Stability limits of tin-based electrocatalyst supports

Tin-based oxides are attractive catalyst support materials considered for application in fuel cells and electrolysers. If properly doped, these oxides are relatively good conductors, assuring that ohmic drop in real applications is minimal. Corrosion of dopants, however, will lead to severe performance deterioration. The present work aims to investigate the potential dependent dissolution rates of indium tin oxide (ITO), fluorine doped tin oxide (FTO) and antimony doped tin oxide (ATO) in the broad potential window ranging from −0.6 to 3.2 VRHE in 0.1 M H2SO4 electrolyte. It is shown that in the cathodic part of the studied potential window all oxides dissolve during the electrochemical reduction of the oxide – cathodic dissolution. In case an oxidation potential is applied to the reduced electrode, metal oxidation is accompanied with additional dissolution – anodic dissolution. Additional dissolution is observed during the oxygen evolution reaction. FTO withstands anodic conditions best, while little and strong dissolution is observed for ATO and ITO, respectively. In discussion of possible corrosion mechanisms, obtained dissolution onset potentials are correlated with existing thermodynamic data.

surface composition 1,3,15,16,[26][27][28][29][30][31][32][33][34][35] . Only in a few works dissolution was actually estimated, e.g. by quartz crystal microbalance [36][37][38] or using radiochemical methods 39 . In these latter studies, however, neutral and/or chloride containing electrolytes, important for photovoltaic applications but not PEM fuel cells and electrolysers, were used. Moreover, only the total but not partial dissolution rates of the respective elements were shown. The current work aims to fill this gap in the existing literature. It presents a detailed study on time-and potential-resolved dissolution of Sn, In, and Sb from FTO, ITO, and ATO in acidic electrolyte in a broad potential window covering the important regions of hydrogen and oxygen evolution. Quantification of the trace amounts of dissolved material and estimation of experimental onset potentials was made possible by using an inductively coupled plasma mass spectrometer (ICP-MS), which was hyphenated to an electrochemical scanning flow cell (SFC). The collected data were used in the description of the tentative dissolution mechanisms based on the available thermodynamics data 40 and in the discussion of the kinetics effects.

Results and Discussion
In order to get a first look on the potential dependence of ITO, ATO, and FTO dissolution a potential program shown as black curve in Fig. 1 was applied to each of the studied electrodes. It consists of a slow potential sweep (5 mV s −1 ) from the open circuit potential (OCP) into the anodic region up to 2.0 V RHE ; a cathodic sweep down to −0.6 V RHE ; and the final sweep from the cathodic vertex potential back to the initial OCP. The current response is coloured in red and the resulting metal concentration in the electrolyte, monitored by SFC-ICP-MS, is depicted in three other colours in the same figure. A standard plot of current vs. potential in Fig. S5 facilitates the interpretation of the electrochemical data. The most significant dissolution is observed during the reduction of the oxides (peak 3) and during the re-oxidation of the formed metal in the following anodic sweep (peak 4). This phenomenon, which holds for all studied materials, can be nicely correlated with the corresponding cathodic and anodic currents. Besides these major processes some dissolution is also observed during the initial contact of the oxides with electrolyte at OCP (peak 1) and anodic polarization (peak 2).
Below we discuss separately dissolution taking places in the peaks 1, 2 and 3, 4. First, however, macroscopic reactions which can be responsible for the observed corrosion are summarized for pH 1 and 1 nM of the equilibrium concentration of the dissolved species in the electrolyte 40 . The reactions are split in three major dissolution processes:   Dissolution during OCP (peak 1) can be explained by chemical dissolution of the oxides as indicated by equations 7-10 with the following tendency for dissolution taken from the equilibrium concentration of the dissolved ions: In 2 O 3 ≫ Sb 2 O 5 ≫ SnO 2 . Note, a direct comparison between ITO and ATO in the experiment is difficult since the portion of In 2 O 3 and Sb 2 O 5 differs (see Table S1). The dissolution signal of Sn (ITO) is significantly higher compared to Sn (FTO/ATO), which indicates destabilization of Sn in an In-Sn-oxide. The reason for this could be intensive dissolution of In, leaving behind prone to dissolution under-coordinated Sn atoms.
Dissolution during anodic polarization (peak 2) cannot be easily assigned to any of the electrochemical reactions presented above 40 . Still, the dissolution increase with potential, presented in Fig. 1, is a sign that the dissolution is rather an electrochemical than chemical process. Only for Sb, dissolution can be assigned using information on thermodynamics of the oxides 40 . Surface states of Sb are expected to be mostly Sb III species while bulk material consists of Sb V 16, 41 . Therefore increased detection of Sb towards anodic potentials could be explained by: To investigate dissolution of ITO and FTO during anodic polarization further, the anodic limit was extended to 3.2 V RHE . Corresponding potential program, current and dissolution signal are presented in Fig. 2, using the same colour code as in Fig. 1. Respective cyclic voltammograms are presented in Fig. S6. The increase in potential causes a steep raise in dissolution of ITO. Additionally, stable during the excursion to 2 V RHE , FTO starts to dissolve at the higher potentials. One can also see that the onset of dissolution coincides with the onset of the OER, clearly indicating a correlation of both processes. As the highest oxidation state is reached for the respective metal cations, oxidation of O -2 to molecular oxygen which is accompanied by metal dissolution can be suggested for the explanation of this effect 39,42 : In order to obtain information on the amounts of dissolved material, the dissolution peaks from Fig. 2a were integrated. Moreover, potentials at which dissolution signal starts to deviate from the blank were indicated (onset potential of dissolution -OPD). Corresponding data is presented in Fig. 2b. Comparing dissolution of Sn from ITO and FTO, it is clear that the former is less stable -the onset of Sn dissolution is lower and the dissolution amount is more than two orders of magnitude higher for ITO. Indium dissolves even more, 30 µg cm −2 correspond to the loss of ca. 45 nm in the film thickness, which is approximately one third of the overall thickness. The amount of dissolved Sn corresponds to ca. 26 nm, adverting an enrichment of SnO 2 during OER, which is in line with findings of Kraft et al. 39 . This might lead to decrease in conductivity explaining the drop in current at 2.7 V RHE . For ITO, assuming that Equations 12 and 13 are operative, 67% of the total current is originating from In and Sn dissolution, while the remaining 33% can be assigned to OER current. In the case of FTO, only 0.013% of the total current arises from dissolution of Sn. Note, in addition to equations 12-14 other dissolution processes cannot be excluded.
Dissolution during reduction and reoxidation (peaks 3, 4) can be discussed on the basis of equations 1-3 and 4-6, respectively. Figure 3 summarizes the amounts of dissolved material in the corresponding peaks and the anodic and cathodic OPDs. Both parameters were extracted from Fig. 1. For FTO and ATO, the cathodic/ reductive OPD deviates from the thermodynamically predicted values of about 400 mV, while for both film and powder ITO this value is about 200 mV (Fig. 3c). It can be suggested that the corresponding oxides in FTO and ATO are kinetically stabilized compared to ITO. It should be noted that the experimental values for OPD and dissolution overpotential (η), shown in Fig. 3c+d, depend on the time scale of experiment and the detection limit of the system (equilibrium concentration close to the electrode is estimated to be on the order of 1 nM) and therefore cannot be used as absolute values; however, comparison between materials is still possible. Hence, a lowered kinetic hindrance for the reductive dissolution of SnO 2 observed in the case of ITO indicates a lower stability of Sn in the oxide with In. A similar trend is seen for anodic/oxidative OPD. Values for η are lower for the anodic dissolution of Sn in ITO (see Fig. 3d).
In terms of dissolved amounts, both In 2 O 3 and Sb 2 O 3 dissolve more during reduction compared to SnO 2 (Fig. 3a). For all materials higher dissolution was observed during reoxidation of the formed metals (Fig. 3b). Especially dissolution of Sn is enhanced, which indicates a lower stability of Sn in an In-Sn alloy. This can be explained by In dissolution triggered destabilization of Sn, since the OPD for In is lower (see Fig. 3b). In other words, started at lower potentials, In dissolution results in formation of high surface area destabilized Sn-rich phases. Similar destabilization is also observed for ATO, clearly seen from Fig. 1, where the onsets of Sn and Sb dissolution in peaks 3 and 4 are well resolved. Here, an additional Sb dissolution peak 3´ appears in the region of anodic Sn dissolution (peak 4). On the other hand, dissolution of Sb triggers Sn depassivation (peak 4′).
Comparing total and dissolution current from Fig. 1, a dissolution efficiency was obtained for each element. For cathodic dissolution, most of the charge is used for reduction of oxides to metals and hydrogen evolution; except for In 2 O 3 , where the dissolution current counts as ~30% of the total current. On the other hand, the current in the oxidative branch is assumed to be predominantly due to dissolution. Corresponding data is summarized in Table 1. During oxidation the ratio of dissolution current to total current was found to exceed 100%, which might be caused by chemical dissolution of monoxides formed by partial reduction of the corresponding dioxides in the preceding reductive scan. Alternatively, electrochemical dissolution of species with other than the used charge number (including zero for neutral species/clusters) can be suggested.

Consequences of dissolution.
The last part of the manuscript is devoted to possible consequences of dissolution for the application of the oxides in electrochemical energy conversion devices. ITO is dissolving rapidly if the potential exceeds the limits of −0.1 V RHE > E > 2.35 V RHE . In between these limits dissolution is lower but still significant. A simple calculation, with assumption that the dissolution rate is constant, reveals that a complete dissolution of a 10 nm film can be reached in just 12 h during OCP. Hence, ITO cannot be recommended for any application involving long stays at OCP in acidic electrolytes. Especially critical is the use of ITO as the catalyst support for hydrogen evolution reaction or electrochemical deposition of base metals.
For ATO, Sb and Sn oxides are relatively stable within 0.36 V RHE < E < 1.1 V RHE and −0.29 V RHE < E < 1.45 V RHE , respectively. Hence, exceeding the stability window of Sb but not Sn, a SnO 2 rich layer can be formed on the surface. This was also recently shown by Cognard et al. 3 who used X-ray photoelectron spectroscopy for analysis of the degraded ATO films in an accelerated stress test (1.0-1.5 V RHE ). Consequences of such surface segregation of SnO 2 should not be underestimated as a core shell structure limits the exchange of electrons 3 . Based on the initial dissolution during 5 min OCP, formation of a 0.2 nm thick SnO 2 film due to preferential dissolution of Sb was estimated. Fabbri et al. 16 also reported on the decrease in Sb amount after potential cycling (0.05-1.6 V RHE ). According to the authors, however, reduction of calcination time and formation of films with a homogenous distribution of Sb lead to formation of a more stable Sb phase. Therefore, synthesis procedure and quality of the resulting films are essential parameters for stability of ATO.
The stability window of FTO ranges from −0.34 V RHE < E < 2.7 V RHE with no indication for any measurable dissolution in between these limits. In contrast to the other studied materials, doping is achieved by replacing oxygen with fluorine atoms instead of cation exchange. In this way less stable oxides are avoided, improving the overall stability of the material. Nevertheless, electrodeposition at very low potentials as well as CO 2 reduction 35 will lead to dissolution of Sn from FTO. Moreover, in chloride containing electrolytes stability of FTO can be an issue 37,38 . Based on our study in ultrapure 0.1 M H 2 SO 4 , FTO shows best stability and is therefore considered as an appropriate candidate for many applications in electrochemistry. At the same time, however, FTO has the poorest intrinsic conductivity, which means that even a small decrease in performance can be detrimental. Further works should be dedicated to the effect of possible F-leaching and related influence on conductivity.

Conclusions
This work provides a detailed investigation of the corrosion of tin-based oxides in acidic electrolyte in the absence of chlorides. All studied oxides, viz. ITO, ATO, and FTO, were unstable at high cathodic and anodic potentials. In terms of relative stability FTO was the most stable followed by ATO. ITO showed poor stability in the whole studied potential region. Dissolution during reduction is most critical for all studied materials questioning the application of these oxides in the cathodic range of potential. This process is assigned to the incomplete reduction of the oxide during the cathodic dissolution. As explanation for the anodic dissolution, surface disturbance due to the oxide lattice decomposition and oxygen evolution reaction was suggested. Based on the stability data, FTO can be considered as the best support.

Methods
Electrochemical tests were performed in argon purged 0.1 M H 2 SO 4 using a scanning flow cell (SFC) connected to an inductively coupled plasma mass spectrometer (ICP-MS) 43,44 . A sketch of the setup is provided in the supporting information (Fig. S1) Table S1. Films and nanopowders of the oxides (ITO, FTO, ATO) were purchased from Sigma Aldrich. As ATO is just available as powder, for the sake of comparison, ITO was as well studied as powder. Before using the working electrodes, the films were rinsed with ethanol and ultrapure water. Nanopowders were supported by a glassy carbon plate. For preparation of an ink, 4 mg of the nanopowders were suspended in 5 ml of 0.001 M NaOH with addition of 20 µL of Nafion-solution (5 w% Sigma Aldrich). After ultrasonic treatment 0.5 µL of the suspension were dropcasted on the glassy carbon plate and let to dry. The dried spots (Ø ~ 1 mm) were rinsed with water and located with the help of a vertical camera attached to the SFC. The measurements on these electrodes were done by placing the spot in the centre of the SFC's opening (Ø = 2 mm) 45 . Current and dissolution were normalized to the electrochemical surface area (ECSA). The geometric surface area was used for films as the roughness factor for sputtered films is assumed to be low. A rough estimation of the ECSA of ATO and ITO powders was done by measuring their capacitance and normalizing the obtained values to the capacitance of the films 46,47 . For the latter, the measured value for ITO was 10 µF cm −2 and the estimated value for ATO was 50 µF cm −2 , in accordance with literature. Also for FTO film measured capacitance was 50 µF cm −2 . A more detailed description can be found in the SI (Fig. S7-S9 and related text).