voltage, i.e., the required application time is almost same when the negative anode potential voltage is large enough and the sulfur-poisoning time is same of 18000 s each other. The influence of sulfur-poisoning time on the recovery time is shown in Figure 13. The recovery time was short (10 min) when the sulfur-poisoning time was limited to 1 h, although the long time of 75 min is needed for the case of long time poisoning (5 h). It is supposed that the long time exposure to the H 2 S containing fuel has increased the thickness of nickel sulfide layer and then the long recovery time was necessary to reduce the inner layer of the thick nickel sulfide. Therefore, in order to reduce the recovery time, the potential shift of the anode electrode should be frequently performed before the growing the thickness of nickel sulfide layer.  Suppression of the sulfur poisoning by sulfide generation during the low temperature region.-The Ni model anode electrode of the SOFC using the H 2 S-containing fuel is poisoned by sulfide formation in the low temperature range for start-stop operation. In order to suppress the sulfur poisoning, the effect of potential control of anode electrode during the start-stop process was investigated. The fresh Ni model electrode was exposed to the H 2 S-containing fuel at 950 • C, and subsequently cooled by the rate of 1 • C/min under the condition of negative voltage of −1.9 V against the reference electrode potential. Figure 14 shows SEM and EDX analysis results of surface and cross section of the anode electrode. Sulfur did not existed in the anode electrode, i.e., nickel sulfide was not formed in the anode electrode. Therefore, the sulfur poisoning by the nickel sulfide formation is suppressed by the application of negative voltage (≤ −1.9 V vs reference electrode potential in O 2 ) to the anode electrode.

Conclusions
The explanation of sulfur poisoning of Ni-yttria stabilized zirconia anode electrode has been investigated by using a 9.31 ppm-H 2 S/H 2 fuel, and was verified as follows.
1) At high temperature around 900 • C, the performance of Ni-based anode is reduced by decreasing the triple-phase boundary as the reaction site of electrochemical oxidation due to physical adsorption and/or chemical adsorption of sulfur. 2) In low temperature range, the Ni particle reacts with sulfur to form Ni 3 S 2 to degrade its performance in a short time period after exposure to the H 2 S-containing fuel.
The performance recovery and the suppression of degraded performance by sulfur poisoning have been attempted by controlling the anode electrode potential.
3) A simple assumed thermodynamic calculation estimates that the controlling the anode electrode potential enables the suppression of the sulfur-poisoning. 4) The nickel sulfide particle in the anode electrode is reduced to metallic nickel by shifting the anode electrode potential to large negative value. By the control of anode potential, its degraded anode performance owing to the formation of nickel sulfide (Ni 3 S 2 ) at low temperature is recovered even under the sulfur-containing atmosphere. For instance, the anode potential at 700 • C in the 9.31 ppm-H 2 S/3%-H 2 O/H 2 fuel should be −1.9 V or lower (vs reference electrode in O 2 ) to reduce the nickel sulfide (Ni 3 S 2 ) to metallic nickel for the recovery of the performance. In order to reduce the recovery time, the potential shift of the anode electrode should be frequently performed before the growing the thickness of nickel sulfide layer. 5) The nickel particle in the anode electrode maintains to be metallic nickel even in the sulfur-containing fuel at low temperature, when the anode electrode potential is controlled to the negative value. Therefore, this sulfur poisoning by the formation of nickel sulfide in the low temperature range for the start-stop operation of SOFC is prevented by the electrode potential control.