Water-Induced Decoupling of Tracer and Electrochemical Oxygen Exchange Kinetics on Mixed Conducting Electrodes

Isotope exchange depth profiling and electrochemical impedance spectroscopy are usually regarded as complementary tools for measuring the surface oxygen exchange activity of mixed conducting oxides, for example used in solid oxide fuel cell (SOFC) electrodes. Only very few studies compared electrical (kq) and tracer (k*) exchange coefficients of solid–gas interfaces measured under identical conditions. The 1:1 correlation between kq and k* often made is thus more an assumption than experimentally verified. In this study it is shown that the measured rates of electrical and tracer exchange of oxygen may strongly differ. Simultaneous acquisition of kq and k* on La0.6Sr0.4FeO3-δ and SrTi0.3Fe0.7O3-δ thin film electrodes revealed that k* > 100 kq in humid oxidizing (16O2 + H218O) and humid reducing (H2 + H218O) atmospheres. These results are explained by fast water adsorption and dissociation on surface oxygen vacancies, forming two surface hydroxyl groups. Hence, interpreting experimentally determined k* values in terms of electrochemically relevant oxygen exchange is not straightforward.

M ixed ionic and electronic conductors (MIECs) are frequently applied on the oxygen side of solid oxide fuel/electrolysis cells (SOFC/SOEC), and some of them, for example La 0.6 Sr 0.4 FeO 3-δ and SrTi 0.3 Fe 0.7 O 3-δ , are even candidates for both SOFC/SOEC fuel side electrodes as well as oxygen electrodes. In such electrochemical cells oxygen is either incorporated from the gas phase into the MIEC or released from the MIEC into the atmosphere. This kind of electrochemical surface reaction can be expressed by Essentially, this is a redox reaction which includes transfer of electrons. The oxygen in the gas O atm may either be present as O 2 molecule or in form of H 2 O, in the latter case hydrogen is produced or consumed in the surface reaction. In equilibrium, both forward and backward reaction of eq 1 take place at the same rate and thus without any net current. An oxygen exchange coefficient can be used in order to quantify the kinetics of such a reaction in dynamic equilibrium. This oxygen exchange coefficient is often determined by electrochemical impedance spectroscopy (EIS). 1−4 Using a small AC voltage, a small perturbation of the equilibrium is introduced and thus a net current can be measured. EIS thus yields an area specific resistance (ASR) of the electrochemical surface reaction (reaction 1) and an electrical oxygen exchange coefficient (k q ) can then be calculated by the relation 5 Symbol c O denotes the oxide ion concentration, z is the charge number (−2 for oxygen ions) and k b , T, and e are Boltzmanns constant, temperature, and elementary charge, respectively.
Another common method for analyzing the oxygen exchange kinetics of MIECs is 18 O isotope exchange and depth profiling (IEDP), yielding a tracer exchange coefficient (k*). Generally, those experiments take place by exposing the MIEC to tracer enriched gas and as long as no voltage is applied during the experiment the net flux of oxygen across the surface remains zero. Upon tracer incorporation 16 The net 18 O incorporation flux and thus also 16 O release flux of this reaction (in oxygen atoms/cm 2 s) depends on the fraction of 18 O (f( 18 O)) in the gas and at the oxide surface by In principle, the two oxygen exchange reactions (reactions 3 and 1) may take place by different atomistic mechanisms. Only if the two reaction parts of 18 O incorporation and 16 O release in eq 3 can be identified as incorporation and release of an oxygen atom by means of reaction 1, the coefficients k* and k q are identical (or at least similar since a small difference due to correlation effects as for diffusion coefficients may still remain). 6 This requires the 18 O incorporation and the 16 O release in eq 3 to be locally unrelated and to include a change of the oxygen reduction state, i.e., an electron transfer. Hence, in principle, the corresponding k factors of tracer and EIS experiments can be rather different as already pointed out in ref 6.
Despite this possibility of strongly different k* and k q values, their equality is assumed in most cases and only very few studies compared the two coefficients obtained upon identical experimental conditions. Similar values of k* and k q were indeed found for La 0.6 Sr 0.4 CoO 3-δ at 400°C 7 in dry O 2 , indicating that the exchange mechanism as well as its rate limiting step is identical in both types of experiments. Some studies showed much faster tracer exchange in humidified (H 2 18 O) atmosphere, e.g., on La 2 Mo 2 O 9 , 8,9 (Ce,Gd)O 2−δ , and (Zr,Y)O 2−δ , 10,11 and BaTiO 3 . 12 In refs 11, 13, and 14 strongly increased tracer exchange kinetics and slightly faster electrochemical kinetics on Pt-YSZ electrodes in humid atmospheres were attributed to fast water dissociation on the oxide surface. Increased gas phase 18 O exchange kinetics were observed on (La,Sr)MnO 3-δ (LSM) and (La,Sr) (Co,Fe)O 3-δ (LSCF) by the introduction of water and related to dissociation of water. 15 From ambient pressure XPS measurements 16 on (Ce,Sm)O 2−δ in H 2 + H 2 O atmosphere, fast water dissociation on surface oxygen vacancies is assumed as well. However, these studies only rely on indirect observations, because the individual rates of electrically driven oxygen exchange and water dissociation on the surface were not quantified and correlated for identical conditions. Accordingly, there remains the question under which conditions k* gives a meaningful measure of the kinetics of the electrochemical oxygen exchange according to reaction 1.
In the experiments presented here, k* and k q are simultaneously determined on one and the same electrode. Impedance spectroscopy was performed during isotope exchange in various 18 O containing atmospheres on different mixed conducting oxides. Hence, quantitative comparison of the exchange coefficients k* and k q became possible. Partly very severe differences with k* > 100k q are found, particularly in humid atmosphere, and a mechanistic explanation is provided.
La 0.6 Sr 0.4 FeO 3-δ and SrTi 0.3 Fe 0.7 O 3-δ thin film model electrodes with an embedded Pt grid as electronic current collector were prepared on yttria stabilized zirconia (YSZ) single crystals with porous LSF counter electrodes. Impedance spectra were acquired during 18 O tracer exchange at 396−418°C, as sketched in Figure 1a. These tracer exchange experiments lasted 10−20 min. Please note that in these experiments the tracer incorporation still corresponds to oxygen tracer exchange without field since the isotope exchange reaction 3 is not significantly influenced by the small AC voltage of 10 mV. EIS is only required to determine k q in addition to k*. Four different atmospheres were used for tracer exchange: (1) 200 The acquired impedance spectra (Figures 1b and 2a) show the onset of a high frequency arc and one or two arcs at intermediate and low frequencies. This is very typical for mixed conducting thin film electrodes and often reported in literature for similar systems, see, e.g., 2−4 and 17−19. In accordance with those earlier studies the following interpretation is made: The high frequency arc is primarily caused by the ohmic drop in the electrolyte (here YSZ) but may also include a contribution from a contact resistance. The arcs at intermediate and low frequencies can be attributed to the electrochemical electrode reactions. For MIEC thin film electrodes three reaction steps have to be distinguished: electrochemical oxygen exchange reaction (cf. eq 1), transport of oxide ions across the thin film and ion transfer between electrode and electrolyte. For materials with moderate to high ionic conductivity, and thus also in our case, ion transport across the thin layer does not significantly contribute to the overall electrode resistance.
Interpretation of the two arcs is made on the basis of the associated capacitances. The large capacitance of several mF/ cm 2 found for all low frequency semicircles can only be explained by the high chemical bulk capacitance 2,19 of the electrode. According to the equivalent circuit model for mixed conducting materials introduced in ref 20, the corresponding parallel resistor has then to be the resistance of the electrochemical surface reaction (in the absence of ionic transport limitation). The smaller intermediate frequency arc, on the other hand, is most probably caused by an interfacial ion transfer barrier, but also the counter electrode may contribute. 2,17,18 As it was done in many earlier studies from different groups, the area-specific resistance of the surface reaction (ASR) was thus determined by fitting the low frequency semicircle to a parallel resistor and a constant phase element 3,4,17,18,21,22 and normalizing the resistance to the active electrode area. Electrical exchange coefficients (k q ) were calculated from the ASR using eq 2, and all values are summarized in Table 1. Interestingly, in all atmospheres and for both materials, the ASR is in the range of 200 Ω·cm 2 (k q ≈ 8 × 10 −9 cm/s), and differences are far less than 1 order of magnitude. This is also in line with the literature data of STF extrapolated to 420°C. 4 Some effects of humidity on the ASR were already reported in the literature 23−25 at higher temperatures and are also found in this study (see Supporting Information).
In Figures 1c,d and 2b, depth profiles of the 18 O isotope fraction f( 18 O) are shown for dry and humid oxidizing as well as humid reducing exchange conditions. All profiles were calculated from raw data by restricting the in-plane integration of the 18 O signals to MIEC regions without the ion blocking platinum current collector beneath. Most of the depth profiles are flat within the MIEC film and some exhibit a distinct step at the MIEC|YSZ interface. Only for LSF in oxidizing conditions (Figure 1c) a clear slope can be observed, indicating low ionic conductivity. This is caused by the small amount of oxygen   ( k* min = 8.3 × 10 −7 cm s −1 ) k*/k q = 1.6 k*/k q = 5.9 k*/k q = 1000 (>240) k*/k q = 860 (>166) a For humid oxidizing and humid reducing conditions, a minimum k* min is also included, representing the k* in the case of the highest possible H 2 18 O fraction of 97.1%. This leads to the k*/k q ratio given in parentheses, e.g., (>76).

The Journal of Physical Chemistry Letters
Letter vacancies of LSF in oxidizing atmosphere, especially at lower temperatures (δ ≈ 0.0001 at 418°C 26 ). For STF in oxidizing and reducing conditions, but also for LSF in reducing conditions, the oxygen vacancy concentration and thus the ionic conductivity is rather high. 26,27 In all humidified atmospheres, the tracer fraction is even close to the atmospheric fraction of H 2 18 O (see below). These profiles were fitted using finite element simulations, considering isotope exchange (k*) and diffusion (D*) coefficients of the MIEC and an interfacial transfer coefficient at the MIEC|YSZ interface (k* int ). Reasons for the slight upward step found for STF in dry air are unknown yet and this was not taken into account. In the simulation, D* of YSZ was fixed to the value calculated from the measured ionic conductivity of YSZ using Nernst−Einsteins relation, i.e., to 2.1 × 10 −8 cm 2 /s at 418°C and 1.0 × 10 −8 cm/s 2 at 396°C. The atmospheric 18 O concentration of 97.1% (provided by the supplier) was assumed in dry oxidizing conditions, whereas a concentration of 80% and 77% H 2 18 O was estimated for humid oxidizing and humid reducing conditions, respectively, by evaluating the 18 O concentration found in the MIEC on top of the ion blocking Pt layer (not shown). In humid conditions, this value of the isotope fraction in the heated gas atmosphere is critical for the k* calculation, due to the high 18 O concentrations in the samples, and thus k* values are prone to errors. Hence, a very conservative estimate of the minimum k* was also made by considering the maximum H 2 18 O level in the atmosphere as defined for 18 O 2 by the provider (97.1%). Those minimal k* values are also given in Table 1. An upper limit of k* cannot be given on the basis of the present data, but any k* value larger than the given ones would be in agreement with all conclusions anyway.
A reasonable D* value of the MIEC could be only determined for LSF in oxidizing atmosphere and the following discussion is limited to k* values. Those are summarized in Table 1. Simulated profiles assuming k* = k q dry are also shown in Figure 1 and Figure 2 (theoretical).
Please note that the surface exchange kinetics of thin film electrodes may vary from sample to sample and also due to degradation. Therefore, the main information is revealed by the ratio of k*/k q obtained on one and the same electrode at exactly the same time. The ratios of k factors in all four atmospheres are summarized in Table 1 and in the bar graph of Figure 3. In dry oxidizing atmosphere, reasonable agreement of k* and k q is found, even though k* is slightly higher than k q . This may have several reasons: The tremendous increase of k* in humid atmospheres (see below) suggests that even traces of residual humidity may enhance k*. The increase of the k*/k q ratio found upon removal of the zeolite filter is an indication for this effect (see "usual oxidizing" column in Table 1 and Figure  3). Moreover, the ASR of the surface reaction (eq 1) may be overestimated since other processes (e.g., in-plane charge transport 28 ) might increase the size of the electrode arc, leading to slightly underestimated k q values.
In humidified oxidizing atmosphere, the electrode ASR slightly increases, which somewhat decreases k q , while the tracer surface exchange coefficient k* becomes 2 to 3 orders of magnitude larger than k q for LSF and STF. This large difference  Table 1 are shown in the bar graph. In addition the minimum ratios of k*/k q in humid reducing and humid oxidizing atmospheres are also included (LSF min and STF min ). The sketches show different oxygen incorporation mechanisms in 18  O, using a neighboring 16 O ion (reaction step iii). In H 2 containing atmospheres also H 2 formation from two surface hydroxyls through a redox reaction step (iv) may follow reaction step ii.

The Journal of Physical Chemistry Letters
Letter strongly indicates existence of different mechanisms of electrically measured oxygen exchange (EIS) and tracer oxygen exchange, in particular since H 2 18 O is the only source of oxygen tracer; H 2 18 O does not directly contribute to the reaction of eq 1 and thus to k q . A decoupling of k* and k q can be expected in the presence of an isotope exchange mechanism that does not change oxygen reduction states. Such a process contributes to k* but not to k q . A likely nonredox reaction is the fast adsorption and dissociation of H 2 18 O on a surface oxygen vacancy, forming two hydroxyl groups (reaction step ii in Figure 3). Here, the oxidation states of O and H remain −2 and +1, respectively. A fast rate of this process is assumed on YSZ 13,14 and ceria surfaces in reducing conditions. 6 The subsequent desorption of H 2 16 O using a neighboring 16 O 2− ion (reaction step iii in Figure 3) can then finalize a tracer exchange according to eq 3 without any electron transfer. Hence, we propose that this nonredox oxygen exchange by water dissociation leads to the strongly enhanced tracer exchange coefficient in H 2 18  O from the solid without electron transfer and thus may lower the measured k* value and may cause an underestimation of the electrochemical oxygen exchange rate of eq 1.
Reducing H 2 + H 2 18 O atmosphere inevitably contains water, so only one ratio of k*/k q is measurable. Also this ratio is very large (>100, see Table 1), and thus we conclude that k* is again not a meaningful measure for describing the electrochemical exchange kinetics by means of eq 1. As discussed for humid oxidizing atmosphere, also here very fast tracer oxygen exchange is possible through water dissociation by reaction steps ii and iii in Figure 3, which are not redox reactions. The much slower electrical oxygen exchange also requires the formation of H 2 and electron transfer to surface OH groups by the redox reaction iv in Figure 3. The very large difference between k* and k q in reducing atmosphere therefore suggests that an electron transfer or the formation of H 2 from surface hydroxyl groups is rate limiting on LSF and STF hydrogen electrodes, as already assumed for ceria. 16 Accordingly, under reducing conditions the very fast tracer exchange still gives valuable mechanistic information on the electrochemical reaction. If water dissociation (reaction ii, Figure 3) was rate limiting in the overall water splitting reaction , the H 2 desorption reaction (reaction step iv, Figure 3) would be faster than the desorption of H 2 16 O (reaction step iii, Figure 3), and the isotope exchange reaction would primarily take place by the reaction involving the electron transfer. Consequently, an agreement of k q and k* would result; k q ≈ k* would thus imply water dissociation to be rate limiting.
Our experiments revealed by direct measurement of reaction rates that this is not at all the case (k* > 100 k q ). Accordingly, we conclude that the rate of water dissociation (hydroxyl formation) and water desorption is much faster than the hydrogen reduction and desorption steps. Therefore, either the charge transfer from the oxide to a surface hydroxyl or the formation of H 2 from two hydroxyl groups is rate-limiting. Methodologically, this means that tracer exchange experiments cannot deliver reliable information on the electrochemical oxygen exchange kinetics when humidity is present in the atmosphere. In such a case, one cannot obtain meaningful estimates of electrical exchange rates from measured k* values, since moving to dry conditions, as for O 2 , is not an option with H 2 being involved.
The absolute k* values in humid reducing and humid oxidizing conditions are very similar, which further indicates that tracer is exchanged through the same mechanism. Interestingly, k q values found in oxidizing and reducing conditions are also very similar. However, this is most likely a coincidence without mechanistic reason behind because strongly different activation energies of the O 2 reduction 4 and water splitting reaction 28 are found, e.g., on STF. It is also important to note that all experimental results were obtained in the temperature range of about 400°C and the behavior may be different at the usually much higher operation temperatures of SOFC electrodes.