The effect of SO2 on the Ni-YSZ electrode of a solid oxide electrolyzer cell operated in co-electrolysis

In this work, the effect of sulphur poisoning of the Ni-YSZ electrode of an SOEC operated in co-electrolysis mode was investigated. Short-term tests with exposure up to 5 ppmv of SO2 were performed at OCV and under polarization ( 0.25 A cm−2). The two-stages degradation pattern observed consisted of an initial fast voltage increase followed by a slower voltage increase similar to that of an SOFC exposed to H2S. Electrochemical impedance spectroscopy and the analysis of the distribution of relaxation times showed that both the catalytic and electrochemical reactions were affected by SO2. After extended periods in SO2-free reactant, only a partial recovery of the performances was observed even when exposure amounted to only 0.5 ppmv of SO2 independently on the current density. A durability test at a constant polarization of 0.5 A cm−2 showed a voltage ‘runaway’ behavior during successive exposures to 1 ppmv and 2 ppmv of SO2. This behavior originated from a drastic increase of the serial resistance, which almost completely recovered when the SO2 supply was cut. This behavior was not observed during exposure to 0.5 ppmv of SO2, suggesting that, in these test conditions, the voltage ‘runaway’ behavior could be avoided at a sub-ppmv level. Successive exposure-recovery cycles were found to weaken the SOEC tolerance to SO2 and a low frequency pseudo-inductive arc was observed in the impedance response during and after the second exposure to SO2.


Introduction
Solid oxide electrolysis cells (SOECs) can be used to store excess renewable electricity into chemical energy by electrolysing steam [1], CO 2 [2], or a mixture of the two referred to as co-electrolysis [3]. The latter is found to be particularly interesting with downstream coupling with methane production due to the high potential for heat integration [4]. Supplying water to the SOEC is relatively straightforward due to its abundancy, whereas supplied CO 2 generally must be separated from a dilute source, such as the atmosphere (~0.04% vol CO 2 ) or a more concentrated one, such as flue gases (10% to 20% vol CO 2 ), or derived from biomass (up to 100% vol CO 2 during fermentation processes) [5]. Impurities present in the CO 2 stream used during the carbon capture process may also be present in the captured CO 2 [6]. The major impurities found in the captured CO 2 are: O 2 , H 2 O, N 2 , NO x , SO x ,CO, H 2 , and CH 4 , [6][7][8]. Traces of sulphur compounds may be particularly problematic for the operation of an SOEC, as solid oxide fuel cells (SOFCs) containing nickel were found to be vulnerable to sulphur compounds such as C4H4S [9], CH3SH, COS [10,11], and H 2 S [12][13][14][15].
Sulphur has been found to impact the electrochemical performance of an SOFC by reducing the catalytic reactions, such as the steam methane reforming (SMR, (1)) and the reverse water-gas shift reaction (RSWG, (2)) [16,17], and by limiting the charge transfer processes at the triple phase boundary (TPB) [18][19][20].
Thermodynamic computations show that nickel sulphide (Ni 3 S 2 (l)) only forms at concentration of H 2 S above 1%; at lower H 2 S concentrations, adsorption of sulphur on nickel is prefered [14,21]. The adsorption process can be described as: where S ad represents the sulphur adsorbed on nickel. Generally, when the sulphur supply is cut, the SOFC performance recovers partially or completely [15,22]. Furthermore, behaviour during poisoning and recovery are dependant on the operating conditions. Generally, an increase of the operating temperature [12] or current density [15] improve the SOFC tolerance to sulphur, whereas an increase of the sulphur concentration [12,22,23] and the presence of CO/CO 2 as reactant compared to H 2 /H 2 O [24] were found to reduce the sulphur tolerance. Although H 2 S is the most commonly used compound for sulphur tolerance experiments on SOFCs, [17] observed a higher rate of deactivation when a nickel catalyst was exposed to SO 2 rather than H 2 S. Residual traces of SO 2 in captured CO 2 may be more critical than a similar amount of H 2 S on the solid oxide cell (SOC). Therefore, operating an SOC in electrolysis rather than fuel cell mode could potentially modify the tolerance of the SOC due to the different reactant composition, the inverted polarization, and nature of the sulphur compound. Few researcher have investigated the effects of impurities on a SOECs. Of these, Kushi [25] found that an SOC's O 2 electrode was more sensitive to SO 2 under electrolysis than under fuel cell operation. Zheng et al [26] investigated the effect of SO 2 on a composite lanthanum strontium cobalt manganite (LSCM) and gadolina-doped ceria (GDC) fuel-electrode during electrolysis of simulated flue gases. No significant impact on the SOEC performance was reported below 15 ppm of SO 2 ; electrolysis activity even improved when O 2 was also present. Ebbesen et al [27] were able to suppress the degradation of an SOEC, composed of a nickel yttria stabilized zirconia (Ni-YSZ) cermet reactant electrode, by cleaning the inlet gases. The degradation observed without the gas cleaning device was, thus, attributed to the presence of impurities in the inlet gases. H 2 S was found in the ppb level in the supply gases (without the gas cleaning device), but the role of H 2 S in the degradation was not clearly assessed.
This work therefore aims at assessing the effects of sulphur on the Ni-YSZ hydrogen electrode of an SOC performing co-electrolysis of CO 2 and H 2 O. First, results obtained from short-term tests performed on SOECs with a controlled level of SO 2 at various current densities are presented. Then results of a durability test (2500 h) conducted in galvanostatic mode under exposure to SO 2 are reported. The effects of SO 2 on the performance and durability of the SOEC are evaluated using electrochemical impedance spectroscopy (EIS) and the analysis of the distribution of relaxation times (DRT).

Experiment description
Hydrogen-electrode (HE)-supported ceramic cells from SOLIDpower originating from the same batch were used to evaluate the impact of SO 2 on a solid oxide electrolyzer cell (SOEC). The cell is composed of a cermet Ni-YSZ HE, a YSZ electrolyte, a GDC barrier layer, and a lanthanum strontium cobalt ferrite (LSCF) oxygen electrode (OE). The cells have a diameter of 60 mm with an active area of 12.56 cm 2 (40 mm diameter). On each electrode, the gas flow is radially distributed from the centre to the border of the cell. A glass sealant is used to guarantee the gas tightness of the set up. The reduction of NiO to Ni is performed at 850 • C by slowly increasing the hydrogen content from 10% vol H 2 and 90% vol N 2 to 100% vol H 2 with a flow rate of 200 Nml min −1 on the hydrogen side and 500 Nml min −1 of air on the oxygen side.
After the reduction step, a mixture composed of 65% vol H 2 O, 25% vol CO 2 and 10% vol H 2 was used on the HE side with a flow rate of 150 Nml min −1 . Addition of argon containing 1000 ppmv of SO 2 was used to control the concentration of sulphur dioxide in the reactant stream. The added Ar and SO 2 always represented less than 1% of the total reactant flow, the dilution effects were thus considered negligible. The OE was swept by air with a flow rate of 150 Nml min −1 . A gold mesh was used for current collection on the oxygen side whereas a nickel mesh was used on the hydrogen side. The tests were performed at 750 • C with a current density ranging from 0 to 0.5 A cm −2 and a SO 2 concentration ranging from 0.5 to 5 ppmv. The SOEC voltage was constantly monitored and its evolution was used as a first evaluation of the effect of SO 2 on the SOEC performance. Electro-chemical impedance spectroscopy (EIS) was used to identify the mechanisms involved. Measurements were taken from 20 mHz to 200 kHz with a sinusoidal perturbation of ± 200 mA (corresponding to about ± 10 to ± 15 mV). Inductance of the lead wires was limited by the use of twisted pair wires and, if necessary, high frequency points were manually removed.
When represented in a Nyquist diagram, as in figure 1, the low-frequency intercept (right hand side) of the impedance spectra represents the total cell resistance (R tot ), and the high frequency intercept (left hand side) the serial (or ohmic) resistance (R s ). The difference between R tot and R s gives the polarization resistance . R s and R p are used to identify and quantify the degradation processes. To get a better insight into the degradation process, the distribution of relaxation times (DRT), γ, of selected electro-chemical impedance spectra was computed using Tikonov regularization [28]. For consistency with the Nyquist representation of the impedance spectra used in this work, the DRT is represented in terms of frequency rather than time. The attribution of the DRT peaks to processes are made according to [29], who performed an extensive experimental sensitivity analysis combined with a dynamic numerical model on SOCs originating from the same provider. Six DRT peaks were identified and are presented in table 1.

Results and Discussion
Thermodynamic equilibrium calculations performed with HSC Chemistry 8.2.0 [30] showed that, with the reactant gas mixture used, SO 2 is not stable at 750 • C under 1 atm and tends to form H 2 S. However, the kinetics of the decomposition reactions was not considered in this computation. The exact chemical form of the sulphur compound present at the hydrogen electrode is, thus, not exactly known. The thermodynamic equilibrium gas composition at 750 • C is 67.2% vol H 2 O, 22.9% vol CO 2 , 7.9% vol H 2 , and 2.1% vol CO, which differs from the sent reactant gas mixture due to the RWGS reaction.

Open circuit conditions
An SOEC, under open-circuit conditions, was exposed to 0.5, 1, 2 and 5 ppmv of SO 2 during a period ranging from 19 to 24h intercalated with a 20 h recovery period during which the cell was flushed with an SO 2 -free reactant. Before the first exposure to SO 2 , the measured cell voltage (V 0 ) was only 2 mV above the theoretical Nernst potential computed using the reactant composition at thermodynamic equilibrium, 0.865 and 0.863V, respectively. This indicated that the reactant flow was close to the thermodynamic equilibrium and that the SOEC sealing was gas-tight. When exposed to SO 2 , the SOEC voltage showed a step-like increase pattern, as presented in figure 2(a). The height of the step, computed from V 0 , and the slope of the step appeared to be correlated with the SO 2 concentration as presented in table 2. After this initial step increase, the voltage steadily increased at a rate of about 20 to 30 mV kh −1 , without any apparent relation with the SO 2 concentration. A schematic representation of the SOEC's voltage evolution during SO 2 poisoning defining the values reported in table 2 is presented in appendix A. When the exposure to SO 2 was stopped, the voltage almost instantaneously started decreasing. The recovery was much slower than the contamination and was never complete, suggesting an irreversible effect of sulphur even after exposure to only 0.5 ppmv of SO 2 . The rapid initial degradation followed by a reduced and constant deactivation and a slower recovery process once the sulphur supply was stopped was in good agreement with observations reported for solid oxide fuel cells exposed to sulphur compounds [9,14,18,23,24]. As the SOEC was not polarized, it was similar as testing an SOFC at high H 2 O and CO 2 content. The initial voltage step was thus attributed to the same process than for an SOFC: adsorption of sulphur on nickel, reducing both the catalytic and electrochemical performances of the SOEC. Analogously to (3), this adsorption reaction can be described as depending whether sulphur is present in the form of H 2 S or SO 2 , where S ad represents the sulphur adsorbed on Ni. As no bias current was applied, solely a modification of the gas composition at the hydrogen electrode would affect the measured cell voltage. It was thus assumed that the catalytic reactions (i.e. RWGS reaction) were limited due to the coverage of nickel by sulphur preventing the feed gas to reach the thermodynamic equilibrium composition. This is in agreement with Hagen [24] and He et al [31], who reported a deactivation of the RWGS reaction when the nickel-based anode of an SOFC operating on syngas was exposed to H 2 S.
The deviation from the thermodynamic equilibrium was then evaluated by assuming that the measured open circuit voltage (OCV) corresponded to the Nernst potential of an effective gas composition between the inlet and thermodynamic equilibrium composition. Using HSC Chemistry 8.2.0 [30], thermodynamic equilibrium gas compositions were computed at varying effective temperatures until the corresponding Nernst potential matched the measured OCV. The difference between the effective and actual operating temperature is then representative of the deviation of the composition from the thermodynamic equilibrium. The observed increase in OCV due to the presence of SO 2 thus corresponds to a thermodynamic equilibrium composition at an effective temperature lower than the actual operating temperature, indicating a reduction of the RWGS reaction. The increase of 12 mV observed after exposure to 5 ppmv of SO 2 seen in figure 2(b) corresponded to a 25 • C decrease of the effective temperature. As the CO Figure 3. Ni coverage and reduction of the CO content with respect to the SO2 content. The reduction of the CO content was computed based on the equivalent composition corresponding to the initial OCV measured. The red data points are extrapolated data. Table 2. Height of initial degradation step measured from the initial and recovered (in between parentheses) voltage, the initial degradation rate, and the final degradation rate, as defined in appendix A, when the SOECs were polarized at a current density j and exposed to various concentrations of SO2. a Linearly extrapolated based on the stabilization times obtained with other SO2 concentration and the initial degradation slope b These results, obtained during the durability test, are discussed in the section 3.2 content of the equilibrium reactant composition originated only from the RWGS reaction, its variation relative to the composition before exposure to SO 2 was used as a quantitative indicator of the modification of the catalytic activity. In figure 3, reduction of the CO content is compared with the surface coverage of Ni using the correlation (6) proposed by Hansen [32] that was originally developed for H 2 S and adapted here by simply replacing the partial pressure of H 2 S by the partial pressure of SO 2 . Further work is needed to assess the validity of this adaptation for SO 2 .
When the cell was exposed to 5 ppmv of SO 2 , the surface coverage reached 0.93, whereas the CO content decreased by only 7.5%. Also, the reduction of the CO content seemed to stabilize below 10%. The effect of SO 2 on the RWGS reaction was thus relatively small, thanks to the high nickel content present in the HE (acting as a catalyst and electro-catalyst), and the nickel current collector (acting as a catalyst). However, the reduction of the catalytic activity in the vicinity of the TPB may still have significant impact on the SOEC performance under polarization due to the limited RWGS reaction reducing the local steam content.
The modification of the electro-chemical performances of the SOEC was assessed by performing EIS measurements after each exposure to SO 2 and after each recovery period. The results, summarized in figure 5, showed that the major degradation occurred between 10 Hz and 10 kHz, whereas no increase of R s was observed. The DRT results (figure 5(c)) showed clearly that the major contribution to the increase of R p occurred on the peak located around 1 kHz, which is associated with the charge transfer processes occurring at the HE. The peak increased drastically and shifted towards a lower frequency, leading to the covering of the peak located around 100 Hz, which is related to the transport processes in the hydrogen and oxygen electrode, making the deconvolution impossible. Even though the measured cell potential mostly recovered from the exposure to SO 2 , the impedance response did not completely recover, suggesting that part of the Ni active sites remained covered. During the recovery following the exposure to 0.5 ppmv of SO 2 , the DRT peak located around 1 kHz kept increasing, despite the decreasing (i.e. recovering) OCV. The current collector, being made of Ni, likely adsorbed sulphur during the exposure period and then slowly released it during the recovery period, thus acting as a sulphur buffer. This may have provided enough time and supply for sulphur to deposit on the active Ni, leading to the observed increase of impedance even after the SO 2 supply was shut off.
The increase of R p after exposure to SO 2 , presented in figure 4, showed a linear dependency with the Ni surface coverage, similarly to the results reported by Hansen [32] for H 2 S poisoning. The polarization resistance obtained after exposure to 0.5 ppmv SO 2 was not considered here as it did not reach a stable state. The multiplicative coefficient of the linear interpolation was in good agreement with the results reported by Madi et al [9] for an SOFC operated on synthetic bio-gas containing various amounts of thiophene. To a smaller extent, the increase of R p after recovery was also found to be linearly dependent on θ s indicating that the retentive effect of SO 2 exposure was likely related to permanently adsorbed sulphur.

Under polarization (0.25 A cm −2 )
A second SOEC was exposed to 1 ppmv of SO 2 during 25 h followed by a 40 h recovery period. The cell was, then, polarized at 0.25 A cm −2 and successively exposed to 0.5, 1, 2 ppmv of SO 2 during a period ranging from 45 to 70 h each time followed by a recovery period where the cell was flushed with a SO 2 -free reactant.
Under polarization, the SOEC showed a step-like degradation when exposed to SO 2 , as shown in figure  2(c), similarly to the OCV case. The initial degradation step height was about two to three times larger when the cell was polarized compared with the OCV case but the time needed to reach a stabilized state was similar regardless of whether a 0.25 A cm −2 bias was applied or not as seen in table 2. However, the final degradation rate was about three to four times higher under polarization than at OCV. When the exposure to SO 2 was stopped, the SOEC voltage decreased in a parabolic manner but the recovery was never complete, as presented in figure 2(d). During the recovery process, after exposure to 2 ppmv of SO 2 , a technical issue with the galvanostat ended the test prematurely, thus only EIS data after exposure to 0.5 and 1 ppmv were available. EIS measurements taken after 72 h of exposure to 0.5 ppmv of SO 2 showed that both the high and low frequency processes were affected, as shown in figure 6(a) and (b). Analysis of the DRT results of the impedance spectra showed that the peaks located around 1 Hz and between 100 Hz and 1 kHz increased, as presented in figure 6(c). The peak located around 1 Hz is sensitive to the reactant composition [29], which was coherent with a reduction of the catalytic reactions due to the coverage of nickel by sulphur [24]. This degradation should eventually stabilize when the equilibrium between the surface coverage and the sulphur content in the reactant is reached. The increase of the peak located around 1 kHz was indicative of a deterioration of the charge transfer processes and a modification of the gas composition. The behavior observed for an SOEC operated in co-electrolysis mode was found to be very similar to the one reported for an SOFC operated under reformate operation [19,20] indicating an analogous contamination process independently of the current direction.
Increasing the SO 2 content to 1 ppmv led to an increase between 100 Hz and 1 kHz; no significant changes were observed around 1 Hz, suggesting that a saturation was reached ( figure 6). When the SO 2 content was raised to 2 ppmv, no further significant changes were observed when compared with 1 ppmv ( figure 6). After the recovery period, the peak located around 1 kHz shifted back to its original relaxation frequency but did not recover completely, whereas the peak located around 1 Hz almost completely recovered.
Similarly to the OCV tests, the relative increase in R p showed a linear behavior with respect to θ s as presented in figure 4(b). The relative increase in R p was found to be smaller than at OCV but this was likely due to the test history. The initial impedance response at 0.25 A cm −2 (figure 4) was not measured on a 'fresh' SOEC but after exposure at OCV to 1 ppmv of SO 2 and a recovery period. The relative increase of R p was, thus, not computed on the same basis and can not be directly compared. The coefficients multiplying θ s in figures 4(a) and (b) represents the sensitivity of the relative increase in R p towards θ s . This coefficient was larger when the SOEC was polarized compared to the OCV case, 1.56 and 1.41 respectively, indicating that the SOEC was more sensitive to an increase of the SO 2 content when polarized. It suggested that the polarization had a negative impact on the SOEC tolerance to SO 2 .
Under polarization, hydrogen was generated, which should reduce the surface coverage of nickel by sulphur, according to correlation (6). However, due to the logarithmic dependency on the hydrogen partial pressure and the low reactant utilization (0.15 at 0.25 A cm −2 ), this effect was likely negligible. In SOFC operation, the current density was found to improve the sulphur tolerance [15,33,34]. It was suggested that a higher oxygen ion flux at the triple phase boundary could help the desorption of the adsorbed sulphur by oxidation, which can be expressed as the reverse of (5). During operation as an SOEC, the opposite effect could be expected, as the polarity and thus oxygen ion flux is inverted. However, prior researchers have shown no [35] or a negative impact [36,37] of the current density on the sulphur tolerance of an SOFC. Further investigation should, thus, be performed to clearly assess the effect of current density on the sulphur tolerance of an SOEC.

The effect on durability
Another SOEC was operated during 2500 h at 0.5 A cm −2 with periodic exposure to SO 2 to assess the long-term stability; the time evolution of the cell voltage and SO 2 concentrations are presented in figure 7. In the first 200 h, the SOEC showed the same step-like degradation pattern as presented before, i.e. a fast degradation (0-15 h) followed by a reduced and constant degradation (15-200 h). When compared with exposure to 1 ppmv of SO 2 at lower current density, the initial degradation rate and the time to stabilization were strongly affected by the current density, whereas no apparent link was found with the height of the initial degradation step and the final degradation rate. At 0.5 A cm −2 , the kinetics of sulphur adsorption on nickel was likely promoted more than at a lower current density.
In the first 11 h, the degradation was only caused by an increase of R p , as shown in figure 8, where the time evolution of R s and R p is represented. According to the DRT of the EIS measurements taken during the rapid degradation, shown in figure 9, the high-frequency impedance increased first, followed by an increase over the entire spectrum until a relative stabilization. The DRT peak located around 1 kHz increased massively and shifted towards a lower frequency, which is likely due to the adsorption of sulphur on the nickel active-sites affecting the electrochemical conversion of H 2 O/H 2 and CO 2 /CO [19,20]. The DRT peak located around 1 Hz first decreased and widened, and then increased while shifting towards a higher frequency. The width of a DRT peak can be related to the spatial distribution of the relaxation time associated to a particular phenomenon [38], e.g. due to inhomogeneities in the micro-structure [39]. During the initial fast degradation caused by sulphur, Rasmussen and Hagen [18], and Hagen [24] observed a peak of the in-plane voltage that was attributed to a redistribution by the current due to the progressive adsorption of sulphur on nickel from the inlet to the outlet. The widening of the DRT peak located around 1 Hz was thus assumed to be the result of the progressive adsorption of sulphur on nickel, which locally reduced the catalytic and electrochemical reactions modifying the reactant composition until the nickel coverage was at equilibrium. The peak located around 25 Hz increased and shifted towards higher frequency, suggesting that the diffusion processes at the HE were impacted. As a consequence of the various peak shifting, the DRT between 10 Hz and about 30 kHz was compressed in a smaller frequency range after the exposure to SO 2 , making the deconvolution of the different processes involved more complicated.
After the initial fast voltage rise, the degradation rate was constant at about 60 mV kh −1 until 200 h, where both R p and R s started rising, as shown in figure 8. The degradation rate then increased to 200 mV kh −1 . After about 400 h, the voltage started rising exponentially due to an increase of R s , whereas R p remained almost constant, thus suggesting the presence of a cumulative effect due to the exposure to SO 2 . After 485 h, the exposure to SO 2 was stopped to limit the voltage increase. After less than 5 h, likely due to the presence of residual SO 2 , R p and R s started decreasing simultaneously. After about 280 h of recovery, only R s and the low-frequency region of the EIS had partially recovered, whereas the high-frequency region of the EIS (charge transfer) remained almost unchanged, as shown in figure 10(a) and (b).
Analysis of the DRT of the impedance spectra, presented in figure 10(c), showed that when the SO 2 supply was shut off, the conversion peak (around 1 Hz) greatly recovered but was still shifted to higher frequency, whereas the low-frequency peak (around 0.1 Hz) widened and shifted towards higher frequency. The charge transfer peak (around 1 kHz) continued increasing while shifting towards higher frequency and getting sharper. The electrochemically active sites were, seemingly affected more permanently by sulphur compared with the rest of the cell, also suggesting, a negative impact of the current density on the SOEC  durability when exposed to SO 2 . During the recovery process, the shape of the low-frequency region of the impedance spectra changed noticeably, especially below 0.1 Hz ( figure 10(b)), which resulted in the presence of a new peak at very low frequency (around 30 mHz), as shown in figure 10(c). The DRT fingerprints between 10 Hz and about 30 kHz remained compressed to a smaller frequency range.
The SOEC was then exposed to 1, 2, 0.5 ppmv intercalated with a recovery period; the resulting SOEC voltage and cell temperature evolution are reported in figure 7. A similar behavior to the first exposure was observed: a fast initial increase of the cell voltage in the first hours after the introduction of SO 2 followed by a slower voltage increase, which eventually started to run away due to a drastic increase of R s (only when exposed to 1, and 2 ppmv of SO 2 ), as shown in figure 8. The height of the initial step-like degradation was not found to be dependent on the SO 2 concentration, whereas the initial degradation rate and the time to stabilization appeared linked to the SO 2 concentration, as presented in table 2.

Voltage 'runaway'
The drastic increase of R s occurred earlier when the SOEC was exposed to 1 ppmv of SO 2 for the second time and even earlier under 2 ppmv of SO 2 . The polarization resistance between about 30 Hz and 1 kHz was the highest during the last exposure, even though the SO 2 concentration was the lowest (0.5 ppmv). It indicated the presence of a cumulative effect caused by the successive exposure-recovery cycles.
The time needed to reach the voltage 'runaway' regime was shorter during the second exposure to 1 ppmv of SO 2 , suggesting a weakening of the SOEC after the first exposure. The voltage 'runaway' regime was reached even more quickly during the exposure to 2 ppmv of SO 2 . When the SO 2 supply was cut, R s and R p again decreased, confirming that the 'runaway' behavior observed was linked to the presence of SO 2 in the reactant. When the SOEC was exposed to 0.5 ppmv of SO 2 , however, the initial voltage increase was also present, but the voltage 'runaway' was not observed even after more than 500 h of exposure. The 'runaway' behavior thus appeared to be dependent on the history of exposure to SO 2 and only occurred above a certain SO 2 concentration.
When an increase in serial resistance was reported for an SOFC exposed to H 2 S, it was found to be irreversible [40,41], suggesting a different degradation and/or recovery mechanism for an SOEC exposed to SO 2 . Two possible mechanisms, discussed in more detail in appendix B, are proposed to explain the reversible increase of R s : • A progressive reduction of the active area due to the deactivation by sulphur of the electrochemically active Ni from the higher (inlet) to lower (outlet) current density regions. Assuming an analoguous behavior to an electrode partial delamination, a comparable increase of R s and R p would be expected [42]. This is thus unlikely to be solely responsible for the observed behavior. • A cumulative effect of sulphur blocking the electrochemically active regions located closest to the electrolyte, possibly from the higher (inlet) to lower (outlet) current density regions. The thickness of the electrolyte being on the order of 10 µm, the extent of the deactivated region would be in the order of 2 µm, considering that R s increased by a factor of 2 to 3.

Low-frequency hook
During the second exposure to 1 ppmv of SO 2 , a pseudo-inductive behavior began to appear at low frequency when the SOEC was exposed to sulphur and during part of the recovery, as shown in figure 11. This low-frequency pseudo-inductive behavior appeared to be more intense at the beginning of the exposures and then lowered in intensity, but did not completely disappear. It then slowly disappeared during the recovery period, as shown in figure 12(a) after exposure to 0.5 ppmv of SO 2 . Figure 12(b) shows that during the recovery, the low-frequency hook seemed to transit from an pseudo-inductive to a capacitive behavior only after the region around 2 Hz mostly recovered. Analysis of the DRT of the impedance measurements (figure 12(c)) indicated that the peak located around 1 kHz first increased and then decreased while shifting towards higher frequency during the entire recovery period, whereas the peak located around 100 Hz appeared to continuously improve and shift slightly towards a higher frequency. Due to the major overlapping of the DRT peaks between about 30 Hz and 1 kHz ( figure 12(c)), the observed peak transformation could originate from the peak itself or the interaction with a neighbouring peak, making the physical interpretation complex. The other recovery periods showed a similar behavior except after the first exposure to SO 2 . The low-frequency hook was not present during the first exposure period and was more intense when the cell was exposed to 2 ppmv of SO 2 than 1 ppmv (second exposure) but less than when exposed to 0.5 ppmv, as shown in figure 11. This suggests that the inductive hook originated from the exposure-recovery cycling rather than from the concentration of SO 2 itself.
After the recovery period subsequent to the exposure to 0.5 ppmv of SO 2 , EIS measurements were performed in co-electrolysis (65% H 2 O, 25% CO 2 , 10% H 2 ) and steam-electrolysis (90% H 2 O, 10% H 2 ) mode. Both showed an inductive hook at 0.75 A cm -2 but not at 0.5 A cm -2 , indicating that the presence of CO 2 was not or only partially responsible for the low-frequency hook, whereas the current density or the overpotential plays a key role. Figure 11. Electrochemical impedance spectra taken at 0.5 A cm −2 during the exposure to (a) 1 ppmv, (b) 2 ppmv, and (c) 0.5 ppmv of SO2 followed by a recovery period without SO2. The time from the beginning of the test at which the impedance spectra were measured is indicated in between parentheses.
The low-frequency hook may be caused by a two-step reaction that involves an intermediate [43][44][45]. Nechache et al [46] observed a low-frequency hook in the impedance spectra of an SOEC operated in steam-electrolysis and attributed it to the deposition of Si-containing impurities (originating from the glass gas sealing) in the HE that limited the electrochemical reduction and transport of H 2 O. They concluded that when a sufficiently high current density is applied, an additional reaction mechanism, with a characteristic frequency between 1 and 10 mHz and involving adsorption of H 2 O, counterbalances the negative effects of Si-containing deposits. In electrolysis operation, the electronic conductivity of the YSZ electrolyte was also found to induce a low-frequency inductive arc [47]. Due to the clear correlation with the exposure to SO 2 , it was assumed that the low-frequency hook originated from the HE but was not related to the deposition of Si-containing impurities. The low-frequency hook observed could then originate from either the activation of an additional reaction mechanism involving an intermediate [46] or electronic conductivity of the YSZ [47], compensating for the deactivation of electrochemical active sites by sulphur. The presence of a low-frequency hook after recovery can be explained by the presence of residual sulphur on electrochemical active sites. A higher current was then necessary to activate the additional reaction mechanism.

Conclusions
The effect of SO 2 in the reactant gas stream of an SOE, with an Ni-YSZ hydrogen electrode operated in co-electrolysis was investigated. Independently of the current density, exposure to as little as 0.5 ppmv of SO 2 13 showed irreversible effects on the SOEC performances, especially in the high-frequency domain of the impedance spectra, indicating a permanent damage done to the electrochemically active nickel. The general contamination process was found to be similar to an SOFC exposed to H 2 S: a fast initial degradation followed by a steady degradation rate. A good correlation was found between the surface coverage of nickel by sulphur and the relative increase in polarization resistance. Analysis of the DRT of the impedance measurements showed that the charge transfer and gas conversion peaks were particularly affected by the exposure to SO 2 confirming that the interaction between sulphur and the nickel present in the electrode reduces the electrochemical and catalytic reactions.
The OCV values were used to estimate the variation of the reactant composition due to the deactivation of the catalytic activity. The estimated amount of CO, calculated via thermodynamic equilibrium, was used as an indicator as CO originates only from the RWGS reaction. The estimated reduction in CO was only 7.5% when the SOEC was exposed to 5 ppmv of SO 2 , corresponding to a nickel surface coverage of 0.93. This suggests a limited effect of the sulphur on the global catalytic activity under open circuit conditions. The reduction of the CO seemed to stabilize below 10%. In future works, gas chromatography, mass spectrometry, or another gas analysis methods could be used to assess the validity of these estimations.
The initial degradation rate was found to increase with the current density and the SO 2 concentration, the time needed to reach stabilization was found dependant on the SO 2 concentration and potentially to the current density, whereas no clear correlation was found for the height of the initial step and the final degradation rate. At a higher current density, the relative increase of the polarization resistance relative to the SO 2 content was found to be greater. This indicates that the current density has a negative impact on the SOEC tolerance to SO 2 ; however, further investigation is necessary to completely assess the effect of the current density.
The durability test performed at 0.5 A cm −2 showed a drastic increase of the cell voltage tens of hours after the beginning of the exposure leading to a voltage 'runaway' when the SOEC was exposed to 1 ppmv and 2 ppmv of SO 2 . This behavior was characterized by a large increase of the serial resistance. When the SO 2 supply was cut, the serial resistance recovered almost entirely. The origin of this behavior is not clear yet but could be related to a cumulative effect of SO 2 leading to the total deactivation of the TPB close to the electrolyte, thus shifting the electrochemically active area further and virtually increasing the electrolyte thickness. The extent of the affected region close to the electrolyte was estimated to about 2 µm. This 'runaway' behavior was not observed during the exposure to 0.5 ppmv even after 500 h, suggesting that it is triggered by the SO 2 concentration and could be avoided by limiting the concentration to a sub-ppmv level in these operating conditions.
After successive exposure-recovery cycles, the SOEC's HE seemed weakened and a low frequency pseudo-inductive hook was observed. This low-frequency hook expanded when the SOEC was exposed to SO 2 , disappeared during the recovery period, and seemed to be dependant on the cell exposure-recovery cycles history rather than on the SO 2 concentration. The origin of this low-frequency hook is not clear yet but could be related to the activation of an additional electrochemical reaction involving an intermediate species or electronic conductivity of the YSZ phase.
This work is expected to provide guidelines for future investigations on the interaction between sulphur and an SOEC's Ni-YSZ electrode. It will, also, help defining a threshold for future real power-to-X applications, especially regarding the required CO 2 quality. Hypothesis 1 A progressive reduction of the active area due to the deactivation by sulphur of the electrochemically active Ni from the higher (inlet) to lower (outlet) current density regions.
As discussed in section 3.1, a higher current density may have a negative impact on the SOEC tolerance towards SO 2 . Regions with a high current density could thus be deactivated first, leading to a reduction of the active area and a concomitant apparent increase of the area specific serial resistance. The current originally flowing through the now deactivated area would then be redistributed towards the cell outlet, locally increasing the current density. This increase in current density would trigger the deactivation process, gradually expanding the inactive regions from the SOEC inlet toward the outlet and leading to the observed voltage 'runaway' behavior. Assuming that a partial deactivation of the active area is analoguous to an electrode partial delamination, a concurrent and comparable increase of R s and R p is expected [42]. During the first voltage 'runaway' (from about 230 h to 500 h on figure 8), R s and R p increased by a factor of 2.1 and 1.1, respectively. A progressive deactivation of the active area is thus unlikely to be solely responsible for the observed behavior.
Hypothesis 2 A cumulative effect of sulphur blocking the electrochemical active regions located closest to the electrolyte, possibly from the higher (inlet) to lower (outlet) current density regions.
The electrochemically active region would be relocated away from the electrolyte, virtually increasing the electrolyte thickness. The serial resistance would automatically rise due to the increased distance that the oxygen ions have to travel to reach the OE and due to the constriction effect caused by the porosity of the HE compared with the dense electrolyte. The polarization resistance would not be greatly impacted, as the electrochemical reactions would not be affected, but only displaced. Once the SO 2 supply was cut, the blocked electrochemically active regions would be recovered, reducing the R s . The virtual increase of the electrolyte thickness was evaluated considering the ratio (M) between the effective ionic conductivity of the YSZ phase present in the HE (σ eff ) and the bulk conductivity of YSZ (σ 0 ), which was estimated to about 0.15 from [48] assuming a volume fraction of YSZ ε ≈ 0.4 [49]. The effective ionic conductivity of the porous electrode is [48]: The area specific resistance related to the virtual increase of the electrolyte thickness, L, becomes: and the serial resistance: where R el s = L el · σ −1 0 is the area specific resistance of the electrolyte, and L el the electrolyte thickness. The contact resistance and electronic resistance in the electrodes are neglected. The thickness of the electrolyte being on the order of 10 µm, the extent of the deactivated region would be in the order of 2 µm, considering that R s increased by a factor 2 to 3.