The surface evolution of La0.4Sr0.6TiO3+δ anode in solid oxide fuel cells: Understanding the sulfur-promotion effect

(LST) anode in SOFCs fuelled by H 2 S-containing methane. This new method enabled us to clearly observe the surface amorphization and sul ﬁ dation of LST under simulated SOFC operating conditions. The ~1 nm-thick two dimensional sulfur-adsorbed layer was on top of the disordered LST, containing e S, e SH and elemental sulfur species. In SOFC test, such “ poisoned ” anode showed increased performances: a ten-fold enhanced power density enhancement (up to 30 mW cm (cid:3) 2 ) and an improved open circuit voltage (from 0.69 V to 1.17 V). Moreover, its anodic polarization resistance in methane decreased to 21.53 U cm 2 , a difference of 95% compared with the sulfur-free anode. Control experiments con ﬁ rmed that once the adsorbed sulfur species were removed electrochemically, methane conversion slowed down simultaneously till full stop.


Introduction
Fuel diversity is, supposedly, one of the most advantageous features of solid oxide fuel cells (SOFCs) [1e3]. The high operating temperature grants direct conversion of economically viable fossil fuels into power cleanly and efficiently. But in practice, theses impurity-containing feedstocks, such as natural gas and biogas with H 2 S contaminant, deactivate many anode catalysts significantly, causing severe coking and/or sulfur poisoning problems [4e11]. Indeed, optimizing the fuel cell operating parameters is a fairly practical way of minimizing the poisoning effects [12]. Replacing the state-of-the-art Ni cermet with alternatives or engineering the microstructures of the anode seems open more opportunities to alleviate such deactivation. For instance, bimetallic alloys [13], simple oxides [14,15], and complex oxides such as doped perovskite oxides were all among the most promising candidates [7,16e19]. Good SOFC performances were also reported when the anode structure was engineered: a typical measure is to add uniformly-dispersed active nanoparticles on perovskite backbones to boost the electrochemical reaction rate [20e23].
While most of the current approaches are unable to fully suppress the sulfur poisoning effect, the extensive studies also revealed an interesting phenomenon: as a notorious catalyst poison, H 2 S, in many cases, can be a catalyst as well (namely, promoter), leading to the catalytic performance increase [24e27]. These seemingly contradictory results reflect the lack of understanding of the interaction between the anode catalyst and H 2 S under SOFC operating environment. However, on one hand, the conventional ex-situ approach analyzing regular "cooled" sample was inappropriate, especially when studying the surface chemistry: the reactions of the surface absorbants and the catalyst can happen readily during the slow cooling process [12]. On the other hand, hydrogen sulfide is highly toxic and corrosive, it is inherently an uneasy task to characterize the anode materials exposed to H 2 S at elevated temperature using either in-situ or operando techniques. Since the pioneering work by Walker's and Liu's groups using in-situ Raman approach [28,29], operando optical and XANES (fluorescencedetection mode) methods were also developed recently to investigate the interactions between adsorbed sulfur species and the catalyst [30,31]. Nonetheless, these methods were surface nonsensitive and used only for the bulk properties examination of the catalyst.
In this work, using a well-tailored quenching test, we managed to characterize the surface structure evolution of La 0.4 Sr 0.6 TiO 3þd (LST) anode in SOFCs fuelled by H 2 S-containing methane. This facile method helps us explore the sulfur promoting effects of methane conversion in intermediate-temperature SOFCs.

Materials and instrumentation
La 0.4 Sr 0.6 TiO 3þd (LST) samples, for both powder and disc, were prepared via a classic solid-state reaction. The disc sample was densified at 1400 C. In the annealing-quenching test, the samples were placed in a sealed quartz tube that was heated up to 850 C under vacuum (<10 À4 mbar). After annealing in H 2 , 0.5% H 2 SeH 2 and 0.5% H 2 SeCH 4 for 30 min, the tube, along with the samples, was immediately quenched to room temperature using ice water within 10 s. Therefore, the possible parasitic reactions during the cooling process were eliminated; both the surface composition and morphology were maintained to room temperature for further measurements. Note that the composition of the simulated fuel atmosphere may vary compared with that in a biased SOFC as a result of electrochemical oxidation. X-ray diffraction (XRD) patterns were obtained on a Miniflex II diffractometer using Ni-filtered Cu Ka radiation. The X-ray tube was operated at 30 kV and 15 mA. Temperature programmed reduction (TPR) measurement in hydrogen was done on a Thermo TPDRO 1100 instrument using a thermal conductivity detector (TCD). 50 mg of LST was placed in the reactor and TPR was measured under 5% H 2 in a N 2 flow (30 mL min À1 ), heating at 5 C min À1 from 30 C to 1000 C. Transmission electron microscopy (TEM) analyses were performed using a JEOL-JEM 2100 microscope. Scanning transmission electron microscopy (STEM) analyses were carried out using a FEI's Tecnai Osiris microscope equipped with high angle annular dark field (HAADF) STEM detector. ChemiSTEM X-ray detection technology was employed to characterize the sample chemistry. The microstructure of SOFC was investigated using a field emission scanning electron microscope (FE-SEM, JOEL 6301F).
Time-of-flight secondary ion mass spectrometer (ToF-SIMS) and X-ray photoelectron spectroscopy (XPS) measurements were performed using an ION-TOF ToF-SIMS spectrometer and a VG ESCA-LAB 250 XPS spectrometer. In ToF-SIMS measurement, a pulsed 25 keV Bi þ primary ion source delivering a 1.3 pA current was used over a 100 Â 100 mm 2 area. Depth profiling was performed by sputtering with a 0.5 keV Cs þ ion gun giving a 25 mA target current over a 300 Â 300 mm 2 raster area. Data acquisition and post processing analyses were performed using the ION-Spec software. For the XPS measurements, a monochromated Al Ka (hv ¼ 1486.6 eV) source was used as the power, the base pressure in the analytical chamber was maintained at 10 À9 mbar. The obtained spectra were analyzed using the Thermo Avantage software calibrated to the C 1s binding energy of 285.0 eV. For curve fitting and deconvolution, a Shirley-type background subtraction and a Gaussian-Lorentzian peak shape were applied. HSC Chemistry 5.0 was used to perform thermodynamic calculations.

The SOFC test rig
The detailed fabrication procedures of membrane electrode assembly (MEA) have been described in our previous work [23]. Briefly, two porous yttria-stabilized zirconia (YSZ) matrix electrodes were initially applied on the opposite sides of a commercial YSZ disc (300 mm thick, 25 mm in diameter) via screen printing and 1200 C sintering. Both electrodes were~35 mm thick. To fabricate the cathode, a barrier layer composed of 10 wt% Gd 0.2 Ce 0.8 O 2 (GDC) was initially infiltrated into one porous YSZ electrode matrix before subsequent impregnation of 16 wt% La 0.6 Sr 0.4 Co 0.2 Fe 0.8 O 3 (LSCF). 16 wt% of LST was infiltrated into the other porous YSZ matrix as the anode. The final 900 C calcination was 2 h to acquire the desired crystal phases of the electro-catalysts.
The MEA was fixed and sealed between a pair of coaxial alumina tubes accommodated in the furnace. Hydrogenation and H 2 S treatment were performed at 850 C in certified 5000 ppm H 2 S þ H 2 (Praxair), respectively. The obtained black LST and sulfuradsorbed black LST were denoted as B-LST and S/B-LST, respectively. The exhaust of the anode tube was guided to a mass spectrometer (Pfeiffer Thermostar GSD 301) for analysis. Electrochemical analyses were performed using a Solartron SI 1287 electrochemical interface equipped with a SI 1260 impedance/gainphase analyzer. The AC impedance spectra were obtained at open circuit with a potential perturbation of ±10 mV and a frequency range from 10 5 Hz to 0.01 Hz.

Results and discussion
3.1. Surface-disordered LST under SOFC operating conditions Fig. 1a schematically shows the steps of the annealing-quenching test. A schematic drawing of the test setup was shown in Scheme S1 in the Supporting Information (SI). Remarkably, after 0.5 h treatment in either H 2 or 0.5%H 2 SeH 2 , the quenched sample became completely dark. In contrast, the regular-cooled LST (cooling rate was ca. 3 C min À1 ) was greyish (see Fig. 1b). XRD patterns in Fig. 1c indicate no bulk phase transformations occurred after such hydrogenation. There was no chemical reaction between hydrogen and LST either, as indicated by the stable hydrogen signal in the H 2 -TPR analysis up to 1000 C (see Fig. S1 in the SI). Therefore, we postulated that the surface reaction caused such discoloration.
Accordingly, we examined the electronic structures of B-LST using XPS (see Fig. S2 in the SI). The binding energies of Ti 2p3/2 and Ti 2p1/2 for LST were observed at 458.1 eV and 464.5 eV, respectively, well correlated with Ti 4þ of SrTiO 3 . Upon hydrogenation-quenching in H 2 or 0.5%H 2 SeH 2 , however, new oxidation state (Ti 3þ ) or peak shift was not observed. This confuted the formation of defects such as oxygen vacancies or Ti 3þ through Hence, excess oxygen in B-LST was sustained which was in good agreement with the literature [35]. This conclusion was also supported by the O 1s spectra in Fig. S2b. The spectrum of LST could be deconvoluted into two peaks at 530.1 eV and 532.1 eV, attributed to TieO and eOH species, respectively. However, after hydrogenation, no eOH species could be found and the new peak (531.6 eV) could be assigned to oxygen excess [36].
The surface reaction associated with the discoloration was finally identified in TEM analysis. The bright field (BF) image in Fig. 1d demonstrates the microstructure of B-LST particles deposited on YSZ support. Their crystallographic phase was determined by the selected area electron diffraction (SAED) pattern in Fig. 1e  ([011] zone axis). The high resolution TEM (HR-TEM) micrograph in Fig. 1f shows the well-resolved lattice features of B-LST. A disordered surface layer,~3 nm thick, could be seen near the edge of the grain as indicated by the arrows. The diffractogram inset also shows a diffuse halo, implying the presence of amorphous phase. This observation was in accordance with the defects image of doped SrTiO 3 materials in the literature [34], and was also found in hydrogenated SrTiO 3 and TiO 2 [37]. Conversely, we did not observe such apparent amorphous shell on grey LST (reduced and regularcooled, see Fig. S3), presumably due to the re-crystallization during the slow cooling process. Thus, we trust that under SOFC anode operating condition (elevated temperature & reducing atmosphere), LST would loss the surface crystallinity, resulting in the discoloration and the formation of a surface disordered layer.

Interactions of sulfur adsorbate and disordered layer
We then examined how black LST interacted with sulfur after sulfidation in 0.5%H 2 SeH 2 . Bulk LST has been proven stable against H 2 S attack widely [7], but their interaction on the surface was not fully clear. We firstly investigated this surface using TEM and energy-dispersive X-ray spectroscopy (EDS). Noticeably, O, Sr, La and Ti elements were uniformly distributed across the entire region without segregations (see Fig. S4). A tiny peak at 2.3 eV that matched up with sulfur was ascribed to the adsorbed sulfur species. Indeed, such sulfur-substrate interaction can lead to the formation of a 2-D sulfide layer on the surface [12]. For instance, the following reaction might occur in this study: However, no binding energy shifts of Sr 3d 5/2 (132.7 eV) and La 3d 5/2 (839.2 eV) were observed after either reduction or sulfidation, suggesting that the surface sulfide layer did not form in this experiment (see Fig. 2, cf. the XPS spectra in Fig. S2). The additional peak of La was assigned to a shake-up satellite caused by the transfer of a valence-band electron to an empty 4f orbital, which occurred with the transition of a 3d electron to a continuum state [38]. This satellite structure was common in trivalent lanthanum compounds including oxides [39]. Fig. 2c compares S 2p core level XPS spectra of B-LST and S/B-LST. After sulfidation, a broad sulfur peak appeared for S/B-LST. The peak at 160.8 eV could be assigned to the chemisorbed eS (S 2À ) bounded with metal centers (it is also possible that the surface oxygen anions were substituted by sulfur anions). At 162.1 eV, the peak was correlated with eSH species (HS À ). The one at 163.7 eV was related to elemental sulfur atoms that were physically adsorbed on black LST. Another detectable peak located at 167.4 eV was attributable to the formation of sulfite (SO 3 2À ) that might be produced due to the slight oxidation of reduced sulfur species during the sample transfer from the reactor to the XPS analysis chamber [40].
The thickness of these sulfur species was determined using the equations developed Carlson and ourselves [41e43]: where I is the enclosed area under the peak, obtained from Fig. 2c and Fig. S2a, d is the photoemission cross section of the emitted photoelectrons, l is the inelastic mean free path (IMFP), D is the atomic concentration, T is the transmission factor determined by the particular machine, K is an apparatus constant and d is the thickness of surface deposits. The calculated thickness d was ca.
1.03 nm and did not vary significantly as a function of sulfidation time (see the parameter values in Table S1). We thus concluded that under SOFC operating condition in H 2 S-containing methane, the LST surface was eventually a mixture of eS, eSH and elemental S. Fig. 3 shows the ToF-SIMS depth profile of S/B-LST sample. A signal of SH À was seen, supporting the XPS result in Fig. 2c. Its intensity reached a maximum initially, implying that the surface layer was rich in sulfur species. As the sputtering proceeded, signals of SH À started to drop whereas those of LaO À and TiO À fragments concurrently began to increase till surpassing that of SH À . This meant that the probing ion-beam has reached the LST substrate. Accordingly, we drew the schematic of the near-surface structure of sulfur-adsorbed B-LST in Fig. 3 (c.f. the TEM and XPS analyses).

Electrocatalytic promoting effects of adsorbed sulfur species
After determining the surface structure of LST in H 2 S-containing fuels, we explored both the structure-activity correlation and the sulfur-promoting effects in SOFC. The operating temperature was selected at 800 C to minimize the intrinsic activity of LST [44,45]. We found that the open circuit voltage (OCV) in methane was merely 0.69 V, far below the Nernst potential, suggesting that methane molecule was hardly activated (see Fig. 4a). It was then not surprising that the maximum power density (MPD) was 3 mW cm À2 . The performance was slightly better in H 2 , showing a MPD of 17 mW cm À2 . However, when SOFC was fed by 0.5% H 2 SeCH 4 (note that H 2 S was to create and retain surface sulfur species), the OCV was greatly promoted to 1.17 V from 0.69 V whereas the MPD rose to 30 mW cm À2 , even exceeding that in H 2 .
In the corresponding impedance measurement, the ohmic resistances of all cells were very close; the semi-circles in the high frequency region (>100 Hz) in all plots were roughly identical, suggesting that the contribution from the GDC-LSCF cathode might predominate (see the overlaid spectra in Fig. S5) [46]. In the lowfrequency region till 0.01 Hz, the spectra differed remarkably, shedding some light on the varied anode electro-catalytic behaviors. B-LST showed a huge charge transfer resistance (R p ) in methane (443.67 U cm 2 ), reflecting its poor activity towards methane oxidation. In comparison, R p value was only 21.53 U cm 2 for S/B-LST anode. This dramatic difference strongly supported the catalytic promoting effects of surface sulfur species. Though less active than conventional transition metal catalysts, they significantly outperformed several perovskite oxide anodes regarding methane oxidation [44].
Intuitively, one would imagine that the possible oxidation of H 2 S may contribute to the enhanced electrochemical performances. In our previous work, we showed such low concentration of H 2 S was unable to power a SOFC by itself [26]. In this work, we monitored the concentration variation of H 2 S as a function of overpotential in Fig. 5. Interestingly, consumption of H 2 S only occurred when cell was deeply biased [12]. Thus, when the voltage was higher than~0.7 V, the surface sulfur was sustained on S/B-LST. Otherwise, H 2 S and SO 2 signals would vary. The generated electrical current in SOFC mainly came from the conversion of methane as CO 2 and H 2 O signals intensified. The presence of COS and CS 2 was attributable to the reaction intermediates of the sulfur promoted methane activation (see below).
Conversely, when the voltage was below 0.7 V, the consumption of H 2 S as well as the production of SO 2 was initiated. Simultaneously, the oxidation of methane was substantially decelerated and the signals of H 2 O, CO 2 , CS 2 and COS were stabilized or even decreased. Because the current was still increasing in step with the degree of polarization (see Fig. 4a), H 2 S became the only contributing fuel. Under this condition (high pO 2 ), the surface sulfur species, particularly those in the vicinity of TPB, must be oxidized according to the thermodynamic prediction [47,48]. We therefore maintained that the actual anode became B-LST. This transformation completed at~0.3 V and was accompanied by the drastically drop of cell performances (see Figs. 4a and 5): in the I-V curve, a huge overpotenital was recorded in the same voltage region (0.3~0 V), and the electric current started to decrease even when the bias was still increasing! This scenario contradicted the commonly observed abrupt current increase when cell was deeply biased, in which the oxygen spillover mechanism caused expansions of the actual triple-phase boundary (see Fig. S6 for details) [49]. Thus, we asserted that the surface sulfur species are indeed the promoters of methane electrochemical conversion. Evidences were also seen during fuel switching where adsorption-desorption of sulfur species was associated (see Fig. S7).

Hypothetic pathway of S-promoted methane conversion
The sulfur species bond with Ti metal center, creating active sites for methane activation [50,51]. This weak interaction (sulfur adsorption on titanate was reversible, but on Ni, the chemisorption was not fully reversible [52]) actually facilitates the CeH bonds activation of methane as revealed by recent studies [25]. The physisorbed elemental sulfur, which was more weakly bound on S/ B-LST, can be even more active according to this theory. The activation procedure was presumably similar to that proposed by Froment and Marks [25,53], the adsorbed sulfur species was denoted as S (ad) in the following equations: Activation: CS 2 would be evolved from methane activation process. The origins of COS, H 2 O and CO 2 might be complicated, a possible route using a Langmuir-Hinshelwood model can be: Oxidation: HeS (ad) þ OeS (ad) / HOeS (ad) CeS (ad) þ OeS (ad) / OC-S (ad) þ S (ad) OC-S (ad) / COS[ OCeS (ad) þ OeS (ad) / CO 2 [ þ 2S (ad) (17) Note that carbon oxidation reaction was usually the ratedetermined step in methane oxidation [53]. This promoting effect became invalid when the S (ad) species were oxidized if the cell voltage dropped to below~0.7 V producing SO 2 . Over-oxidation: The promoted activation of methane by sulfur was also supported by the thermodynamic calculation data shown in Fig. 6. In comparison with the thermal activation of methane yielding eCH 3 and eH, the sulfur-assisted process decreased the energy barrier significantly from 3.07 eV to 0.43 eV at 800 C, assuming the formation of eSCH 3 and eSH species.

Conclusions
In this study, we understood that LST anode underwent significant surface changes during SOFC operation in H 2 S-containing methane. A surface shell of amorphous disordered layer was observed, causing discolorations of the catalyst. The surface sulfur layer was~1 nm thick, consisting of eS, eSH and elemental sulfur species. The combination of the disordered LST shell and sulfur layer promoted the electrochemical oxidation of methane in SOFC.