The electrode model system Pt(O2)|YSZ: Influence of impurities and electrode morphology on cyclic voltammograms

Solid State Ionics 180 (2009) 1019–1033

Contents lists available at ScienceDirect

Solid State Ionics j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / s s i

The electrode model system Pt(O2)|YSZ: Influence of impurities and electrode morphology on cyclic voltammograms E. Mutoro a, B. Luerßen a, S. Günther b, J. Janek a,⁎ a b

Institute of Physical Chemistry, Justus-Liebig University Giessen, Heinrich-Buff-Ring 58, D-35392 Giessen, Germany Department of Chemistry, Ludwig-Maximilian-University München, Butenandtstr. 11, D-81377 München, Germany

a r t i c l e

i n f o

Article history: Received 20 January 2009 Received in revised form 8 April 2009 Accepted 17 April 2009 Keywords: Cyclic voltammetry EPOC Impurity Model system Morphology NEMCA Pt PLD Silicon Silica Spillover YSZ

a b s t r a c t This study is focused on the influence of impurities and electrode morphology on cyclic voltammograms (CV) of one of the most prominent solid state electrode systems, Pt(O2)|YSZ, exemplifying the difficulties in unequivocal interpretation of CV in the solid state in general. By investigating differently prepared electrodes— either by Pt paste or pulsed laser deposition (PLD)—with and without an Si containing additive, the impact of both effects can be separated. For characterisation of the sample SEM, XRD, EDX, Tof-SIMS, XPS, and XPEEM have been used. We demonstrate that the presence of impurities can change the shape of CV and even cause peaks, a fact which has not been considered so far. The processes which theoretically can cause a CV peak in the electrode system Pt(O2)|YSZ are discussed. We reconsider the information unambiguously obtainable from CV studies, and we comment on the controversial questions of the formation of interface Pt oxide and the appearance of spillover oxygen in CV studies. A compact commented survey of literature on CV studies of the system Pt(O2)|YSZ is given. © 2009 Elsevier B.V. All rights reserved.

1. Introduction The high temperature oxygen electrode is perhaps the most important electrode system in solid state electrochemistry, and the Pt(O2)|YSZ (yttria stabilised zirconia) electrode can be considered as the corresponding prototype material system, comparable to the Pt(O2)| H2O electrode in aqueous electrochemistry. Among the electrochemical methods for the study of electrode processes, CV (cyclic voltammetry) plays an important role as one of the simplest and fastest methods without the requirement of expensive equipment. Correspondingly, one finds numerous CV studies of Pt|YSZ electrodes in the literature which, unsatisfactory, are controversial in a number of points [1–36]. Although the first CVs have been published more than 25 years ago, major questions are still not answered, as most of the experiments have been conducted on poorly defined electrodes with respect to composition, morphology, and microstructure. Up to now, two main problems were not always properly addressed: (a) Electrode microstructure: In liquid state studies, the use of single crystal surfaces improved our understanding of electrode

⁎ Corresponding author. E-mail address: [email protected] (J. Janek). 0167-2738/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.ssi.2009.04.012

kinetics considerably and is a standard there, at least after the introduction of flame annealing of electrode surfaces by Clavillier [37]. On solid electrolytes, single crystal electrodes were not available until recently, when Beck et al. demonstrated that virtually single crystalline epitaxial Pt film electrodes on YSZ single crystals can be prepared by PLD (pulsed laser deposition) [38]. The electrochemical characterisation of these electrodes is part of this paper. (b) Impurities: Reproducible electrode studies in the liquid state rely on the use of pure components with a minimum impurity level. So far, the corresponding influence of impurities on the kinetics of (high temperature) electrodes on solid electrolytes has mostly been neglected, especially in the discussion of CV results. This factor is also addressed in the present paper. In particular, we attempt to distinguish between impurity and morphology induced effects on CV. To this end, we investigated model electrodes and varied systematically only one sample parameter, e.g. the tpb length or the impurity content. Knowing that impurities can influence the electrode kinetics, we chose to study the impact of glass, mainly containing silicon, as Si represents an ubiquitous contamination in the Pt|YSZ system [39–41], and it is already known to have an impact on electrochemical processes [40,42] and on the formation of Pt oxides

1020

E. Mutoro et al. / Solid State Ionics 180 (2009) 1019–1033

[43]. In addition, usually glassy sintering aids are added to commercial Pt pastes. In general, the basic question which processes in the system Pt(O2)| YSZ can theoretically cause a CV peak at all is not answered consistently in the literature—e.g. the influence of spillover oxygen which is important for the so called EPOC (electrochemical promotion of catalysis)—and thus we will comment on that. Another important aspect for interpreting CVs is the question whether a Pt oxide forms at the interface Pt|YSZ during anodic polarisation. Even though there is no direct, i.e. spectroscopic, evidence for such an oxide, it has frequently been used for explaining features in CVs [4–6,14,15,25,30,32–34]. To our knowledge no complete survey of literature on CV studies of the system Pt(O2)|YSZ has been published so far. Hence, we prefix a compact commented review, emphasising open issues and disagreements in literature. 2. Commented survey of CV literature 2.1. Investigated systems 2.1.1. Electrodes As working electrodes (WE) mostly sintered (Tmax = 1073 K – 1573 K, sintering time t = 10 min–2 h) Pt pastes [4–8,10,11,13,15,16,18,20–24,35] and screen printed Pt pastes [9,30] were investigated, respectively. In two of these studies the electrodes were polarised during sintering [5,9]. Often point electrodes [1,2,12,14,19,29,31] or sputtered electrodes [10,13,16,32,34] were used, and rarely other preparation techniques like spray pyrolysis [27] or microlithography [13,22] were applied. More recently a cermet electrode has been studied [32–34]. The counter (CE) and reference electrode (RE) were usually made up of Pt paste [1–5,7,10,15,21,22,24,31,36] or prepared by the same method as the WE [10,30,32,34], alternatively Pt foil [14] or gauze was used [21]. Frequently the electrode was made of Ag [11,13,17,18,25,27] or Au [6,20,23] paste. Especially Ag with its low melting point is much more mobile at high temperature compared to Pt which may influence the results. A solid state Me|MeOx reference electrode has been reported by Shoemaker et al. [9]. 2.1.2. Electrolyte Mostly 8 mol% [4,6,7,9,11–14,17–21,23,24,27,31,32] or higher ((9.0 [3], 9.5 [36], 12.3–13.3 [11,13,17]) mol%) yttria-doped zirconia, i.e. fully stabilised cubic YSZ was investigated. Only in one case 4 mol% doped material was used [1]. If there is a comment on that at all in the majority of studies the electrode was polycrystalline [1,2,6,11–13,17,18,20,23,25, 31–34], rarely single crystalline [7,11,13,16,17,36]. 2.2. (Ex situ) electrode characterisation Only in a few cases additional information on the electrodes was presented or referred to: the electrode microstructure as obtained by SEM [4,27,32–34,36], the size of electrode crystallites by XRD [27], the electrode composition by XPS [4], the gas exposed WE surface area determined by an isothermal surface titration technique [4] or an estimation of the tpb length and contact area by reflecting optical microscopy [7]. The majority of the available studies rely only on electrochemical methods. Therefore, correlations between sample characteristics and CV behaviour cannot be formulated in retrospect reviewing the literature. 2.3. Electrode geometry Except in a few studies [15,26,28] a three electrode—WE, CE and RE—configuration was used. The relative size and the geometrical arrangement of the three electrodes influence the electrochemical measurement [44,45], and thus the resulting CV [10,21]. Once not a

point contact WE is used, a symmetrical arrangement of WE and CE on both sides of the solid electrolyte ensures a uniform current density at the WE. The RE should be placed close to the WE [44,45], which was realised in many studies [7,11,17,21,27,36]. However, very coften the RE was placed on the backside next to the CE [3,4,6,8,13,16,20,22–25,30,32–35] or even other arrangements were chosen [5,9,15] so that these results have to be compared with care. 2.4. Influence of sweep rate The sweep rate v influences the shape of the CV in two ways: Firstly, a vertical shift of the forth and backward scan—in the following named vertical hysteresis ΔI (see Fig. 19a)—occurs due to the (un)charging of the electrochemical double layer capacity cd. This hysteresis is proportional to the sweep rate (ΔI ~vcd). Secondly, also the peak currents Ip increase with increasing v and—depending on the reversibility of the system—the peak potentials may shift. In addition, at higher sweep rates usually different peaks in close vicinity cannot be resolved which results in one broad and less well defined peak. A vanishing sweep rate v → 0 Vs− 1 finally corresponds to a steady state measurement [19] and thus at very small sweep rates ((0.05 [12], (0.001–0.1) [31], 0.01 [29]) mVs− 1) no peaks appear. However, slow changes in the electrochemical performance, e.g. due to electrode activation or slow morphology changes, can be observed at low sweep rates. In general, peaks are recorded at higher v values, typically between 1 mVs –1 and 200 mVs− 1 [1–7,9,11,13,14,16,18,20,22,23,25,30,32,34,36]. The effect of the sweep rate on the shape of the CVs or the peak currents was investigated in a number of experiments [4–6,13,16,19,22,30–32,34]. However, conclusions out of these studies are not very clear as Ip was found to be proportional to v [4] as well as proportional to v1/2 [5,6,30] or even a relation in between [22]. By choosing only one sweep rate it has to be assured that the required information is really obtained. In general, a systematic variation of the sweep rate is recommended. 2.5. Influence of temperature Studies have been performed at temperatures between 473 K and 1273 K, either at one fixed temperature or in a temperature range [4– 7,9,10,14–16,21,22,25,35,36]. The effect of T on CV has also been studied in detail [4–6,9,14,15] and the area of the peaks (in oxygen atmosphere) usually passes a maximum and vanishes with increasing T. The reported temperatures differ extremely (Tmax = 700 K–935 K, Tvanish = 845 K– 1170 K) and some authors report peaks even at temperatures as high as 1273 K [1–3]. However, the observed behaviour has been explained by spillover oxygen [4] as well as by the stability of Pt oxide [5] and thus is no evidence for one of these alternatives. Therefore, in this study we chose T = 723 K which results in clear peaks, assures oxygen ion conductivity of YSZ and avoids temperature induced morphology changes of the Pt electrode. In addition, 723 K allows a direct comparison to recent studies [30,32–34]. 2.6. Influence of pressure CVs were measured in air (or p(O2) ~ 2·104 Pa) [1,2,4–6,11– 13,16,19,21,24,29,30–34] or at a reduced oxygen partial pressure (p(O2)min = 1·10− 7 Pa) [3,4,6–10,13,14,16,18–20,22,23,25,30,35], in N2 [5,9,14–16], He [4], or pure O2 [5,7,9,14,19,27]. The influence of p(O2) on CV and the cathodic peak current was investigated in several studies [4–9,13,16,19,20,22,23,30,34]. In general, it was found that the main cathodic peak shifted to more negative potentials as p(O2) was increased [5,7,9,22]. 2.7. Reversal potentials The anodic and cathodic reversal potentials Vr,a, Vr,c strongly influence the shape of the recorded CVs. The maximum voltage range

E. Mutoro et al. / Solid State Ionics 180 (2009) 1019–1033

being studied was between ± 1.5 V [13]. Detailed studies can be found in several papers [5–7,14,25,30,34]. By systematically varying the anodic reverse potential [5–7,14,25] corresponding peaks of a forward and its reverse reaction have been determined. Usually the main cathodic peak shifted to more negative potentials by increasing the anodic reversal potential [5,6,25]. In one study [14] the direction of this peak shift was reversed above 823 K and once also no peak shift at all has been observed [30,34]. 2.8. Anodic behaviour Most authors report that the CVs rather show one [4,16,22,30] or two [5,7,32,34] poorly defined waves or shoulders than a distinct anodic peak. Often also no feature at all was observed in the anodic regime [11,13,16,18,21,25,29,32,34]. Only in a few studies a pronounced peak [3,6,9,10,14,32,34,36] or a peak and a small shoulder [10,14] appeared. Different explanations for this unsymmetry in peaks have been proposed: Vayenas et al. assumed that oxygen excorporation O2− → Otpb competes unsuccessfully with direct chemisorption of gas O from the gas phase to the tpb (O2 → Oad → Otpb) [4,6,23]. Jaccoud et al. proposed that the formation of PtOx species is also represented by an increase of the anodic current density in the CV and not only by a peak [30,32]. They give as reason a limited amount of Pt-O species available for reduction, but an unlimited amount of oxygen in the YSZ for oxidation [32]. 2.9. Cathodic behaviour Most CVs showed at least one cathodic peak [1,2,4–7,10,11,13, 16,18,22,25,30,32,34] and sometimes an additional but less pronounced shoulder which always was observed in close vicinity to the main peak or even interfering [5,8,9,14,16,18,20,22,23,30,32,34,35]. A few measurements only showed a minor shoulder [21] or even no cathodic feature at all [3,16,21,24,27,32,34], which we suggest to attribute to high temperature (1273 K [3], 1123 K [24]) or the low reversal potentials [19,21]. However, sometimes there is no evident reason why no peak appeared [16,27]. Some authors give an explanation for the absence of a cathodic peak: a high value of residual capacitive current [27] or the electrode morphology (sputtered electrode after heat treatment) [32,34] have been proposed. However, these explanations are not valid in general for all systems presented in literature.

1021

polarisation was observed by a number of authors [6,8,18,20,22,23, 30,33–35] usually in the field of EPOC studies. However, frequently two peaks were also recorded during normal CV [5,8,9,14,16,18,20,22, 23,30,32,34,35] which questions whether the occurrence of two peaks relates to the anodic pre-polarisation at all. Of course, as the number of peaks also depends on other parameters, e.g. the sweep rate, one has to take care by comparing different results. However, more than two peaks only appeared after anodic pre-treatment [8,20,22,23,30,33– 35]. All additional peaks were found at more negative potentials and by increasing the holding time th the peak potentials usually shifted in negative direction [6,8,16,20,23,30,33,35], except in one study—which might be only an error in the description [18]. A correlation between the type of electrode and the appearance of additional peak(s) cannot be identified which suggests that not only the preparation method and the electrode morphology respectively but also other parameters, e.g. contaminations, play a role. 2.11. Positioning of the baseline For a quantitative analysis of CVs, especially for the determination of peak areas or peak currents, the position of the baseline is important as it strongly influences the results. Unfortunately, only a few comments on the procedure being used are found in the literature—a fact already noted by Breiter [14]. Usually the baseline is drawn by eye and in only three studies such a baseline is depicted explicitly at all [4,6,14]. Recently Jaccoud et al. proposed a different approach [32]: The steady state current—which is attributed to the redox step O2– = 1/2 O2 + 2 e–—is subtracted from the measured CV resulting in a residual CV in which the baseline is now the axis of abscissae. On the one hand this is a more systematic approach, on the other hand we like to note that the double layer capacity cd causing a vertical hysteresis ΔI of the CV is not considered, and thus the determined peak areas and peak currents are too large by this value. In addition, this method can only be used if the characteristics of the electrode do not change during polarisation, i.e. for the gas tight Pt films in this study it cannot be applied. The uncertainty in positioning the baseline may also be one reason why frequently contrary results have been reported, e.g. concerning the question of reversibility or peak areas. Therefore, we avoided inconclusive statements caused by the problem of the baseline positioning and—if used at all—depict the baselines as dashed lines (Fig. 7a). 2.12. Correlation of features in CV to electrochemical processes

2.10. Cathodic behaviour after anodic polarisation at a constant potential Vh for a holding time th Reviewing the literature (Table 1) the appearance of cathodic peaks in CV after anodic polarisation for a holding time th differs strongly. A system which showed no peak during normal cycling, did also show no peak after anodic polarisation [16]. In some studies there was only one peak visible with and without anodic pretreatment [7,16,25]. In contrast, the appearance of a second peak after anodic

Fig. 4 shows the processes that may occur during electrochemical polarisation influencing the shape of CVs. All authors agree that the reaction O2– = 1/2 O2 + 2 e– along the tpb (Fig. 1a), including several intermediate steps, is described by a Butler-Volmer-type I–Vcharacteristics [6,14,27,29]. Concerning the peaks in CV different interpretations are offered: Often a formation of adsorbed oxygen (Fig. 1b) or a Pt oxide (Fig. 1c) at the interface Pt|YSZ (Pt + x O2- = PtOx + 2x e−) or the electrode surface (Fig. 1d) is discussed. Also the

Table 1 CV after anodic polarisation. Reference [6] [7] [8,20,23,35] [16] [16] [16] [18] [22] [22] [25] [33,34]

Vh|V 0.3 0.3/0.4 0.8 0, 5 0, 8 / 1.2 0.5 0.5 0.5 0.5 0/0.1 / 0.2 0.08–0.12

th/s 5–800 x N 180 0–100/0–600 0–200 20–300 20–200 0–200 0–100 0–50 300 1–1.2·105

T/K, p(O2)/Pa 553, 100 573, 105–1013 673, UHV/5·10− 4 773, N2 773, N2 673, air 773, 10− 6–0.01 773, 5·10− 3 673, 0.01 633, 10% O2 723, 2·104

Electrode (preparation method) Sintered paste Paste, differently sintered Sintered paste Sputtered Sintered pastes/microlithography Sintered pastes Sintered paste Sintered paste Microlithography Sintered paste Screen printed, sintered

Number of cathodic peaks th N 0

th = 0

2 1 2 (indicated), 3 (–4) visible 1 0 1 2 2 2 (–3) (0), 1 2–3 (– 4), depending on th

1 – – 1 0 1 1 (+ small wave) 1 1 1 1–2

1022

E. Mutoro et al. / Solid State Ionics 180 (2009) 1019–1033

Fig. 1. Possible electrode processes indicated by a)–k) at the Pt|YSZ interface influencing CV (details see text).

appearance of spillover oxygen (Fig. 1e) was frequently considered. Schouler et al. interpreted the cathodic peak as a result of surface oxidation of the electrolyte (Fig. 1f) [3]. In addition, results of experiments with slow sweep rates v, e.g. current loops, have been explained by dynamic interface morphologies [12,29] or surface migration of electrode material [31] (Fig. 1g). However, we like to note that these rather slow processes can also be important for the interpretation of fast cycling CV if measurements on one sample take a long time, e.g. for extended anodic polarisation times th. However, there are additional processes having an impact on CV which will be demonstrated by our experiments. A process, which so far has not been discussed to influence CVs, is a change of the electrode morphology due to bubble formation (Fig. 1h). This effect can be caused by an oxygen pressure built-up underneath extended dense electrode areas during anodic polarisation [50]. In addition, the segregation and/or migration of impurities (Fig. 4i), which can also form oxides at the interface (Fig. 1j) or surface (Fig. 1k), has not been considered so far.

or Oδ–, more strongly bonded) during anodic polarisation which they considered as the origin of the EPOC [35]. They also introduced the assignment of a second cathodic peak in CV after anodic polarisation to this spillover oxygen [4]. Henceforward, in several studies [6,8,20,23,30,32–35] an additional cathodic peak has been accounted to this special oxygen (Table 1). By doing so the corresponding peak area is supposed to be proportional to the tpb length and can be determined [4,11,13,16,22] leading to tpb lengths of approximately 235 m per cm2 [4] geometrical electrode area. But there are also observations which are in disagreement with the appearance of a peak caused by this spillover oxygen, e.g. in different studies only one cathodic peak appeared after anodic polarisation [7,16,25] (Table 1). Breiter et al. [14] interpreted the shapes of waves involving a two electron transfer and did not support a Pt-O- [6] species as the more strongly bounded intermediate.

2.12.1. Pt oxides The appearance of Pt oxides—which are usually not exactly specified, and thus denoted as PtOx or simply as oxygen species [18]— has been considered in many studies [4–6,14,15,25,30,32–34]. Mostly the Pt|YSZ interface is assumed to be the location of the oxide formation [18], either in the form of one monolayer (ML) of adsorbed oxygen [14,25,30] or several MLs of (an) oxide(s) [5,25,30]. A surface Pt oxide on the electrode has also been proposed to cause a peak in CV [6,15]—as argued in the discussion we do not agree to this. When peaks have been assigned to Pt oxide formation/reduction [5,7,14,15,25,30] the peak area should be proportional to the contact area Pt|YSZ and has been determined [7] by peak integration as well as the amount of oxide resulting in different values: 5–8 ML [7], 1 ML [15] or first 1 ML and subsequently several (5–20) MLs [30,32–34]. In a few studies more than one peak was explained by Pt oxide formation—either by more than one oxide species [5], the adsorption of oxygen in two layers and the formation of a surface PtOx phase [15] or by a differentiation of the first oxide layer at the interface Pt|YSZ and bulk oxide [30,32– 34]. It is important to note that direct evidence for Pt oxides by any spectroscopic or -microscopic techniques in situ or even ex situ has not been reported.

Mostly, the reported studies contain only minor or no information on the impurity content of the investigated materials. Only in one paper [4] it was claimed that XPS showed no impurities, in another paper the WE was assigned as ‘pure Pt’ without supporting information [3]. Most authors indicate the source of the Pt paste [2– 7,9,11,15–17,19,21,22,25,33,34], sometimes noting that it was an unfluxed one (without glass sintering aids) [2,6,7]. Considering the electrolyte, there is even less information given, only in one report the elemental composition was investigated and Al found to be present [16]. In [10] it is mentioned that the YSZ was polished with SiC paper, which might have caused Si contamination on the surface. We can safely assume that polycrystalline YSZ has a higher concentration of impurities (from sintering aids) than single crystals. Except in the early studies of Isaacs and Olmer [1,2] the effect of electrode surface impurities (Au, Pr, Bi) on the kinetics of the oxygen reaction was not investigated or considered in the interpretation of results. They correlated the electrode composition to the observed currents in CV. To our knowledge there is no consideration on the generation of peaks in CVs of the system Pt(O2)|YSZ by impurities.

2.12.2. (Back)spillover oxygen Vayenas et al. proposed the generation of an oxygen species (referred to as backspillover oxygen) with specific properties (O2– [4]

Already the early studies gave evidence for a strong influence of electrode morphology on CV. Chao et al. recognised that the shape of the CV depended on the sintering conditions of the electrodes [5].

2.13. Influence of impurities

2.14. Influence of morphology

E. Mutoro et al. / Solid State Ionics 180 (2009) 1019–1033

1023

described [4,30]. We note that all equations for a quantitative analysis of CV have been theoretically deduced for liquid systems with twophase electrodes and that we have to be careful in transferring them directly without critical analysis to three phase boundaries in solid state electrochemistry. 3. Experimental As there are different namings in literature—e. g. the method is mostly called CV [3–6,8–11,13,14,16,18,20–25,27,30,32–35], but also programmed CV [30,34], linear potential sweep [12,19,29] or, which is not quite correct, LSV (linear sweep voltammetry) [7,15,17,26,28,31]— we refer to the nomenclature depicted in Fig. 2 [46,47]. 3.1. Sample preparation

Fig. 2. Definitions: VWR = applied voltage, Vr,a/Vr,c = anodic/cathodic reversal potential, Vh = (anodic) holding potential, t = time, th = holding time, v = sweep rate, thicker line: 1st cycle of a “normal” CV.

Kenjo et al. investigated the influence of the microstructure by using differently annealed porous paste electrodes [7]. Thus, point electrodes were often used to avoid the influence of the electrode morphology [1,2,14], however, ignoring the micromorphology of the point contact. Later, a change in the Pt|YSZ interface morphology [12,14], an increased tpb length [21] or contact area [29] after CV measurements was observed resulting in the proposal of a dynamic interface morphology [19]. More recently, Nielson et al. observed Pt surface migration and growth of dendrite like Pt structures using point electrodes [31]. In studies of electrodes with different electrode morphologies due to different preparation methods always changes in the CV shape occurred [16,22,32], and recently the effect of microstructure on CV was investigated in a more systematic study by using differently prepared electrodes (cermet, sputtered and sputtered, heated) [32]. Comparing all publications we cannot identify a simple correlation between the method of electrode preparation, the resulting morphology and the shape of the CV or the appearance of peaks, respectively. One reason might be the interplay of morphology and impurity effects—as already pointed out by Olmer and Isaacs [1]— which cannot be separated without additional information. Probably the occurrence of the different possible processes (Fig. 1) also depends on the electrode morphology, e.g. the ratio between contact area and tpb, and thus a comparison of different types of electrodes is even difficult from a qualitative point of view. In general, the effect of bubble formation and changing morphologies on CV (see “Influence of Morphology on CV”) should be considered if electrodes with extended dense and gas-tight areas are investigated, e.g. if sputtered film electrodes are used [32]. 2.15. Comparison to liquid electrochemisty In many studies similarities to CVs of Pt electrodes in aqueous electrolytes have been pointed out [5,14,15,25,30,32,34] and also main differences between solid state and liquid state CVs have been

We used (111)-orientated YSZ (9.5 mol% Y2O3) single crystals (CrysTec GmbH, thickness 1.5 mm or 0.5 mm, average roughness b0.5 nm; no information on impurity content available). The working electrodes (WEs) were prepared by two different methods: sintering Pt paste (preparation parameters and used abbreviations are given in Table 2; the paste was deposited by a brush as screen printing did not result in a more reproducible morphology) or PLD as described in [38]. For investigating the influence of impurities we used different fluxless Pt pastes (for identification see Table 2), and we added glass particles (an EDX (energy-dispersive X-Ray Spectroscopy) analysis showed only Si and O) into the paste, which is a typical sintering aid. To exclude an impact of a slightly different sintering process, we usually compare experimental results of electrodes which have been sintered identically at the same time. For the examination of morphology effects we compared identically prepared paste electrodes with a different amount of solvent, and partly porous PLD films to a gas tight one. Therefore, we increased the tpb length of the initially dense films by scratching the surface. For CE and RE we always used Pt paste electrodes. Both electrodes have been placed symmetrically and the RE next to the WE. The sample areas varied between 3.8·10− 5 m2 and 2.7·10− 4 m2. For practical reasons, all current densities I·A− 1 refer to the macroscopic area A of the WE—as measured by using an image analysis software— and neither to the actual electrode|electrolyte contact area nor the tpb length. Thus, a comparison of current densities of PLD electrodes to paste electrodes is not straightforward. 3.2. Sample characterisation We used different methods for structural, compositional, and morphological sample characterisation which will be presented at the beginning of the ‘Results and discussion’ section. 3.3. Cyclic voltammetry The CV measurements were carried out in air at ambient pressure using a scanning potentiostat (IMG, Zahner electronics or Solatron, Model 1286). We observed the influence of the temperature on the main cathodic peak which passed a maximum peak current, as frequently observed [4–6,9,14,15], in a temperature range between 673 K and 1023 K. For a more detailed CV study we always chose a

Table 2 Abbreviations for sintered paste electrodes (composed of 4 signs denoting the kind of paste, their thickness, the addition of glass, and the sintering parameters). Paste A B C D

dmc2–6402100, 319/00 Ferro–54021001, 206/04 Ferro–64021015, 229/08 Heraeus–C3605S, 3674548666

Thickness

Glass additive

Sintering parameters

↑

Thick electrode

–

Without

α

↓

Thin electrode/more solvent

+

With

β

With 7 K/min to 673 K, holding for 2 h, with 7 K/min to 1123 K, holding for 30 min With 1.7 K/min to 1273 K, holding for 5 h

1024

E. Mutoro et al. / Solid State Ionics 180 (2009) 1019–1033

specific contaminations the information that there are impurities is more important than a detailed quantification. Even a bare YSZ(111) single crystal without Pt electrodes showed an accumulation of impurities in the outer surface region after annealing (10 h at 1273 K)—which will be most probably also present at the electrolyte|electrode interface—as shown in the Tof-SIMS depth profile of Fig. 3. Si was exclusively located in the outermost surface layers and its concentration rapidly decreased in the depth compared to other impurities like Ca.

Fig. 3. Tof-SIMS depth profile of an annealed YSZ single crystal.

temperature of 723 K. If not the first CV cycle is of interest, we show the “steady state” behaviour which represents a nearly identical shape of CV, independent of the number of cycles. In some cases we show a cycle averaged over a few “steady state” cycles. 4. Results and discussion 4.1. Sample characterisation 4.1.1. Morphology We investigated electrodes before and after CV by scanning electron microscopy (SEM). The sintered paste electrodes were composed of a porous network [36,50], where the porosity and the microstructure depended on the preparation parameters. These electrodes showed no change in morphology after CV. In contrast, the PLD films were virtually dense with a minimum tpb and thus bubble formation took place during anodic polarisation due to their gas tightness. A detailed study of this phenomenon, including SEM images, has been published previously [50] as well as typical SEM images before electrochemical treatment and an estimation of the electrode film thickness [36,38]. Small SEM images are depicted next to the respective CVs.

4.1.4. Interface Pt|YSZ High resolution transmission electron microscopy (HRTEM) analysis [38] including EDX and electron energy-loss spectroscopy (EELS) [71] showed for the PLD model electrode films on YSZ(111) an atomically sharp, semicoherent interface without any diffusion or segregation across the interface Pt(111)|YSZ(111) before electrochemical treatment. In addition, we investigated the Pt|YSZ interface of a dense PLD sample after prolonged (t = 1 h) anodic polarisation in order to clarify whether a Pt oxide formation takes place—as a minimum tpb length and a high oxygen flux towards the electrode should favour an oxidation reaction. Therefore, we used the SPELEEM (spectroscopic photoemission and low energy electron microscope) of the nanospectroscopy beamline at the synchrotron ELETTRA [48,49] applying the X-ray photoelectron emission microscopy (XPEEM) and micro-spot X-ray photoemission spectroscopy (micro-XPS) as operation modes. The Pt film was detached from the YSZ crystal and flipped so that the back of the electrode, which before was the interface to YSZ, became the surface. To preserve an eventually formed thin oxide layer, the sample was kept in air, transported to the beamline at ELETTRA where the electrode was separated, and immediately introduced in the vacuum chamber. All measurements were performed at room temperature without annealing or sputtering prior to the analysis to avoid any modification of the freshly exposed interface surface. The base pressure during the measurements ranged between 1.0·10− 9 mbar and 2.7·10− 10 mbar. Pt oxide is not thermodynamically stable under these conditions, but does not decompose at T b 370 K [51]. In addition, Pt oxide layers are stabilised by a very slow kinetics of oxygen diffusion to the surface [43]. Fig. 4a shows a photoelectron spectrum acquired from the back of the Pt electrode using a photon energy of about 180 eV. The doublet of

4.1.2. Structure The structure and the orientation were determined by X-ray diffraction (XRD). Unlike the paste electrodes with orientations comparable to a statistically orientated powder sample [36,50], the PLD electrodes were (111)-orientated on (111)-orientated YSZ. An XRD pole figure analysis showed that the PLD Pt films were virtually single crystalline after annealing [38]. 4.1.3. Impurities Information on the composition and the impurities was obtained by EDX and time-of-flight secondary ion mass spectrometry (TofSIMS). With respect to the detection limits of EDX—e.g. the detection of small amounts of Si besides Pt is not possible—the Pt|YSZ systems showed no contaminations. Tof-SIMS spectra of the four different pastes sintered on YSZ (A–D, Table 2) revealed different amounts of contaminations in each paste, however due to the absence of calibration standards and matrix changes in the outermost impurity region no quantification can be offered. A GDMS (glow discharge mass spectrometry) analysis provided by Ferro of a commercial Pt paste without sintering aid showed approximately 600 ppm Fe as main impurity and several additional contaminations, e.g. Pd, Rh, Al, Cr in the (50–100) ppm range. The Si content has been declared as less than 50 ppm. Anyway, as our CV studies do not allow a correlation to

Fig. 4. a) Identified and fitted Pt 4f and Si 2p peaks obtained by XPS on the back of a Pt electrode after polarisation (details see text); b) XPEEM images of the back of this film acquiring photoelectrons from the Pt 4f and Si 2p core levels on a 40 µm wide area and the Si 2p/Pt 4f intensity ratio image.

E. Mutoro et al. / Solid State Ionics 180 (2009) 1019–1033

the Pt 4f core level and the Si 2p peak was clearly identified. The binding energy Eb of the Pt 4f peak was determined as Eb = 71.1 eV. This value corresponds to the metallic Pt (Eb = 70.9 eV–71.2 eV [52]), whereas any Pt oxide appears at higher binding energies, e.g. EPtO b = 2 72.3 eV or EPtO = 74.1 eV [51–54]. Close to the Pt Auger transition b (Ekin ≈ 64 eV) the Si 2p peak was identified with a binding energy of about Eb = 101.2 eV. Si 2p binding energies obtained from elemental Si in its reduced state are reported between 99.0 eV and 99.5 eV, whereas binding energies for SiO2 range between 103.3 eV and 103.7 eV [52]. The observed value of Eb = 101.2 eV correlates well with a thin SiOx film. Finster et al. reported a similar Si 2p binding energy of 101.4 eV when producing an about 0.3 nm thick SiOx film on a Si single crystal [55]. They oxidised Si single crystals either in air or applied thermal or anodic oxidation or even oxygen ion implantation. Depending on the thoroughness of the oxidation process, they found that the Si 2p peak shifted gradually up to Eb = 103.9 eV corresponding to a thick SiO2 layer on top of the Si single crystal. Similar energy shifts of the Si 2s peak were observed when oxidising a 3 at.% Si–Pt alloy [56]. In this study the Si 2s peak shifted up to almost 3 eV towards higher Eb when oxidising the Si–Pt alloy sample, and the Pt 4f peak remained unchanged at the binding energy obtained from the clean alloy sample, which was almost equal to the binding energy of pure bulk Pt (Eb ~ 71.1 eV). We detected exclusively platinum, silicon and oxygen by XPS/XPEEM on the Pt electrode. Therefore, we can unequivocally interpret the observed Si 2p and Pt 4f peak energies as results of a thin SiOx film on the back of the Pt electrode, i.e. at the polarised Pt|YSZ interface. The XPS results (Fig. 4a) allow a rough estimation of the Si coverage at the interface. Comparing the relative peak areas of Si 2p and Pt 4f and considering the different cross sections σ (at hv = 180 eV: σSi(2p) = 3.5 Mbarn, σPt(4f) = 2.0 Mbarn [57]) we obtained an element ratio Si/Pt of 0.13. To obtain information on whether the interfacial SiOx is laterally homogeneously distributed, we used XPEEM. Fig. 4b depicts two XPEEM images of the back of the Pt electrode, one tuned to the energy of Si 2p photoelectrons (Eb = 101.2 eV) and the other to Pt 4f ones (Eb = 71.1 eV). Both images display a 40 µm wide area providing almost identical contrast. The black areas represent bubbles which have been formed during anodic polarisation [50]. Since the surface of the flipped electrode was bent at these positions with respect to the unmodified flat surface and the X-rays illuminated the sample under a very grazing angle, the areas of the bubbles appear dark in contrast to the flat interface surface which is well illuminated. As all contrast arised from topography, the extracted spectra did not differ much from the one displayed in Fig. 4a, and the intensity ratio between the Si 2p and the Pt 4f peak did not vary significantly at different places of the imaged area. However, on a several hundred micron length scale

1025

such a variation could be determined resulting in element ratios Si/Pt of 0.10–0.13. These values correspond in a homogeneous alloy to 9–10% Si. If the Si segregation is restricted only to the outermost atomic layer [40] the Si concentration is slightly higher. For an evaluation of this Si concentration increase, we assume that the contribution of every underlying Pt layer is reduced by the factor of exp((aPt/√3)/IMFP)=0.58 (lattice constant aPt of Pt=3.92·10− 10 m, inelastic mean fee path IMFP of electrons with ~100 eV is 4.2·10− 10 m [66]). Thus, the contribution of the Pt volume is (1 − 0.58)− 1 = 2.4 times higher than the contribution of the top Pt layer and the Si coverage in this top layer results in 0.22 ML–0.29 ML. In a specific area a much more pronounced lateral intensity variation was detected, as visible in Fig. 4b. Close to a big bubble and around smaller bubbles an about 12 times increased relative Si 2p/Pt 4f intensity ratio was obtained. Extrapolating this high intensity (Si/Pt ≈ 1.20–1.44) we estimate a local Si coverage at the interface of approximately 1 ML. However, the amount of Si at the interface might be even higher if a part of the Si oxide remained on the YSZ during detaching the Pt film from the electrolyte which is also reasonable with respect to the SIMS measurements (Fig. 3), as we did not detect any other impurities than Si by XPEEM/XPS. 4.2. Cyclic voltammetry 4.2.1. Which processes can cause a peak? In general, a diffusion-controlled redox reaction (R = O + ne–) can cause a peak in (liquid) CV. Fig. 5 shows concentration-distance profiles during a diffusion-controlled (DR = DO) oxidation of R (concentration cR marked as dotted red lines) to O at a planar electrode. The resulting current in CV is proportional to the slope of the c – x-profile for the electroactive species at the electrode surface [47]. Looking at the dotted lines in Fig. 5a (∂c/∂x)x = 0,t passes a maximum with proceeding time t corresponding to a current peak in CV. In Fig. 5a (continuous lines) also the c – x-profile of the product with increasing time t is drawn in for two different cases: firstly (I, continuous blue lines) for an insoluble O in the electrode material and secondly (II, white lines) for a soluble one. In addition, a peak can appear if the product R does not diffuse away from the electrode| electrolyte interface and thus finally blocks the forward reaction (Fig. 5b). There is no consistent opinion in literature about which processes cause a CV peak. Therefore referring to the processes depicted in Fig. 1, we discuss their potential of causing a peak (Table 2) in the following: as demonstrated experimentally [3,16,21,24,27,32,34] the reaction O2– = 1/2 O2 + 2 e– does not cause a peak which shows that no diffusion–limitation occurs. A Pt oxide or any impurity oxide at the Pt| YSZ interface can cause an anodic peak if the product will constrain further reaction (Fig. 5b). A depletion of O2– in the YSZ is, in agreement

Fig. 5. Concentration-distance profiles with time t during a diffusion-controlled oxidation of R to O, I) product insoluble in the electrode (continuous blue lines), II) product soluble in the electrode material (continuous white lines); a) DR = DO, b) DR NN DO → 0. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

1026

E. Mutoro et al. / Solid State Ionics 180 (2009) 1019–1033

with [32], not likely. A cathodic peak should appear because the transport of O2– out of the (Pt) oxide will be diffusion limited, which is reasonable. A surface oxide on top of the electrode (Pt|O2 interface) should not cause a peak, only an indirect influence due to a different diffusion coefficient of Oad on the oxidised electrode material is possible. This is in disagreement with [6,15]. An O2– spillover species (which is strongly adsorbed at the electrode surface and does not desorb) [4] also cannot cause a peak because no redox reaction takes place at all. An Oδ– spillover species (O2- → Oδ–) should not lead to an anodic peak because the reaction O2– → Oad (→ O2) also does not cause one and thus no concentration depletion of O2- occurs in the YSZ. It is also implausible to assume that spillover oxygen cannot diffuse away from the tpb blocking reaction sites. A cathodic peak can only appear if the incorporation of spillover oxygen is diffusion-controlled which is only possible if DO,spillover bb DO,ad holds. Theoretically, spillover oxygen should be generated at any sample having tpb, and thus, experimentally obtained CVs without any features [3,16,21,24,27, 32,34] are strong arguments against peaks due to spillover oxygen. Hence, a spillover species does not necessarily cause a peak in CV— contradicting to several interpretations [4,6,8,20,23,30,32–35]. A surface oxidation of the electrolyte [3] is improbable as YSZ only contains Zr(+ IV) and Y(+ III), thus we do not expect a peak. Morphology changes due to surface migration of electrode material occur on a larger time scale and thus do not influence one single CV scan. Similarly, the bubble formation cannot cause additional peaks, only an indirect influence due to a change in the oxygen activity inside the bubbles compared to the ambient oxygen pressure or the increased tpb length is possible. In addition, it has to be considered that the reactions may involve multistep charge transfer reactions which are more complex and can result in two separated peaks out of one process [58]. 4.2.2. Influence of morphology on CV Our experiments confirmed the well-known influence of the electrode morphology on the shape of CV [1–2,5,7,12,14,19,21,29,31]— e.g. the directly observable difference of paste and PLD electrodes (cp. Figs. 13 and 16). In particular, we investigated two Pt electrodes made of the same paste apart from a different amount of solvent. The more dilute paste and thus thinner D↓-β (cp. Table 2) electrode showed two distinct peaks, an anodic and a cathodic one (dotted red line in Fig. 6). In contrast, investigating the thicker electrode, two hardly visible peaks appeared at a slightly shifted peak potential (continuous blue line in Fig. 6). Small sweep rates v, e.g. v = 25 mVs– 1, even resulted in no peaks at all. As only one sample parameter was varied, we are sure that no other effect than the morphology has to be accounted for the differences in the cycling behaviour. The most noticeable feature of CVs of dense PLD electrodes was a large cathodic peak. Fig. 7a demonstrates the influence of increasing the tpb length by successively scratching the Pt film after every experiment (from I to IV). The current density increased, mainly in the cathodic region. The cathodic peak potential was not shifted and the

Fig. 6. CVs and SEM images of a thick D↑-β electrode (continuous blue line) and a thin D↓-β (dotted red line) electrode. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 7. CV of PLD films: a) dense (dotted blue line) electrode where the tpb was successively enlarged (I–IV) by scratching; baselines drawn in as dashed red lines; b) dense electrode at the beginning of a measurement (dotted green line, “steady state1”) and after several experiments (continuous green line, “steady state2”). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

peak area (assuming the red dashed base lines) decreased. Integration resulted in the charge densities q A– 1 listed in Table 4. In addition, the increase of the tpb length which is not associated with an enlargement of the cathodic peak area (Table 4, see also Figs. 7b and 8b), clearly demonstrates that the process corresponding to this peak—or at least its main component, assuming an unresolved superposition of more than one peak—does not take place at the tpb. Thus, the interface Pt|YSZ is most likely to be the reactive place (for further discussion see also ‘Interpretation of CV of Porous Paste and Dense PLD Electrodes’). The decreasing cathodic peak area in Fig. 7a may be explained by an influence of the electrode morphology on the fraction of oxygen ions transported through YSZ consumed by the different processes depicted in Fig. 1. At a completely dense film electrode the interfacial reaction, causing the main cathodic peak, is preferred; as more and more tpb is generated an increasing part of the total amount of oxygen is excorporated during anodic polarisation there, and thus less interface oxide is formed. Hence, the peak area corresponding to the reduction of this oxide decreases. This may also be a reason why PLD electrodes showed a very large cathodic peak compared to all porous paste electrodes where the main part of the oxygen is pumped through the sample. Comparable to the scratching, but less pronounced, the shape of the CV also changed after several experiments starting with a dense

Fig. 8. a) First negative scan after variation of holding time th at Vh = 0.5 V, b) “steady state” cycles before (“1”) and after (“2”) investigations.

E. Mutoro et al. / Solid State Ionics 180 (2009) 1019–1033

1027

Table 3 Processes causing peaks in CV (details see text). Processes

in Fig. 1

Potential of causing a peak

O2– = 1/2 O2 + 2 e(Pt) Oxide at the Pt|YSZ interface (Pt) Oxide at the Pt|O2 interface Spillover oxygen O2– Oδ– (0 b δ b 2) Surface oxidation of YSZ Morphology changes Bubble formation

a) b), c), j) d), k) e)

No Yes No No Not necessarily No No No

f) g) h)

PLD film due to an increase of the tpb length by bubble formation [50] (Fig. 7b). 4.2.3. The question of spillover oxygen As already discussed (Table 3e), from the theoretical point of view the presence of spillover oxygen does not necessarily cause an additional peak in CV. In order to investigate the impact of spillover oxygen on CVs, measurements were carried out after polarisation at a fixed anodic holding potential Vh for different holding times th. Here, the first cycle is of interested as it is directly influenced by the anodic pre-treatment. Fig. 8a depicts experiments starting the first measurement with a dense PLD film. On the one hand, longer th led to an increased current density and on the other hand, the main cathodic peak was extremely magnified. Additional experiments or further increasing th, finally resulted in decreasing peak areas or even the peak vanished due to the strong increase of the cathodic current density. In general, the behaviour of the dense PLD film electrodes strongly depended on the history of the sample because of the changing electrode morphology during CV [50]. This was also indicated by a changed steady state CV at the beginning of the measurement (Fig. 8b “1”) compared to the “steady state” after several experiments (Fig. 8b “2”). As porous electrodes do not show a polarisation induced morphology change in addition to the spillover process, their investigation is more instructive to get an insight into the influence of spillover oxygen on the shape of CV. However, the shape of the first cycle after anodic pre-treatment differed from the investigated Pt paste samples—a fact which is well known from literature (cp. Table 1). In general, by choosing short holding times th only the first cycle—or more precisely its cathodic region—was influenced (Figs. 9 and 10a), and already in the 2. cycle the “steady state” was reached. The shape of these steady state cycles was independent of th (green dotted line Figs. 9 and 10a). However, long th affected more than one

cycle (Fig. 10b, th = 15.4 h), which has not been noted in literature so far. This shows that during long anodic polarisation a process took place which cannot be reversed completely in the comparably short time—e.g. 20 s for a potential range of 0.5 V and ν = 25 mVs– 1—of cathodic polarisation within one cycle. Increasing th always led to a shift of the cathodic peak potential in negative direction (Figs. 9–11) as frequently observed [6,8,16,20,23,30,33,34]. Usually the area of the primary cathodic peak increased only slightly or reached a maximum value (Figs. 9 and 10), while an additional broad peak, or the vertical hysteresis ΔI of CV (definition see Fig. 19a) respectively, increased continuously (Figs. 10 and 11). Sometimes the overlap of this second process with the primary cathodic peak resulted in one broad wave (Fig. 11a). To sum up, in contrast to a number of previous observations [6,8,18,20,22,23,30,33,34], not necessarily additional peaks arose due to anodic polarisation, but rather already existing ones were amplified (e.g. cp. Fig. 10a). It is also worth mentioning that some electrodes showed an irreversible change of the electrode characteristics, which was indicated by a different “steady state” cycling characteristic at the beginning of the measurement (always labeled “steady state 1”) and of the used sample after several experiments (“steady state 2”), e.g. depicted in Figs. 7b, 8b, 10c. However, there were also samples showing no change in the steady state behaviour within the investigated time period (e.g. Fig. 11b).

Fig. 9. Variation of holding time th of anodic polarisation at Vh = 0.5 V before cycling and magnification of the cathodic peak, B↑-α electrode, dotted green line corresponds to the ‘steady state’ cycling behaviour.

Fig. 11. D↓-β electrode, a) detail of the first negative scan after variation of the holding time th at Vh,a = 1.0 V, b) no change in “steady state” cycles due to polarisation experiments.

Fig. 10. CVs of a A↑-α electrode after different th, a) magnification of the cathodic region, dotted green line: “steady state” cycles; b) th = 15.4 h, 4 cycles, c) “steady states” of two measurements with th = 5 min after different number of previous experiments.

1028

E. Mutoro et al. / Solid State Ionics 180 (2009) 1019–1033

Further conclusions can be drawn by the investigation of glass containing porous paste electrodes (Fig. 12). Fig. 12a shows the first negative scans of CVs after anodic polarisation at Vr,a = 1.0 V for varying th. Even at th = 0 s (green line) two distinct peaks appeared (labelled p1 and p2) which is a specific characteristic of the Si contaminated paste, as it was not observed on different Pt pastes without glass additive. An irreversible change in the electrode characteristic occurred during these measurements, and thus repeating the experiments (Fig. 12b) with the used sample resulted in different shapes of the CVs. In both data sets the area of the (blue marked) p1-peak—apart from the question of the positioning of the baseline—did not increase much or even remained at a constant value, unlike the (red marked) p2-peak which strongly increased. The p2-peak may be a superposition of more components, but substantial evidence for that is missing. However, the area of the enlarged cathodic p2-peak clearly exceeded all values of peak areas of porous paste electrodes without glass. We assume that the second cathodic shoulder, which some “pure” Pt pastes (e.g. Fig. 10a) show, has the same origin. We like to point out similarities between these results of a glass containing electrode and published work [8,20,23,30,32–35] of systems after anodic polarisation. Neophytides and Vayenas et al. [8,20,23,35] showed CVs of porous Pt paste electrodes on polycrystalline YSZ after different th (0 s–600 s) at Vh = 0.8 V (v = 50 mVs– 1), however in contrast to our investigations, under UHV conditions. They even found two peaks at th = 0 s which agrees to our presented results (Fig. 12). With increasing th mainly the second peak grew, and even a third and potentially a fourth one can be identified. The results are somewhat comparable to Fig. 12b in the same time regime. Neophytides et al. interpreted the first peak to a weakly bonded highly reactive O state and the second one to a strongly bonded O (back)spillover state. The additional peaks/shoulders have not been assigned. On the basis of our findings these far-reaching conclusions are not justified. Jaccoud et al. [30,32–34] investigated screen printed (Pt [30] or Pt/ YSZ [33,34]) paste electrodes on polycrystalline YSZ at the same T and p(O2) as in this study, varying the holding time th between 1 min and 200 min [30] or even up to 2000 min [33,34]) at a holding potential Vh of 0.1 V. Starting with th = 1 min they already found two anodic peaks during the first negative scan similar to the measurements of Vayenas et al. [8,20,23,35] and our experiments with a glass containing electrode (green line in Fig. 12a). In their study, due to increasing th the area of the first peak (comparable to the p1-peak in Fig. 12) soon saturated, while the second peak (comparable to p2) grew identically to the observed behaviour here. Finally, this second peak reached a

Fig. 12. First negative scans after variation of holding time th at Vh = 1.0 V investigating a glass containing D + β-electrode; a) first series of experiments, top down th = (0 (green line), 1, 5, 30, 60, 100, 500, 1000, 3000, and 5900) s, b) repeated series of experiments with used sample, top down th = (0 (green line), 1, 5, 10, 60, 300, 600, 900, 1800, 2700, 3600, 5400, 7200, and 18000) s. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 13. Variation of the anodic reversal potential Vr,a, Vr,c = –0.5 V of identically prepared samples using two different pastes; a) A↑-α, b) B↑-α.

maximum value, and after 120 min a third peak became visible increasing up to the maximum investigated th. The emergence of a fourth cathodic peak is suspected. Despite minor experimental differences (vJac. = 10 mVs– 1, vhere = 20 mVs– 1, Vh, Jac = 0.1 V, Vh,here = 1.0 V), the qualitative results are identical in the comparable period of th. Jaccoud et al. assigned the first peak to an ML of Pt oxide at the interface Pt|YSZ, the second one to oxygen backspillover, and the third one to growth of Pt oxide layers at the Pt|YSZ interface towards the bulk Pt during anodic polarisation. Comparing our experimental results of an electrode with (Fig. 12) and without additional Si (e.g. Figs. 9 and 11—for the additional peak in Fig. 10 see also discussion to Fig. 13; a D↑-β-electrode, see Fig. 14, also did not show two peaks after anodic polarisation treatment before CV) only the first one showed more than one peak and a comparable behaviour as described by Jaccoud and Neophytides et al.. In contrast, we conclude that the additional peak(s) are not an intrinsic characteristic of the “pure” Pt| YSZ system and assign the p2-peak—and probably also the broad additional peak in Fig. 10—to the reduction of a Si-containing oxide phase at the Pt|YSZ interface. Thus, experiments which so far have been explained by spillover oxygen [8,20,23,30,33,35] can also be accounted to Si—or other contaminations. In view of the many contradictory results in the literature on this topic, we prefer the explanation of contaminations causing additional CV peaks after anodic polarisation. 4.2.4. Influence of impurities on CV using Si as example Having a closer look to the influence of small amounts of impurities on CV, we consider results in Fig. 13 of two identically prepared porous electrodes from two different pastes (A↑-α, B↑-α) having the same structure (XRD) and a comparable microstructure (SEM), but a different composition of contaminations (ToF-SIMS). Despite their similarity,

Fig. 14. CVs of electrodes with (D↑ + β, continuous blue line) and one without (D↑-β, dotted red line) glass.

E. Mutoro et al. / Solid State Ionics 180 (2009) 1019–1033

both samples show different peak features and cycling behaviour. This explains also the many different results in literature. In addition, the impact of the chosen potential range becomes apparent. At a smaller anodic reversal potential (Vr,a b 0.3 V) the CVs of both pastes showed quantitatively the same features: One clear cathodic peak and one hardly visible shoulder along the positive scan (marked with an arrow). However studying the A↑-α electrode and increasing the potential range (Vr,a N 0.3 V), one or even two additional broad peaks appeared and overlapped with the first one. This evidenced that at least one additional diffusion controlled process took place. As this process is only visible at one electrode, it is not considered to be an intrinsic feature of the system Pt(O2)|YSZ, but rather caused by contaminations. Although the B↑-α electrode showed no additional, comparable peak at higher VWR, an increase of the vertical hysteresis in the same voltage range (-0.2 V b Vr,a b -0.5 V) appeared (Fig. 13a). Thus, it might also be interpreted as the same effect as shown in Fig. 13b, but less pronounced. The systematic variation of the reversal potential also offers the possibility to obtain information on corresponding processes of forward and their reversal reactions: the less pronounced anodic shoulder corresponded to the main cathodic peak, as it started arising exactly when the anodic reversal potential reached the marked shoulder. No additional anodic feature, which is linked to the broad overlapping peak or the vertical hysteresis, was observed. We also like to point out that the shape of CVs with increasing Vr,a was often comparable to a series of CVs with increasing anodical holding time th, e.g. Figs. 13a and 9. To further prove the influence of impurities on the shape of CV, we compared electrodes prepared by the same Pt paste with (D↑ + β) without (D↑ - β) added glass particles (Fig. 14). The intentionally Si contaminated electrode clearly showed additional peaks and a strongly decreased current density. As expected, a change in the chemical nature of the reactive sites—here due to the presence of Si— caused an effect on the electrode kinetics. This is in agreement with Isaacs and Olmer [1,2] who recognised a change in current densities due to a different impurity content. We like to note that the glass-free D↑ - β electrode (red dotted line in Fig. 14) never displayed two cathodic peaks throughout all experiments with different sweep rates, anodic polarisation times before CV and different morphologies. In contrast, the glass additive caused a distinct second cathodic peak (Fig. 15: p1 and p2) which occurred at all sweep rates. In contrast to the glass-free Pt electrode, a change of the “steady state” shape appeared after several experiments

Fig. 15. Variation of sweep rates (inside out: v = (8, 19, 41, 78, 128, 160) mVs– 1) investigating a glass-containing D↑+β electrode; a) first experiments, b) repeated experiments after several CV measurements including anodic polarisation before cycling (Fig. 12a).

1029

Fig. 16. Variation of anodic reversal potential, Vr,cc = – 0.5 V, dense PLD film electrode, b) magnification.

including anodic polarisation (Fig. 15): A shoulder was added at negative potentials (p4) and a corresponding one at anodic potentials (p5)—which may also be adumbrated at high sweep rates in the first series of experiments (Fig. 15a). These changes after long time polarisation on porous electrodes (Figs. 10c, 12, 15) may also correspond to impurity effects as they rather occurred at contaminated samples, where at the same time no morphology change took place. A C-type electrode (cp. Table 1) with glass also showed additional peaks compared to the ‘pure’ C-electrode and as well a reduced current density. The fact that small peaks, which may origin from impurities, can overlap with larger ones is demonstrated in Fig. 16b. A closer look after variation of the anodic reversal potential investigating a dense PLD electrode, revealed that there was an additional very tiny peak (p2), which was usually hidden under the main peak. In general, impurity induced and intrinsic peaks can overlap and one has to be very careful by drawing quantitative conclusions when investigating non-highpurity samples, as done frequently [4,6,7,30,32–34]. Relating a peak at a certain potential exclusively to Si impurities is not straightforward since a number of phases can occur, including complex ones composed of more than one impurity species, different oxidation states, and crystal structures. The corresponding peaks can also overlap leading to one broad feature with a different peak potential. Furthermore, a thin and potentially nanocrystalline interface layer has another chemical potential than an extended bulk phase which also causes a peak shift. Thus, bearing in mind the difficulty of placing the RE perfectly, it is not surprising that the peak potentials vary for different “Pt|YSZ” electrodes. Let us consider Si contaminations and their influence in solid state electrochemical systems a bit deeper: In general, Si impurities appear to be an omnipresent problem since even high purity electrolyte and electrode materials are often Si-contaminated, as specified in the following: 4.2.4.1. YSZ (Si). Annealing YSZ in air at temperatures between 1073 K and 1773 K generated a silicon-rich surface layer [59–61] which was even observed with the purest YSZ available [40]. Our SIMS depth profile (Fig. 3) also confirms the accumulation of impurities in the surface region of an annealed YSZ single crystal used in this study. Brongersma et al. showed that under working conditions of solid oxide fuel cells (SOFCs) the surface of YSZ is covered by a ML of contaminations, followed by an yttria segregated layer [41]. 4.2.4.2. Pt (Si). Si impurities are often found in high purity Pt single crystals [39]. In general, impurities tend to segregate to interfaces and

1030

E. Mutoro et al. / Solid State Ionics 180 (2009) 1019–1033

surfaces, as their accumulation reduces the free enthalpy of the system. Silica segregation in Pt and its influence on the structure [68] and the oxide formation is well known in the field of surface science. Diebold et al. investigated surface segregation of silicon as a trace impurity of a few ppm in Pt(111) [65] and described the formation of an ultrathin surface film of platinum silicide during annealing at T = 750 K–1100 K. Strong Si segregation in Pt containing a small amount of Si, either as intentionally alloyed material or as an impurity, has been observed when oxidising Pt samples at elevated temperature [56,66]. The origin of SiOx—either from the electrolyte or the electrode material—found at the interface Pt|YSZ in our experiments cannot be specified. Its presence appears to be unavoidable, which is a fact that has not been properly addressed in literature up to now. 4.2.4.3. Pt|YSZ (Si), temperature treatment. It is also well known that Si-rich phases occur at the interface between YSZ and an electrode at elevated temperatures [62–64]. Diffusion bonding of Pt foils to a YSZ single crystal at high temperatures in air led to pockets of Ca–Y–Zrsilicate glass at the interface whereas at lower oxygen activities and the same temperature a clean interface was obtained [63]. 4.2.4.4. Pt|YSZ (Si), electrochemical treatment. Our previous scanning photoemission microscopy (SPEM) measurements also suggested a segregation of Si to the surface of a dense PLD Pt electrode on a YSZ single crystal during electrochemical polarisation [67]. The current XPEEM measurements showed that Si is not homogeneously distributed along the YSZ|Pt interface after anodic polarisation, but rather locally enriched around bubbles and thus along the tpb. This site specific Si enhancement evidences an electrochemically driven effect, not only caused by a segregation due to the temperature treatment. The impurity accumulation at the tpb is in good agreement with a study of the YSZ|Ni interface, where a so-called “rim ridge” of contaminations appeared along the tpb of a Ni point electrode at elevated temperatures [64]. But there is less information on the influence of Si and other impurities on electrode kinetics. Ridder et al. described a decreasing oxygen exchange due to accumulation of impurities in the outermost surface layer of YSZ [40]. An increased electrode and electrolyte resistivity due to the presence of an SiO2 has been investigated by Bae et al. for the case of oxide electrodes on ceriabased electrolyte [42]. Summing up, even small unintentionally existing concentrations of impurities of less than 1 ML [40] can cause pronounced effects on the shape of CVs by influencing the current density, peaks, shoulders or the hysteresis. Considering that peak areas are particularly sensitive to the presence and amount of impurities, we suggest that some (contradicting) results reported in literature can be ascribed to them. 4.3. Reproducibility problems and contradicting results in literature as a result of the interaction of impurity and morphology effects As demonstrated, peaks in CV are affected by the electrode morphology or the presence of impurities and the interaction of both complicates the interpretation. Without sufficient characterisation samples which appear at the first view similar can provide contradicting CV data. Thus, it is also not straightforward to compare different results in literature due to samples and experiments in which more than one influencing parameter has been varied or not specified exactly. In general, there are no comments on the sequence of experiments or the electrochemical and thermal history of the samples in literature. However, these factors can influence the shape of CV due to time dependent changes in electrode morphology or the impurity distribution and properties, as demonstrated (Figs. 7b, 8, 10c, 12, 15). Variations in the results deduced from peak areas—e.g. the number of interfacial oxide layers [7,15,30,32–34]—can arise from overlapping

impurity peaks on the one hand, or faradaic processes which are not diffusion controlled and thus do not cause a peak, on the other hand. A detailed interpretation of CVs which attributes specific reactions and microscopic processes to peaks and shoulders [4,6,7,30,32–34] should also include changes in the morphology and impurity distribution within the system. Due to the complexity, such a model has not yet been proposed. 4.4. The question of Pt-oxides In general, different Pt oxides have been described [70] and thermodynamically their formation is possible under the experimental conditions [51]. A Pt oxide at the Pt|O2 interface is very likely to exist, if the oxygen surface coverage on the Pt(111) electrode exceeds 0.75 ML [72]. However, as this oxide does not influence CV (see ‘Which processes can cause a peak?’), we focus on Pt oxide at the interface Pt|YSZ. The oxidation state and structure of such a Pt oxide will depend on the sample properties (atomic structure at the interface, ratio of interface contact area Pt|YSZ to tpb) and the electrochemical treatment (oxygen activity, applied voltage). After formation of a first ML at the interface, subsequent growth of the oxide into the bulk electrode requires a Pt-O site rearrangement [30] or a diffusion of oxygen to the reacting interface. If an oxide has been formed during e.g. several hours of anodic polarisation, this layer is also stabilised by a very slow kinetics of oxygen diffusion to the surface [43] and thus it is very unlikely that the oxide is completely reduced during the comparably short time of the first negative scan. This argument explains why after long anodic polarisation more than one CV cycle is needed to reach a “steady state” (Fig. 10b). The formation of any other impurity induced oxide phase will lead to the same effect. The existence of Pt oxide at the Pt|YSZ interface during electrochemical polarisation has already been proposed years ago [5,7] and is assumed in recent studies [30,32–34]. But to our knowledge, there is still no definite direct in situ or even ex situ experimental proof for the existence of Pt oxide(s) in the system Pt|YSZ during electrochemical polarisation. The difficult accessibility of the interface with experimental methods is a challenge, especially the lack of suitable in situ methods, the small amount of a potential oxide, as well as questions about its ex situ stability and the influence of impurities or the interface morphology on the oxide formation. We did not detect any oxidised form of Pt by ex situ investigations with XRD on any sample (bearing in mind that only film thicknesses above approximately 1 µm can be detected), and also no Pt oxide was found by XPS/XPEEM investigations of the interface of a Pt|YSZ system with a dense film electrode after pronounced anodic polarisation. On the one hand, it is well known that the presence of impurities like Si in Pt leads to their preferential oxidation, often even preventing the formation of any Pt oxide [43]. On the other hand, the detection of Si impurities on a Pt surface is not easy using common surface sensitive techniques [65]. This task is even more difficult at the buried Pt|YSZ interface, thus Si contaminations were often not observed, although present. Hence, considering that a completely Si-free Pt|YSZ system cannot be prepared, it is rather unlikely that a Pt oxide formation takes place— contradicting to the conclusions in several papers [4–6,14,15,25,30,32– 34]. Another evidence for the formation of an impurity oxide, instead of a Pt oxide, is the main cathodic peak in CVs which sometimes appeared at temperatures even as high as 1273 K [1–3], where the existence of a Pt oxide is unlikely. Of course, we cannot exclude that on the YSZ crystal at first a layer of an impurity oxide is formed and in addition on top a region with oxidised Pt (PtOX). As stated in Table 3, the presence of interface Pt oxides should cause at least one cathodic peak in CV. Thus, the interpretation of CV is hampered by the ambiguity that peaks can appear either in high purity systems due to Pt oxide or in commonly prepared systems due to impurity oxide formation and reduction.

E. Mutoro et al. / Solid State Ionics 180 (2009) 1019–1033

1031

Fig. 17. Variation of cathodic reversal potential, Vr,a = 1.5 V, D↑ + β-electrode after several experiments including two series of anodic polarisation before CV.

4.4.1. Vertical hysteresis of CV and double layer capacitance cd The experimentally observed vertical hysteresis ΔI of CV resulted for some samples in very high cd values compared to literature (deduced by impedance spectroscopy, e.g. 20 µFcm– 2–70 µFcm– 2 for porous paste electrodes [69]), e.g. to 2000 µFcm– 2 (Fig. 7). This indicates that only a part of the experimentally observed vertical hysteresis can be assigned to the double layer (un)charging. Jaccould et al. [30] supposed that additional faradaic processes,, not specified by the authors, take place leading to an increased current density. By changing the reversal potentials, we found that some CVs show an additional vertical hysteresis instead of a peak (e.g. Fig. 13a) above a certain Vr which cannot be attributed to an increase of cd. Some of the investigated porous electrodes exhibit a more distinct anodic peak (Figs. 17 and 18), and here the phenomenon clearly becomes apparent. The variation of Vr,c revealed two different processes under the anodic peak (marked continuous blue and dotted red), although no single scan showed separately resolved peaks. As an anodic peak (continuous blue lines) already appears before Vr,c reached the main cathodic peak (cp. Fig. 17) the vertical hysteresis of the CV in the range between -0.4 V b VWR b 0.1 V cannot be attributed only to charging of the double layer capacity cd (nonfaradaic process). In addition, at least one faradaic process, involving charges transferred across the electrode|electrolyte interface [46], also takes place (cp. Fig. 19a, grey shaded area). This first (continuous blue) anodic peak shifted to more positive potentials while Vr,c was decreased. As soon as Vr,c reached the cathodic peak an additional second anodic peak (dotted red lines) at less positive potential arose, and the primary (continuous blue) anodic peak was only visible in an enhanced vertical hysteresis in the anodic region. A further decrease of Vr,c caused again in a peak shift to slightly more positive potentials.

Fig. 18. Variation of cathodic reversal potential Vr,c; magnification of the anodic peak, D↓-β electrode.

Fig. 19. Interpretation of CV of a) porous paste electrodes, b) dense PLD film electrodes (details see text).

As these anodic peak components are more pronounced at the electrode with glass additive (Fig. 17) compared to the one without (Fig. 18), this may be related to Si—which is, of course, not a clear indication. 4.4.2. Interpretation of CV of Porous Paste and Dense PLD Electrodes We describe the CV as a result of the superposition of different processes (I–VII) and the resulting shape as the enclosing line (continuous red line in Fig. 19). Two simplified and partly different interpretations of the experimental results are offered in the following, one for porous paste electrodes and one for dense PLD film electrodes. 4.4.2.1. Process I (dotted green line in Fig. 19). The Butler-Volmer-type behaviour is assigned to the reaction O2– = 1/2 O2 + 2 e–, which is consistent with several reports [6,14,27,30,32–34]. The oxygen exchange takes place at the tpb and thus an increase of its length should result in an increased current density. This applies for the scratching of a dense film electrode (Fig. 7a) as well as for the formation of bubbles (Figs. 7b and 8). Another experimental result, which confirms this attribution, is the reduced current density of samples containing glass (Fig. 14), as Si reduces the rate of the oxygen reduction/oxidation reaction [40]. In addition, this reaction is the process which should always occur during electrochemical polarisation, and there are some samples [3,16,21,24,27,32,34] which showed no peaks at all.

1032

E. Mutoro et al. / Solid State Ionics 180 (2009) 1019–1033

4.4.2.2. Process II (dashed blue line in Fig. 19). A part of the observed vertical hysteresis ΔI can be assigned to the (un)charging of cd [46].

Table 4 Cathodic peak areas of the dense and scratched PLD electrode (Fig. 7a), q denotes charge and A the WE area.

4.4.2.3. Process III (indicated by a grey shaded area). The residual part of the vertical hysteresis corresponds to yet unknown faradaic processes which are not diffusion controlled and therefore take place without causing a peak in CV (Figs. 13a, 17, 18).

Electrode

Named in Fig. 7a

q A– 1/C m– 2

Dense PLD film Scratched PLD film (increasing tpb length)

I II III IV

44.7 39.7 25.5 15.2

4.4.2.4. Process IV: (dotted line with less dots in Fig. 19b). The dense films showed virtually no increase of the current density in the cathodic region (Figs. 7, 8, 16) as the oxygen incorporation is only possible at a very small tpb length. In contrast, excorporation of oxygen during anodic polarisation can occur due to bubble formation and thus the current density proceeds as expected. It also agrees with the development of Butler-Volmer-type characteristics by increasing the tpb length (Fig. 7a).

Pt|YSZ. In contrast, an increased contact area at a porous paste electrode caused a reduction of the peak area and the current density (Fig. 6) which suggests a process located at the tpb. This is qualitatively in agreement with the observations by Vayenas et al. [4] leading to the conclusion that the tpb length for porous electrodes can be estimated out of the anodic peak area.

4.4.2.5. Process V and VI (hatched and double hatched areas in Fig. 19a). For the porous paste electrodes we found different overlapping peaks. One anodic peak (grey shaded) corresponds to a process which only contributes to the vertical hysteresis (Figs. 17 and 18). An additional anodic peak at smaller potentials (double hatched, process VI) was assigned to the first cathodic peak in a negative scan (Figs. 17 and 18). Samples with glass showed at least (Fig. 15a) an additional second cathodic peak at more negative potentials and also an increased anodic peak, which may be different to the two already mentioned anodic features. Some of the pastes without glass additives also had a less pronounced second anodic feature (Fig. 13b) coinciding with contaminations in the paste. Thus, we assign the hatched peak couple (process V) to the formation and reduction of a Si-containing phase at the Pt|YSZ interface. In addition, the glass containing electrode system showed an additional peak couple after several experiments (Fig. 15b), which is not drawn in here. As stated before, based on the CV measurements and an ex situ sample characterisation, we cannot attribute these peaks to specific microscopic reactions. 4.4.2.6. Process VII (hatched areas in Fig. 19b). The corresponding electrochemical process to the main cathodic peak does not take place at the tpb as an increase of the tpb length did not increase the peak area (Fig. 7). Thus, for the dense PLD electrodes it is not possible to determine the tpb length by using the cathodic peak area as proposed in [4] and done in [11,13,16,22]. Excluding the tpb as location of reaction the 3 two phase boundaries have to be considered: the interface YSZ|O2 is improbable (e.g. in PEEM measurements under even lower p(O2) and same voltages no surface reduction of YSZ was visible), reactions at the interface Pt|O2 do not lead to a peak in CV (Table 3), thus the reaction takes place at the interface YSZ|Pt. A reduction of an oxide there is most likely, if one excludes the reduction of Zr or Y. However, the observed (0.22–0.29) ML of a Si oxide film cannot explain the pronounced cathodic CV peak of the dense PLD electrodes (e.g. Fig. 7a) as the peak area corresponds to a higher amount of charge (Table 4). This can indicate that most of the oxide film remained on the YSZ substrate when the film was detached and only a minor part could be detected by XPS. Most likely, the big cathodic peak is a convolution of more than one component (e.g. p2 in Fig. 16b) where only one corresponds to the reduction of interfacial SiOx. It is possible that we observe a combination of reduction of an impurity oxide and a Pt oxide. The variation of Vr,a showed that the small broad anodic peak corresponded to the main cathodic peak, both representing a peak couple (Fig. 17a). We cannot rule out that the cathodic peak observed in CV on the dense PLD films and the porous ones is caused by the same electrochemical process. In fact, the experimental results rather suggest different processes: as argued, the underlying process of the cathodic peak of PLD electrodes takes place at the two phase boundary

4.4.3. Unambiguous conclusions from CV measurements and their generality CV investigations without other simultaneous in situ characterisation methods do not allow clear conclusions on the microscopic processes during electrode polarisation. However, by comparing selected experiments it is possible to determine the sites where the electrode reaction takes place. Once the baseline is placed the peak area offers information on the amount of charge. Knowing the underlying reaction and having information on overlapping peaks quantitative statements can then be easily given. As both requirements are uncertain, we consider the charge related to a peak as a reasonable estimate (Table 4). Corresponding peaks for a certain reaction and its reverse can be determined unambiguously due to a variation of the reversal potentials―if no adverse overlap of peaks occurs. However, it is not clear whether explanations, holding for one specific electrode, can be transferred in general to all other Pt|YSZ samples due to different microscopic processes which are favoured at different electrode morphologies or impurity compositions. Serious is the fact that two peaks in CVs of two different samples arising at the same peak potential do not necessarily have to belong to the same reaction. In addition, investigating two different samples, a specific reaction, e.g. a redox process at the interface Pt|YSZ, does not implicitly need to appear exactly at the same peak potential. For example, Pt oxide formation and reduction may occur at slightly different peak potentials depending on the interfacial microstructure, the orientations of the materials or the presence of impurity atoms. Thus, we believe that the shape of a measured CV results from a complex combination of many processes (Fig. 1) including the influence of other elements than Pt, O, Zr and Y. The direct influence of the single processes on CV is not clarified yet. Depending on the specific electrode characteristics―which are not always known exactly―each of these processes is pronounced differently or can even be suppressed completely. 5. Conclusions We demonstrated that the presence of impurities within the system Pt|YSZ as well as its electrode morphology influence the shape of CV, either by changing the current density or by the (dis) appearance of peaks. Usually the effects of impurity and morphology interact, thus the interpretation of features in CV is not straightforward. If two samples differ in more than one influencing parameter―which usually holds for a different preparation method or even different starting materials―a comparison of the experimental results is difficult. As many processes (Fig. 1) can change the shape of a CV, a complete sample characterisation with special focus on contaminations is recommended. In addition, our results indicate that the fraction of oxygen ions, migrating through the YSZ, which is consumed by one of the different possible microscopic processes

E. Mutoro et al. / Solid State Ionics 180 (2009) 1019–1033

depends on the sample characteristics. Hence, the CVs of different Pt electrodes might need a different interpretation. With respect to reproducible results the time dependent change in the electrode characteristics plays an important role. All above mentioned facts are reasons for partly contradicting results in literature. Our experiments aimed at separating morphology and impurity effects and clarifying some open questions: • The vertical hysteresis of CV is composed of two parts: the (un) charging of the electrochemical double layer capacity cd and faradaic processes which do not cause a peak or shoulder. • The formation of Pt oxides at the Pt|YSZ interface is rather unlikely or at least not the only oxide which is formed. On the one hand―in agreement with the literature―we did not find any direct experimental proof of the existence of Pt oxide(s) after electrochemical treatment by XRD and XPS/XPEEM. On the other hand, we detected a nm thick region on the YSZ surface contaminated with different impurities (Ca, Si, Al, Na, etc.) by SIMS. As the phenomenon of impurity accumulation at interfaces in solid state electrochemical systems [62–64] and the fact that small amounts of Si or other impurities on pure Pt can prevent a Pt oxide formation [43] are well known, we do not support the existence of a Pt oxide at the Pt|YSZ interface―even though we cannot definitely exclude it. • Moreover, contaminations like Si seem to be an intrinsic problem when using YSZ as solid electrolyte and Pt as electrode material [59–61], and thus the influence on electrochemical measurements like CV is significant. Si accumulation at the tpb appears to be an electrochemically driven effect and not only caused by segregation at high temperature, because it is accumulated at the tpb. The intentionally glass-contaminated Pt paste electrode showed a CV behaviour very similar to the one described by Vayenas et al. and Jaccoud et al. after anodic polarisation. They explained one of the peaks with (back) spillover oxygen [8,20,23,30,32–35]. As the glass-free electrode did not show a comparable behaviour, we assign these peaks to the presence of contaminations rather than to spillover oxygen. Finally, we conclude that our current understanding of electrode kinetics in the solid state is still in its infancy. Unlike in liquid state electrochemistry, the role of a surface science based approach to electrodes has yet not been generally accepted. We believe that major advances in solid state electrochemistry will only be possible if future studies concentrate on high quality (model-type) sample preparations (like e.g. [38]) and combining electrochemical measurements with spectroscopic and microscopic in situ techniques (like e.g. [73,74]). Acknowledgements We thank A. Locatelli for his support at the Syncrotrone ELETTRA (Trieste, Italy) and S.O. Steinmueller for support in SIMS measurements. E. Mutoro is grateful to the FCI (Fonds der Chemischen Industrie) for a scholarship. Financial support by DFG within the project Ja648/10-1 is also acknowledged. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14]

H.S. Isaacs, L.J. Olmer, J. Electrochem. Soc. 129 (1982) 436. L.J. Olmer, H.S. Isaacs, J. Electrochem. Soc. 129 (1982) 345. E.J.L. Schouler, M. Kleitz, J. Electrochem. Soc. 134 (1987) 1045. C.G. Vayenas, A. Ioannides, S. Bebelis, J. Catal. 129 (1991) 67. T. Chao, K.J. Walsh, P.S. Fedkiw, Solid State Ionics 47 (1991) 277. J. Yi, A. Kaloyannis, C.G. Vayenas, Electrochim. Acta 38 (1993) 2533. T. Kenjo, Y. Yamakoshi, K. Wada, J. Electrochem. Soc. 140 (1993) 2151; T. Kenjo, H. Matsuyama, Y. Yamakoshi, Denki Kagaku 61 (1993) 758. S.G. Neophytides, C.G. Vayenas, J. Phys. Chem. 99 (1995) 17063. E.L. Shoemaker, M.C. Vogt, F.J. Dudek, Solid State Ionics 92 (1996) 285. G. Fafilek, K. Leeb, M.W. Breiter, Solid State Ionics 86-88 (1996) 1415. S. Völkening, Diploma Thesis, University Hannover (1996). L. Bay, T. Jacobsen, Solid State Ionics 93 (1997) 201. J. Poppe, Diploma Thesis, University Hannover (1997). M.W. Breiter, K. Leeb, G. Fafilek, J. Electroanal. Chem. 434 (1997) 129.

1033

[15] B. Zachau-Christiansen, T. Jacobsen, L. Bay, S. Skaarup, Solid State Ionics 113-115 (1998) 271. [16] M. Rohnke, Diploma Thesis, University Hannover (1999). [17] J. Poppe, S. Völkening, A. Schaak, E. Schütz, J. Janek, R. Imbihl, PCCP 1 (1999) 5241. [18] R. Imbihl, J. Janek, Solid State Ionics 136-137 (2000) 699. [19] T. Jacobsen, B. Zachau-Christiansen, L. Bay, M.J. Jorgensen, Electrochim. Acta 46 (2001) 1019. [20] C.G. Vayenas, S. Bebelis, C. Pliangos, S. Brosda, D. Tsiplakides, Electrochemical Activation of Catalysis, Springer, 2002. [21] A.C. Co, S.J. Xia, V.I. Birss, Proc. of Symposia on Ionic and Mixed Conducting Ceramics, Fall Meeting of the Electrochemical Society, Inc., San Francisco, 2002. [22] B. Luerßen, PhD Thesis, University Giessen (2003). [23] C.G. Vayenas, S. Brosda, C. Pliangos, J. Catal. 216 (2003) 487. [24] X.J. Chen, S.H. Chan, K.A. Khor, Solid State Ionics 164 (2003) 17. [25] L. Bultel, C. Roux, E. Siebert, P. Vernoux, F. Gaillard, Solid State Ionics 166 (2004) 183. [26] W.-F. Zhang, P. Schmidt-Zhang, U. Guth, Solid State Ionics 169 (2004) 121. [27] O.S.N. Sum, E. Djurado, T. Pagnier, N. Rosman, C. Roux, E. Siebert, Solid State Ionics 176 (2005) 2599. [28] P. Schmidt-Zhang, W. Zhang, F. Gerlach, K. Ahlborn, U. Guth, Sensors and Actuators B 108 (2005) 797. [29] T. Jacobsen, K.V. Hansen, E. Skou, J. Electrochem. Soc. 152 (2005) A2203. [30] A. Jaccoud, G. Foti, C. Comninellis, Electrochim. Acta 51 (2006) 1264. [31] J. Nielsen, T. Jacobsen, Solid State Ionics 178 (2007) 1001. [32] A. Jaccoud, G. Foti, R. Wüthrich, H. Jotterand, C. Comninellis, Top. Catal. 44 (2007) 409. [33] A. Jaccoud, C. Falgairette, G. Foti, C. Comninellis, Electrochim. Acta 52 (2007) 7927. [34] A. Jaccoud, PhD Thesis, EPFL (2007). [35] C.G. Vayenas, C.G. Koutsodontis, The Journal of Chemical Physics 128 (2008) 182506. [36] E. Mutoro, B. Luerßen, S. Günther, J. Janek, Solid State Ionics 179 (2008) 1414. [37] J. Clavilier, R. Faure, G. Guinet, R. Durand, J. Electroanal. Chem. 107 (1980) 205. [38] G. Beck, H. Fischer, E. Mutoro, V. Srot, K. Petrikowski, E. Tchernychova, M. Wuttig, M. Rühle, B. Luerßen, J. Janek, Solid State Ionics 178 (2007) 327. [39] R.G. Musket, W. McLean, C.A. Colmenares, D.M. Makowiecki, J.W. Siekhaus, App. Surf. Sci. 10 (1982) 143. [40] M. de Ridder, A.G.J. Vervoort, R.G. van Welzenis, H.H. Brongersma, Solid State Ionics 156 (2003) 255. [41] H.H. Brongersma, M. de Ridder, A. Gildenpfennig, M.M. Viitanen, J. Europ. Ceram. Soc. 23 (2003) 2761. [42] J.M. Bae, C.C.H. Steele, Solid State Ionics 106 (1998) 247. [43] M. Salmerón, L. Brewer, G.A. Somorjai, Surf. Sci. 112 (1981) 207. [44] M. Nagata, Y. Itoh, H. Iwahara, Solid State Ionics 67 (1994) 215. [45] J. Winkler, P.V. Hendriksen, N. Bonanos, M. Mogensen, J. Electrochem. Soc. 145 (1998) 1184. [46] J. Bard, F. Faulkner, Electrochemical Methods: Fundamentals and Applications, John Wiley & Sons Inc., 2001. [47] P.T. Kissinger, W.R. Heineman, Laboratory Techniques in Electroanalytical Chemistry, 2. Edition Marcel Deccer, 1996. [48] A. Locatelli, A. Bianco, D. Cocco, S. Cherifi, S. Heun, M. Marsi, M. Pasqualetto, E. Bauer, J. Phys. IV France 104 (2003) 99. [49] A. Locatelli, L. Aballe, T.O. Mentes, M. Kiskinova, E. Bauer, Surf. Interface Anal. 38 (2006) 1554. [50] E. Mutoro, S. Günther, B. Luerßen, I. Valov, J. Janek, Solid State Ionics 179 (2008) 1835. [51] M. Peuckert, H.P. Bonzel, Surf. Sci. 145 (1984) 239. [52] J.F. Moulder, W.F. Stickle, P.E. Sobol, K.D. Bomben, in: J. Chastain (Ed.), Handbook of X-ray Photoelectron Spectroscopy, Perkin Elmer Corp., Eden Prarie, 1992. [53] C.R. Parkinson, M. Walker, C.F. McConville, Surf. Sci. 545 (2003) 19. [54] C. Puglia, A. Nilsson, B. Hernnas, O. Karis, P. Bennich, N. Martensson, Surf. Sci. 342 (1995) 119. [55] J. Finster, D. Schulze, F. Bechstedt, A. Meisel, Surf. Sci. 152/153 (1985) 1063. [56] P. Legare, L. Hilaire, G. Maire, Surf. Sci 141 (1984) 605. [57] J.J. Yeh, I. Lindau, Atomic Data Nuclear Data Tables 32 (1985) 1. [58] D.S. Polcyn, I. Shain, Anal. Chem. 38 (1966) 371. [59] A. Bernaski, K. Kowalski, A. Sadowski, J. Phys. Chem. Solids 63 (2002) 233. [60] A.E. Hughes, S.P.S. Badwal, Solid State Ionics 10/41 (1990) 312. [61] A.E. Hughes, S.P.S. Badwal, Solid State Ionics 46 (1991) 265. [62] D. Gozzi, M. Tomellini, A. Bianconi, M. Fanfoni, J. Electroanal. Chem. 198 (1986) 53. [63] M. Backhaus-Ricoult, M.F. Trichet, Solid State Ionics 150 (2002) 143. [64] M. Mogensen, K.V. Jensen, M.J. Jorgensen, S. Primdahl, Solid State Ionics 150 (2002) 123. [65] U. Diebold, L. Zhang, J.F. Anderson, P. Mrozek, J. Vac. Sci. Technol. A 14 (1996) 1679. [66] S. Tanuma, C.J. Powell, D.R. Penn, Surf. Interf. Anal. 11 (1988) 577. [67] H. Fischer, PhD Thesis, University Giessen, in preparation. [68] W. Thale, U. Korte, G. Meyer-Ehmsen, Surf. Sci. Lett. 276 (1992) L19. [69] A. Mitterdorfer, L.G. Gauckler, Solid State Ionics 117 (1999) 203. [70] T.B. Massalski, Binary Alloy Phase Diagrams, 2 EdASM International, 1991. [71] V. Srot, M. Watanabe, C. Scheu, P.A. van Ake, E. Mutoro, J. Janek, M. Rühle, EMC 2008, Materials Science, vol. 2, Springer-Verlag, Berlin Heidelberg, 2008, p. 369. [72] S.P. Devarajan, J.A. Hinojosa Jr., J.F. Weaver, Surf. Sci. 602 (2008) 3116. [73] B. Luerßen, J. Janek, R. Imbihl, Solid State Ionics 141–142 (2001) 701. [74] J. Janek, B. Luerßen, E. Mutoro, H. Fischer, Sebastian Günther, Topics in Catalysis 44 (2007) 399.